Report No. 38665-ET Ethiopia Urban Labour Markets in Ethiopia Challenges and Prospects (In Two Volumes) Volume II: Background Papers March 2007 Poverty Reduction and Economic Management Unit Africa Region) Document of the World Bank CURRENCYAND EQUIVALENTUNITS (as o fNovember 2006) Currency Unit = EthiopianBirr ETB) US$1 = Br8.7 FISCAL YEAR July 8 - July 7 WEIGHTSAND MEASURES Metric System ABBREVIATIONS AND ACRONYMS ADLI Agricultural DevelopmentLed Industrialization AEMFI Association o f EthiopianMicro Finance Institutions ALMPs Active Labour Market Policies A L M S Addis Labour Market Survey CEM Country Economic Memorandum CGE Computable General Equilibrium CLFS Child Labour Force Survey CSA Central Statistical Agency DG Director General EBDSN Ethiopian Business Development Services Network EEF Ethiopian Employers Federation FCSC FederalCivil Service Commission FeMSEDA Federal M S E DevelopmentAgency GDP Gross Domestic Product GOE Government o f Ethiopia I C A InvestmentClimate Assessment IFPFU InternationalFood Policy Research Institute ILO InternationalLabour Organization IMF InternationalMonetary Fund IVCAGs IntegratedValue Chain Analysis Groups KILM K e y Indicatorso f the Labor Market LFS Labour Force Survey LMTC Labor Market Technical Committee M A M S Marquette for MDG Simulations MDAs Ministries, Departments and Agencies MDGs MillenniumDevelopment Goals MFI Micro finance Institution MOE Ministry o f Education MOFED MinistryofFinanceandEconomic Development 11 M O L S A Ministryo fLabor and Social Affairs MTEF Medium-Term Expenditure Framework MTPRS Medium-Term Pay Reform Strategy MSEs Micro and Small Enterprises NGO Non-Governmental Organization OECD Organizationfor Economic Cooperation and Development PASDEP Plan for Accelerated SustainedDevelopment to EndPoverty PND Private Sector/NGOs/Donors PPA ParticipatoryPoverty Assessment PRSP Poverty Reduction Strategy Paper ReMSEDA Regional M S E Development Agency SASE Selective Accelerated Salary Enhancement SOEs Statement o f Expenses SMEs Small and Medium Enterprise SSA Sub-Saharan Africa TVET Technical and Vocational EducationTraining UBEUS UrbanEmployment and Unemployment Survey UN UnitedNations UNECA UnitedNations Economic Commission for Africa VAT Value-added Tax WED Wellbeing inDeveloping Countries Vice President Hartwig Schafer Country Director Ishac Diwan Sector Director Sudhir Shetty Sector Manager Kathie Krumm Task Leaders Jeni Klugman and Caterina Ruggeri Laderchi ... 111 ACKNOWLEDGEMENTS We would like to thank our primary counterparts, the Ministry of Labor and Social Affairs (MOLSA), and inparticular H.E.Woz. Zenebu Tadesse, State Minister, and Ato Tamiru Jeba, Head, Department of Planning and Programming. The Central Statistical Agency (CSA) provided essential support in accessing and understanding key data sources, and especially Woz. Samia Zekaria, Acting Director General, At0 Yasin Mossa, Acting Deputy D.G, Economic Statistics, At0 Yakob Mudesier, Acting DeputyDirector General of Operation Methodology and Data Processing, Ato Mekonnen Tesfaye, Head, Social Statistics Department, and Ato Kassu Gebeyehu, Manpower, Wages and Salaries Statistics Analysis and ResearchExpert Team Leader, Social Statistics Department, CSA. The Ethiopian Labor Market Technical Committee (LMTC), chaired by MOLSA, served as focal point for collaboration. Through a series of consultations they contributed to the agenda and reviewed background papers. The LMTC includes members from MOLSA, Ministry o f Finance and Economic Development (MOFED), Ministry of Education (MOE), Ministry o f Trade and Industry, Central Statistical Authority (CSA), Ethiopian Federal Civil Service Commission (FCSC), Ethiopian Employers Federation (EEF) and Confederation o f Ethiopian Trade Union (CETU). This study also benefitedfrom the close coordination with the ILO, both in Geneva and inAddis. A joint two-day workshop was heldinDecember 2005 to discussthe emergingresults of the ILO "Flexicurity Study" and this study. The team would like to thank inparticular Sabina de Gobbi, George Okutho, Lany Kohler, Zafar Shaheed, Muhammed Muqtada, Marva Corley, and Ato Bedada Urgessa. Generous funding from the German Trust Fund program on "Job Creation, Core Labor Standards, and Poverty Reduction" is gratefully acknowledged, as i s the support o f Achim Johansen, Deputy Head of Division, German Federal Ministry for Economic Cooperation and Development (now at BDI) and Hein Winnubst, First Secretary, Development Cooperationat the Embassy inEthiopia. Support i s also acknowledged from the Swedish and the DanishConsultant Trust Funds. The team was composed of Jeni Klugman and Caterina Ruggeri Laderchi (task team leaders), Emily Kallaur, Eyerusalem Fasika, Taye Mengistae, and Simon Commander. Background papers were prepared in collaboration with the team by Abebe Shimeles, Alula Pankhurst, Arne Bigsten, Bo Rosendhal, Feleke Tadele, Hans Lofgren, Martha Getachew, Mohammed Mussa, Niels-Hugo Blunch, Peter Madsen, Philippa Bevan, and Rahimaisa Abdula. Angela Cipollone and Martin Cumpa provided excellent research assistance. Thomas Pave Sohnesenand Jacob Ladenburg also provided research support. Administrative support was ably provided by Dora Harris, Lucy Kang'arua, Marjorie Kingston, Monica Wachera Ndungu, and SenaitKassa Y i h . The team benefited from the overall guidance of Ishac Diwan and Kathie Krumm. Useful suggestions and substantive inputs were received from a range of colleagues, and we would especially thank Louise Fox and the Africa Labor Market Analysis Group, Arvil Van iv Adams, Amy Luinstra, Bill Maloney, Furio Rosati, Gary Thiesen, Gordon Betcherman, Irena Omelaniuk, Jean Fares, Jee-Peng Tan, Keith Hinchliffe, Jemal Mohammed Omer, Lorenzo Guarcello, Luc Christiaensen, Magdi Amin, Marito Garcia, Mesfin Girma Bezawagaw, Gizaw Molla, Bob Rijkers, Jennefer Sebstad, and Pieter Serneels. Finally, but not least, we are grateful to our peer reviewers, Pierella Paci and Tamar Manuelyan-Atinc, for valuable advice. V CONTENTS 1 URBANLABOURMARKETSINETHIOPIA:CLARIFYINGTHEMETRICS 1 2 YOUTHEMPLOYMENTETHIOPIA ............................................................................... 33 3 A PORTRAIT OF INFORMALITYINETHIOPIA:DIAGNOSTICS AND POLICYIMPLICATIONS ... ................................ IN .66 4. TIME TO MOVE ON? INTERNALMIGRATIONETHIOPIAAND ITSLABOUR IN MARKET IMPLICATIONS ........................................................................................................ 110 5 MOBILITY . AND EARNINGS ETHIOPIA'S IN URBANLABOURMARKET: 1994-2004 ......138 6. A SIMULATION PERSPECTIVE ETHIOPIA'SLABOUR ON MARKETUp TO 2015 ..............167 7 MONITORING . LABOURMARKET DEVELOPMENTS:PRIORITIESTO OPERATIONALIZEAN EFFECTIVELABOURMARKETINFORMATION SYSTEM .............................................. 176 Box 1PrimaryLabour Market Data Sources used for this Study....................................... ListofBoxes Box 2 Rural Labour Markets inEthiopia............................................................................ 1 Box 3 Definitiono fKey Indicators used inthis Study....................................................... 2 Box 4 Voices o f Ethiopia's Youth.................................................................................... 4 ...................................................... 38 Box 6 Employment Exchange Services inEthiopia ......................................................... Box 5 Youth Civil Society Organizations inEthiopia 55 Box 7 Case Study o f a Vocational School: Hope Enterprise............................................ 57 Box 8 Subjective Well-being andEmployment Status..................................................... 59 77 Box 10 Informal Brokerage inthe Grain Trade................................................................ Box 9 Dynamic Female-OwnedBusinesses inEthiopia .................................................. 85 Box 11The Addis Ababa ReMSEDA .............................................................................. 87 88 Box 12 China: Reaping the Potential o f Internal Migration for Sustained Growth and ........................................................................................................... Box 13 Data Sources Used for this Study....................................................................... Poverty Reduction 112 115 Box 15 Perceptionon Constraints to Internal Migrationinthe PPA.............................. Box 14 Understanding the Dynamics o fResettlement inEthiopia ................................ 117 Box 16 Methodological Difficulties inQuantifying Migration...................................... 118 Box 17 Push and PullFactors inRural andUrbanAreas -Qualitative Evidence .........120 126 Box 18 Networks o f MigrantsInformation Flows and Job Search ................................ 130 Figure 1: Decomposition o fPopulation 15+ by Labour Market Status.............................. List of Figures 4 Figure4: Household Size Distribution. Urbanvs.Rural areas. 1999/00 (share ofthe age Figure3: PopulationPyramids.Urbanvs.Rural. 2004.................................................... Figure 2: Decomposition o fYouth (15-25) by Labour Market Status ............................... 5 37 ..................................................................... Figure5: Age-Specific FertilityRates. 2000 .................................................................... group that lives inhouseholds o f each size) 37 39 Figure6: Nationwide School Enrollment Trends. 1967-2005 .......................................... Figure 7: Supply o f Skills inthe UrbanPopulation. Males andFemales. 2005 ...............40 40 Figure 8: Trends inKey Labour Market Indicators for Different Age Groups. by Gender (inpercent) ........................................................................................................................ 41 Figure9: Number o fHours Workedper Week. 2003 (currently employedpopulation) .42 ......................................................................... Figure 12: UrbanYouth Unemployment Rates inSelected African Countries................45 Figure 11:Underemployment byAge Cohort: 1999 and2005 ........................................ Figure 10: Youth Employment Elasticity 44 46 Figure 13: Educational Attainment o f UnemployedYouth. 2005 .................................... 48 Figure 14: MedianandMeanUnemployment Duration (inmonths) byGender. 1999 and 2005 ................................................................................................................................... 48 Figure 15: Global Self-Employment as percent o fNon-Agricultural Employment. Relative to GDPper capita. Rural and UrbanAreas (PPP. constant 2000 US$) Figure 16: Demographic Composition o f Informal Sector Operators. 2003 .................... ..............68 Figure 17: Trend inthe EducationLevelo fInformal Sector Workers............................. 72 Figure 18: Reason for Choosing Current Activity. by Gender ......................................... 73 74 Figure 19: MedianMonthly Wages. Formal vs .Informal o f Informality .................................................................................................................... Employment-Broad Definition Figure 20: Sectoral Composition o f Informal Employment by Region, 2003 ..................76 77 Figure 21: Enterprise Location and Reason for Location by Gender ............................... 80 Figure22: What arethe 3 most DifficultProblemsCurrently Affecting the Operationo f Figure 23: The Rural Share o fPopulationDeclines at HigherIncome Levels ..............111 your Activity/Enterprise? .................................................................................................. 91 Figure 24: Agriculture's Share o f Labour Declines as Countries Develop..................... 111 Figure 25: Share o fUrban Women o fReproductive Age who are Recent Migrants,by Region............................................................................................................................. Figure26: Share ofWomen who are Recent Migrants, byAge group andRegion .......119 119 Figure 27 :Migratory Flows at the Regional Level........................................................ 120 Figure28 :Migration DestinationbyRegiono fOrigin for 6 Regions .......................... 120 Figure29: A Percentages o fUrbanWomen ofReproductiveAge who are Recent Migrants, by City Population Size and Location, Rest o f the World ............................. 121 Figure 30: Percentages o fUrbanWomen o fReproductive Age who are Recent Migrants inEthiopia, byCityPopulationSize andLocation ......................................................... 121 Figure 31:Gender Ratio (medwomen) by Age and Migration Status 123 Figure32: Reasons for Migration for Recent Migrants, byGender ............................... ........................... Figure 33 :Reasons to migrate, by areas........................................................................ 123 123 Figure 34: Mainreasons for move (single movers only 13+) ......................................... 123 Figure 35 : Educational distribution o fmigrants andnon-migrants............................... 124 Figure36: recent migrantsgender ratio (medwomen) by educational achievement .....124 Migration......................................................................................................................... Figure37: Employment Rates for Recent UrbanMigrantsbyGender and Reason for 128 Figure38: UHSESParticipationandUnemployment Rates (percent) ........................... 149 Figure 39: UHSES Employment Rates by Sector (percent) ........................................... 149 . Figure 40: Formal Employment Rates by Sector (percent) ............................................ 149 Listof Tables Table 1 Trends inKey Labour Market Indicators inUrbanEthiopia (percent) ................3 Table 2: Trends inUrbanActivity Rates. byAge Group and Gender (percent) ................6 Table 3: Trends inActivity Rates by Age and Gender (percent) ....................................... 6 Table 4: Supply o f Skills andActivity Rates. Total Population (15+) (percent) ................7 Table 5: Supply o f Skills and Activity Rates. Total Population (15-24) ............................ 7 vii Table 6: Supply o f Skills andActivity Rates. Total Population (25+) (percent) ................8 Table 7: Supply o f Skills andActivity Rates . Only (25+) (percent) ...................8 Women Table 8: Cohort Profile o f Education for the Economically Active Population 15+ by Table 9: Cohort Profile o f Education Population 15+ by Gender (percent). 2005 .............99 Gender (percent). 2005 ....................................................................................................... Table 10: Activity Rates by Gender and Marital Status. (25+) (percent). 2005 ...............10 Table 11: Trends inActivity Rates byR e g i o f i l l i l(percent) Table 12: Trends inEmployment Rates. byAge Group and Gender (percent) ................11 ......................................... Table 13: Trends inEmployment Rate by Education and Gender 15+ (percent) Table 14: Trends inemployment rate by education and gender 15-24bercent) ..............12 ............. 12 Table 15: Trends inEmployment Rate by Education and Gender 25+ (percent) .............13 13 Table 16: Trends inthe Education Composition o fEmployment. by Gender (15+) ............................................................................................................................ Table 17: Trends inEmployment byR e g i o f i l l i l(15+) (percent) ................................. (percent) 14 14 Table 18: Composition o f Employment by Type o f Employer. (15+) (percent). 2005 .... 15 Table 19: Education Levels by Employment Type (15+) (percent). 2005. ...................... Table 20: Trends in Sectoral Composition o f Employment by Gender (percent) ............1516 Table 21: Trends inUnemployment Rates by Age and Gender (percent) ........................ 17 Table 22: Unemployment Rates by Age and Gender 15+ (percent) ................................. 18 Table 23: Profile o fthe Unemployed byDuration and Gender (percent) ........................ 18 Table 24: Unemployment Rates by Education and Gender. 15+ (percent) ...................... 19 20 Table 26: HeckmanRegressions on Earnings. sample 15+ LFS 2005 ............................. Table 25: Unemployment Trends by Region andAge Group (percentage) ..................... 24 per month) in2005 (inpercent) ........................................................................................ Table 27: Mean andMedianEarnings for the Employed (15+) by Educational Level (Br 25 Table 28: Earnings for Employed (15+) by Industrial Classification (Br per month) in 2005 ................................................................................................................................... 26 Table 29: Earnings for Employed (15+) by Occupational Status and Type o f Employer (Brper month) in2003 ..................................................................................................... 27 Table 30: HeckmanRegressions on Earnings. Sample 15+ LFS 2005 ............................ 28 Table 31: OLS Regressions on Earnings. sample 15+ LFS 2005..................................... 31 Table 32: HIVPrevalence byAge Group (percent). 2003 ............................................... 39 Table 33: Activity Rates by Age. Gender. and Education Level. 2005 ............................ 43 Table 34: Employment Status by Age. 2005 .................................................................... 43 Table 35: Ratio o fYouth to Adult Unemployment by Region. 2005 Table 36: Unemployment by EducationLevel andAge. 2005 (percent unemployed) ....47 ............................... 47 Table 37: Youth Time Use inUrbanAreas. Selected African Countries (percent o f youth) ........................................................................................................................................... 49 Table 38: Primary Activities o fUnemployed Addis Ababa Youth inthe Previous Week. * 2006 (percent) ................................................................................................................... 50 Table 39: Prevalence o f Informal Employment by Sector (informal as percent o ftotal employment inthe sector) ................................................................................................. Table 40: Reason for Choosing Current Activity. byPrevious Activity. 2003 74 Table 41: Sectoral Composition o f Informal Sector Survey. 2003................................... ................ 71 Table 42: Results o f Cluster Analysis. Informal Sector Survey. 2003 ............................. 79 83 ... V l l l Table 43: Median andMeanMonthly Sales by Gender of Operator andAge o fFirm. 2003 ................................................................................................................................... 84 Table 44: Sourceof Initial Capital for Informal Sector Enterprises. 2003....................... 93 Table 45: MedianStartUp Capital by Sector and Ownership Status Table 46: Starting a Business: Comparison of Ethiopia. Sub-SaharanAfrica & OECD .94 ............................... Table47: ProbitEstimations; Marginal Effects.............................................................. 97 Table 48: Cluster Analysis. MedianValue and StandardDeviation shown...................101 102 Table 49: Probit Estimation: Likelihoodof Intentionto Discontinue Current Activity (marginal effects) ............................................................................................................ 103 Table 50: Probit Estimation: Operators Reporting eachProblem as among "Three Most Difficult Problems. (marginal effects) ............................................................................ 105 Table 51:Probit Estimation: Likelihood of Intending to Discontinue Current Activity/Enterprise .......................................................................................................... 107 Table 52: Probit Estimation: Operators Reporting eachProblem as one of the "Three Most DifficultProblems Affecting the Current Operations Table 53: Determinants of Migration-RecentMigrants inUrbanAreas Only ............108 ........................................... Table 54: Determinants of Employment -RecentMigrants inUrbanAreas Only ........135 Table 55: The determinants of Wages byMigrant Status and Gender........................... 136 137 Table 56: UHSES Sample Age Distributiono fthe Sample All waves andAge Groups . ......................................................................................................................................... Table 57: UHSESDistributionby Labour Market State and Year. Wave by Wave ......144 145 Market States. 1994-2004............................................................................................... Table 58: Maximum Likelihood Estimatesof MultinomialLogit Models of Labour 147 Table 59: Maximum Likelihood Estimation of Probit Model of Selection to Public Sector Employment .................................................................................................................... 154 Table 60: Maximum LikelihoodEstimation ofProbit Model of Selection to Informal Sector Employment......................................................................................................... Table 61:Maximum Likelihood Estimation ofProbit Model of Unemployment ..........155 156 Table 62: Earnings and Selection into Private andPublic Sectors: 2004....................... 158 Table 63: Earnings and Selection into the Wage Employment and SelfEmployment: 2004 Table 64: Transitions Probabilities Between 1994 and 1997. age group 15-64.............159 ................................................................................................................................. Table 65: Transition Probabilities Between 1994and2000. Age 15-64........................ 162 162 Table 66: Transition Probabilities Between 1994and 2004. Age 15-64........................ Table 67: Transitions Probabilities Between 1994 and 1997. age group 15-64 163 Table 68: Transition Probabilities Between 1994 and2000. Age 15-64 ........................ .............162 163 Table 69: Transition Probabilities Between 1994 and2004. Age 15-64........................ Table 70: Transition Probabilities o f 1995 RespondentsBetween 1995 and 1997........ 163 Table 71: Transition Probabilities o f 1997RespondentsBetween 1997 and2000 Table 72: Transition Probabilities o f2000 RespondentsBetween2000 and 2004 ........164 ........164 164 Table 73: MultinomialLogit Model of Transition Out Public Sector Employment: 1994- 2004. Base Outcome i s being out ofthe Labour Force. Age group 15-65 ..................... 165 Table 74: Multinomial Logit Model ofTransition Out of InformalSector Employment: 166 Table 75: Simulation Definitions.................................................................................... 1994-2004....................................................................................................................... Table 76: Simulation results: Macro. Production. andMDGIndicators........................ 169 170 ix Table 77: Simulation results: Wages. Wage Gaps. andEmployment ............................ 173 Table 78: Possible List of Labour Market Indicators ..................................................... Table 79: Additional Labour Market Indicators ............................................................. 185 189 X 1. URBANLABOURMARKETSINETHIOPIA: CLARIFYINGTHE METRICS 1.1 This chapter provides an assessment of urban labour market indicators focusing on the level and recent evolution o f activity, employment, and unemployment rates, as well as wages. It provides detail on the broad trends synthesized in vol. I.The analysis relies on the 1999 and the 2005 Ethiopia Labour Force Surveys (LFS; urban data only). Box 1 provides a description o f the data used. The Urban Biannual Employment Unemployment Surveys (UBEUS) surveys used elsewhere in this report have not been usedto calculate trends due to doubts o n their overall comparability with the LFS surveys arising from differences in sampling and possible seasonality effects. Further work on understanding labour market seasonality, particularly in relation to seasonal migration from rural areas, is needed. As discussed also inchapter 7, new plans to holdthe UBEUS survey o n a yearly basis offer a chance to enhance the comparability of the survey over time. Box 1:PrimaryLabourMarketDataSourcesusedfor this Study 1999 National Labour Force Suwq (LFS) conducted in March 1999 by the Central Statistical Agency (CSA). The survey was nationally representative and covers all zones except 6 in the Somali region and 2 inAfar. The data are representative at the national, urban, rural and regional level. The survey covers the population aged 10 and over, recordingtheir labour market status as well as a variety o f socio-demographic variables such as age, gender, migration status and education. The survey didnot collect wage data. The survey covers 31,859 households inurban areas, 2,5 18 o f which are inAddis. 2005 National Labour Force Survev (LFS) conducted in March 2005 by the CSA. The survey was nationally representative and covers all zones except Gambella region (with the exception of Gambella town) and 6 zones inthe Somali region and 3 inAfar, mostly inhabited by pastoralists. The data are representative at the national, urban, rural and regional level. The survey covers the population aged 10 and over, recording their labour market status as well as a variety o f socio- demographic variables such as age, gender, migration status and education. The survey instrument has been updatedwithrespect to the 1999version, the most notable change beingthat it includes wage data. Other changes, including an updating o f the statistical sampling frame and a more precise characterization o f urban areas which allows a break down o f major urban and other urban areas have also been made. The survey covers 29,623 households in urban areas, 21,420 inmajor urban areas. An important caveat on the figures reported from this survey is that, at the time o f writing, the data have not been fully released (inparticular, it i s not possible to identify which observations belong to the same woreda), so that with the information provided it is not possible to control for the clustering o f observations in the sample. This implies that standard errors and therefore confidence intervals are underestimated, affecting the statistical significance o f trends and of regression coefficients. 1 1.2 The LFS survey instruments are similar at least as far as key indicators are concerned, except that wages were not collected by the 1999 LFS. However, the analysis i s constrained because It is not possible to estimate correctly the confidence intervals for the 2005 LFS on the basis o f the information released, leading to some caution in the interpretation o f changes. 0 The identification o f some educational categories (such as "beyond general education") which were affected by the 1994 educational reforms might be imprecise despite attempts at specifying the equivalence o f different grades across the two systems.' 1.3 Before turning to substantive issues, it i s also worth stressing the heterogeneous nature o f "urban" as an analytic category. The CSA defines urban areas as "localities with 2,000 or more inhabitants" though in practice (i)all administrative capitals (regional, zonal and wereda capitals), (ii) localities with urban dweller's associations not included in(i) and (iii) localities not included ineither (i) (ii) inhabitants are all or whose primarilynot engagedinnon-agricultural activities are considered as urban. 1.4 Ina major step forward towards unpackingthe heterogeneity which characterizes the urban hierarchy, the 2005 LFS for the first time allows the identification o f major urban areas and other urban areas. Being able to distinguish between smaller centers whose economy is likely to be tightly integrated with adjacent rural areas, from those o f larger places which have more complex economic roles i s likely to provide very useful insights, particularly because o f the differences between the urban and the rural labour markets. The definingfeatures o frural labour markets are synthesized inBox 2 below. Box 2: RuralLabour MarketsinEthiopia Rural labour markets in Ethiopia, as elsewhere in Africa, require the consideration of broadeI forms o f labour exchange than those traditionally captured by labour force surveys. These range from labour exchange (e.g. sharecropping), non-market based exchanges (such as tributary labour), to adjustments to family size (including adding wives, children or extended family) (White et al. 2006). Broader analytic tools, beyond standard surveys, are needed to understand these. Most Ethiopians working in agriculture are self-employed, although land pressures and other factors suggest that hired labour might be on the rise. Insome areas, cultural practices and local technology result in some households without adult males hiringlabour, leasing land or entering into share-cropping arrangements. Rural productivity i s particularly low in agriculture, where 80 percent of the labour force works yet only 45 percent of GDP is generated (Bigsten, et al. 2003). Open unemployment i s low, but underemployment i s pervasive. Ethiopian students begin with 8 years o f primary education then proceed to the first cycle o f secondary education the completion o f which is considered "General Education". Before the 1994 reforms the completion o f this cycle was at grade 12, while now it occurs at grade 10. By 2005 the first cohorts in the new systemhadreached the labour market. 2 1999 2005 1 Male 12.3 Female 17.7 iVote: * ~ i ~ ~ id~i f fc~ar efrom 1999 figure at 95 ~~ ~~ l ~ percent, though results. Source: For alt tables unless ~ ~ i ~ ~d LFS ~~ ~ ~ ~~ ~ d~ i ~ (urban only),OWTI calcul~t~ons~houldbe l ~ t c ~ r with ~ d e ~ caution (see Box 1). 3 Inthisreportwe adopt the KILM(Key lnd~ca~orsofthe Labour~ a r k e~t ~e ~ i ~~~e ~~ i~olby ~~e d on thc ILO, unless o ~ e ~specified. i ~ e There are ~ ~ ~names for each indj~a~or~ we use ~ r ~ n t which ~n~erehan~cabl y: Activity rate or Labour Force P ~ ~ t i c i ~ Rate~ ar~P a t o ~ ~ irate - ~&e shareaof the ~ ~ ~ t ~ ~ ~ o p ~ ~ aagedot5+ either engaged in, or a ~ a ~ to ~ b ~ e ~ r t ~ n ~ undg~~ke~o d ~ ~ct~vitjes. thus c ~ ~ e It c a ~ t u r the idca of labour supply for all p ~ o d ~ ~a tc it ~~ e~ taccord in^ to the 1993 UN ~ ~ s i ~ s y ~ ~ ~ m of ~ ~ ~ i Accounts,~ though to the exten1 that some sectors (e.g. i n f o ~ aactivity or where o n a l en~rgdom~na~e~ fu'uilyrefllectedin~ are not ~ ~data,~~abours u~ ~is u l~~~~ cn ~ t ~ s ~ ~ ~ o ~ such ~ ~ ~ l oRelevant~to eE ~ ht i i,~the r~e~c ~ ~ ~that~ "`inosni ~ ~ a whereo theslabour ~ ~ t s ~ i ~ ~ ~ ~ ~ ~ a r iks glargely u n o r ~ ~ uherc d ~ ~ n ~ ~labour a b ~ o ~is,~ ati othe~time, ~ n ~ d e ~or whcre.,thc ~ a ~ e labour force i s largely ~ e I ~ " thec crirenonof seeking work" may be relaxed. 'X'hcCSA ~ ~ 1 ~ ~ ~ ~ ~ d e ~ ~is~basedionoa ``rchxd ~ ~ ~ ~ whichj leadsn to' h~~ ~ hune~~Ioyrnen~ However ~ ~ i ~ o e r rates, inthis studywe adoptthe i ~ t e ~ ~ ~ i coonran ~la~~~~ee~ i nbased~onoa11threecriteria, l b ~ ~ n 1 1 ~ 1.6 Figures 1 and 2 (011thc youth and the adult ~ o p u l ~ t~i e~sn~~e c t i ~provide~a r e l ~ further look at these trcnds by d ~ ~ ~the ~working age p ~ p ~ l intot mutually ~ i n g ~ ~ o ~ excl~rsivec a ~ e ~ A~d~~~sst .~ can~be~made bne t ~ ethe~"passive u ~ e ~ ~(who ~ y ~ ~ ' ~ ~ o e ~ l ngnot having ajob andbeinga far work do not report that they are ajob) and the "activ who are without a jab, avail actively ~ ~ a r c ~ iThe~p~ . i n ~ o n ~ i ~ ewithdthe ilia r e tile ~ ~ dep ~ ~ ~ nition used (see Box 3). ~ 0 ~ ~ e Figure1: Decamp of P ~ ~ # ~15+ byaLabourItl ~ j n __---- Inactive 0% 20% 40% 60% 80% 1000~ 4 1.7 These d ~ s a ~ ~ c ~ arevealn sa shift in the compos~t~o~of the working age t ~ o po~ulat~onbctwcen 1999 and 2005, with passive u ~ i e m p ~ od~ e ~ct ~ in favor o f ~ ~ ~ ~ n ~ ~ i p ~ o }andi eeven~ more of ~ n ~ c t i ~ Thist }shirt. was ~ a ~ ~ cvisible for rthe ~ ~ n ~ ~ ~ u ~ ~ l 2%. For adults and youth, active ionof Youth(15-25) by Labour h'l Status 2005 DActive 1999 kqerrployed 0% 20% 40% 60% 80% 100% 1.8 A c ~ i vrates~appear to be stable at around 65 percent. In~emat~ona~co~paris~n$ ~ t ~ L ~ ~ ~ randn~d~t ~ ~ ~2c~s0hsuggest a 4 e~ n ~that ~ t ~ ~activityprates~for iolder groups are ~ i c ~ ~ p a ~ with~those of Uganda and Kenya. but no~iccab~ylower for youth (the 15-24 a b e ailable data to detail trends by age and gender, ~ d ~ cfcvef, ~ ~ ~ ~ ~ l a r. As shown in vity rates havereniai e net effect ofa dec ti and an increasc Eo different trends have charac age cohorts: while among young people increased p ~ r ~ ~ c i ~ 3 a ~ o n e n by ~ o n ~ has c ~ n ~ ~ for decreased~p~a ~di c ~ ~ abyj omen, e n ~ ~ t ~ i a ~ o otdcr groups there has been an increase in ~ ~ g ~driven by increases in~ ~ ~ fcmalc ~ a ~ ~ c j p a t i o ~ ~ . 5 Table 2: Trends in UrbanA ty Rates, by Age Groupand Gender &e 1 ._ 1 ~ 1999 2005 Age goup 15 $. 11Male 56.4 f 5 1 . V f Female 11 47.8 f I 49.9* f Age group 2 S 1 with caution. ds by cohort (Table 3) shows that there were ~ ~ g n ~declinesnfart melt age ~ ~ a 20-25, and s ~ g n ~increases~for ~ ~ ~ a t ~ ~ o ~ 20-25, and 49-59 years old, Table 3: Trends io Activity Ratesby Age and Gender ~ p ~ ~ c e n t ~ 75.6 73.2" 61.3 65.5" 55.6 71.7" 63.0 679- I.It ~ ~ Levef. Tables 4 to 6 6present the~d i s ~ ~ ~o~f education in the total ~ ~ t ~ o n ~ ~ ~ 0 p o ~ ~ l a tbyo cohort alongside trends in activity rates. Table 4 (age 15";) slzows the j ~ ~ increasing educat~or~a~ profile of ~ t ~ i The ~~ r ~o ~p ofos those~who are illiterafc ~ ~, ~ ~ decreased by 1Ipercent from 1999 to 2005, and more ~ e o havc c~o ~ ~ ~ lprimary p ~ e t ~ d (~ea~h.ed grade 8) and general education (and ~ ~ y ~ As expected, activity sates tend to n d ~ . rise with e d u ~ a t i ~Ievels,a ~though. ~ ~ s a g ~ rby ~age shows nthat for youth the ~ ~ ~ a ~ i ~ distn~u~iQn i s bipolar, with higher activity rates at both ends of the ~dLicationd ~ ~ t r ~ ~ ~ t ~ ~ n . 1.12 L ' at trends, activity rates categories. There are jncrease~far tlmse and significant drops for post ~ ~ j ~~I ~a u~ ~y ae ~~~ ca ~thi~, ~ j ~ g stable, Tables 5 and 6 show that these changes have been driven by increased activity rates for young pcoplc with some p ~ educa~~on,and by decreased activity rates for ~ a ~ ttiosc with c ~ ~ ~ l egeneral e d u c a t i ~ ~both youth and a d ~cohoits. t e d in l ~ 1.13 Tables 5 and 6 also show the effects of the e d u c a ~ ~ oe~ i~~ l ~ a OII~ sthe~ o ~ c o ~ ~ ~ oof~thei youtho axid the adult cohort. By 2005, three ~ ~ i ~ ~ of the~young~had e ~ s at least 4 years of e ~ ~ ~ c awhile ~for, the oIder cot~oft,despi~es ~ ~ ~ a progress in ~ i o ~ n e d ~ d u c a ~this~was true for S I percent of the ~ o p ~ l a tonly.~ However, betu.een 1999 i o ~ i o and 2005, the sharc of those with at least 4 years o f ~ ~ ~increasedyby 6 p e r c e ~ t a ~ e a r points. Table 5: ~ ~ pof Skills~ Activity Rates, Total ~ ~ ~ ~(15-24)a t j o ~ p l and ~ Skills 1 1999 1 2005 1 1999 1 2005 be rntcrpretrdwithcaution, 7 1.14 Tables 6 and 7 ~~lustra~egender d~ffere~ces c o ~ ~ `rvomcn's ~~rfo~mance by p ~ ~ ~ g with the average for their ~ o u ~ . Table 6: S ~ pof~Skills~and Activity Rates, Total F ~ ~ ~(253-) (percent) o ~ l I ~ t ~ I itre catching up with men in terns of tes i s more ~ ~ r butk theefemale~pop~ ~ ed than the male. Note also that adult ~ F o n i efi~ ificantly less active than thc average, The snialles small but g ~ ogroup~o ~~ ~ ~nwith at~leastesome h ~ o~ ~ Table 7: Supply of Skills and Activity Rates ~~~e~ - Oniy (254-1 (p I.16 The gtndcr co~pa~isonf u ~ela~oratedby Table 8. This shows the differ~n~ is h ~ ~ cohorts' educatio~alou~co~esterms of illiteracy and average years o f schooling for in those who went to school. For all age groups active ~ o ~ ~ d~u ce~ t n~outcomcs are o n ~ 1~ ~ less so for y ~ ucohorts, e ~ ~ ~ 8 All I Male Female 1 Average Average Average Age % Years of % Years of o/o Yearsof Group Illiterate . Schooling IIliterate Schooling Illiterate Schooling 55-40 53.5 7.6 f 32.2 7.7 83.4 6.7 601- 64.6 6.7 1 49.5 6.8 93.6 5.2 1.17 Contrast~n~ Tables 8 and 9 stiows that the shac of illiterates in the ~ o ~ ~ cohorts is higher ~~o~~ the active than the overall ~ o p ~This i s~ to~be iexpcctcd ~ ~ ~ ~ , people who arc: illit e are more likely t y to bc at: school. However, inolder co s are more likely to Table 4 stiosvs a striking ~ j ~in ~ ~ r ~ es~between~older c and~~~~~~e~wonten, with an ~ ~ of91l percent~ani rate ~ ~ r ~ ~ ~ 15 percent among 15-20 the ~ ~~ ~ u cpa ~ j o tr~ a ~ ~ ~ ~ ~ A11 I Male Female Average Average Average % Yearsof 94 Yearsof % Years of Illitcrate Schooling IIIiterate Schooling Illiterate Schooling 9 1.18 en's lower activity rates are ely to be c o ~ e ~ a ~withd e gender roles and their h o ~ ~ ~ e and~childcare~responsibilj~~es.An irzdireet way of ~ ~ a k n ~ g these issucs is to focus on di by marital status (T ns, n ~ a ~wornd ~ e h less likely to be econ active thm those: never married. At the upper end of the ed~eationa1s p e c ~ ~(those m with Beyond General and Higher Ed~cation~,however, the n i a ~ catch up with the ~ d u n ~ a ~ ~ reflecting partly the fact that these women arc older but also the high e d , opportu~itycost for staying out of the tabour market and h ~ ~ s e ~factors o l d correlated with higher ed~cation(e.g. ater wealth, likely a~ailab~~ityof child care). ~ o ~ who n e are Divorcedi'sc ~ ) a r a tic ~ ~ed~are cons~sten~ly more likely lo be econon~~~altyactive. This does not hold, however, for ~vomenwith at least Some highere d ~ c ~ t i o ~ . ast the distinct activity p virtually all lev niorc likely to be active if they are y o ~ ~ g117erl are e r 1120 y to be in school, and since activity rates increase up to the 35-40 year old group Is0 ~ u ~ ~ ethat S s ~ a ~ untilthey~ a g have ajob and are thus abs'ieto support a family. Or, convcrsdy, that men who are married are morc likely to become active, beca~seofthe needto support a family. The ljke~i~~ood of activity is reversed when inen are DivorcediSeparare~iWicfo\Yed,possib~yalso because they are older on average than ~ ~ naen,e and di~~ereas~~igage is corrclated with d e c ~ ~ nactivity. i n ~ Table 10: Activity Rates y Gender and Marital Status, (2S-t) ~ ~2005 ~ ~ ~ Male Female Never Di\orced/ Never Divorced/ Married 'Iarried Separiltcd/\C idowed Married SeparatedNidowed 1.20 ~ ~ Activity~rates have~r e ~ ~ i n ~ e d stable in large paxts of the ~ ~ ~ ~ ~~~~~~~~~~ country, though s i g n i ~ c aincreases were recorded in Tigray and Addis, and s i ~ n i ~ ~ c ~ n ~ ~ ~ t decreases in.Orontia and ~ a ~ ~ b~e l T~ 11). a ~ b ~ ~ Table 11:Treadsin Activity Ratesby R ~ g ~ ~ ~~/p~~l rl ~i le n t ~ f t Kegioai'Killil 1999 2005 1.2I ~e~ress~ona~ia~ys~sprovides a profile of the inactive. ~ ~ ~ a c tdecreases i ~ ~ i t y with age and is higher for those with some ~~ucation. On avcragc w o ~ i i ~arejust as likely as n men to be inactive, but gender ~ a ~ ~ c r s - ~ac~i n~~ o lessensatheneffects of ~ d u c a ~ ~ o ~ ~ ~ in ~~icr~asingactivity rates. Further, there are ~ ~ effectspof family composition ~ o ~ ~ and gcndcr on ~ n a & ~ ~For rexample, a larger raumber of children youn~erthan 10 in ~ ~ t ~ , the ~ouseholdis ~ s o c ~ ~ withd higher activity rates, but the effect is reverse^ for t e ~ o ~ i e Further the number o f female chi n , ~ ~ i ~ c ct o\~~s ~ s~with~female children . i tyc ~ This s ~ ~ that trad~t~o~~a~ ~ e ~ ~ s useho1~division of 1.22 ~~~~~~~a~~ u ~ r ~ ~i ~ ~ dgj~v ~are~niore llikely to be inactive once t n y d, ~ ~ s ~omeedu~atjon.As ~ ~ a ~ to~be p a~~ ~~c ~~ ~ ly afor~those with ~ i strong l ~ ~ ~ o ~ ~ ~ general education, axid beyond ion - i ~ ~\it.ho~are likely~to be~still ~ v ~ ~ ~ ~ ~their studies - ina~ ~ pcars to bc~driven by school a~ t ~ ~ This d ~~ ~ ~n . t ~ ~ is indeedcon~m~edby lookingat results for the 25+ age group, for which most e d u c a t ~ ~ ~ ~ a ~ variableshave no ~ ~ g ~effect ~ ~with respectto the illiterate group. Note howeverthat c a ~ t the highly s k ~ ~ aree 30 percent less likely to be inactive (more than SO percent in non- ~ d Addis urbanareas, while inAddis thc effect is i n s ~ ~ n ~ ~ c a n t ~ . 1.23 A ~ g ~ e e~ na i~~eI o ~ratestremain unchan~e~between 1999 and 2005, ~ h o ~ g h e ~ there are varia~~o~isbetween cohorts ~ e ~ n ebyd age, gender, edu~a~~onlevel and geogra~h~callocation. This section explores these differences in more detail, and then reviewschanges inthe ~o~positiono f e ~ ~ ~ l o ~ ~ e n t . I1999 I 2005 Ape goup 15+ Table 13: Trends in p l ~Rate by~Education and Gender 153.(percent) ~ ~ n ~ All Rlale FeiiiaIe Einploymcntrate 1999 2005 1999 2005 I999 200.5 1.26 ~ o c ~ s i nang youth (Table 14), as already seen far activity rates, the ~ l l i t e r ~ ~ e s have higher e ~ p l o ~ e rates (indeed higher than most other categories). n t Strong increases are recorded in the e ~ p l a ~rat@of the iIl~terat~(driven by women) for i e ~ ~ those with grades 1-4 and for those with c o n i ~ ~ ~ general e d ~ c ~ t ~`Those. with. t e d ~ n beyond generat e d ~ c a ~ ~have~lower e ~ p l o ~rates. t o n ' e ~ Table 14: Trendsin e ~ ~ rate~by ~ ~~ ~o candagender~15-24~~~rcent) ~ t ~i e~ ~ ~ All Male Female Employmentrate 1999 2005 1999 2005 1999 2005 Note: * S ~ ~ i ~ ~ different~ from 3999 figure at 95 pcrcent, ~ ~ a n l y h results should ~be o ~ ~ ~ ~ ~withecaution. r ~ ~ ~ ~ d 1.27 A ~ o ~ad~lts(Table 15). little change is found between 1999 arid 2005. Unlikc i g the casc afthc young, the illiterate ~ o ~ u l ahas~by ~Par~the lowest e t o ~ ~ ~ratc-58o ~ ~ c n ~ cQni~aredto 72 percent for Grade 1-4 c o ~ ~ ~and ~ e s l ~90 pcrccnt for those with some ~ i g ~~e r d ~I ~ a ~ ~ o ~ Tabie 15: Trendsin~ ~Rateby E ~~ u c a and~Gender 25+~(percent) ~ ~ ~ o ~ ~ ~ vote: * ~ ~ ~ i ~ cdifferentl from 1999 figure at 95 percent, ~~~u~~results should he a ~ t y i ~ ~ ~ ~ rwitht caution e e d 1.28 In line with the overaH decline in illiteracy and ~ ~ supply of ~ w skills, nthe ~ e d u c a ~ ~ o ~c oa ~ p o s i t j a ~ofi @ ~ ~ ~ l o ~i si cehnat~ g ~ nwith l g ~ lower education levels a ~ c Q ~ far~a smaller share. As Table 36 shows, the share of the e ~ ~ ~~oo ~~ eu d~ a t ~ o ~ ~ n ~ n ~ 13 that i s iIIjEerat~has declined from 32 to 25 ercent, while workers with at most 4 years of s c h o o l ~ represent now only 42 percent of the e ~ i i p ~ o(against 47 perccnt in 1999). n ~ y ~ d AI1 Male I Female Employment rate 1999 2005 1999 2005 1999 2005 49.8 51.6 cantly drfferent from 1999 percent, thoughresults should be ~ ~ ~ e ~ ~ ~ ~ e d 14 mcnf, Despite a r e l a ~ ~ stable l ~ ~ ~ e rate, some sectors haw grow rtance in terms of their share of enf has grown ni r c ~ ~ ~ atemis) in ~ o ~ s ~ ~ u c ~ ~ o n , g e cs and other se ~~~c~~~~ and tt' gas red the highest perc drops, ~ o ~ i acen c all o f the decline inthe h ~ ~ e ~ ~ r e sc t~~~~~~~andnrmost o f the increase in e d ~ ~ a ~ ~ o n ~ ~o tys , h ~ a and~social work. Note that cros~-sect~una~data, however, may conceal ~ h ~ ~ ~ n g ~ h ~ ~ rsectorst as jobs~ have keen d ~ ~ t ~ and new ones have been created (e.g. in ~ ~ i o ~ e d ~ a ~ ~ ~ ~ d ~ talso~Chapter~3,-id.1).e ~ ~ n s e 1.33 To s ~ p ~ l ~ thisi e~ne s~e r ~ ~ t ithen ~cowelates of e ~ o ~ p ~ have ~been ~ ~ e t i ~ i v c ~ t ~ ~usingeregression andpis. The analysis offers some ~ n t e r einsights~in~ a ~ d ~ ~ i p a ~ i ~ uonathe effect of gcndcr. ~ n ~ ~ ~givenstheir lower~e 3 ~y p ~ r ~ ~ n I rates, all e ~ ~ ~ ~ other t h equal, adul~women~are less Iike3yto be em~loyGdthan men, ~ a ~ i ~ ~inl a r ~ y ~ ~ ~ Addis. ~ ~ ~ e r a ~oft igender with other correlates of e ~ p l o ~ e reveal a strong o n ~ n t 16 reduct~onin the p r o ~ aof~e~ i p ~~o ~~i i~fornf~ ~ ~ nwith you~zgerch~~dre~i(aged less ~ e t i e n than 10; note that for men the effect is si~ni~&ant positive); such an effcct is and pa~~cularlystrong in Addis as co~paredto other urban areas of the country. Another i n t e r e s ~ ~d~~fcrcnccbetween Addis and other cities is a s i ~ i ~ c a rcduct~onin the t i ~ n t ass~~~ationbetween increasing age and i n c r ~ ~ej n~~p ~ ofor w o~~i~~n~.~ n a l ~ y ~ ~ e having sonic education ( e 5-8 to having c o ~ p l e ~ egen d Ied#cat~on~lowers employme ity with respect to h a ~ y ~ ~educat~on(also for those no ~ g 5+>* This is an klding 8s these e ~ ~ & a t ~categories had no significai o n a ~ t on the results tivity for this age group, ~~~~~~~~~~~~ 1.34 ~ n e m p l o ~ i c which is the subject of Chapter 4, vol. E', n t ~ decreased by 13 per~enta~epoints betweerz 1999and 2005 ~ ~21).~ b ~ e Table 21: Trends in ~ ~ n e r n p ~ ~Ratesbyt Age and Gender ~~e~~~~~~ y ~ e n Age group 15.- I 'i'otat sample 14.8 12.3 hrm: * ~ i ~ n i ~ ~d~fferentfrom I999 figure at 95 a n t ~ y percent, though results s h ~ ube~i ~ t ~ ~ ~ r e t ~ d l with caution. 1.35 Age and Gender. The decline in u n e ~ p l o ~ eappears fo have been driven by a n t drop in youth u n ~ ~ ~ p l and~lower ~u , ~ e ~ ~forlofder ~ ~ ~ i y~ ~ o e ~ o males, Still, u ~ ~ er u n e ~ i i p ~ o ~thann tadults. ~ i ~ o ~ e ~ i ~ to experience u n e ~ p ~ o ~ e ~ i t . ent ratcs by age cohorts (Table 22) shows that the 25 age group for bothmen andwomen. ~ o ~have~ n z n for all aye groups less than 50 y this crossover point. Although # n e ~ p l is ~ ~ ~ ~ ~ o r e l ~ t i v lowyamong the oldest age g~oups,it remains signi~can~8 percent for those ~ ~ at aged over 60. Labour market ~ ~ i c ~ p ~ i i s tstillocn o for this age group (the activity ~ ~ ~ ~ ~ rate for males aged GO+ is 60 percent, ~ h o u ~ forh~ o it is~ only 26 percent). Between ~ n I999 and 2005, signilicaxit decreases were recordedin youth ~ n e m p l o ~ ~white,other e n ~ cohorts faced stable ~ ~risks, with few exceptions, ~such as wornen aged SO- e ~ ~ ~ SS for whom the ~ ~ e n ~ ~rate inercascdnbyt45 percent. l ~ ~ e Table 22: ~ n e r n ~ 1 ~Rates~by~Age~and Gender I 5 1(percent) 5 . ' ~ t Table23: Profileof the ~~e~~~~~~~ $5.' Duration and Gender ~ p e r c e ~ ~ ~ AH I Mate Female Employment 1999 2005 1999 2005 1999 2005 rate 18 1.38 Level. ~ d ~ ~plays~a ~~ o ~ ~irole in cxplai a ~ ~ i ~ ~ ~~~~~~~~~ of ~ n e ~ p ~ o ~(Table t241, which is highest among those who have some ed~cat~oIi. e t 1 ~o~~ the i~l~terate pop~latio~i, can perhaps least afford not to work, ~ u.ho ~ e ~ ~ ~ a ~ averaged 8 ~ e r c ~in n t2005. ~ n e ~ i p ~climbseamong those with basic and ~ e n ~ r a ~ ~ ~ n t ~ ~ n i auntil it peaks at General E d ~ c a ~ ~ oThose tvith at least sotm higher e d ~ ~ a ~ ~ Q ~ r y n , had an avera~e~ n e r n p l o rate ~ ~ eof~9~percent, tvith no ~ ~ ~change over~ the n ~ ~ f ~ period except for ~ ~ ~ o ~ ~ ~ . Table 24: Un~mpl#smentRatesby E ~ ~and~Gender, IS+ (percent) ~ a ~ j ~ I I Total 1999 1 2005 I1 Male FemaIe 1999 1 2005 1999 I 2005 6 7.5 9.2 22.1 8.5 6.8 1$39 ~~~~~~~~~~.r Regional s ~ ~ a in ~u #~n e n ~ pisls~j ~~ j ~ c(Tablet ~ a ~ ~~~~~~~~~ 25). The large cities-Addis Ababa and Dire Dattfa-have the hest Smaller urban ccntcrs have a prafrle more similar to that sal areas, where urid er than open ~ ~ i e ~prevails. Tigray, ~ ~ ~ o ~ ethe~fourth~highest r i for the ~ e ~ a t i ~ ~ ~ lgap between (high) youth and (closer to a ~ ~ e adult~ e ~ large y r ~ unen1p~o~1ent.In Addis youth ~ n e r n p l o ~i~s e70~percent higher than for the older t Goho~s,~ ~ hisithec small observed gap.. ~ 1 Dire Dawa 1 21.8 1 f8.3 I 34 I 30.4 I 36.4 f 13,9 f 1.40 Regional u n e ~ p ~ o trends are~revealing, Xn ~ a ~ ~ Harari, ~and~ Tigray ~ ~ e n b e ~ , ~ i ~ ~ ~increases thi y o ~ ~~ nh e ~ ~ p lwere ~ ~ e c a n ~ ~ recorded,n white for aider cohorts the t only s i ~ ~ ~ incrcasr: twas recorded in Oroniia. c a ~ The only s ~ ~ i ~ c decline in a n t u n c n i ~ l o ~was~observed inAddis, for youth. ~ n t 1.41 Analysis o f the d e ~ ~ ~ of~u ~i e~n t~ sp l o ~allowstus to i d e n t the~ effects o f i ~ e n ~ ~ 25+, Having child beingun~n~ployed 1.42 Thc role o f mple aged 1 5 1 the l~kel~ho~d o fbeingu 11 levels of e d ~ ~ a ~ j o ~ above grade 4, with the exce 11 dccrcascs that r o b a b j ~ ~ tIn.Addis, how ~ not s ~ ~ n ~ ~ c a n ~ . uschold ~ ~ i s ~ i ~e ~ p l ~ ~ o ~ with~ at ,~ i greater likelihood that the ~ n e ~ ~ p l willebe living tvith other adults aid ~ o t e n t i ~ l o y ~ b ~ ~ a d w ~ n nThisr ~effect i s s i ~ i ~~ ~t ~ce ~~ i for ~~ e~ d ~ e ~ ~n a\voml~~i. 1.43 R e s t ~ ~ cthe~ analysis EOthe age group 15-25 shows rhat therc are no s ~ g n i ~ ~ a n t t n ~ d i f f c r c ~ ~ cby gender, with the exception o f Addis, tvhere xvomen are less likely to be s ~ n e m ~ ~ o y eFurther, the effect of age on u n e ~ p l ~is~at~least~10 times h ~ ~ lfore ~ d . e t i this age group (over this group age spm) than for the 25+, The effects ofeduca~ionon the ~ ~ o b ~ b iofi t~y l ~ ~ ~arem much stronger in~Addis as~opposed to other urban p l o ~ ~ areas (where grade 1-4 and some higher education have no s ~ g n ~ ~ ceffect). a n t 20 Wages andEarningsin20055 1.44 The 2005 LFS includes data on earnings from the main occupation. This type o f data is generally difficult to collect, with concerns about underreporting and the imputation o f non-monetary components o f compensation. Because the compensation o f the self-employed and unpaid workers in the family business are not reported, earnings also provide an incomplete picture o f returns to work. Nevertheless, earnings data are essential to understanding the structure o f the economy (looking at compensation for different sectors, skills, activities) and to linkinglabour market analysis with poverty. 1.45 The Annex (Table 27) presents a profile o f wages by education and type o f employer for the formal and informal sector (informal is here defined as employees o f enterprises who meet the three criteria o f having no book o f accounts; no license; and fewer than 10 employees, as well as anyone who is a domestic employee, self-employed, an apprentice, an unpaid family worker, or who is only paid in kind; see also section 2, Chapter 3 this volume). Inorder to assess the level o f wage compression in each sector, both mean and median wages are shown. Mean wages are highest for NGO workers, while medianwages are highest for government. The lowest mean andmedian wages are domestic workers, with the private sector a distant second. 1.46 The median salary level for the most highly educated across sectors is quite similar, around Br 800-1,000, with the exception o f the NGO sector where it i s much higher at Br 1,350. In other words there appears to be a going rate for the highly educated, at least across their main employers (the Government, the private sector and parastatals). For the most highly educated, the largest difference between the mean and the median wage is for the "employee-other" category, and for the informally employed. 1.47 The Government pays more than the private sector at almost all levels o f education. The difference in median wages i s greatest for those with non-completed general education (who are paid twice as much by the government) and for illiterate workers (those who are illiterate and working inthe private sector earn about 60 percent o f what their counterparts working for the government earn). The gap is less for workers with primary education. 1.48 Government workers are paid on a comparable scale to NGO employees, with the most glaringdifferences being that those with some higher education are paid on average only about 70 percent o f what their NGO counterparts earn. Incontrast the illiterates are paid 30 percent more. 1.49 Table 28 in the Annex presents a profile o f wages by education and industrial sector. With the exception o f the hotel and restaurant sector, this shows relatively homogeneous wages for the highly skilled. Illiterates earn extremely low average wages (Br 122 a month, or about 14 dollars), but can expect twice as much in more capital intensive sectors such as in the social sectors, transport and communication and some 'In what follows the terms "wages" and "earnings" are used interchangeably. 21 types o f manufacturing. Those with an intermediate skills profile, such as those who have completed grade 8, find the highest earnings in transport and telecommunications, the public service, public utilities andthe financial sector. 1S O Finally, Table 29 provides a profile o f earningsby occupational status andtype o f employer. Senior officials and managers appear to receive the highest pay in the NGO sector, while parastatals, the Government and the private sector offer on average very similar earnings (though the median i s lower in the private sector).6 In the case o f professionals, NGOs offer almost 50 percent more than the government sector, while differences between government andparastatals are 30 percent. 1.51 The observed differences in earnings reflect the different productivity levels in different sectors, different employers' compensation policies, and different individual characteristics. Wage regressions help us to understand how observed characteristics affect earnings. 1.52 The main findings (Table 26; see also Annex Table 30) are that: Average returns to education increase with the educational level. Taking the illiterates as the reference group, the impact o f education on earnings ranges from 26 percent for those with grade 1-4 to 130 percent for the highest skilled. Women are paid 22 percent less thanmenwith equal personal characteristics even when controlling for selectivity. Note that women are more likely than men to self-select into the workforce according to their personal characteristics (e.g. because o f their skillprofile andmarital status), and low wages are systematically related to the probability o f not working as employees. For men, in contrast, the selection mechanism has an opposite effect, with men whose characteristics are more likely to result in lower wages being more likely to enter the labour market as employees.' 6The extent to which non monetary elements o f compensation are reflected in these estimates is unclear. In particular, the Government and the parastatals might offer additional benefits in terms of pension benefits, job security etc. which are unlikely to be captured by these estimates. 'Different specifications have been tried to account for the fact that individual choices are affected by their own characteristics. The pattern of results i s similar across the OLS and Heckman specification though controlling for selectivity attenuates, as expected, the effect o f the independent variables. O f note is that in our Heckman specifications the effect o f selectivity i s negative, suggesting that people whose expected earnings are higher than those offered by the market do not take up employment as wage earners (the only category for which we have recordedearnings). This contrasts with (the so-called "search hypothesis") that unemployed people only accept ajob if the earnings are higher their reservation wage. This finding might reflect a greater likelihood to earn more inother types o f employment (e.g. self-employment) or reflect the so-called "crowding out" hypothesis (Nicaise, 2001). Such hypothesis suggests that a group which is relatively underrepresented among wage earners but with a high-wage earning potential (e.g. because highly educated) displaces lower-wage earners by taking lower-skilled jobs. Inaddition, those who remain at their jobs accept lower pay inorder to avoid displacement. Insuch view a negative Lambda impliesthat lower wages are the price for higher chances o f employment. 8Nicaise (2001) finds similar gender patterns. 22 0 Women face a different structure o f returns than men - returns are higher for all levels o f education above completed general, for higher education they are 147 percent against 132 percent for men. 0 There i s only a minor difference inthe returns of parastatal workers vis-&vis civil servants (except for women in parastatals). Interestingly, controlling for individual observed characteristics, the private sector earns 35 percent less on average than civil servants (this penalty i s higher for women - 57 percent - and in urbanareas outside Addis). 0 Differences between Addis Ababa and other regions are small but significant for the largest regions. Somali and Gambella enjoy higher earnings, while SNNP, Amhara, Oromia, Tigray and Benishangul andhave lower earnings. 0 Differences in the structure of returns also emerge when splitting the sample betweenAddis and other urban areas. Inparticular, inother urbanareas, returns to education are significant and higher thaninAddis for all levels o f education. 23 k E I - Table 26: HeckmanRegressionson Earnings, sample 15+ LFS2005 All Male Female Addis Non-Addis Female Age squared Informal educ. Grade 1-4 educ. IGrade 5-8 educ. I(0.04) I(0.04) I(0.03) I(0.04) NonCompleted education I0.789** I0.626** II(0.04) 0.505** I0.617** I0.738** (0.07) (0.06) (0.05) (0.06) (0.06) Complete general educat. 0.805** 0.769** 0.953** 0.588** 0.968** IBeyond (0.03) (0.04) (0.04) (0.03) (0.04) general education 1.077** 0.999** 1.157** 0.996** 1.139** I(0.03) I(O.05) I(O.05) I(0.06) I(O.05) Hieher education I 1.296** I1.316** I1.474** I1.158** I1.412** ITrainine Parastatals IPrivate emDlovees NGOemployees (0.07) (0.08) (0.12) (0.08) (0.09) 55,423 24,278 31,145 16,178 39,245 Dependent variable: log o f monthly earnings. Reference categories: "Illiterate" (educationa 1 attainment), "Addis Aballa'' (region), "Male" (gender), "Government Employee" (employment status). Standard errors i.nparenthesis. Regional controls not shown. *significant at 5 percent **significant at 1percent. 24 v1 N 01 01 0 ;c, "co 32 27 Table 30: Heckman Regressionson Earnings,Sample 15+ LFS2005 (0.03) (0.02) 0.066** 0.064** (0.00) (0.00) , I -0.001** -0.001** (0.03) (0.04) 0.322** 0.488** (0.03) (0.04) 0.617** 0.738** (0.06) (0.06) 0.588** 0.968** (0.03) (0.04) 0.996** 1.139** (0.06) (0.05) 1.158** 1.412** (0.04) (0.05) I 0.009, 0.254** (0.04) (0.03) -0.021 -0.084* (0.03) 0.075 (0.05) -0.958** (0.04) -0.636** (0.09) -0.016 (0.04) 0.198** (0.03) (0.04) (0.05) (0.04) Amhara -0.254** -0.161** -0.411** -0.137** (0.02) (0.03) (0.03) (0.03) Oromia -0.193** -0.154** -0.265** -0.080* (0.02) (0.03) (0.03) (0.03) Somali 0.250** 0.198** 0.107* 0.316** 28 All Male Female Addis Non- Addis I(0.04) I(0.04) (0.05) (0.04) Benishan -0.122** -0.053 -0.167** -0.163** (0.03) (0.04) (0.04) (0.03) SNNP -0.264** -0.221** -0.355** 0.266** (0.02) (0.03) (0.03) (0.04) Gambela 0.175** 0.158** 0.155** 0.044 (0.04) (0.05) (0.05) (0.04) Harari -0.057 -0.108** -0.019 0.152** (0.03) (0.04) (0.04) (0.04) Dire Dawa 0.020 0.028 -0.007 (0.04) (0.04) (0.05) Constant 4.405** 3.916** 4.088** 4.450** 3.999** (0.13) (0.11) I(O.11) (0.10) (0.11) -0.289** -0.370** (0.02) (0.02) -0.011** -0.003** 0.000 (0.00) (0.00) (0.00) -0.258** 0.258** 0.342** (0.02) (0.02) (0.02) -0.020** -0.107** -0.088** (0.01) (0.01) (0.01) -0.025* -0.033** -0.054** (0.01) (0.01) (0.01) 0.019* 0.028** 0.049 ** (0.01) (0.01) (0.01) (0.01) (0.01) Informaleduc. -0.063 -0.192** 0.052 -0.018 -0.113 (0.05) (0.07) (0.07) (0.06) (0.07) Grade 1-4 educ. 0.056* 0.137** -0.009 0.095* 0.015 (0.03) (0.05) (0.04) (0.04) (0.04) Grade 5-8 educ. -0.339** -0.247** -0.419** -0.201** -0.409** 29 Non- Addis -0.027 (0.04) 0.042 (0.05) -0.153** (0.04) -0.139** (0.04) -0.500** (0.04) -0.180** (0.04) -0.230** (0.05) -0.166** (0.05) -0.115* 1 (0.04) 0.000 Dependent variable: log of monthly earnings. Reference categories: "Illiterate" (educational attainment), "Addis Ababa" (region), "Male" (gender), "Government Employee" (employment status) N o n married" (man ed), "Non trained" (training). Standard errors in parenthesis. Significance c selection equation (** next to Lambda) from Wald test of independent equations. *significant at 5 percent;**significant at 1percent. 30 Table 31: OLSRegressionson Earnings, sample 15+ LFS2005 OLSregression on dependentworkers aged 15 andmore Male Female Addis Non- Addis -0.363** -0.341** -0.372** (0.02) (0.01) 0.064** 0.047 ** 0.063** 0.065** (0.00) (0.00) -o.ooo** -0.001** -0.001** (0.00) (0.00) 'W 10.281 ** 0.302** 0.133* 0.391** (0.06) (0.06) , (0.05) 0.264** 0.175** 0.295** (0.02) (0.03) (0.03) Grade 5-8 educ. 0.420** 0.381 ** 0.267** (0.02) (0.02) (0.03) NonCompleted education 0.599** 0.555** 0.385** (0.02) (0.04) (0.04) Complete general educat. 0.861** 0.940** 0.580** 0.998** (0.02) '(0.03) (0.03) (0.03) (0.02) Beyondgeneral education 1.061** 0.977** 1.151** 0.807** 1.150** (0.02) (0.03) (0.04) (0.04) (0.03) Higher education 1.364** 1.318** 1.443** 1.207** 1.444** (0.02) (0.03) (0.04) (0.04) (0.03) Training 0.310** 0.328** 0.220** 0.254** 0.334** (0.02) (0.03) (0.02) (0.02) Parastatals -0.085** -0.009 -0.200** -0.027 -0.086** (0.03) (0.04) (0.04) (0.03) -0.236** -0.572** -0.241** -0.344** (0.02) (0.02) (0.02) (0.02) 0.183** -0.064 0.191** 0.076* (0.03) (0.04) (0.04) (0.03) Domestic employees -0.874** -0.668** -1.068** -0.703** -0.963** (0.02) (0.03) (0.03) (0.03) (0.02) Other emlovees -0.566** -0.480** -0.768** -0.266** -0.639* * I(O.04) (0.04) (0.08) (0.07) (0.04) Tigray -0.14 1 ** -0.052* -0.259** -0.021 (0.02) (0.03) (0.03) (0.05) 0.071 0.098 0.048 0.202** (0.05) (0.07) (0.06) -0.156** -0.404** -0.150** (0.02) (0.02) (0.05) 31 Dependent variable: log o f monthly earnings (missing values were recoded to zero). Reference categories: "Illiterate" (educational attainment), "Addis Ababa" (region), "Male" (gender), "Government Employee" (employment status) "Non married" (married), "Non trained" (training). Standard errors in parenthesis. *significant at 5 percent; **significant at 1percent. 32 2. YOUTH EMPLOYMENT INETHIOPIA9 Highlights Labour market supply trends pose major challenges, especially for youth. As the educational profile o f younger generations in urban areas i s rising, labour market participation is likely to increase, since better educatedpeople are more likely to be economically active. Coupled with a growing youth cohort, this implies a needfor acceleratedjob creation, and an improvement inthe quality o f available jobs to alleviate a skills mismatchurrently, more than half o f youth work in the informal sector, often as unpaid family workers. The youth unemployment rate is about double that of adults, and underemployment i s frequently a problem as well, though not disproportionately so. Youth labour market status i s more volatile than that o f adults (though adult women also experience larger fluctuations than do adult men). This suggests youth may be more vulnerable to changing economic conditions, which i s consistent with the expectation that thosejust entering the labour market will be the first to experience difficulty ifjob creation slows. The most recent available data, from 2005, show encouraging trends in youth employment outcomes, but sustained growth o f decent employment for young people will be required for many years if Ethiopia i s to meet the MDGsand its other development objectives. While broad-based economic growth offers the most important route to sustainablejob creation, other interventions can improve employment opportunities. These include measures to enhance access to education, credit, and information, as well as second-chance programs like basic skills and literacy training. Empowering young women will also be central-although the gender gap i s shrinking, young women are still less likely than young men to be employed (and also less likely to be inschool). The school to work transition, whichis longinEthiopia compared to other countries, is even more protracted for women. Job quality also tends to be very low; about one- fourth of employed young women are engagedas domestic workers. Increasingopportunities for women by raising their human capital and by ensuringa favorable policy environment will have major payoffs for growth and poverty reduction. Although these challenges are daunting, the growing youth population also offers an unprecedented opportunity to boost growth. Eventually, as the demographic transition progresses, Ethiopia's dependency ratio will fall, meaning that the active population has fewer dependents to support. This will offer a window of opportunity for growth acceleration, as it did in parts of East Asia. Making sound investments and policy choices now will be essential to ensuringthat Ethiopiacanreap the benefits ofthis demographic dividend. 9 Portions o f this chapter draw on Getachew and Kallaur (2005), a background paper prepared for the World Bank study "Youth inAfrica's Labour Market." 33 1.Introduction 2.1 The global youth population, which exceeds one billion, i s larger than ever before-and the youth cohort is growing faster in the Sub-Saharan Africa (SSA) region than anywhere else. There are currently about 200 million people aged 12-24 years in Africa, and by 2030 the continent i s expected to catch up with the youth population o f East and South Asia inabsolute numbers (World Bank 2006a). Africa's youth are living in rapidly changing societies, and although in some respects they enjoy greater opportunities thantheir parents did, they also face serious challenges, including highrates o f unemployment and underemployment, HIV/AIDS and other health risks, gender discrimination, and in some areas the persistence or legacy o f violent conflict. Because the share o f youth in the population has not yet peaked in Africa, these challenges will continue to mount. 2.2 Becoming economically independent i s one o f the hallmarks o f the transition to adulthood. Indeed, finding a decent job offers the surest hope for millions o f young people to lift themselves out o f poverty, and to break the cycle o f intergenerational transmission o fpoverty, since many youth are also parents. Yet inAfrica, many children are already active in the labour force, meaning the entry into employment happens prematurely. Among other consequences, this limits the opportunity for schooling and can therefore have a permanent negative impact on employment outcomes (UCW 2006). 2.3 Even those youth who do acquire formal education frequently experience great difficulty when entering the labour market. The youth unemployment rate indeveloping countries tends to be substantially higher than that o f adults. Based on data from 60 developing countries around the world, Fares, Montenegro, and Orazem (2006) found that the typical youth spends about 1.4 years in unemployment or temporary jobs before settling into stable work, andinEthiopiathe transition i s much longer for educated youth. Andemployment alone does not necessarily signal apositive outcome, since a largeshare o f youth are in unpaid (for example a family business) or low-paying jobs with limited prospects, and frequently in unhealthy or unsafe conditions (Denu, Tekeste, and van der Deijl2005). 2.4 Employment opportunities for youth are partially determined by individual characteristics such as educational achievement, health status, and personal networks. Yet prospects are also to a large extent influenced by the prevailing economic and demographic conditions. On the demographic front, cross-country analysis shows that the youth unemployment rate tends to increase as the share o f youth in the labour force increases (Fares, Montenegro, and Orazem 2006). This suggests that as the "youth bulge" expands, young people may experience increasing difficulty in finding employment. Ethiopian data suggest that this effect i s much more pronounced for less educated youth, who comprise a large share o f youth and are already among the less well off. Downturns inoverall economic demand also affect youth prospects. A deterioration in adult employment prospects tends to be associated with disproportionately worse outcomes for young people-and again, this particularly affects those with the least education (World Bank 2006a). 34 2.5 This chapter presents key trends related to the well-being of youth and their labour market participation and outcomes in Ethiopia, and describes relevant aspects o f Ethiopia's institutional environment. The focus on youth as a distinct group rests on several points: 0 There are more youth (inabsolute and relative terms) than ever before inEthiopia, signaling a major demographic shift inthe labour force, which particularly affects the prospects o f young people as well as having broader implications. 0 The current generation o f Ethiopian youth has higher levels of humancapital, and thus different career expectations and potential than its predecessors; yet job creation has not kept pace in quantitative or qualitative terms. The Participatory Poverty Assessment recently conducted by the GOE found that youth unemployment i s a major contributor to vulnerability in urban areas (MOFED 2005). 0 Youth represent a politically as well as demographically significant group that will have a key role in shaping the country's development path and democratic transition. A young workforce has the potential to accelerate productivity growth and spur better governance, but without adequate opportunity to contribute productively, youth can turn to risky behavior with possible implications for the incidence o f crime, druguse, andHIV/AIDS, for example (World Bank 2005d). 0 Ethiopia i s experiencing rapid societal changes that disproportionately impact new labour market entrants. These changes include urbanization and a falling fertility rate in cities and towns that changes the nature o f family life, increasing career opportunities for women, the HIV/AIDS epidemic, andrising technological adoption, the policy implications ofwhich need to be better understood. The government i s increasingly focusing on the needs o f the growing youth cohort, evidenced for example by the establishment o f a Ministry o f Youth and Sports (in 2001) to ensure that issues affecting young people are approached in a coordinated, multi-sectoral manner. 2.6 This chapter begins by briefly summarizing key demographic and social indicators, thenproceeds to key labour market indicators, including activity, employment, and unemployment rates. It provides an overview o f existing policies and programs targeted at youth inEthiopia, and ends with some concluding remarks. The chapter uses the International Labour Organization (ILO) definition o f youth (age 15-24) to allow for international comparability. The prevalence o f child labour in Ethiopia (perhaps the highest in the world) also has important implications for youth employment outcomes, but is not dealt with inthis study; see UCW (2006) for a detailedtreatment. 2.7 The data analysis in this chapter depends mainly on the 1999 and 2005 Labour Force Surveys (LFS) and 2003 and 2004 UBEUS," as well as the 2006 Addis Ababa 10The Urban Bi-annual Employment Unemployment Surveys (UBEUS) were conducted in October 2003 and April 2004 by the CSA. These surveys are the first 2 waves o f a series o f cross-sections intended to 35 Labour Market Survey (ALMS); see Chapter 1 o f this volume for more details. Unless otherwise indicated, the chapter refers to urban youth only, consistent with the urban focus o fthis study. 2. DemoPraDhic Shifts 2.8 Ethiopia's population has grown rapidly in recent decades and was estimated at 71 million people in 2004, only about 15 percent o f whom live in rural areas (CSA 2004a)." The population is predominantly young, with children and youth (0-24 year olds) comprising well over half. Ethiopia's low life expectancy (44 years)12means that persons aged 65 and above represent less than 3 percent o f the total population. Youth have grown from about 14 percent o f the population in 1984 to about 20 percent currently. According to Central Statistical Authority (CSA) projections, this proportion i s expected to remain roughly constant as the youth population grows in absolute numbers from about 15 million in2005 to 26 million in2030. 2.9 Inrural areas, 45 percent of the population is under 14 years; the proportion in urbanareas is smaller, but still over a thirdo fthe population (35 percent). Youth account for a slightly larger percent o f the population in urban (23 percent) than in rural (20 percent) areas. The population pyramids below (Figure 3) illustrate the striking difference between the demographic makeup o f rural areas, where the fertility rate is about 6.4 percent, and urban areas, where it is about 3.3 percent. In the exceptional case o f Addis Ababa, the total fertility rate o f 1.95 percent is below replacement and comparable to the rate inindustrial countries (World Bank 2005a). However, population growth i s supplemented by an influx o f migrants, a large share o f whom are youth (see Chapter 4 in this volume; for an extensive discussion o f population issues, see World Bank2006b). take place twice a year. The series was interrupted and there is debate on whether it should now become annual. These surveys are representative o f urban areas nationally and by region, and they cover all zones o f the country with the exception o f three zones in Afar and 6 in Somali. The questionnaire used is very similar to the one adopted by the 1999 LFS, one notable exception being the inclusion o f wage data. UBEUSdata are notusedinChapter 1due to report have notbeenusedto calculate trends due to doubts on their overall comparability with the LFS surveys arising from differences in sampling and possible seasonality effects. The most recent census was conducted in 1994. l2 2001 UNPopulation data (cited inWorld Bank 2005a). 36 Figure 3: PopulationPyramids,Urban vs. Rural, 2004 10-14 15+ H 1014 75+ 60-64 65-69 65-69 55-59 60-64 55-59 45-49 50-54 50-54 40-44 45-49 35-39 I 40-44 30-34 35-39 25-29 30-34 20-24 25-29 15-19 20-24 10-14 15-19 10-14 0-4 5-9 0-4 5-9 - 20% 15% 10% 5% 0% 5% 10% 15% 20% Source: CSA 2004a, own calculations. 2.10 In the absence of age-disaggregated poverty headcounts, we can only indirectly gauge the monetary well-being of youth. An abundance of literature confirms that poverty tends to be positively correlated with household size (for example, Bigstenet al. 2003). Figure4 presents household size by age cohort (the left graph shows urban areas, and right graph shows rural areas). Inurban areas, it is clear that youth are less likely than adults to live in small households; only about 11percent of 20-24 year olds live in households o f 1 or 2 members, while about 50 percent live in households o f 6 or more. Teena ers, unsurprisingly,seem more likely than young adults to still live intheir family home.P3 2.11 Rural youth are more likely to live in small households, presumably because they are starting their own families inlodging separate from that of their family (although this pattern may be changing due to land scarcity in densely populated rural areas). It i s possible that the transition to living independently is more protracted inurban areas. To contextualize these data, Box 4 presents some qualitative evidence o f youth well-being. Figure 4: HouseholdSize Distribution, Urbanvs. Rural areas, 1999/00 (share of the age group that lives inhouseholdsof each size) 18% - I 16% - ;1 5 14%- 2 2 % - -15-19 10% - --t15-19 -+- 8 % - -C-20-24 20-24 2I 6 % - 4% - 2% - 0% 7 1 2 3 4 5 6 1 8 9 1 0 + 1 2 3 4 5 6 7 8 9 1 0 + # of HouseholdMembers # of HouseholdMernbers Source: CSA 199912000, own calculations. l3"Teenagers" refers to the 15-19 year old group, and "young adults" to the 20-24 year old group, while "youth" and "young people" are usedto refer to the entire 15-24 year old cohort. 37 Box 4: Voices of Ethiopia'sYouth In 1999,aspart of qualitative background workfor the 2000/01 World Development Report, the World Bank conductedfocus groups withpoor people around the world to learn more about the challenges they werefacing, in their own words. Thefollowing is excerptedfrom the Ethiopia country report. View of youthfocus group members in the urban site of Dessie Zuria Wereda (urban community of about 10,00Opeople,part of Dessie town, in the highlands north ofAddis Ababa): "We have nojob andwe are dependent on our families. We have no food to eat. Our problem o f food can be solved ifthere i sjob opportunities or employment. For instance they said there was a Mennonite mission project and 50 persons were employed here in constructing stone dams through food for work inwhich number of people benefited from 3 kilos of wheat paid them per day. This was a great thing for us. However, this project closed and we missedthis advantage. Besides this, there i s a credit facility through kebele. However, most o f us can't participate since the conditions ofparticipating are above our means." Urban site of Ada Liben Wereda(part of Debre Zeit, a town withpopulation of 72,000, about 45 kmfrom Addis Ababa): The youth see very little, ifany, opportunity for them inthis community. They see their parents or other adults thrown or forced out o f their jobs for reasons they cannot comprehend. They are at a loss as to what they could do. They seem to have no means to bringabout changes in their lives. There are no social or political groups that cater to their needs. As a result many are desperate and demoralized. Addis Ababa Sites: Among the youth focus group, unemployment i s the main cause of poverty. Unemployment i s causedby populationpressurewhich i s a result o f migration and population displacement. There are, the group thought, no new job opportunities for the young generation. The group believed Ieducational that the impact of poverty was hunger, exposure to disease, theft, drug addiction, and lack o f opportunities. The issue of drug addiction was raised by the young focus group only. Source: Rahmato and Kidanu 1999. 3. HumanDevelopment 2.12 Health. Evidence from the 2004 Welfare Monitoring Survey shows that 5-25 year olds, on average, enjoy the best health o f any age group, as might be expected (CSA 2004b). However, high fertility rates and reproductive health challenges put young women at risk, and young people are disproportionately likely to contract HIV/AIDS. Due to the positive correlation between poverty rates and household size, the fact that many women start families at an early age is cause for concern (Figure 5). While the peak childbearing years for women nationwide are ages 25-29 (ages 30-34 in urban areas), fertility i s high among 15-24 year olds, with about 16 percent o f 15-19 year old women already having a child or being pregnant (the likelihood o f this i s about twice as highinrural as inurbanareas). Given that 15-19 year old girls have double the risk of maternal mortality that 20-24 year olds face, according to international data, this poses a major threat to young women (ICRW 2005). 38 2.13 Youth have the lowest percentage o f satisfied demand for family planning methods among women o f reproductive age-only 9 percent o f demand among 15-19 year olds i s satisfied, and 17 percent among 20-24 year olds, relative to 18-23 percent for other age groups (these are national figures; in urban areas, overall met demand i s 59 percent, versus 10 percent in rural areas, but age group breakdowns are not available) (World Bank 2005a). The prevalence o f female circumcision is very high (about 80 percent inurbanas well as rural areas). Nationwide, a majority o fwomen support female circumcision, although the share is lower among youth: about 59 percent o f 20-24 year olds approve, compared to 67 percent o f 45-49 year olds (although women inrural areas are more than twice as likely as those in urban areas to approve o f the practice; World Bank2005~). Figure 5: Age-Specific Fertility Rates, 2000 300 250 200 +-Overal -+-Urban 150 +-Rural 100 50 0 15-19 20-24 25-29 30-34 35-39 4-44 45-49 Age Source: World Bank2005a. 2.14 As shown in Table 32, HIV prevalence peaks among the population aged 15-34, with the urbanprevalence rate far surpassing the ruralprevalence. However, the Ministry o f Health (2004) reports that the rate i s leveling off incities, while it continues to grow in rural areas. Among some groups, rates are much higher; one study recorded the prevalence among female sex workers at 73 percent in certain urban areas (UNAIDS 2006). The rapid spread o f the epidemic i s a major cause o f death among youth, and i s expected to reduce life expectancy overall. Table 32: HIV Prevalenceby Age Group (percent), 2003 I 15-24 4.3 8.6 125-34 I 12.5 II II I 3.9 8.1 35-49 10.3 3.6 6.3 Total II 12.0 4.1 IIII 8.2 I Source: Ministry of Health2004. I 2.15 Education. A major supply-side push by the Government has significantly increased Ethiopia's traditionally low levels o f school enrollment, particularly in basic education (Figure 6 shows total enrollment for urbanand rural areas combined). Gender 39 disparitiesare substantial, but less so inurbanthan inrural areas. While primary (grade 1-8) enrollment is approaching parity in urban areas, girls only comprise 36 percent o f first cycle secondary enrollment, and 27 percent o f second cycle secondary enrollment (GOE2005). Figure6: Nationwide SchoolEnrollmentTrends, 1967-2005 / 9000 8000 I 7000 -a G [3 6000 Gr. 1-4 - 5000 d3e4000 3000 4 2000 1000 0 Source: Government of Ethiopia Education Statistics Annual Abstracts, 1994-2002, cited in World Bank 2005b; also GOE 2003,2004a, and 2005. 2.16 The investment in human capital i s already becoming visible in the skills profile o f the labour force, especially for youth, and can be expected to impact job opportunities for the upcoming cohorts o f young people. As shown in Figure 7, the illiteracy rate among young males i s less than a third that o f adult males; about the same i s true o f females, although illiteracy is much more common at all ages for women. The differences are strikingat other levels as well: the percentage o f male youths with a "non- complete general education" i s about 2.5 times as high as the same share o f male adults (or 1.5 times for women). The pattern reverses at the upper levels since youth pursuing higher education are likely to still be inschool. Figure7: Supply of Skills inthe UrbanPopulation,MalesandFemales,2005 1 50 , 60 45 . 15 40 7 50 2 P 3 5 - 40 3 O - lld rn Age 15-24 2 5 - .Age 15-24 0 Age 25+ 30 0 Age 25+ 20 P 10 0 Source: LFS 2005, own calculations. 40 4. LabourMarketParticipationandEmulovment 2.17 Patterns o f labour market participation (activity) and employment rates vary significantly by age and gender (Figure 8). Figure8: Trends inKeyLabourMarketIndicatorsfor Different Age Groups, by Gender (inpercent) 80 ...................x #:50- M 70 - +Activity Age 15-24 60- --c-ActivityAge25+ c .- .A.. EmploymentAge 15-24 s ...?e..EmploymentAge25+ sL' ::I 40- -.rn.-UnemplAge 15-24 3 0 - -.----X +.- ._.-. , 0 1999 2005 100 90 80 2 70 ______. 160 -Activity Age 25+ % x.. ................ *X ..-A-..EmploymentAge15-24 50 Y ..-x.. EmploymentAge25+ 22 g 4 0 A.. .................A .-.rn.-UnemplAge15-24 30 *.-----.-----' -. -UnemplAge25+ X 20 e 10 0 1999 2005 Sources: LFS 1999 and 2005, own calculations. 2.18 Activity and employment rates are substantially lower among women overall, although the increases observed in female activity and employment over the period are statistically significant for both youth and adults. Activity among young men has declined, however, a change that i s also statistically significant and to at least some extent associated with rising school enrollment. Meanwhile, unemployment declined significantly for both male and female youth (see vol. 11, Chapter 1, for more discussion and analysis o f key trends). The wide divergence between the labour market status o f male youth and that o f adult men i s consistent with the fact that educational participation 41 i s higher among males than females (young women are more similar to their adult counterparts in terms o f labour market participation, because they are less likely than young mento be inschool rather than working). 2.19 Further disaggregation reveals that young adults are, unsurprisingly, much more likely than teenagers to be economically active. Again, there are significant gender differences. Teenage males are about half as likely to be active as young adult males, while teenage females are about 60 percent as likely as the 20-24 year group to be active. As shown in Figure 9, the number o f hours worked by females in urban areas peaks among youth (age 15-19), and declines fairly steadily thereafter. Beginning with the 20- 24 year old age group, males work more hours on average (while females presumably have a higher domestic work burden). Figure 9: Number of HoursWorked per,Week, 2003 (currentlyemployedpopulation) g 50 40 f 30 0 s 20 +Male I" 10 0 1 I I I , I I I I I I I I 10- 15- 20- 25- 30- 35- 40- 45- 50- 55- 60- 65+ 14 19 24 29 34 39 4.4 49 54 59 64 &e group Source: UBEUS 2003. 2.20 Among adults, labour market participation increases steadily with education, from 62 percent activity among illiterate adults to 95 percent among adults with at least some higher education (Table 33). The pattern i s much less straightforward among youth, at least partly because some are still in school. Ethiopian students begin with 8 years o f primary education (subdivided into basic primary, Grades 1-4, and general primary, Grades 5-8), then proceed to the first cycle o f secondary education (Grades 9-10>, the completion o f which is considered "General Education"-r they have the option o f applying to a Junior Technical and Vocational Educational Training (TVET) program. Students who pass an exam at the end o f Grade 10 are eligible for preparatory secondary (Grades 11-12), and then can compete for a place in a tertiary institution. Those completing Grade 10 who are not admitted to preparatory secondary, or who prefer, can compete to enroll ina medium-level TVET program (World Bank 2005e). 2.21 Unlike adult activity rates, youth activity rates by educational level display an inverted-U shape, with the highest rates at the ends o f the spectrum (illiterate and higher education), and the lowest rates for non-complete general education. The gender differential in participation decreases dramatically at higher education levels, especially for youth, reaching near parity after Grade 5-8. A similar but less pronounced pattern i s 42 at cond~tjonsfar female p ion in the labour cult to compare since young Table33: Activity Ratesby Age, Gender, and Education Level, 2QOS Age 15-24 I Age 25+ Male I Fernate 1 TOM Male f 1 Female I Total o f c r n ~ i ~ differse markedly between the youth and a y ~ ~ ~ ~ a r a s ~ ajobs.l ~ ~ ithe~adtilt pop~~l~tion, arc much more likely than tt'omct?to ~ a ~ ~ i n men ublic sector job, but the l~keli~~oodyouth is b ~ o a gender neutral. Private for d ~ ~ e m p ~ ~ ~~ iwih~i nincludes both format and ~ n f o wage~jobs) is somewtiat niare c ~~ ~ a c o ~ amr n~o ~youth than adults, t ~ ~ omore ~soi for men than women in both age ~~ ~ u ~ groups. ~ ~ ~ oun ~ ed~na ~g~ theacategory of ~ m p l o as andotnestic worker. ~ ~ e ~ ~ ~ ~ ~ ab^^^ the same ~ e r ~ e n ~ aofg eyouth and a ~ u l ~work in the ~ ~ ~ fsector, abutl the s o ~ c ~ ~ ~ i ~of co m ~i ~l o~~~varies~d r a ~ ~ a t ~ c a ~ l y ~ s e ~n i ---. . . . --I- Age 25+ Male IAge 15-24 Female `rota1 `Male I Female I Total 2.23 U ~ p a ~ d family work occupies about a quarter of E t h ~ o ~youth, compared to i a ~ just 6 percent o f adults. Nearly half of adults are self-e~pl~y~d, compared to less than a third of youth. This suggests that the j ~ ~sector serves~ as ~an j l~ i p o ~ entry~ point o ~ a n 43 to the labour market, but that starting a business i s not within reach o f all young people. Moreover, even though women account for just over half o f adult self-employment, young men are more likely to be self-employed than young women. This suggests that a 4 common path for women may be to enter the labour market as an unpaid or domestic worker, later becoming self-employed, while young men are somewhat more likely to enter the labour market as a self-employed worker, later to transition into the public sector or other wage work. 2.24 Youth generally earn lower wages than adults, which may be explained by their limitedwork experience: median youth wages per month are only about 30 percent o fthe adult median (the youth mean is about 40 percent o f the adult mean). Because these wages are not adjusted for time worked, they do not allow for an exact comparison of hourly wages. However, they do indicate that employed youth on average earn total wages barely above the lower poverty line and significantly below the upper poverty line.14 The medianwage o f Br 120 (about US$14, or under US$1 per day) helps explain why most youth continue to live with their families. 2.25 Well educated youth tend to fare better. The average monthly wage for youth with at least some higher education is 63 percent that o f adults, while illiterate youth make only about half the average wage o f illiterate adults. Education also helps explain the effect o f larger shifts on youth employment. As shown in Figure 10, youth employment elasticity with respect to their share in the population i s estimated to be negative and significant for those with little or no education. This means that less skilled youth are more vulnerable to fluctuations ineconomic conditions and to supply pressures from changes intheir cohort size (World Bank 2006a). At the same time, unskilled youth benefit relatively more than their skilled counterparts from increases in labour demand (as proxied by a rising adult employment ratio). 4 - U e (2) Elasticitywithrespect to changes Q, -LE$$ E>I 2 - Y O O n O - - I 8, Q, .E x z 6 .= '05 - 2 - U 0 - 4 - W u) -6- changes in youth cohort size Youth educational attainment Source: World Bank2006b. l4The lower poverty line is estimated at Birr 110 per month, and the upper poverty line at Br 163 per month (or about US$13 and US$19, respectively). These estimates are unweighted averages from the povertylines for urbanareas aroundthe country inWorldBank (2005~). 44 2.26 The extent o f underemployment provides another perspective on labour market outcomes. Youth tend to be disproportionately underemployed in developing countries, but the evidence for Ethiopia shows that youth underemployment is more a rural than an urban phenomenon (Denu, Tekeste, and van der Deijl 2005). Overall, about 36 percent o f men and 31 percent o f women in urbanareas say that they are "available and ready to work more hours," indicating that they are underemployed (on a time basis; this o f course does not capture whether their skills are fully engaged). The lower prevalence o f underemployment among women presumably reflects their heavier domestic work burden. 2.27 Inurbanareas, underemploymentpeaks among 30-34 year olds, who are intheir prime working years and also likely to have dependent children, but young adults are already near peak rates (for teenagers, underemployment i s somewhat less common). The general trajectory shown in Figure 11 suggests that underemployment is not particularly associated with youth, meaningthat it does not represent a disproportionately large share o f disguised unemployment among youth relative to adults. The more strikingfinding is the dramatic decrease across the board-for men and women, old and young-in the extent o f underemployment since 1999. This encouraging finding provides evidence o f a more robustjob market. Figure11: Underemploymentby Age Cohort:1999and2005 -Male - - - 99 -Male05 _-.*-*-Female99 - Ferrrale05 O I 15- 20- 25- 30- 35- 40- 45- 50- 55- 60+ 19 24 29 34 39 44 49 54 59 Age in Years Source: LFS 1999 and 2005, own calculations. 5. Unemplovment 2.28 Ethiopian youth face much higher unemployment rates than do adults, and rates are very high in absolute terms as well (Figure 12). However, the adult unemployment rate i s also high (11percent). The situation in Tanzania provides a contrast. About 40 percent o f young adults inDar es Salaam, Tanzania arejobless (Kondylis and Manacorda 2006), outstripping the youth unemployment rate in Addis Ababa (34 percent)-but the adult male unemployment rate inDar es Salaam is negligible. This comparison suggests that while unemployment in some countries like Tanzania may be seen as a transitory 45 phenomenon," in Ethiopia it i s a structural issue that disproportionately impacts youth. Moreover, unemployment rates in Ethiopia are persistently high-the prevalence o f youth unemployment in urban areas more than doubled between 1984 and 1994, and has hovered around the same level since then (Denu, Tekeste, andvan der Deijl2005). Figure12: UrbanYouthUnemploymentRatesin SelectedAfrican Countries , - - . I I I 90 80 70 6 60 50 4 40 $ 30 20 10 0 Source; Leibbrandt and Mlatsheni (2004), except Ethiopia, UBEUS 2004, own calculations; "youth" defined as ages 15-24. Note that the years vary. 2.29 Overall, Ethiopia's youth unemployment rate i s approximately d o d e the a d t rate, though the differential varies by region (Table 35). Youth fare the worst in comparison to adults in Gambela and Harari, where the youth unemployment rate i s almost three times the adult rate; Afar has the smallest ratio (1.3). In Addis Ababa and Dire Dawa, where unemployment rates (for both groups) are the highest in absolute terms, the differential i s about average. Women at all age levels except those aged 50 years and older have a higher likelihood o f unemployment than their male counterparts (see Chapter 1). This i s unusual in that in most African countries, male youth unemployment rates tend to be higher (Denu, Tekeste, and van der Deijl2005). l5Itis also possible that inother countries youth unemployment is a new structural problem. 46 .___ ._. ' Region I Ratio 1 2.30 ~ ~ u c a t j oplays an inipo n in e ~ p ~ ua n~ e~ ~~ n~ ~As sftown in~ ~ o ~ ~ ~ . Table 36, ~ ~ ~ e ~rates are ~ ~ amongbetter e ~ ~ c a t >~uth- ~ l e n t e d cnt) is found a ~ o yo ~ n "beyond general" e ~ ~ ~ a ~ j o n , youth have the lowest rate (9 ~ e r c ~ ~Over time, however, e ~ ~ ~ ~sccnis~too pay ofE t ) . a t n among adults, those with SQIW higher e ~ ~ c have~ thc lowest u n e ~ ~ l orate~(5e ~ ~ t a ~ ~ n ~ percent). The ~ ~ ~ ~ i - s ~ ihowever, still seem to be a concern: 17 percent l l e d , u n ~ m for~adults with a~general ~education.~ This raises ~ u e s ~ i oabout a possible ~ o ~ ~ n s skills ~ ~ s ~~ a ~ ec labour~supply~and de~nand,and alsonabout the extent of "ituxury" h ~ ~ u n e ~ ~ ~ o ~ ~ ~ niftthose, who arc ~ i ~ e ~ p l o are - ~ e . y e dthose who can afford to be so. This is ~ i ~ c ~f~u ~sts~~~ ed~r ~while n~ o~~ irhere that u ~ e ~ ~ I orateseare tfairly high o n ~, ~ n even among illi~era~eadults (8 percent), who are ~ z i ~ h~~~yi l j ktoe "choose" ~ ~ u n e ~ ~ p ~ o ~ ~ ~ e n ~ . 2.31 Thc typical d u r ~ ~ i oofn youth ~ n ~ ~varies widelyo by~country, eand ~ ~ ~ ~ in t Ethiopia tends to be p r o r ~ ~ ~particularly for youth. A ~ ~ o rtodanalysis by Fares, e d , j ~ ~ , and Orazenz ( 2 ~ ~omo ~ g from ~scfioof ta~ \vork in ~ ~~ ~ i t ~andl ~ o ~ ~ ~ 47 for around 5 or more year b, ~ o ~ p ator 3e years in Uga ~ rindabaut 1 year in C6te ri~?eragef about 1.4 years. o 2.32 The goodnewsis that, when looking at the populrition as a whole, ~ ~ i ~ m p ~ o du~~tion~ ~ ~ ~is si ~o ~~~ikaFigure I. 3 presents median(left graph) andmean fright in ~ n g ~ r ~ ~a~ i ~~ ~ d~i p ~~ o~in~tmonths,n~by,gender and age group, and suggests several ~ ~ ~o i t kcy c o ~ ~ ~ u s i o~~nse. n ~ p l oduration ~ e ~ t ~ F i 14) is~ ~~h er ~for all ~ ~ g for ~ ~ k ages, and ~ ~ r o as e n as inen. Al~houghwomcn still ~ ~ p ~ ~longercspells of u n e ~ ~ ~ ~ a ~ ~ e n ~ ~ well e n e than men, the medianduration for young adults (age30-24) has s~~~~ s ~ ~ n j ~froma ~ ~ 1 ~ c about 2 years in 1999 to about I year in 2005 for both males and females. Youth (especially young women) do have longer p e ~ o d ~~ i ~ ~ ~ ~than psime~age e ~ t of p l o ~ g $hailin 1999,~ u ~that~the ~~ ~~ tmayu ~~ n~ j~ ~ ~ t 48 estive ofthc case o f e youth face, since some may be too ing at ~ob1~ssness~i.e. percethe 8 s t ~ d e nmay provide a better ~ ~ y_ ), Table37 compares youth time use inEthiopiaIO ries; in general, ~ t ~ ~ i oyouth seem to be inthe middle of the spectmni on p j a ~ ~ the various ~~dicators. higher percentage of youth are only in school (46 ~ e r c e nthan A t ~ in several other countries, such as ~ o z a m b i q ~(30 percent) and ~ ~ a (41 percent). e n ~ a A~out28 percent of ~ t h ~ ayoutha in urban areas were jobless in 2001, meaning ~ i ~ ~ neither in school nor employed igher than in ~ ~ ~(24i pd~ rac e n but~lower than t ~ several other countries, i n c ~ u ~fornexample Kenya (37 percent}. i ~ Table 37: Youth Time Use in UrbanAreas, SeXected African ~ ~ u n t r j (percent of youth) e s 58,6 14.8 11.6 15.0 26.6 48.3 20.1 18.5 11.2 29.7 34.8 28.3 14.4 22.6 37.0 46.1 26.4 13.5 14.1 27.6 37.0 16.9 2.3 43.3 45.5 27.3 36.2 19.6 16.9 36.5 39.6 37.1 16.1 5.4 21.4 55.1 14.8 2.8 26.4 29.1 29.8 20.2 36.1 12.8 48.9 29.3 25.4 8.6 31.8 40.3 40.6 31.7 2.3 21.9 24.2 osis o f the problem. if ~ ~ e ni s caxlcen ~ p ~ ~ ~ ~ ~ ~ t ilies and can afford to take tim ot be a pressingissue froni a poverty re duct^^^ perspectiye. On the ent is high a ~ the o ~poor,~and ~if inactivity actually ers who have given up searching ions. Another p ~ ~ s iisbthat the~jobless ~ ~ l ~ ~ work but are not actively ~ e a r c for~ajobn could be ~ ~ ~ ies, such as ~ t ~ lwithia~ ~ p fanii~ybusiness. To better u n ~ e r s t a nthe ~ a ~of~the ep r ~ b ~ e insights into how the ~ r ~ n , ~ ~ ~ ~ p ~l ~or oyaeddefined4.e. ~ ~ c l u d i ~those not actively searching for ~~y i g e ~ n ~ ~ o ~spendntheir time are needed. i e t ~ 2.35 The Addis Labour Market Suwey ~ A c ~~~ ~ ~~ sbysSthisostudy~(see vol. i ~n ~ 1, Chapter 4 for deta~ls~ainted to gather ~ual~tatjvei n ~ o ~ ~to~help in diagnos~n~ i o n the puzzle o f E ~ ~ ~ i o ~high' sIcvel of ~ ~ n e ~ ~ p l coexisting twith a high poverty i a o ~ ~ e n 49 headcount, Nearly all d by the ALMS reported that their primar~source of sust , and only half said that the other rnenxbersof their hotlse uickly as possible," as d ea ~ ~for a ~goodjob or ~ a t ~ ~ ~that the ~ ~ i e ~ ~ iyo y e dn a ~ ~ otodwaif and tend r 11-off. On the ather hand, the vast majority of youth said they would "any available job," as osed to only paid ~ ~ p l o withe the ~~ o v e ~ n ~ e nprivate firm. In general youth seem to be ~ n or t pess~ra~is~i~about their chances of finding attrac~iv~e n i p l o ~ ~56npercent said it ~vould ~ ~ : be very d ~ f ~andu23~perccnt said it would be d i f ~ cto~find ""work that you would be c ~ l ~ ~ ~ r ~ lto~take." This evidence, along with the higher ~ro~erasity i n g o f the ~ ~ e ~ a ~to o y e d ~ have a n e ~ a subjective percep~jo~xabout their well-being (see Chapter 3 in this ~ ~ ~ e volume), u ~ d e ~the ~ n r~ ~thatuyouth itre~~ ~ ~ ~~ ptbecauseythey prefer leisure. i ~ l ~ e d 2.36 The ALMS included a niodule on a c ~ ~ v ito iobtain i n f ~ ~ a t i oon~ how the ~ e ~ a ~ ~ i eare~actuallyl spo ~ ~ ~ ~days, This ~ o ~ uincluded a list af actiFiitics l e inc~uded"selling products p~oducedby ho~~ehold~ ~ i ~andeb b~bsu'~~n ~ ~ einputs,n g b ~ ~ l i ~ ~errands for busiraessesiproducfivea c t j ~ofi ~friends ~ i ~ g ~ ~ ~or relatives,"' but oraly a ~ i ~ g ~ irau~iib~r ~ i ~ l elisted these kinds of ac~~v~ties~ o ntheir primary tasks. ~ g l~teres~i~~gly,the activity profile of ~ ~ e m p ~ oadults y e d is broadly similar, Table 38: Primary Activities of ~ n e Addis Ababa Youth inthe~PreviousW'eek," ~ ~ ~ ~ ~ d 2006 ~~~~~~~~~ *Percent ranking each activity one of their 3 most time cons^^^^^^ a ~ t ~ ~in' thetpast~week, Activities with less than 3% of responses ~ ~ s are omitted. S~jurce:ALMS 2006. 2.37 If youth unemployment in urbanEthiopia is as high as the data and qualitative research suggest, a comparison with the Middle EastNorth Africa region (MENA), which is also struggling to absorb a young and increasingly urban labour force, i s instructive. In the MENA region as a whole, more than a fourth o f the active youth labour force i s unemployed, though there are large differences across countries (particularly between labour-importing and labour-exporting countries) (Kabbani and Kothari 2005). Open youth unemployment rates in MENA are as high as 28 percent in Egypt and Saudi Arabia, ranging up to 53 percent in Algeria. Inboth Africa and the Middle East, the youth cohort has been growing, whereas it has been shrinking in other regions o f the world. In order to absorb youth entering the labour market, MENA will have to generate 47 millionjobs inthe next decade (Keller andNabli 2002). At the same time, the share o f youth in the MENA population (22 percent) peaked in 2005, and is projected to decline to 15 percent by 2040, thus gradually easing pressure on the labour market, while in Ethiopia the share o f youth will not begin to decline for many years (Kabbani andKothari2005). 2.38 The situation in MENA reflects that even with improving average educational attainment faster than any other region in the world in recent decades, supply-side measures alone are insufficient to facilitate adequatejob creation, particularly in the face o f major demographic pressures. The comparisonbetween the economic miracles inEast Asia andMENA could not be starker. Inthe former, productivity per worker rose quickly inthe 197Os, while inMENA countries productivity growth tended to be small or even negative inthe 1990s. 2.39 From a policy standpoint, what lessons can be learned from the MENA experience? Keller and Nabli (2002) point to several drags on growth inMENA: a large public sector; governance-related problems such as regulatory capture by powerfbl public and private business interests; an inward-oriented trade regime; incomplete financial sector reform; and an ineffective property rights system. Their conclusion i s that improving employment prospects depends on tackling such issues-which are broadly relevant to Ethiopia as wellrather than solely addressing labour market issues per se.l6 6. Youth EmdovmentPrograms: InternationalExDerience "In cooperation with developing countries, develop and implement strategies for decent and productive work for youth" - Targetfor MDG Goal 8, Develop a Global Partnershipfor Development 2.40 Around the world, as youth have become increasingly better educated and more likely to leave rural areas for urbancenters, the issue o f youth employment has become more salient. This has suggested the need for policies that directly address youth participation in the labour market (Leibbrandt and Mlatsheni, 2004). Thus far, however, international evidence o f best practices inprograms targeting youth i s scanty, particularly in comparison to programs designed for young children, where the knowledge base is much deeper. The World Bank's 2007 World Development Report, Development and the l6See World Bank (2006a): "Ethiopia: Accelerating Equitable Growth" for extensive discussion o f options for increasing growth. 51 Next Generation, which focuses on the multifaceted transition to adulthood, seeks to galvanize interest in improving the evidence base for youth-oriented policies and programs. Youth employment is one o f the key areas where more research on successful strategies is needed, and i s a relatively politically neutral topic that can be used as an entry point to facilitate discussion on other challenges youth face, as well as on labour policies in general (Miller 2003). The UN Secretary-General's High Level Panel on Youth Employment, convened in 2001, recommended that efforts to boost youth employment focus on four principles: 0 Employability: Invest in education and vocational training for young people, and improve the impact o fthose investments. 0 Equal Opportunities: Give young women the same opportunities as young men. Entrepreneurship: Make it easier to start and runenterprises to provide more and better jobs for young women andmen. 0 Employment Creation: Place employment creation at the center of macroeconomicpolicy. 2.41 The majority o f resources andprograms related to improving youth labour market outcomes fall under the heading o f employability, which includes all categories o f formal and informal education, as well as employer-provided training. Improving formal educational attainment is generally seen as the most effective way to boost the quality o f the labour supply. World Bank (2006a) found that in Burkina Faso and Uganda, for example, estimated earnings for secondary schooling and beyond are higher than returns to primary education, and that returns to education have been rising over time. This seems to be particularly true for growing economies-in Nigeria, for example, returns fell over the same period, which the study attributes to Nigeria's stagnant economy. This underscores the complexity o f identifying drivers o f youth employment outcomes, and the lack o f a magic bullet. Moreover, education does not necessarily lead to better employment outcomes ifthere are significant skill mismatches or market failures. 2.42 Despite the undisputed centrality o f formal education to better employment outcomes, many African youth never attend school, or drop out early, meaning that second-chance educational programs also have an important role. Interventions to keep young people in school longer could include conditional cash transfer programs, which would provide households whose youth stay in secondary school with a stipend (such an approach has beenused inMexico andBrazil, for example). 2.43 Knowles and Behrman (2005) provide a thorough review of the literature available on economic returns to investing inyouth. They review 41 programs, including inter alia employment-related interventions, and find that programs aimed at both improving the quality o f formal education, and programs for adult basic education and literacy training, have comparable or better economic returns than investments in other sectors. However, they provide a strong note o f caution about the global evidence base: 52 ...many studies that seem to confidently present estimates o f associations as if they are measuring causal effects probably are misleading because they do not deal well with the estimation problems ... Furthermore, almost none o f these studies seriously and persuasively consider the differences between private and social impacts, costs and rates o f return that i s at the heart o f the efficiency motive for policies. And many do not address explicitly the distributional motives for policies. Thus this review concludes that, though there are many studies related to investments in youth, what we know with confidence from the existing literature about causal effects and about policy motives for investing in youth is strikingly,perhaps shockingly, limited (p. 151). 2.44 Strengthening linkages between education and work i s also thought to increase the relevance o f training opportunities and improve employability. Schemes that combine work and school in developed countries (e.g. Germany) are impractical in low income countries like Ethiopia, but improving opportunities withinexisting institutions- such as the apprenticeship system-could yield returns. For example, helping master craftspeople to improve their own skills and increasing their access to new technologies could result in better opportunities for apprentices. A voucher program in Kenya illustrates such an approach; however, one review concludedthat the private sector would be better suited to implementing this kind o f program than the Government (Godfrey 2003). Informal education programs such as literacy training, like other interventions, work better when they are carefully targeted, demand-driven and contextually appropriate, or perhaps even work with local businesses. Some literacy programs in Ghana and Senegal include follow-up efforts to ensure participants' skills are maintained (World Bank 2006a). 2.45 Reducing gender disparities will be central to ensuring equal opportunities, though disparities along other dimensions are also important (improving opportunities for disabled, orphaned, or HIV-positive youth, for example). To the extent that males have disproportionate access to opportunities (e.g. for education or credit), increasing access to opportunity for females may have a higher economic rate o f retum-particularly inareas such as literacy training for adults (Knowles and Behrman 2005). A secondary school scholarship program for rural girls inBangladesh transferred funds directly into the girls' bank accounts; their eligibility depended on school attendance, remainingunmarried, and academic performance. Eventually, the enrollment rate o f girls in areas covered by the program outstripped that o f boys. In many countries, microfinance organizations target women specifically; there appears to be scope in Ethiopia to improve gender equity via microfinance, as discussed invol. 11, Chapter 3. Programs for entrepreneurs can also help alleviate the disproportionately high hurdles that women face in building a successful business; for example, one qualitative study by DFID o f a training program inDire Dawa found it to be empowering for women and to have a positive impact on their income. 2.46 Facilitating youth entrepreneurship, since the informal sector accounts for most employment creation inlow-income countries, can be an important strategy infacilitating the school-to-work transition. Interventions can be geared toward increasing access to credit, information, or training programs, or helping young entrepreneurs navigate 53 government regulations (e.g. on finding a workplace)-or they may combine these and other activities. Ingeneral, integrated, multi-sectoral approaches are believed to be more effective inpromoting entrepreneurship in a sustainable way. Uganda's "Youth Truck" program, implemented with the assistance o f GTZ, sends a truck equipped with information on how to start simple, small businesses to rural areas, and also provides some skills training. The truck later returns to the same villages to track progress and improve the likelihood o f sustainability (though no formal evaluation has yet been conducted). South Africa's youth fund program provides an illustration o f public and private organizations working together to improve youth access to credit (World Bank 2006a). 2.47 Economic growth leading to employment creation in general offers the surest route to better jobs for youth, especially since economic slowdowns disproportionately affect youth employment, as mentioned earlier. This underscores the need for a healthy investment climate and outward orientation that allows for trade and investment to expand. More narrowly, according to World Bank (2006a), directly creating jobs via wage subsidies and other financial incentives for firms that hire youth has been effective if well targeted and designed to create new rather than displace existing employment. However, so far these strategies have not been tested in low income countries, and are probably impractical in countries with very large informal sectors. Thus employment creation strategies in countries like Ethiopia are more likely to be akin to safety net programs rather than programs for youth seeking formaljobs. 2.48 One successful example o f a public works program for youth seems to be Senegal's Agence d'ExCcution des Travaux d'IntCrCt Public (AGETIP), which trains and provides temporary jobs on public infrastructure projects to unemployed youth. One benefit o f public works programs i s that they help in determining a target group for complementary programs (such as in training or job placement) that could help boost long-term employability. After the first seven years, an evaluation found that the number o f construction firms had grown five-fold, and that 35,000 person-years o f employment had been created. However, the evaluation underscored that success with such programs i s contingent on good targeting, transparency, and sound management. 7. Youth-relatedPoliciesandPropramsinEthiopia 2.49 Key policies in Ethiopia are directly related to youth employment, including the National Youth Policy, Labour Law, EducationPolicy, and HIV/AIDS Policy, as well as some insights from the global literature on youth. The policies governing Micro and Small Enterprises are also highly relevant and are covered in Chapter 3 o f this volume. Within this policy framework, public, private, and non-governmental organizations operate a variety o f programs aimed directly or indirectly at boosting youth employment and income-generating opportunities. These interventions can be loosely grouped into four categories: (1) education and skills training; (2) business development support; (3) employment matching services; and(4) multi-sectoral interventions. 2.50 National Youth Policy. The newly-formed Ministry o f Youth and Sports (MOYS) formulated Ethiopia's first National Youth Policy inMarch 2004 with the broad 54 objective o f encouraging the active participation of youth (defined as those aged 15-29 years) in the economic, social and cultural life o f the country, and to support democratization and good governance (see Box 5 on youth civil society organizations). The basic principles of the policy are to ensure that youth will be active participants in andbeneficiaries o f democratizationand economic development activities; to bringabout unity; to allow youth to organize themselves to protect their rights and interests; and to build capacity (e.g., via skills training). The policy addresses a wide range of issues, rangingfrom HIV/AIDS to environmentalprotection andsocial services (GOE2004b). Box 5: Youth Civil Society OrganizationsinEthiopia There are about 4,000 organizations in Ethiopia that work on socio/cultural/economic issues relevant to youth, though more than 90 percent of these organizations focus on HIV/AIDS. One example o f a youth CSO in Ethiopia tackling multisectoral issues is the Addis Ababa Youth Association (AAYA), which was established in 1998 and aims to encourage volunteerism; solve the economic and social problems o f youth; empower youth; and stem the spread o f HIV/AIDS among youth. The Association gets direct support from the regional government, including office space and vehicles from the Addis Ababa City Government, though it i s not considered a government organization. The AAYA estimates that it has around 80,000 members (roughly on the order o f 10percent of the city's youth population), and members pay a monthly fee o f Br 0.25 cents (about US$0.03). The three focus areas of the AAYA are health, human rights/democracy, and MSEs. Inthe health sector, the association concentrates on HIV/AIDSby providingpeer education and facilitating a youth dialogue about HIV prevention; it has organized about 200 anti-AIDS clubs. AAYA members participated in the May 2005 election process by raising awareness among voters and providing more than 400 election observers. To support MSE development, the association provides skill trainings and facilitates access to credit (it became a shareholder inAddis Ababa's largest MFI,the Addis Credit and Saving Institution S. C., to gain a voice inthe management and policies o f the MFI). The AAYA i s also working to strengthen the network of similar youth associations around the country. * * * In June, 2006 the World Bank office in Ethiopia organized a forum for dialogue, called "Fostering Youth Economic Empowerment," inpartnership with youth CSOs. Youth participants inthe forum indicatedthat fundingis scarce for activitiesother than HIV/AIDS-relatedprograms, and that there i s a need to buildcivil society capacity to work on youth economic empowerment. They also noted a lack o f networking among youth, and a lack o f awarenessof opportunities for young people; in parallel, they identified a lack o f youth participation in decision making, and a lack o frecognitiono fyouth by society and the Government as active agents o f development. Participants proposed a variety o f initiatives that the Government, private sector, and non-state actors could take to address these weaknesses. Inparticular, they emphasizedthe need for greater involvement o f youth in designing, implementing, and monitoring development programs, and including youth as a partner in solving societal issues. Expanding access to microfinance and improving opportunities for young people in the private sector (via apprenticeship programs and fair recruitmentprocedures) were also mentioned. Strengtheningnetworks o fyouth organizations and facilitating access to information, including through experience sharing programs, were suggested as ways to empower youth, alongside general capacity building for youth organizations. 55 2.51 The policy was officially launched in September 2004, but has not yet been fully publicized, although at the time o f writing the Ministrywas working on a three-month radio publicity campaign. The MOYS and its regional bureaus have the responsibility o f coordinating, integrating, and evaluating the policy's implementation. Both the strategic and action planare still under preparation. 2.52 Labour Law. Ethiopia's Labour Law was revised in 2003 to ensure that worker- employer relations are governed by certain basic principles; to guarantee the rights o f workers and employers to form associations; and to strengthen and define labour administration. It applies to the entire labour force, though some specific provisions are particularly relevant for youth (of course those working inthe informal sector are outside theprotectiono fthe Labour Law). For example: Article 29 states that in the event o f a workforce reduction, the employer in consultation with trade unions/representatives shall give priority according to workers' skills and productivity. In the case o f equal skills and productivity, those having the shortest length o f service in the undertaking, and those with fewer dependents, should be laid off. Since youth are more likely to fall into these categories, this provision may bemore likely to affect them. 0 Article 48 discusses Apprenticeships, and allows for contracts to be formed with those at least 14 years old. The chapter includes the contents o f the contract, obligations o f the parties, termination o f a contract and certification. Since regulations on formal and informal apprenticeships are the responsibility o f the Ministry o f Labour and Social Affairs (MOLSA), apprenticeship training centers are required to have a contract agreement with relevant Bureaus o f Social Affairs (BOLSAs) inorder to ensure that they are inconformity with the Labour Law. 0 The Labour Law prohibits employment o f those under 14 years, as well as employment o f young workers for activities that would endanger their life and health. The Law generally stipulates a maximum workday o f 8 hours, or 48 hours per week; however, Article 90 states that the normal workday for young workers should not exceed seven hours. 0 Employers are prohibited from employing young workers for night work (between 10 p.m. and 6 a.m.), overtime work, weekly rest days and public holidays. 2.53 Ethiopia does not have an "employment policy" per se (Le. a policy with the explicit aim o f encouraging job creation), or a minimumwage law. Active labour market programs include employment exchange services (see Box 6). 56 Box6: EmploymentExchangeServices inEthiopia Through its regional bureaus (BOLSAs), MOLSA offers employment exchange services to jobseekers and employers. To register with employment services, jobseekers give basic information about themselves and receive an ID card. When employers post vacancies, the BOLSA sends themnames of relevantjobseekers to consider. There are also plans for BOLSAs to begin offering career counseling services (pilot activities are underway), and to develop materials(e.g. films on career options). Although legislation was recently amended to require employers to report vacancies for which they are soliciting applicants (though actually hiring through employment services is optional), administrative capacity i s insufficient to enforce this requirement. Jobseekers may register with their local BOLSA, butthis is strictlyvoluntary. However, According to MOLSA data, there are unfilled vacancies; this is attributed to a lack of appropriately qualified workers, employers' desire for workers with substantialexperience, and the negative employment servicesdata cannot beconsideredrepresentativeoflabour supply or demand. Therole of employmentserviceshas changedsignificantly. Under the Derg, use ofthese services was obligatory, and jobseekers were centrally allocated to different institutions, with priority given to the placement of 12th grade graduates on the basis of year of graduation. University graduates were allocatedjobs with the government, and attempts were made to place vocational school graduates as well. Inaddition to the BOLSAs' employment services, there are 49 Regional Public Employment Services Offices that do similar job-matching activities. There are also private employment agencies, governedbythe PrivateEmployment Agencies Proclamation. 2.54 Education Policy. In1994, a new education policy that dramatically changed the education system and included a major supply-side push on TVET to facilitate the school-to-work transition. The Government recently developed its third Education Sector Development Program (covering 2005/06-20 10/11) in order to continue implementation o f its Education Policy. Before 1994, primary school included grades 1-6; junior secondary grades 7-8; and secondary school grades 9-12. In grade 12, students took a school-leaving exam in order to pursue higher education. However, few enrolled in higher education, while the majority o f school-leavers were left without any readily marketable professional or technical skills. 2.55 The new education policy aims to change this picture by focusing on producing a skilled labour force, rather than a large cohort o f relatively unskilled secondary school graduates. Grades 1-8 are now considered primary school, and grades 9-10 the first cycle o f secondary school. Both levels provide general academic education. A national exam i s given upon the completion of grade 10, and those who score well are promoted to the second cycle o f secondary school, or grades 11-12, which is considered college/university preparatory. Those who do not score well enough to continue in secondary school have the opportunity to pursue formal TVET, which takes from one to three years. One and two year training programs (known as "10+1" and "10+2") are considered "certificate level," while three years o f training ("10+3") is considered "diploma level." Note that in addition to formal TVET programs for those who finish grade 10, the MOE operates adult and non-formal education programs for out-of-school children (7-14 years o f age) and youth and adults above 15 years o f age, basically focusing on literacy, numeracy and the environment. 57 2.56 A pilot tracer study o f TVET is underway to gather information on graduates, including their employment status, to see ifthe new system i s effective. The Ministry o f Education (MOE) i s also working on a study, incooperationwith the German Agency for Technical Cooperation (GTZ), o f the projected demand for mid-level humanresources to better understand current skill gaps in the labour force and thus to inform education policy. Also, the MOE has formed a stakeholder network, which includes employers, to help prepare the TVET curriculum. For the last six years, the MOE has offered training every summer for TVET teachers in order to improve the quality and practical relevance o f its programs. Still, it believes that a major challenge is better understanding the demandfor the skills taught inTVET programs. The tracer study recognises the need to ensure that the programs are not entirely supply-driven, but that they respond to the changing needs o fthe market. 2.57 The scale-up in TVET represents one o f the most visible and concerted public sector efforts to improve youth employment. Presently, there are 25 fields o f specialization in public TVET institutions and 16 in non-governmental ones.l7 From 2000/01 - 2004/05, the number o f TVET institutions nationwide increased fiom 48 to 199, and enrollment grew fiom 9,000 to 106,000. However, enrollment in the second cycle (college preparatory) o f secondary education fell sharply in the 2001/02 school year, presumably because o f the change in the education system. There were about 33,000 students enrolled in grade 11 in the 2001/02 school year, compared to about 176,000 the previous year. Despite rapid growth, by 2004/05 the total number o f TVET students (106,000) plus Grade 11-12 students (92,000) was still far less than the total number of Grade 11-12 students in 2000/01 (271,000) (GOE 2005). Meanwhile, enrollment in grades 9-10 has been increasing steadily, putting additional pressure on post-grade 10 options, and the government intends to continue the expansion o f the TVET program to accommodate tens o fthousands o fnew students. 2.58 Before graduation, TVET students participate in apprenticeship programs organized in collabouration with private and government enterprises. Moreover, the M O E works incollabouration with different ministries and institutions inorder to design relevant TVET curricula (GTZ is also highly involved in support to TVET, and works with employers to define their needs, with the aim o f influencing the TVET curriculum). For instance, the Ministry o f Health i s involved in the program for health extension trainees; the Institute o f Leather and Leather Products for three leather technology trades; the Hotel and Tourism Training Institute for hotel and tourism trades; and the Ministryo f Agriculture curriculum guidelines were developed for agricultural training institutes by agricultural professionals. Despite these efforts, there i s room for improvement. The annual abstract report o f the M O E (2005) acknowledged that most TVET graduates at all levels did not meet employers' expectations, mainly due to the prevalence o f theoretical instruction instead o f an emphasis on workplace and labour market requirements, and the lack o f life-long learningkkills upgrading. 2.59 Moleke et al. (2004) analyzed the issue o f TVET responsiveness to labour market needs, and found that the absorption rate o f TVET graduates is very low, due mainly to a 17TVET figures inthis paragraphare nationwide(rural andurban). 58 lack o f employment opportunities and a low quality o f training. The skill level o f trainees was also found to be a major problem, particularly among the 10+1 graduates, which is aggravated by equipment shortages for practical training, and by the use o f outdated equipment and machinery. Moleke et al. also concluded that the institutional structure o f the TVET system results in poor coordination o f policy and activities. Demeke, Guta, and Ferede (2005) also identified a number o f constraints related to the quality o f teachers and the adequacy o f in-service programs. Of the 5,000 teachers in government and non-government TVET in 2004/05, 60 percent had a Diploma or lower level o f education (GOE 2005). 2.60 Despite these problems and the tendency o f the international literature to be either ambivalent or pessimistic about the impact o f TVET, there are reasons to be open minded about the program in Ethiopia. There i s anecdotal evidence o f some schools demonstrating positive results (see Box 7 for an example). Also, regression analysis by UCW (2006) found a significant positive correlation between "training" and employment, although the data do not enable controlling for possible endogenous selection into the program. Hence the results may be upwardly biased, but suggested that training could increase the likelihood o f employment by about a quarter. Wage regressions done for the present study (see Chapter 1 o f this volume) also point to a significant and positive effect o f training." Further research in this area, given the significance o f the program, would be very useful. Box 7: Case Study of a Vocational School: Hope Enterprise Hope Enterprisewas establishedthree decades ago as the first indigenousNGO, and i s somewhat of arole model since it not only trains its studentsbutalso helps them findjobs through its Office of Career Planning and Job Placement. The organization's strategy i s built around what it calls `Five LaddersofHope': Basic Care, or fulfilling people's basicneeds; Education, which i s provided to needy children from pre-school to secondary level in three states of the country-Addis Ababa, Desse andGambella; Competence, or giving youth marketable skills-students are enrolled in Hope's Vocational Training Centerswhere they receiveskill training for 1,2, or 3 years ina field of their choice, andare apprenticedinparticipating organizations; Sufficiency, or placing graduates in jobs through the assistance of Hope's Office of Career Planning and Job Placement-Hope also has a start-up fund available on a rotating basis to those who want to start abusiness, or are unableto findwork; and Value maturity, meaningthat peoplerecognizetheir duties andrights. With respect to Competence, Hope's vocational training centers have programs in electrical technology, auto mechanics, general mechanics, construction technology, information and communication technology, office management and secretarial science, woodwork, metal work, plumbing, cosmetology, international hotel and home service, ceramic production, poultry, vegetable gardening and apiculture. Sponsoringa student in one of the skill training fields, on average, i s estimatedat US$30per month. l8Also, it i s not possible from the data to determine where and what sort o f "training" the survey respondent participated in-it could refer to any kindo f formal or informal vocational education. 59 Hope prides itselfon the employabilityof its students, as over 85 percent of its graduates secure professionaljobs, while the rest are self-employed. The Addis Ababa branch currently has 194 students, and the branch's Office of Career Planning and Placementhas contacted 96 companies seeking apprenticeship positions for the prospective graduates. Inaddition, the branchhas looked for jobs for 126 graduates-f these, 107 trainees graduated from Hope's vocational training school, while the rest attendedother programsbut were financedbyHope. O fthese graduates, 19 were set up inthe businessoftheir choice while the remaining90percentgotjobs. Source: Getachew andKallaur 2005. 2.61 Ethiopia's tertiary gross enrollment ratio has risen dramatically inthe span o fjust a few years, from 0.9 percent-among Africa's worst-in 1999, to 2.5 percent in 2004. This places Ethiopiajust behind countries such as Ghana (3.1 percent) and Uganda (3.4 percent) (World Bank World Development Indicators). From 1996/97 to 2004/05, the number o f students enrolled inhigher institutions more than quadrupled, from 39,600 to 191,165 (including TVET 10+3 students; GOE 2005). Oftotal tertiary students, about 72 percent were pursuingan undergraduate degree, while 26 percent were Diploma students (TVET 10+3) and 2 percent were postgraduate students. About 24 percent o f tertiary students are women. 2.62 Alongside public institutions, private colleges are also expanding in the country, mainly in Addis Ababa. However, the cost is prohibitive for many youth (tuition fees generally range from Br 2,500-3,500, or US$288-403, per year) (World Bank 2003). The low unemployment rate among adults with tertiary education suggests that demand for skilled labour is high; however, the rapid expansion inuniversity enrollments will lead to an unprecedented influx o f highly skilled new graduates to the labour market, and it will be important to monitor how quickly they can find employment that utilizes their skills. 2.63 HIV/AIDSPolicy. The HN/AIDS Policywas designed in 1998 inresponse to the alarming spread o f the epidemic, albeit more than a decade after the first reported AIDS cases. It contains several provisions relevant to employment, including that no person should be forced to undergo an HN screening for job recruitment purposes, unless the nature o f the occupation requires it to do so. It also outlines the rights o f people living with HN/AIDS with respect to access to employment and associated privileges, educational and/or training facilities and public facilities, and states that people should not be subjected to discriminatory practices. 2.64 In 2003, the HIV/AIDS Prevention and Control Office developed a Mainstreaming Guideline to provide both conceptual and practical guidance and information on how government sectors should respond to the threat o f HIV/AIDS inthe workplace. The guideline emphasized mainstreaming HIV/AIDS awareness into routine operations o f all Federal Ministries and organizations in order to encourage prevention. MOLSA in particular was requested to incorporate HN/AIDS awareness in its development plan, strategies and policies. In addition to this mainstreaming function, MOLSA was asked to undertake studies on the HIV/AIDSimpact on women, youth and children from various perspectives, and to coordinate and assist relevant organizations in eliminating HIV/AIDS. MOLSA i s supposed to give guidance on employment procedures and the labour law in order to prevent mandatory pre-employment and periodic medical checkups for HIV/AIDS, and to develop and disseminate a national 60 workplace HIV/AIDS code o f conduct. Through its Children and Family Affairs department, MOLSA i s also required to establish and strengthen `Youth Anti-AIDS Clubs' andpeer-to-peer leadership forums to combat the epidemic. 8. Conclusions 2.65 With the creation o f the MOYS, the major effort on expansion o f skills-based education (TVET), and investment in MSE development (see Chapter 3 o f this volume), the GOE has demonstrated an interest in and commitment to tackling the challenges faced by youth. It takes seriously the problems o f the school-to-work transition and is attempting to formulate practical interventions that make this process easier. Given the huge demographic pressures on urban employment impliedby population growth and in- migration, the attentionto this issue could not bemore timely. 2.66 The evidence suggests several main conclusions and policy implications. Average youth unemployment (15-24 years) in urban areas i s much higher than that o f adults, and is drivenprimarilyby 20-24 year olds rather than teenagers, who are also less likely to be economically active. The informal sector employs 81 percent o f working youth (compared to 66 percent o f working adults), and youth are frequently unpaid family workers rather than entrepreneurs or paid employees. Our survey suggests that unemployed youth are pessimistic about their jobs prospects. In general, the school-to- work transition is relatively protracted in Ethiopia, though welcome new evidence from the 2005 LFS suggestsrecent improvements. Likewise, underemployment maybeon the decline as well. However, incomes remain low for youth and opportunities are far from equal-the likelihood o f finding a goodjob varies widely by gender, geographical region, disability andhealth status, and other factors. 2.67 Despite rising human capital, most Ethiopian youth still receive relatively little education, and it i s not yet clear whether the changes in the secondary education system are producing a more appropriately skilled and adaptable workforce, suggesting the need for further researcwstudies in this area. Current TVET/upper secondary school capacity i s woefully insufficient to meet demand (much less to educate the entire relevant age cohort), as recognized via the continuing push to expand. Yet increasing human capital supply is only part o f the challenge, since the current generation o f youth-better educated than its predecessors-needs to find a way to use its skills. 2.68 A deeper and broader skill base among youth is indispensable to Ethiopia's growth and development prospects-far from being a liability, youth aspirations for higher value-added jobs are precisely what i s needed. To ensure that the skills youth acquire are relevant and practical in the job market, education and training programs- both formal and informal-need to increase their demand-side orientation, as some are already attempting to do. Strengthening links between educational programs and employers themselves could further these efforts. Youth also need better access to credit and information, and coordinated approaches will probably work best. Since some programs along these lines already exist inEthiopia, analyzing their effectiveness to date and drawing lessons learned should be a priority to inform policy makers. The context- 61 specific knowledge that thorough evaluations could generate would provide invaluable insightsabout how to increaseprogram coverage and effectiveness. 2.69 In a broader sense, as emphasized in other chapters of this study, economic growth will be the most important determinant o f employment outcomes for youth. Since young people comprise the bulk o f new entrants to the labour market, sluggish job creation will disproportionately impact them-but at the same time, today's relatively well-educated youth can also drive growth if given the opportunity. With a conducive policy environment that fosters private sector expansion, accompanied by effective programs to improve youth employability, Ethiopia can leverage the "demographic dividend" providedby its growing youth population to accelerate development writ large. 62 References(Chapter 1and 2) Bigsten, Arne, Bereket Kebede, Abebe Shimeles, and Mekonnen Taddesse. 2003. "Growth and Poverty Reduction in Ethiopia : Evidence from Household Panel Surveys." WorldDevelopment 31.1:87-106. CSA (Ethiopian Central Statistical Authority). 2004a. Ethiopia Statistical Abstract 2003. Addis Ababa: CSA. . 2004b. Welfare Monitoring Survey 2004: Analytical Report. Addis Ababa: CSA. . HICES (Household Income Consumption and Expenditure Survey). 1999/00. Addis Ababa: CSA. Demeke, Mulat, FantuGuta and Tadele Ferede. 2005. "Towards A More Employment- Intensive and Pro-Poor Economic Growth In Ethiopia: Issues And Policies." Addis Ababa University. Denu,Berhanu,Abraham Tekeste, andHannahvan der Deijl. 2005. "Characteristics and Determinants o f Youth Unemployment, Underemployment, and Inadequate Employment in Ethiopia." Employment Strategy Paper, ILO. http://www.ilo.or~public/english/emplovmen~~ett/publ.htm. Dercon, Stefan and Pramila Krishnan. 2000. "Poverty in Rural Ethiopia 1989-95: Evidence from Household Panel Data in Selected Villages' in A. Booth and P. Mosley (eds), The New Poverty Strategies: What have they Achieved? What have We Learned? Macmillan. Dercon, Stefan and Pramila Krishnan. 1996. "Income Portfolios in Rural Ethiopia and Tanzania: Choices andConstraints." The Journal ofDevelopment Studies, 32.6: 850-875. Fares, Jean, Claudio E. Montenegro, and Peter F. Orazem. 2006. "HOW Youth Faring are in the Labour Market? Evidence from Around the World." Background paper for the 2007 World Development Report. Getachew, Martha, and Emily Kallaur. 2005. "Youth Employment in Ethiopia: Overview and Inventory of Existing Policies and Programs." Drafbackground paper prepared for the Ethiopia Urban Labour Market Study. World Bank, Washington, D.C. Mimeo. Godfrey, Martin (2003). "Youth Employment Policy in Developing and Transition Countries - Prevention as well as Cure." World Bank Social Protection Discussion Paper Series Oct. 2003. No. 0320. 63 GOE (Government o f Ethiopia). 2003. Education Statistics Annual Abstract, 2002-2003. Addis Ababa: MinistryofEducation. . 2004a. Education Statistics Annual Abstract, 2003-2004. Addis Ababa: MinistryofEducation. .2004b. NationalYouth Policy. Addis Ababa: MinistryofYouth and Sports. . 2005. Education Statistics Annual Abstract, 2004-2005. Addis Ababa: MinistryofEducation. ICRW (International Center for Research on Women). 2005. "Too Young to Wed: Education andAction Toward EndingChildMarriage." Washington, DC: ICRW. Kabbani, Nader and Ekta Kothari. 2005. "Youth Employment inthe MENA Region: A Situational Assessment." Social Protection DiscussionPaper 0534, HumanDevelopment Network, World Bank, Washington, DC. Keller, Jennifer andMustapha K.Nabli. 2002. "The Macroeconomics of Labour Market Outcomes in MENA over the 1990s: How Growth has Failed to Keep Pace with a BurgeoningLabour Market." Washington, DC: World Bank. Knowles, J and Behrman, J. 2005. `The Economic Returns to Investing in Youth in Developing Countries: A Review ofthe Literature." Washington, DC: World Bank. Kondylis, Florence and Marco Manacorda. 2006. "Youths in the Labour Market and Transition from School to Work in Tanzania." Draft Background Paper for the World BankYouth inAfrica's Labour Markets study, Feb.2006. Leibbrandt, Murray and Cecil Mlatsheni. 2004. "Youth in Sub-Saharan Labour Markets." Paper presented at the "African Development and Poverty Reduction: The Macro-Micro Linkage," Somerset West, SouthAfrica, October 13-15,2004. Miller, Steven. 2003. "Youth Employment and its Implications for the Youth Population Explosion in the Developing World's Cities." In Youth Explosion in Developing World Cities, eds. Blair A. Ruble, Joseph S. Tulchin, and Diana H.Varat, with Lisa M.Hanley, 47-61. Washington, DC: Woodrow Wilson International Center for Scholars. Ministry of Health. 2004. Fifth Report: AIDS in Ethiopia. Addis Ababa: Ministryof Health Disease Prevention and Control Department. htt~://~.etharc.or~aidsineth/publications/AIDSinEth5th En.pdf. MOFED (Ethiopia Ministry o f Finance and Economic Development). Forthcoming. "Ethiopia Participatory Poverty Assessment 2004-05," by Frank Ellis and Tassew Woldehanna. Consultant Report. Addis Ababa. 64 Moleke, Samuel Feyisa, Betelehem Gebru, Alemayhu Asfaw and Asnake Welderaguel. 2004. "HIV/AIDS andEmployment." International Labour Organization. Nicaise, Ides. 2001, "Human capital, reservation wages andjob competition: Heckman's Lambda reinterpreted." AppliedEconomics, 33:309-313, Rahmato, Dessalegnand AkliluKidanu. 1999. "Consultations with the Poor: A Study to Inform the WDR 2000/01 on Poverty and Development (National Report, Ethiopia)." Addis Ababa: World Bank. UCW (Understanding Children's Work). 2006. "Child Labour and Youth Employment: EthiopiaCountry Study." Rome: UCW. UNAIDS 2006. 2006 Report on the Global AIDS Epidemic. Geneva: UNAIDS. http://www.unaids.org/en/HIV data/2006GlobalReport/default.asp. White Howard, Jennifer Leavy, MwilaMulumbi, Grant Mulengaand Venkatesh Seshamani (2006) "Rural Development and Labour Markets inAfrica: Preliminary Findings from aResearchProject inNorthern Province, Zambia," Institute of Development Studies, Brighton, DiscussionDraft. Woldenhanna, T. and A. Oskam. 2001. "Income diversification and entry barriers: evidence from the Tigray region ofnorthern Ethiopia." FoodPolicy 26: 351-365. World Bank. 2003. Higher Education Development for Ethiopia: Pursuing the Vision. Washington, DC: World Bank. . 2005a."Ethiopia: A Country Status Report on Health and Poverty." 2 vols. Washington, DC: World Bank. . 2005b. "Education in Ethiopia: Country Status Report." Washington, DC: World Bank. , 2005c. "Ethiopia: Well Being and Poverty in Ethiopia." Washington, DC: World Bank. . 2005d. Children and Youth: A Resource Guide for World Bank StafJ: Washington, DC: WorldBank. . 2005e. "Education in Ethiopia: Strengthening the Foundation for SustainableProgress." Report No. 34352. Washington, DC. . 2006a. "Youth in Africa's Labour Market." 2 vols. Draft, June 2006. Washington, DC: World Bank. . 2006b. "Capturing the Demographic Bonus in Ethiopia-The Role of Equitable Development and Demographic Actions." Draft, June 2006. Washington, DC: World Bank. 65 3. A PORTRAIT OF INFORMALITYINETHIOPIA: DIAGNOSTICSAND POLICY IMPLICATIONS Highlights The informal sector provides more than half the employment inurban Ethiopia. Iti s the home o f many o f the working poor, who cannot afford to be openly unemployedyet earn too little to lift themselves out o f poverty. As such understandingthe characteristics o f workers and firms inthe informal sector i s essential to understanding the labour market as a whole, and to designing policies that foster growth and poverty reduction. Yet a simple "dualist" model o f the economy-formal vs. informal-is inadequate because of the heterogeneity of the informal sector. Most informal enterprises are concentrated at the low-value-added end o f the spectrum, which results in intense competition among relatively undifferentiated firms. In addition to market saturation, small firms often have trouble accessing capital and workplaces. As firms grow, they become more likely to experience investment climate problems, such as unpredictable regulations or harassment by officials. Informal networks are central to the way firms operate, since entrepreneurs tend to find many of their customers among their personal contacts, or via word o f mouth. Most informal enterprises seemto perceive few benefits to formalization. Women and youth face specific challenges. Women are disproportionately located in the informal sector; female-operated enterprises are frequently home-based, and concentrated in low productivity sectors. Household and child care responsibilities may restrict their ability to reach customers and to build human capital, while societal norms about appropriate activities for women can limittheir entrepreneurial potential. Women tend to enter into business with a lower level of education and less start-up capital than their male counterparts, and they earn less over time. Youth are more likely than adults to work inthe informal sector, and within the informal sector, youth (unlike adults) are nearly as likely to be unpaid family workers as they are to operate their own enterprises. Policy priorities with respect to the informal sector should include improving the investment climate for microenterprises, and adopting an evidence-based approach to active interventions. Removing constraints on access to workplaces, improving microfinance regulations to encourage greater competition in the sector, and ensuring a predictable policy environment and freedom from harassment would give the sector space to innovate and grow. There also appears to be scope for scaling up existing initiatives to provide training, credit, and other business development services to firms and would-be entrepreneurs, if evaluations demonstrate that they have Dositive results. 66 1.Introduction 3.1 As in other developing countries, the informal sector plays a central role in Ethiopia's labour market, accounting for an estimated 71 percent o f urban employment. It therefore serves as a vital source o f income-and a safety net-for the poor, who comprise somewhere inthe range o f 46 to 70 percent o f the urbanpopulation (according to "lower" and "upper" poverty lines; World Bank 2005). It i s particularly important for women, who represent half o f the informal sector (relative to only 35 percent o f the formal sector), with many engaged inproductive activities closely intertwined with their domestic responsibilities. 3.2 The informal sector is a supplier o f goods and services to the population, and a source o f labour and other inputs to formal sector enterprises, and represents approximately 40 percent o f Ethiopia's GNP (Doing Business Database). Informal firms are concentrated inmanufacturing and trading o f goods and services. Consisting largely o f low-productivity microenterprises usually with no employees, informal firms appear to be severely constrained by low purchasing power inthe economy and lack o f financing. Some do manage to grow into successful small businesses, while for others, informal work merely enables survival. As an engine o fjob creation, the informal sector's share o f the urban economy could increase as urbanization continues and the labour force grows. Creating an environment that allows the sector to flourish is therefore critical for both growth andpoverty reduction goals. 3.3 Our interest in the informal sector is motivated by both its significance in Ethiopia's urban economy and its special relevance to the poor. While the prevalence o f informality ina country's economy tends to decline as per capita GDP grows (Figure 15), recent studies indicate overall growth indeveloping countries inthe past decades (Becker 2004). Ingeneral, informal sector employment i s believed to be less desirable from the worker's perspective, due to the lack o f legal protection, lack o f pension/medical/vacation benefits, potentially unsafe working conditions, and so on. On the other hand, some workers may prefer informal work to their other options (such as subsistence agriculture or unemployment), may appreciate the greater flexibility to cope with familyresponsibilities, or may choose informality to avoidtaxation andregulations. 3.4 There are a host o f policy-related questions vis-&vis the informal sector. By definition it exists outside the legal framework for business activity, yet operates within the institutional realities o fthe country, andmakes a substantial contribution to economic output and employment. If informality i s correlated with lower levels o f economic development, is formalization good for growth? Should the rise in informality observed inthe last fifteen years or so inEthiopiabetaken as an encouraging sign ofdynamism, or a worrisome indication o f labour market segmentation and lack o f productive employment? In Africa in general, despite rising awareness o f the key economic role played by the informal sector, governments still tend to treat it "as separate from other sectors o f the economy with little recognition o f the complementarity and interconnectedness o f the informal and formal sectors" (Tripp 2001 p. 12). In Ethiopia, official government policy recognizes the informal sector's vital role, but entrepreneurs still confront obstacles on the ground. To what extent could government action alleviate 67 these constraints? This chapter exploits a variety o f survey data and marshals relevant information to help answer these questions. Figure15: GlobalSelf-Employment as percentofNon-AgriculturalEmployment,Relative to GDP per capita, RuralandUrbanAreas (PPP, constant2000 US$) 1980/1990 1990/2000 100% + loo%, I 90% Mali 80% .Mali 80% 70% b h i opia U -&4 % 60% SO% 40% 30% 20% 10% 0% 0 SO00 10000 15000 20000 25000 30000 0 SO00 10000 15000 20000 25000 30000 GDP Per Capita GDP Per Capita Data Sources: ILO 2002; World BankWDI. 3.5 The informal sector defies simple description because o f its extreme heterogeneity. The sector comprises the self-employed (some o f whom are highly successful entrepreneurs, while others barely eke out a living), employees o f informal firms, "homeworkers" (home-based industrial outworkers), and workers at formal firms without legal protections or permanent contracts. l9 While there have been several competing schools o f thought on the appropriate characterization o f the informal economy, all apply to some degree, depending on the segment being considered. Thus the definitions should be considered complementary rather than mutually exclusive: the dualist school argues that the informal sector i s "comprised o f marginal activities" unrelated to the rest o f the economy, which does describe some o f the survival-oriented work; the structuralist school, characterized by the beliefthat firms inthe informal sector are "subordinated economic units" that are "inextricably connected" with formal firms, accurately describes part o f the picture (e.g., outworkers); and finally, the legalist school contends that firms choose the informal sector to avoid onerous licensing requirements and regulations, which probably describes some o f the more entrepreneurial operators (Chen 2005 p. 4-6). 3.6 This chapter serves two main purposes: to provide a snapshot o f informality in Ethiopia's urban labour markets, including its size and characteristics; and to assess the constraints to improving opportunities in the sector. Section 2 provides an overview o f measurement issues and the informal sector, and an estimate o f its share o f employment in Ethiopia. Sections 3 and 4 use survey data to assemble a portrait o f the informal sector, including demographic characteristics o f the operator, self-reported reasons for choosing the given line o f work, sector o f activity, and basic firm characteristics. These The informal sector also comprises illicit activities. This chapter is only concerned with activities that in their own right are legal, but that are not formally licensed by the authorities. For more details on the definition o fthe informal sector, see Section 11. 68 sections attempt to illuminate the "black box" o f the informal sector. Section 5 builds on this description to look further into the heterogeneity o f the sector, and assess to what extent informal firms are dynamic versus survival-oriented businesses. Given the informal sector's importance to the economy and poverty reduction, Section 6 considers what constraints are restricting income and productivity growth, with implications for government policy, and Section 7 concludes. 3.7 Like the rest o f this study, the chapter focuses only on urban areas, where the formalhnformal distinction i s most relevant, we limit the focus to the population aged over 15 years, which excludes the 15 percent o f urban children aged 5-14 who are working (UCW 2006). 2. Measuring the Informal Sector: Data and Definitions 3.8 Measuring the size o f the informal sector i s no easy task. Ethiopia has an unusually rich array o f labour market data, but informal sector activities are difficult to capture. Workers may change their place o f work frequently, and may have an incentive to hide the true extent o f their activities and income for fear o f being forced to pay taxes or back rent on government-owned housing, for example. Choosing an appropriate definition o f informality also poses difficulties. There are arguments for basing the definition around official registration and compliance with tax and other regulations, but this would potentially exclude some very large firms. It would also include licensed microenterprises with few o f the trappings one would associate with formality. In practice there is a large gray area betweenthe formal and informal sectors. It i s probably better to think o f a continuum, with subsistence activities such as gathering and selling firewood on one end and established firms, such as shops and service businesses operating inpermanent buildings with fixed payrolls, on the other. 3.9 The Government defines informal employment at the firm level-Le. anyone working for an informal enterprise is considered to be inthe informal sector, and anyone working for a formal enterprise is considered formal. The official Ethiopian definition o f informality encompasses enterprises that meet all o f the following three criteria: no book o f accounts; no license; and fewer than 10 employees. Being "licensed" essentially means becoming a formal business; an intermediate step, "registering," means becoming known to the authorities, but still informal (ILO and MTI/WAD 2003). By this definition, according to the 2005 LFS, 38 percent o fworkers are informal. 3.10 The official definition is quiterestrictive and may understate the size o f Ethiopia's informal sector in international comparisons. It is therefore useful to do another estimation based on a looser definition. Inaddition to employees o f enterprises who meet the three criteria above, this would include anyone who is a domestic employee, self- employed, an apprentice, an unpaid family worker, or who is only paid in kind. Since only a handfbl o f the self-employed are in professional occupations (such as law, architecture, etc.), the scope for error o f inclusion seems limited. 69 3.11 Table 39 shows the application o f the looser definition, which indicates that the informal sector comprised about 71 percent o f urban employment in 2005. This is more consistent with what would be expected at Ethiopia's level o f income (the African average for urban areas i s over 60 percent; Becker 2004). The prevalence o f informality varies considerably; the wholesale/retail trade and primary production sectors are almost exclusively informal (as measured by the prevalence o f informality among workers inthe sector) while the reverse is true for the financial sector. Interestingly, under this definition, the informal share o f men employed in "Manufacture o f Food, Tobacco, and Beverage" rises to 66 percent (from 17 percent), given the addition o f unpaid family workers and those paid inkind. Gender differences are substantial. For example, inthe "Manufacture o f Food, Tobacco, and Beverage" category, about 90 percent o f female workers are informal, compared to 67 percent o f male workers. A similar difference is observed inthe "Hotel and restaurants" category. 3.12 This chapter relies mainly on the Urban Informal Sector Survey (ISS) conducted in 2003 by the Central Statistical Authority (CSA), using the official definition of informality (n = 14,791). Inthe areas chosen for the survey, enumerators went door-to- door asking households whether any member operated an informal activity (per the official definition)-if so, they were added to the list o f eligible households from which the sample was then derived. The methodology therefore ensures coverage o f very small home-based activities. But while the survey is comprehensive and representative o f the informal sector as a whole, it does not offer comparative evidence on formal activities. Moreover, only operators (not employees) o f informal firms were interviewed, meaning that the analysis is at the firm rather than individual level. But since most firms do not have employees, firm income and challenges tend to be synonymous with those o f the operator. 3.13 Several other data sources are used. The CSA's 2003 firm-level Distributive and Service Trade Survey (DSTS) is not restricted to informal firms per se, but since only about 1percent o f the sampled firms keep a full book o f accounts, and less than 1percent has 10 or more employees, for practical purposes it i s an informal sector sample. The survey covers about 12,500 firms in urban areas drawn from a two-stage stratified sample; the firms represent three sectors: wholesale trade, retail trade, and service trade. 3.14 The ALMS commissioned by the World Bank in 2006, which was run on 282 households randomly selected from the Addis Ababa subsample o f the Urban Panel Survey (see vol. 11, Chapter 5), provides additional insights. The survey included a module o f questions for those who reported that self-employment was their primary or secondary income-generating activity (for a total sample size o f 135 firms). This module gathered information on some key firm characteristics and behavior, as well as the constraints faced by the operator and the social networks on which the operator relies. These respondents are not representative o f the self-employed population as a whole because o f the sampling frame. However the questions probe qualitative issues not previously explored, so the results usefully suggest some hypotheses for future research. The results are also interestingbecause the self-employed respondents include licensed as well as unlicensed firms (proxying informality), allowing us to compare responses betweenthese two categories. 70 Table 33: Prevalence of ~ ~ lby Sector ~ j ~~ fa5e percent~oft~ ~ l ~ ~ ~ r~ total yment in the sector) LFS2005 Male I Female 1 Total of Vehicles, Personal& ~ ~ u ~ ~ h o ~ d 71 3. Profile of Informal Sector Operators 3.16 Gender, Age, and Education. Women account for about 60 percent o f the informal sector," and represent a larger share o f informal sector operators than men for all age cohorts (Figure 16), with the smallest gender differential observed for the youngest and oldest cohorts. About half o f operators are in the 25-44 year age group; about 10 percent are youth (ages 15-24). Although human capital is low inthe sector, it appears to be increasing for younger groups. Figure 17 compares the educational composition o f the informal sector using available data from 1999 to 2004, which shows illiteracy declining, and higher levels o f grade school and general education. The rising educational profile could be a positive sign consistent with increasing human capital in general, or a negative sign that skilled workers are facing increasingdifficulty ingetting a formaljob. Both explanations could be true to some extent. Figure16: DemographicCompositionofInformalSector Operators,2003 20% I +M a l e +Female Data source: Informal Sector Survey2003, owncalculations. 3.17 Sources of Income. Most informal sector operators have just one main income- generating enterprise-according to the ISS, only about 25 percent have a secondary job or activity. The share varies by sector, with about 40 percent o f those working in agriculture having a second income-generating activity, but only 15 percent o f those working in construction. Women are slightly more likely than men to have a second activity (27 versus 23 percent). Because income diversification i s a common livelihood strategy o f the poor inAfrican countries (Ersado 2005), it i s relevant to ask how Ethiopia compares to other countries. Bigsten and Makonnen (1999) found that urbanEthiopians more frequently rely on a single source o f income than their counterparts in Kampala, and attributed this to the socialist legacy. The hypothesis that Ethiopians may have 2oAs mentioned, measuring the informal sector presents a challenge. According to the "broad" definition o f informality usedby this study, inthe 2005 LFS women comprise half o f the informal sector (though this also means that they are disproportionately informal, since they comprise only 35 percent o f the formal sector). However, women account for 58 percent o f informal workers under the official government definition o f informality, which i s more consistent with the 60 percent found using the ISS data (2003). 72 ely undiversified ~ncon~e-ear~ orted by the ALMS, in w ~ ~ ~ g ~ ui ~ ~i o~f elallre respo~dc~~ts n (not only those in the ~ n f o ~sector) had CJ a l secondary occupat~onof m y kind. There could be s ~ g ~ idiversi~~ c~a t j oover the ~ c i ~ course of a year, ~ ~ r ihs ~ i~f ~ ~tougauge with the a ~ ~ ~ ~data, bbute since most ~ h l t ~ a l i n ~ o f~i m~s aoperate about I t ~ o ~ of h ~ ~ the syear (see below) seasonal~tymay be r e ~ a ~limited.eMore researchin this area could be fntitfut. ~ ~ ~ ~ y n 3.18 1s: inthe ~ ~ ~sectorois t~y pai ~ ~ ~ l y f ~~~~~~ eniploymc~~ last resort. hi re of ucstion "why did you choose this activity from others`?" the ~ ~ eshare of workers reported that "it li s Nphat Ican~ ~ ~ ~ e ~ ~ ~ a f ~ o r ~~nves~~nent ~ s ~ ~ lneed~d,'~ ``1 have no a ~ ~ e ~sourcei of income.'* Other ~ or ~ r ~ e resportse options that would suggest a niore " ~ o ~ choice~ ~nc~udcd r ~ ~ 3 ~ ~ ~ ~ "family tr~~~ti~~n,'3 act~~~~ty,"and " a c ~ ~bringsyhigh incomc." "1 like the ~ ~ t The same question cd in the ALMS sur elicited somewhat more positive respanses, but b ~ e ~ k i ~ ~ g dC3a.n the data t compare the self-crnplo j n f o ~ aoncs, l ~ 3.19 Looking at the ISS results by the worker's e ~ ~ ~status prior~to the ~e u~ ~ tc n ~ ~ o ~ activity (Table 40) reveals that those who r r ~ ~ ~ sfrom i ~ u~~~e m pe~~o ~ nweret the most n e n likely to have made an ~ 6 i n ~ o choice. ~ However,' the small share of those l ~ ~ ~ ~ y ~ r a ~ ~ s i ~ fromn ~ nng e ~ p ~ (4 percent of youth and 3 percent of adults) suggests ~ o u o ~ ~ e ~ ~ the luxury o fbeing u ~ e n ~ p l o y c ~ . al sector e cnt. 73 Table 40: Reasonfor ChoosingCurrent Activity, by PreviousActivity, 2003 Figure 18: Reasonfor C i ~ ~~ ou~ riActivity, by Gender ~~ ~e ~ ~ 3.21 This pattern contrasts tvith evidence from Mexico, which may mean that the ~ n ~sector ~in a~poor country like ~ t ~ x ~isoquite diff~rentfrom that of middlc- o l p j ~ income countries. When survey respondents in Mexico were asked about their reasons for leaving their last job, more than 60 percent cited voluntary reasons for becoming self- employed; overall a total o f about 60-70 percent o f informal sector employment in Mexico can be considered voluntary (Maloney 2004). At the same time, informality can be associated with both pecuniary and non-pecuniary benefits. Tripp (2001) reports on an interview with a Ugandan woman who is part o f an unregistered savings club; according to her, there are advantages to remaining informal: We work so well informally... The minute we become formal we will collapse... The group i s based on trust, mutual confidence, flexibility. You do what you want, the organization i s yours. What would we do if we registered? We would have to have a location, an office, and we can't afford that. We would have to get registered and do the proper paperwork. Who would have the time to go around and do all that? We are all working women.. . Wejust want things nice and simple. (p. 9) 3.22 Moreover, as Maloney (2004) explains, "arguing that workers are voluntarily informal does not, o f course, imply that they are not living in poverty, only that they would not obviously be better off in the formal jobs for which they are qualified" (p. 1160). Even if in Ethiopia workers are engaged in their current activity because they have no other options, a formal sector job may not necessarily imply an improvement in status. Of course, perceptions o f the pros and cons o f informal sector work must be highly dependent on where one is on the continuum o f earnings, working conditions, access to services, and other factors. Available data on earnings among the illiterate and Grade 1-4 populations-which comprise the majority o f informal sector workers- suggest that median wages in the informal sector are substantially lower than those in formal employment (Figure 19). However, when removing domestic workers from the dataset, the difference between the wages for unskilled formal and informal workers becomes very small. For illiterate workers (excluding domestic workers), the median monthly wage i s Br 70 (formal) and Br 80 (informal). Likewise, for grade 1-4 workers, the median monthly wage i s Br 100 (formal) and Br 112 (informal). This suggests that unskilled employees o f private enterprises are not necessarily worse off because they are inthe informal sector per se. Broadly speaking, the nature ofthejob rather than the legal status o f the enterprise appears to determine wage levels. Yet all else being equal, the evidence suggests that people prefer formal employment to informality, perhaps because they value the greater security that comes with formal employment (Box 8). 75 Figure19: MedianMonthlyWages, Formalvs. Informal Employment-Broad Definitionof Informality loo0 900 I a 700 CI E"3 600 - 500 - 3 400 - 300 - - 200 - I Data Source: LFS 2005, own calculations. Note that the LFS data only capture the subset of informal workers who work for private organizations/NGOs, or as domestic workers (excluding the self employed, unpaid family workers, employers, apprentices, and members o f cooperatives). 3.23 Regional differences. Doing justice to Ethiopia's enormous geographical diversity is well beyond the scope here. Figure 20, however, says much about the geographical variation in the sectoral composition of the informal sector (only the two main categories, manufacturing and tradeihotelsh-estaurants, are shown). Gender composition also varies considerably-the share o f women in the sector i s as low as 49 percent inAddis Ababa, for example, and as highas 66 percent inAmhara. 76 Figure20: SectoralCompositionofInformalEmploymentbyRegion, 2003 10Manufacturine 60 w so s wE"f40 30 p c 20 10 0 Data source; Informal Sector Survey 2003, own calculations; Dire Dawa data not available. Box 8: SubjectiveWell-beingandEmployment Status Are people necessarily worse off in the informal sector than the formal sector? Answering this question is fraught with difficulty-not least interms o f defining well-being. For instance, many informal workers may have lower wages than their formal counterparts. However, they may actually prefer their work because of the flexibility it offers to combine work and childcare responsibilities, or perhaps because it allows them a measure o f independence. Therefore we cannot assume that objective measures of labour market outcomes capture an individual's satisfaction with his employment status-yet understanding how individuals view their jobs has important implicationsfor policy. The ALMS module on subjective well-being allows usto compare how different kindsofworkers view their status relative to others. This i s not to suggest that labour market status i s the sole or even most important determinant of welfare and happiness, but it i s reasonable to argue that it plays a large role. The table below divides respondents into 5 basic categories: inactive, unemployed, employee, and formal and informal self-employed (based on whether the businessi s licensed). By far the formal self-employed seem to be the most satisfied group, followed by employees, and then the informally self-employed and inactive population. The unemployed are the least satisfied. All things considered,how satisfiedwouldyou say you arewith your life?" Source: ALMS 2006. 77 Using probit estimations we looked further into the data. Controlling for key demographic characteristics, 21 we used a more disaggregated range of employment status variables to determine whether there are statistically significant associations between subjective well-being indicators and these states. The kinds o f employment status (which are mutually exclusive) include: unemployed; government worker (civil service or parastatal); private organization worker22;informal wage worker; NGO; formal self employed (those with a license); and informal self employed (those without a license). There i s also a dummy for "professional," which means that the individual has a professional occupation, whether they work for the public or private sector. The reference group i s the inactive population. When asked "How well-off financially do you think you are compared to your friends, neighbors, and other people you know personally?," government and private organization workers, and the formal self-employed, are significantly more likely than the inactive to say they are somewhat or much better off. The unemployed are significantly less likely to say they are better off (see Annex 1, Table 47). However, there i s no statistically significant difference between the responses of the informal workers and informal self-employed, and the inactive population. The same regression analysis on the question "Right now, how satisfied with your life do you think you are compared to your friends, neighbors, and other people you know personally?" results in precisely the same pattern. While the problems inherent in"subjective" indicators (e.g. Ravallion and Lokshin 2002) limit the reliability of this analysis, the results are consistent with the hypothesesresultingfrom other data and sources, and reinforce the notion that most people, given the choice, would prefer a formaljob to informality. 4. ProfileofInformalEnter~rises'j 3.24 Sectoral Composition. Most urban informal activities fall under the broad headings of manufacturing (45 percent) and trade/hotels/restaurants (38 percent) (Table 41). The most significant gender difference is in manufacturing, which employs the majority o f women in the survey (55 percent), but only 30 percent o f the men. M e n are more likely to be engaged in "Other Social Activities" (19 percent o f all men, compared to 5 percent o f women) and are also much more likely to work in construction, transport, andreal estate thanare women. Men andwomen inthe survey are equally likely to work inthe trade, hotels, andrestaurants sector. Education level varies somewhat by sector of activity as well. Levels o f illiteracy are generally high (71 percent of those engaged in manufacturing; 63 percent of those inthe trade/hotels/restaurants category). However 81 percent o f those in the construction sector fall in the Grade 1-5 cohort, and 77 percent o f those working intransport are inthe Grade 5-8 category. Interestingly, national accounts data and other sources confirm that the construction sector has been growing rapidly in urban areas, and the transport category also seems to be more dynamic than the larger sectors (see below). ''Age, 'I age squared, years o f education, years of education squared, gender, and whether currently married. Technically we cannot separatethe informal and formal employees o fprivate f i ; however, we assume that the bias i s fairly small, because several important categories o f informal workers are listed separately in the survey: domestic worker; apprentice; contract worker; member of cooperative; and unpaid family worker. Therefore we group these ina separate "informal wage worker" category. 23Data and analysis in this section are drawn from the Urban Informal Sector Survey (2003) unless otherwise specified. 78 Table Ji: Sectoral Go positionof Infor 32.5 Firm ~ ~ r. How hetero~e~eousthe informal sector`? It seems that in~ ~ ~ irs i ~ ~ ~ ~ fact m a ~ informal ente~risesshare similar observable c~aracte~st~cs.The typicat Eim y inthc ISS: ** IKceps s a sole prvpr~etors~i~ppercent); (99 no book of accounts (only 1 percent keeps partial accou~its); # Sells on cash temis only (rather than credit) (95 percent); # Operates 1Imonths out ofthe year; little v ~ a t ~ across sectors s ~ i ~ ~ e~s~t sm ~ t e d o n , althou~hthis could be related to time o f year the survey was ( ~ a n ~ abouty th~ ~ ~ r esses operate 12 ~ ~ o n perh s ~ ** year; 2 percent), or has be~~Feen and 4 1 * Most enip~o~eesare unpaid f a ~ members (46 perce~~t)?though many are as more than4 c ~ ~ ~ o ~ande s ; e ~ ~ y ~ e ~ p ~workers~(32 percent^; only 10 ercent are p e ~ ~ ~ ploy^^^ r a r e n t 3 2 6 The majority of ~ u ~ ~ ~ ~areshs~e~si e - b a(53 ~pcrcent), which is ~aturaIlyhigher e s e in a ~ ~ c ~(64 ~percent), but even higher in ~ i a n u ~ a c r L(88r ~percent). G ~ n e ~ ~ ~ ~ y ~ ~ u t e ~ n ~ location ctioicc stems to be associated with the type o f activity: 83 percent of rans sport e ~ t ~ ~ have ae "`moblfe location." ~ s s Tradeihotelsitestaurants are mainly t i o ~ ~ ~ - ~ a ~ e d (29 percent) or in m open space (54 percent); these e ~ ~ ~ ~arc also thesmost likely to r ~ s e have a pemanent stritc (5 percent). Of t who work at home or in a pemianent structure outside the home, r y 45 percent own and 45 percent ther 10percent usethe s ~ ~ cor free,r e ~ ~ 3.27 Not s e ~ ~ facation~ varies ~si ~ ~ s e The ~ ~ j perated enterprises~arc home o ~ ~ dcdicatcd space, while men are more likely to work in a p e ~ or~~ e ~~ ~ o~ r aer y ~ ~ s t ~ c ~outside the home, on an open space in the street, or without a fixed locatjon ~ r c 79 (mobile). Nearly half o f women chose their location because "owner lives here/convenient to look after children." This underscores the limitations on women's mobility and activities resulting from their responsibilities in the home. Location also varies according to the combination o f gender and sector. Among men who are home- based, 26 percent work intrade/hotels/restaurants, and 67 percent work inmanufacturing; o f home-based women, 19 percent work in trade/hotels/restaurants, and 77 percent in manufacturing. Men are more likely to have chosen their location for a business reason (near customers, competitors, or raw materials). Figure21: EnterpriseLocationandReasonfor Locationby Gender I 50 7 45 45 40 40 35 35 c8 0 Male 25 20 0 15 0 15 I O 10 5 5 0 0 1 Source: Informal Sector Survey 2003, own calculations. 5. Analvsisof EnterwiseDvnamism 3.28 Based on the foregoing discussion, the informal sector seems to be characterized by a low level o f productivity and dynamism-but the informal sector is heterogeneous by definition, since it includes many types o f firms and workers. We therefore attempt here to see where there are important discontinuities. There i s no doubt that the sector harbors both subsistence workers living well below the poverty line (firms that could be loosely categorized as the "lower tier" o f the informal sector; Cunningham and Maloney 2001), as well as entrepreneurs who have grown substantial businesses (the "upper tier"). A key question, therefore, is whether the "upper tier" is a small minority or a significant part o f the sector. Given data limitations, it is not possible to assess productivity, or whether these firms are actually growing-but the assumption is that the upper tier has more potential for growth than the survival-oriented enterprises. Based on the ISS, most informal firms have characteristics that suggest a low level o f dynamism-they are mostly sole proprietorships with no employees, and generally either home-based or conducted inthe open air. However, looking only at averages could mean overlooking a significant number o f dynamic firms. In short, because the informal sector is so heterogeneous, it would be useful to subdivide it into smaller categories and compare the relative size o fthese categories. 3.29 We follow Cunningham and Maloney (2001), who use cluster analysis to identify these unequal tiers o f firms ina large dataset o f Mexican firms, inorder to determine the 80 size o f the "upper" and "lower" tiers o f the informal sector. They choose cluster analysis because it "permit[s] the segmentation o f the market to emerge from the data with a minimumo fimposedprior structure" (p. 132). Controlling for Characteristics ofthe firm and operator, entrepreneurial dynamics (inter alia plans to expand; initial financing; financing problems), and several indicators o f participation in formal market and legal institutions, the clustering results in five categories o f firms. The least successful cluster comprises only about 17 percent o f the firms. At the same time, about 70 percent o f all firms entered the informal sector voluntarily. Inshort, the authors conclude that "though the resulting map o f the self-employed sector i s complex, only a relatively small fraction o f firms appears to correspond to the traditional dualistic view o f informal self- employment as a disadvantaged sector o f a segmented market" (p. 132). 3.30 Using cluster analysis we divide the ISS sample into two groups based on 15 variables representing characteristics o f firms and their operators. 25 These include education level o f the operator, why he/she chose the activity, whether the firm operates away from the home, whether it operates 12months o f the year, andhow many workers it has (see Annex 2, Tables 48-50 for the full results). This enables us to assess, for example, whether better educated operators have stronger firms-in other words, whether positive characteristics such as higher human capital and agency are associated with indicators o f firm dynamism. Most importantly, it provides a rough approximation o f the size o f an "upper" tier o f informal firms inEthiopia. After dividingthe sample into two cohorts using the clustering technique, we examine the key features o f the resulting groups, including gender, sector o f activity, reason for choosingthe activity, andothers. 3.3 1 Table 42 presents the results, noting that the "upper" tier comprises only 4 percent o f the sample. It is important to underscore that the clustering has been done using the data available, which may not accurately capture the key characteristics distinguishing these firms; if unobserved features are more important. With this caveat, however, the table suggests that the dynamic segment o f the informal economy i s quite small. Average monthly sales, calculated by averaging the firm's reported typical sales in a peak month and typical sales in a slack month, drive the breakdown o f the groups. The difference in sales i s by far the most striking contrast, clearly defining the upper and lower tiers o f the informal sector. Lower tier firms have median sales o f Br 215 (US$25) and mean sales o fBr 336 (US$39) per month, or inthe range o f a dollar a day (gross).26 Uppertier firms have mediansales ofBr 2,850 (US$332) andmean sales ofBr 3,392 (US$392), roughly a one order o f magnitude difference. Note that sales rather than profits are reported, 25 The data are not ideal for this type o f analysis as they offer clear tradeoffs between characteristics o f the firms usedandsample size. Due to the number ofmissingvalues for some questions, 9,200 responses were available for the cluster analysis (of the total survey sample o f 14,910). 26 Currency conversions use an exchange rate o f Br 8.58 I/'US$l, which was the exchange rate at the end o f 2002 (the Informal Sector Survey was conducted in January 2003). The sample was capped at monthly sales (peak or slack month) o f Br 10,000 after inspection o f the outliers suggested they were likely due to measurement error. 81 implyingthat the income o f the operator could be significantly lower than these figures (especially inthe case o f firms with additional workers).27 3.32 Differences between the two groups are generally in the direction expected, but the magnitude o f these differences is small. The upper tier contains a higher proportion o f males (43 percent compared to 39 percent o f lower tier operators), and upper tier operators are better educated, as evidenced by the higher median years o f schooling (3 versus 0 years) and prevalence o f vocationalkechnical training. The upper tier has a larger proportion o f tradehotelshestaurants firms and a smaller proportion o f manufacturers, suggesting that there may be more attractive business opportunities inthe former. There are also relatively more transport firms in the upper tier, suggesting that this niche activity may offer growth potential. Upper tier operators are somewhat more likely to report a larger amount o f startup capital, to work year round, to keep some accounts, and to sell on credit. Another sign o f the greater dynamism i s the higher likelihood o f choosing the place o f work for a business reason (e.g. close to customers or competitors), rather than because it is convenient for looking after children, or based on affordability. 3.33 The small differences between the groups' mean firm characteristics are more randomly distributed than one would expect ifthere were in fact a large dynamic segment within the informal sector. In short, the data suggest that Ethiopia's urban informal sector i s fairly homogenous with respect to key observable characteristics, and is survival-oriented. Comparing the standard deviation o f characteristics inthe two groups also shows fairly little difference in within-cluster heterogeneity, with the exception o f the average sales indicator (see Annex 2). 21 The survey contains information o n expenses, theoretically allowing for profitability analysis, but inspection o f the data revealed inconsistencies that suggested simply using sales figures would yield more reliable results. These figures are therefore best understood to indicate fmsize. 82 Cluster Profie (percent unless indicatedotherwise) Lower Upper Number ofobservations (n) 8,835 365 Percent o f total entemrises 96 4 Median monthly sales (Br) 215 2850 Meanmonthly sales (Br) 336 3392 Sectoral Composition Agriculturehlining and Quarrying 2 1 Construction 2 1 Manufacturing 44 35 Transport 1 3 Trade. Hotels. and Restaurants 39 48 Other 12 12 Total 100 100 Characteristics of Operator Male 39 43 Median years o f schooling (yrs) 0 3 Hadvocational/technical training since leaving school 3 7 Activity prior to current enterprise was formal employment 9 10 I Enterwise Characteristics hercent) I I I Keeps partial book of accounts 1 3 Sells on credit 4 11 Works ina permanent or temporary structure outside the 6 5 home Operates 12 months of the year 63 68 Chose workplace because it i s near customers, competitors, or 47 55 raw materials (rather than convenience or affordability) Has at least one worker (other than operator) 17 19 Amount of startup capital i s >Br 1000 11 14 Reason for Choosing Activity Likethe activity/activity bringshighincome 14 13 Small investment neededho alternative source of income 82 84 Future Plans of Enterprise Plan to hiremore workers 10 11 Plan to expand (of those who plan to continue enterprise) 66 62 3.34 As an additional check, further iterations o f cluster analysis divide the data into three and four clusters. The four cluster version is more interesting. Again the groups are mainly distinguished by their median average sales per month: Br 5,520 (less than 1 percent of the sample); Br 2,450 (3 percent); Br 866 (14 percent); and Br 168 (81 percent); in this version, there appears to be a larger contingent o f more successful firms (including the first three groups, just under 20 percent o f the sample i s dynamic). Yet even in this breakdown, the percentage choosing the activity for voluntary reasons was 83 about the sa'fsle for a11 grau seems to be that the "dyti arter of the informal sector. it is notable that thc lowest ~ o n ~ salesyfigure meed above, Br 268 (about h ~ US$L per day}, is close to the a p ~ r o ~ upperapoverty tine for ~ ~ ~ ~ e r areas o f aBr 163 ~ ~ per ~ ~ o which~implies~it i~s p r~o ~ a balreasonable ~ ~ ~ a ~ ~ r e ~ ~ ~ i ~ . ~ 4 n h ~ 3.35 This analysis does not mean that there are no dynamic b ~ s i ~ e s in~the i ~ f o ~ a ~ s s scctor-there are of course ~~a~~success stories. One study o f fcmale ~ n ~ r e p ~ ein~ i e ~ r ~ E t ~ ~ o pp ~ ao ~a lgroup o f ~ ~ s i n e s ssome started with low levels of c ~ p i ~ athat grew e ~ e ~ , l , le ~ ~ ~for thc ~ ~ i ~ ~ firms studied had an nitdian and nieltri Table 43: Median and Mean ~ 1 0Sales by Gender of~~ p e r a ~ o r ~ t ~ ~ and Age of Firm, 2003 ''~ n ~ r uavusage ~ i ~ cpoverty lines for urbanareas as reported inWorld Rank (2005). ~ ~ from the ~ 24Thotigh note again that the figures from the ISS are of average sales rather than Income, sa they mag ~ ~ ~ ~overstateywages. a l I Box 9: DynamicFemale-OwnedBusinessesinEthiopia In2002 the L Oinitiated a study of female entrepreneursinEthiopia, and "in particular on the factors that facilitate or inhibit the growth of their businesses" (p. xiii). The study profiled 123 licensed SMEs that were significantly larger than the vast majority inthe Informal Sector Survey (assets of at least Br 10,000). The operators and firms tended to have more o f the characteristics that are associated with dynamic businesses: 66 percent of the operators had at least secondary education; also, the main reasons cited for choosing to start a business were "to be self- employed" (45 percent) and "family tradition" (21 percent), while only 14 percent said "no other altemative/small investment required." The percentage o f those who "chose" their activity willingly therefore seems to bemuchhigherthan inthe ISS. According to the entrepreneurs, finding a workspace was the biggest constraint at startup (cited by 42 percent). Most rent their space, and the "relatively highrents pose critical problems for them and can hinder their expansion and diversification" (p. xv). Like most firms inthe Informal Sector Survey, these entrepreneurs largely relied on personal savings and loans/gifts from relatives for startup and working capital; 33 percent said that the lack o f a credit facility was their biggest startupproblem. According to these successful firms, access to formal bank credit i s very difficult, particularly due to collateral requirements. Borrowing from M F I s is also a problem because of the small size o f loans offered, and "the inconvenient lending and repayment arrangements" (p. xv). The study found that problems in securing access to credit threatened the sustainability o f many new firms. Source: ILO2003. 3.36 Networks and Cooperation in the Informal Sector. The informal sector literature is increasingly intertwined with research on social capital and informal networks and institutions in general (Barr 1998). This literature contends that informal business is conducted via personal contacts and word o f mouth, and requires mutual trust, since activities are unprotected by regulatory and legal oversight institutions. Personal networks can give entrepreneurs access to customers, markets, information, and technology, and the lack o f contacts can be a handicap. Moreover, the survival-based enterprises o f many informal sector operators are often nearly indistinguishable from their household and family responsibilities (especially for women), and as such are governed by basic social norms o fmutual assistance. 3.37 A s Tripp (2001) argues, microenterprises frequently "adopt operating principles that defy conventional neoclassical assumptions that business people simply are self- seeking individualists single-mindedly bent on maximizing their profits" (p. 4-5). Indeed, 15 percent o f the enterprises covered by an Ethiopian Development Research Institute (EDRI) Micro and Small Enterprises Survey in 2003 indicated that growth was not their primary objective; the survey report points out that this could indicate a `"survivalist' attitude among operators" (Ageba and Amha 2004 p. 64). For example, small firms may share business ideas with each other, charge poor customers less, and help their "competitors." Such behavior is driven by "norms o f reciprocity, mutuality, and fairness," largely because the operators "share a common perception that their survival is contingent on that of others in their community" (Tripp 2001 p. 5). Among the self-employed respondents to the ALMS, more than half said that they often or sometimes "share business ideas and discuss problems with others who do the same kind 85 o f activity." About two-thirds o f the entrepreneurs know at least half o f their customers, and 14percent say that they charge very poor people lower prices thanother customers. 3.38 Research on Ethiopia reveals evidence o f the importance o f personal contacts and trust to the functioning o f the informal sector, and indeed to the formal sector as well. The linkages between formal and informal enterprises, and the importance o f informal institutions, are illustratedby the grain trade (Box 10). Another commodity critical to'the economy-seed-is traded largely via informal sociaVfamily networks. The seed trade is sometimes linked with participation in iddir, an indigenous Ethiopian institution with a central role in communal life. Originally a funeral society, iddir has recently become a multi-functional institution, including serving as a forum for discussion o f farming issues, and trading o f seeds: ....exchange o f seeds i s effected in the form o f bartering, gift, borrowing, and of course sale. ... The rate o f exchange o f one type o f seed with others depends on the importance o f a seed (cash crop, food, or both) in question and thus the exchange would not necessarily be based on a one-to-one ratio. A farmer also obtains seed in return o f his labour he provided for others. It is this living tradition o f mutual interdependence that sustains local seed supply.. , (Seboka andDeressa 2000, p. 250) The practical realities o f the seed market, therefore, reflect the principle o f "generalized reciprocity," and support the notion that the behavior o f informal sector enterprises may not be adequately explainedbythe norms o fprofit-maximizing behavior. 3.39 Furtherresearchusingqualitative methods could delve into questions about social capital, networks, and firm behavior. One interesting question is to what extent vulnerability and the lack o f strong social networks are related-presumably they are closely intertwined, and this could be an important (and somewhat overlooked) determinant o fpoverty. For example, it is striking that 44 percent o fthe self-employed in the ALMS report that ifthey are unable to attend to their enterprise on a given day, there i s no one they can ask to look after it for them. This percentage i s actually higher (a statistically significant difference) among the unlicensed firms (where informal networks mighttheoretically bevery active) thanthe licensedfirms. 86 Box 10: InformalBrokerageinthe GrainTrade Dueto the lack of formal regulatory, inspection, and brokerage services, Ethiopian grain traders are reluctant to deal with buyers and sellers they do not h o w personally, or have a strong connection to (e.g. via kinship). Traders who make sight-unseen agreements on the purchase of a certain quality and quantity o f grain riskbeingcheatedby an unscrupulous trading partner. To mitigate this risk, an efficient and sophisticated informal institution o f brokerage has developed inAddis Ababa, which isthe hubofthe country's roadnetwork andthus acts as a clearinghouse for trading grain among far-flung regions. About 40 brokers are permanently located in the city's grain market, and they perform a variety o f functions-most importantly matching buyers and sellers in exchange for a flat-rate commission, which guarantees that the seller will be paid and the buyer will receive what he expects. Brokers also inspect and assign quality "grades" to each bag o f grain they trade, and "carry out an implicit biddingprocess that results in a single daily Addis Ababa spot price for eachtype and origino f grainprevailing inthe Addis Ababa market" (p. 41). Brokers can also be engagedto mediate disputesthat arise among traders. Inshort, brokers leverage their wide range o f personal contacts and reputations for integrity to "facilitate impersonal or anonymous market exchangebetween traders despite the constraints imposed by the market environment" (p. 83). Since brokers and grain traders are supposed to have licenses, it i s not clear what percentage would qualify as informal sector operators-perhaps the majority would not. This impliesthat a major activity taking place in the "formal" sector relies heavily on an informal institution to mitigate riskand improve efficiency. Moreover, few traders resort to the formal judicial system for dispute resolution, preferring instead to work within the grain trade's informal norms o f conduct and turning to brokers when an objective mediator is needed. Inthe process, brokers function as a critical link in a value chain that starts with (largely informal) farmers and ends with (presumably, often informal) retailers selling to consumers, possibly in a different region than the region of the grain's origin. Moreover, enteringthe brokerage business i s highly dependent on social capital, since it requires trust and personal contacts; "thus, 60 percent o f the brokers surveyed had a parent in the grain trade, and 40 percent had a parent in grain brokerage. Grain brokerage businesses pass fkom generation to generation, gradually transferring trust from father to son, who acquires the father's clients" (p. 42). As a result, there are fairly high informal barriers to entry, and relatively few brokers. Source: Gabre-Madhin 2001. 6. Constraintsto DoingBusiness 3.40 Policy Environment. Since 1992, the policy environment has become more hospitable to entrepreneurship. Small business activity was powerfully repressed under the Derg. Inthe more open environment after the Derg's fall, the rate of new business formation accelerated a trend that was reinforced by the structural adjustment program, which led to the movement of some public sector employees into self-employment (Ayele 2006). In 1997, the Government announced the National Micro and Small Enterprises Development Strategy, which allows for license requirements to be waived for MSEs (those with capital under Br 5,000), and aims to improve access to credit (GOE 1997). The strategy recognizes the key role of the sector injob creation, and states that it 87 i s "a quick remedy for [the] unemployment problem," but that this will necessitate "a direct intervention and support o f the government" (p. 7-8). 3.41 Among other points, the MSE strategy outlines ideas for implementing training programs, helping MSEs export their products, and rebalancing government support toward smaller enterprises (given the traditional bias toward supporting large firms). It also emphasizes that the private sector itself will play the most important role in promoting the MSE sector, alongside the efforts o fNGOs and industrykrade associations. Interms o factive programs, the Federal MSEDevelopmentAgency (FeMSEDA) andits regional counterparts (ReMSEDAs) provide training, credit facilitation, and other services to entrepreneurs, with activities varying by region. The Addis Ababa ReMSEDA inparticular has been very visible, and its programs are being considered for scale up to other parts of the country (Box 11). Box 11:The Addis Ababa ReMSEDA The Addis Ababa MSEDevelopment Agency was established in2002, andhas served as a model for other regions. It has reached around 200,000 people, of whom about 50 percent are engaged inthe informal sector. About 90 percent ofthe 91,000 licensedenterprisesthe agency serves are engaged in retailinghervice activities. The agency targets unemployed youth (including some WET graduates) andunemployedwomen (particularly female householdheads), and assists both new entrants to the labour market as well as those wishing to promote existing businesses. Each kebele inAddis Ababa has five extension agents working for the agency. Before forming the agency, the city government conducted a needs assessment and prioritizedthe following sectors: textiles and garments; food processing; wood and metal works; and construction and municipal activities (such as dry waste collection and car parking services). Around 11,000 enterprises were also surveyed to determine key constraints, which were found to be a lack of working space, technology, start-up capital, input access, information, and adequate managerial skills. Based on these identified constraints the city government put together a package of services including: 0 WorWtrudeprernises: The ReMSEDA has constructed kiosks aroundthe city (the kiosks are blue and have become quite recognizable evidence of the agency's efforts), which are provided to entrepreneurs with existing businesses. For new businesses without working premises, the agency provides spaces in large work sheds/buildings. These work spaces are intendedto be usedtemporarily; once they are somewhat established, entrepreneurs are expected to move elsewhere. 0 Credit: The city government i s the largest shareholder in the Addis Credit and Saving Institution, S.C. This institution started with a capital o f Br 10 million (around US$115,000), which has grown to Br 100 million (around US$11.5 million). It has loaned Br 106 million (around US$12.2 million), but i s not able to satisfy the ever-increasing demand. To this effect, the agency is working with 7 other M F I s to extend the provision o f credit. In order to obtain a loan, entrepreneurs have to meet certain requirements (i.e., have a viable business plan) or have collateral, which can take several forms (for example, tools provided by the agency can be used as collateral, or for clients that work for city housing project offices, part o f their salary can be divertedto repay the loan). 0 Marketing: Marketing know-how among informal sector operators, particularly youth, tends to be low. The central agency's marketing department i s working to organize staff at the kebele level to provide marketing support, but current capacity cannot satisfy demand (there i s 88 opportunity in the textile sector, for example, to export internationally, but support to entrepreneurs i s needed). Activities include conducting analyses o f individual businessesinorder to offer advice; in each kebele, in a 5 month cycle, 10-15 businesses can be analyzed and provided withrecommendations. a Business Development Services: The office works with bilateral organizations, including GTZ, in order to provide business development services (BDS). In a 5-month cycle, a BDS facilitator at the kebele level can analyze about 5-10 enterprises in order to help identify wealmessesand strengths, andto help indevelopinga business plan. a Technology: The agency provides technology-related advice and i s trying to develop efficient and cheap means o f production that could be introduced to private training centers; it attempts to be selective in the technologies it promotes, to ensure they are context-appropriate (current plans call for introduction o f weaving and pottery technologies, as well as mobile kitchens). Italso provides clients with tools. a Information and Counseling: Clients can approach kebele offices to obtain information about licenses, credit, etc., butthe lacko fphone service inkebele offices i s a constraint. a Licenses/Legal Support: The office helps clients obtain licenses, and assists them in the legalprocess o f accessingcredit. Businesseswith paid-up capital of up to Br 5,000 (US$575) can register their business but do not need a license, while businesses with greater capital require a license. The agency has a prominent place in the City Government, and its Bureau head is a cabinet member o f the City Council. It works with a variety o f stakeholders, including the Land Authority, various MFIs, and some private colleges and NGOs (such as Plan-Ethiopia and Propride Ethiopia) in order to provide services. It has created about 150,000 jobs through its activities, o f which 25-30,000 were temporary jobs. A number of capacity- and market-related challenges confkont the agency. Staff and resources are currently insufficient to meet demand; providing training sessions can be very costly, as clients are often unable to attend if all expenses (transportation and food) are not provided. Managerial and marketing skills among beneficiaries tend to be low, and this i s an important constraint on the growth of their enterprises, which are often operated without books o f account or a strategic plan. Raising awarenessamong actors inthe privatehon-governmental sector of the critical role MSEsplay inthe economy is also a challenge. burce: Interview with Addis Ababa SMEProgramofficial, June 2005. 3.42 The environment for informal firms is nonetheless challenging. Of a panel o f Ethiopian experts convened to provide input for the UN Economic Commission for Africa's "Governance Profile o f Ethiopia," only 22 percent believed that the Government "always/usually recognizes the importance o f the informal sector," while 42 percent and 36 percent believed the Government "sometimes" and "rarely/never" recognizes its importance, and "provides no encouragement towards its development" (UNECA 2004, p. 8). About 60 percent o fself-employedrespondents to the ALMS saidthat they hadnot noticed an improvement in access to working premises or to credit in the previous 3 years, despite the government's efforts in those areas. Unlicensed firms were less likely to have noticed improvements, suggesting that the benefits of reform may be reaching more formal than informal firms. Three-fourths of firms in the ALMS have not sought any outside help in solving business problems, most often because they "do not know 89 where to go for help." Among firms that have sought help, unlicensed firms are much less likely to report that they received the help they needed. 3.43 Qualitative research and anecdotal evidence suggest that some government policies, or the interpretation thereofby local officials, can dampen informal activity. For example, according to one study, "it is commonplace to witness police harassment o f informal sector operators, including women, on the roadsides" (Zewde and Associates 2002). It seems that government action is needed to ensure that actual practices o f officials are inaccordance with stated policy. 3.44 Key Constraints. This section exploits the large amount o f available survey data, complementedby qualitative insights from the literature, to assess the relative importance o f key constraints facing informal enterprises. It relies most heavily, however, on the 2003 ISS and 2003 DSTS, both o f which asked respondents to indicate the three biggest problems currently facing their businesses. We used regression analysis to estimate the likelihood that various demographic and firm characteristics are associated with identifying certain business constraints. The findings are referred to below where relevant, and the actual regression results appear in Annex 2, Tables 48-50 (ISS) and h e x 3, Tables 50-51 (DSTS). 3.45 The main problems cited by respondents to the ISS and DSTS, shown in Figure 22, demonstrate remarkable consistency. Chief among them are lack o f demand and lack o f working capital. Problems with capacity to provide goods and services are important (perhaps an encouraging sign, since it suggests the presence o f demand), as well as difficulty in accessing a workplace. "Harassment by government bodies" was a response option only in the DSTS, where it appeared as a significant issue. Health and family- related problems were important in the ISS (the latter was not an option in the DSTS). These problems can be aggregated into three broad issue areas: the demand for goods and services, worker skills and assets, and the investment climate, which are assessed in turn below. 3.46 Demandfor goods and services. "Lack o f demand" is a major complaint o f micro entrepreneurs, who seem to suffer from intense competition among relatively undifferentiated firms providing low value added goods and services. While the number o f new businesses created suggests there are low barriers to entry, the fact that many business fail to grow or fail entirely signals that firms are often mired in subsistence activities with little prospect o f raising the operator's standard o f living. According to the EDRI Micro and Small Enterprises Survey, years after their formation "52 percent o f those [firms] which started with 1, 2, or 3 workers still have the same number o f workers.. .. only 26 percent [of microenterprises] have managed to grow.. ,. [and] at the same time, a large number o f firms that started as small enterprises have `contracted' to microenterprises" (Ageba and Amha 2004, p. 17). Moreover, a large share of firms are price-takers-i.e. selling commodities or standard services at the going rate (ALMS)- which limits their potential for increasingprofit margins or developing a larger customer base. EBDSN (2006) notes a degree o f market saturation by similar products, and that gains by firms that attempt to innovate are quickly erased as others imitate their success. 90 Weak purchasing power in the economy naturally limits demand as well, although economic growth rates have been positive inthe years since the 2002/03 drought. Figure22: What are the 3 mostDifficultProblemsCurrently Affectingthe Operationof your Activitymnterprise? 80% 60%. ~~ [Ilss_ 50% - d- 40% - 30% - 0DSTS 20% - 10% - 0% I I I I Source: Own calculations. Data note: Response options from the two surveys are very similar, though with slight wording differences. Some o f the categories have been aggregated in the graph to facilitate comparison (note that the "no problem" and "other problem" categories are omitted, along with a few others which hadnegligible responses). The full category labels, inthe order presented inthe graph, are as follows. The wording in the ISS i s listed, with the DSTS wording in parentheses where it differs from the ISS: 1) Lack ovinadequate market (lack o f profit; lack oflimited market); 2) Shortage o f working capital; 3) Limited capacity to produce, trade, give service/lack o f raw material (limited capacity to trade or give service/shortage o f goods or commodities); 4) Lack o f working place (lacklinconvenience o f working place); 5) N/a in ISS (Harassment from government bodies); 6) Barriers on free movementlgovernment regulations/too muchbureaucracy to obtain license (Barriers on free movementllack o f clarity o fregulations/Bureaucraticproblems to obtain license); 7) Health problem; 8) Inadequate skilVProblems with workers; 9) Unable to copy with family, social responsibility/Credit to friends, relatives (N/a inDSTS). 3.47 Lack o f demand seems to be a more frequent complaint among firms on the less dynamic end of the spectrum. According to the ISS regression analysis, manufacturing firms are significantly more likely to complain o f lack o f demand. This supports the results of the cluster analysis, which suggested that manufacturing firms are less dynamic than those in tradehotelshestaurants and some o f the smaller sectors. Older operators and those who reported that they chose their activity because they had no other option were also more likely to report a weak market for their offerings. About half o f respondents to the Micro and Small Enterprises Survey thought that domestic competition had increased a lot, and another 43 percent thought it had increased slightly, since the reforms o f the early 1990s (Ageba and Amha 2004). 91 3.48 The Small Scale Manufacturing Survey (which covers rural and urban areas)30 asks respondents to further explain the "lack o f market" that many cited as a key issue. Ofthose reporting it as a problem: 2 percent are "unable to compete with foreign product on quality" 7 percent are "unable to compete with foreign product on price" 6 percent are "unable to compete with local product on quality" 0 27 percent are "unable to compete with local product on price" 0 58 percent said "other" The share o f respondents concerned with domestic price competition and "other" problems supports the argument that there is a glut o f undifferentiated products on the market. There could also be problems with market access, given the importance of personal networks to finding customers, and the mobility constraints that many female informal workers experience. Although this survey i s not representative o f the urban informal sector as a whole, it suggests that the fundamental problem is related to the low level o f economic development ingeneral. 3.49 Workers'skills and assets. While general market conditions are key to explaining low incomes in the informal sector, workers' own capabilities-their skills and assets- are another central piece o f the puzzle. Although few firm operators self-report inadequate skill as one o f their biggest problems, qualitative studies frequently mention low skill levels as part o f the explanation for low productivity in the sector. Access to credit, on the other hand, i s probably the most common firm-reported problem across both the quantitative and qualitative literature. For example, the recent Participatory Poverty Assessment (PPA) identified "lack o f education and skills, and inability to start- up self employment enterprises due to lack o f savings or credit" as the key factors causing vulnerability inurbanareas (MOFED, forthcoming, p. xxviii). 3.50 The human capital base is very low in Ethiopia, given that about 27 percent o f employed adults inurban areas are illiterate, and the illiteracy rate rises to 37 percent for the informally employed (own calculations from LFS 2005). Analysis o f the Ethiopian Urban Panel Survey (see vol. 11, Chapter 5) revealed that while the better educated members o f the labour force are relatively unlikely to be informal, within the informal sector those who have more education tend to have better outcomes. As an illustration o f the importance o f raising the skill level o f informal workers, about three-fourths o f respondents to the EDRIsurvey said that they are "willing to pay fully for or share inthe cost o f training" for themselves (Ageba and Amha 2004). The types o f training they indicated would be useful, in order o f importance, were 1) marketing management; 2) skills training; and 3) business management training. 3.51 Poor health can be another limiting factor. A small but significant share (6 percent) o f informal sector operators report that a health issue i s their most difficult problem (ISS), and these operators are concentrated at the small/low productivity end o f the spectrum, underscoringtheir vulnerability. 30The SSMS i s composed primarily (85 percent) o f firms engaged ingrain millservices. 92 3.52 Most informal firms are dependent on their own savings or gifts/loans from relatives and friends to meet working capital needs. As shown in Table 44, only a very small percentage o f firms acquire capital via formal lenders; in contrast, monetary help from relatives that does not have to be repaid is a significant source o f initial capital (23 percent). The formal microfinance industry was created by the legislation passed by the National Bank o f Ethiopia (NBE) in 1996, with the objective o f facilitating access to financial services for the poor. There are a total o f 25 licensed MFIs in Ethiopia (licensing is required for MFIs to operate legally), and they exist in all regions except Somali, Afar and Gambella. Table 44: SourceofInitial Capitalfor Informal Sector Enterprises,2003 Source ofInitialCaaital Percent Own savings 46 Borrowing from frienddrelatives 24 Donation/assistance/inheritance from 23 friendshelatives Borrowing from bankhndividuals on 2 terms/MFIs Assistance from gov't/NGOs 1 Other 4 Source: Informal Sector Survey 2003, own calculations. 3.53 According to the Association o f Ethiopian Micro Finance Institutions (AEMFI), the number o f loan clients nationwide grew from just under 500,000 in Dec. 2001 to more than one million inDecember 2004. However, the NBE estimates that less than 10 percent o f demand for microfinance services i s satisfied. Inpart, the gap is due to lack o f interest from the private sector (only one MFI i s partly owned by a private company). Significant reforms to the regulations governing MFIs have already been undertaken, including some liberalization o f interest rates and loosening restrictions on maximum loan sizes, but there may be scope for further beneficial reforms. Gobezie (2005) studied microfinance regulations inEthiopia and argues that a "definite time table" is needed for eventual elimination o f the current prohibition on foreign investment inthe microfinance sector, to allow for increased competition, knowledge transfer, and subsequently better choices and services for clients. Moreover, the mandated minimum rates MFIs must pay on savings accounts limits their ability to operate as viable businesses, which in effect means reduced financial services for those inunderserved areas. 3.54 According to the regression analysis, migrants3'are 5 percent more likely than non-migrants to report that a shortage o f working capital i s one o f their 3 most difficult current problems. Those who reported that they chose their current activity because o f the small investment needed/lack o f alternative were 6 percent more likely to report a shortage o f capital. Sectoral dummies have a very large impact: firms in the trade/hotels/restaurants and manufacturing sectors are both about 40 percent more likely than the base case (service sector) to cite a shortage of working capital. Problems with working capital seem more likely to affect small firms; regression results from the DSTS 31Defined as respondents who have lived inthe town for less than 10years. 93 indicate that variables related to firm size-keeping accounts and having more employees-are correlated with a somewhat lower likelihood o f reporting cash problems. 3.55 Female-owned businesses start far more modestly than male-owned firms, as shown inTable 45, which lists the median startup capital by major sector (DSTS). Inthe retail category (55 percent o f the total sample), the median startup capital i s about twice as highfor male-owned enterprises. Among service firms (37 percent o f the sample), thq difference is a factor o f 8. However, among the small group o f wholesale firms (there are 99 female-owned wholesale firms in the sample), female-owned firms actually have a substantial advantage-but these are likely to be formal rather than informal firms, considering their size. The percentage o f microfinance loans held by women varies widely-some MFIs specifically target women, and women account for 60 percent or more o f the lending portfolio, while other MFIs lend well under 40 percent to women,32 suggesting that there are opportunities to improve gender equity relatively easily through the existingMFIstructure. Table 45: MedianStart Up Capital by Sector and Ownership Status Ownership Wholesale Retail Service Male Sole Proprietorship 5380 2000 2450 Female Sole Proprietorship 8500 850 300 Not Sole Proprietorship 8060 3000 2400 Source: Distributive and Service Trade Survey 2003, own calculations 3.56 Investment climate. Issues related to government regulations andpolicies are less frequently cited by firms than lack of capital and lack o f market, but they are significant, andare especially interesting as they are more easily influenced inthe short term. These issues include rules and regulations, taxes, access to land and workplacehuildings, and guarantees on property/work rights. Access to infrastructure services (e.g. electricity, water) could potentially also be a problem for small firms, but this is not referenced inthe available survey data. 3.57 Harassment o f informal sector operators appears to be an important concern, though it i s unclear how widespread the problem is. Nearly 40 percent o f ALMS respondents agreed or strongly agreed with the statement that "government and local administration officials make it difficult for me to conduct my business activities." The DSTS included "harassment by government bodies" as a response option when asking firms to name their three biggest current problems. 3.58 Regression analysis, controlling for other factors, suggests that retail firms are 11 percent more likely than the base case to say harassment i s a problem. Although only 9 percent o f operators report this as one o f their three problems, those who do cite harassment as an issue are 8 percent more likely to say they plan to discontinue the firm's current activities. This suggests that harassment is a relatively severe problem for firms that do experience it (interestingly, hotels and related businesses are 16 percent more likely than the base case to say harassment is a problem). On a related note, lack o f 32See http://www.bds-ethiotia.netlfinance/loans-micro.html. 94 policy predictability appears to be a major issue. In the MSE Development in Ethiopia Survey, about a third o f microenterprises reportedthat they have to deal with `unexpected changes in rules, laws or policies which materially affect their enterprise,' and about 40 percent "do not believe that the government adheres to its announced policies and rules" (Ageba and Amha p. 60-1). 3.59 Accessing a workplace proves a significant obstacle for many small enterprises, especially in Addis Ababa. Regression analysis on both the ISS and DSTS revealed a statistically significant association betweenproblems with finding a workplace and living in the capital. Also, new migrants (defined as those who have lived in the town of residence for 10 years or less) tend to have more difficulty in finding a workplace (in the ISS regression, 4 percent greater likelihood o f citing a problem with access to a workplace). Firms in the tradehotelsh-estaurants sector are also more likely to have workplace problems relative to the base case (services sector). 3.60 Key constraints. First, gender o f the operator is frequently statistically significant inexplaining which constraints firms report. Female owned enterprises seem less likely to report problems with policy or harassment, but more likely to be constrained by family-related issues. This i s consistent with the fact that female-owned businesses tend to be more subsistence oriented, and are more frequently home-based. Regional differences are also significant. Inparticular, firms in Addis Ababa seem to have more difficulty with investment climate constraints, especially with accessing a workplace, but also with respect to harassment by government bodies and dealing with policy issues. On the other hand, Addis Ababa firms have less trouble in terms o f capacity (i.e. accessing goods and services) than firms inother towns. 3.61 Reported constraints vary widely by firm size (as proxied by sales). For example, inthe DSTS, the mediumannual income offirms reporting aproblemwithharassment by government bodies (Br 5,400) is about three times higher than that o f firms reporting a shortage o f working capital (Br 1,700). This suggests that cash constraints are more salient for smaller firms, while larger firms are more likely to face problems with the investment climate. The median income o f firms citing a "lack oflimited market" is Br 2,600, which i s close to the median for all firms, suggesting that this i s a problem faced by firms o f all sizes. 3.62 Although the regression results cited here focus onjust one o f the questions inthe surveys-naming the three biggest current problems-data from other questions (the biggest problems at startup andbiggest constraint to expansion) point to the same general conclusions. They are omitted for the sake o f brevity, but suggest that certain key issues are persistent and affect firms at all stages. However, responses to a question about the assistance most needed from the government vary widely. This indicates that there is no magic bullet to solving the problems o f informal firms. Overall, economic growth and empowering the individual-via improved access to education and productive assets and capital-are probably the most important factors for improving the business climate. However, there are steps that the Government could take in the short term to make the investment climate more conducive to higher incomes inthe informal sector, as discussed below. 95 3.63 What are the policy priorities, given the myriad challenges faced by informal firms? The difficulties in crafting active policy interventions to stimulate entrepreneurship are illustratedby Ethiopia's Investment Incentives Scheme (11s). Ayele (2006) assesses the reintroduction o f this program inthe 1990s-first started inthe 1960s but suspended for about 15 years under the Derg, it provides for duty-free import o f capital goods and income tax exemptions to new businesses that meet certain criteria. The 11s aims to influence the location (with a view to encouraging investment in relatively undeveloped regions) and industry choice (preferring `pioneer' industries) o f new businesses. Ayele finds, however, that these incentives had virtually no impact on entrepreneurs' decisions about when, where, and inwhat line o f work to form a business. Because Ethiopia's entrepreneurs tend to be sole proprietors with limited financial resources, their choices regarding business formation are largely determined by their experience and training, and where they live. Entrepreneurs who are in a position to choose a suitable location (beyond their home area) are influenced primarily by the business environment, including access to infrastructure andmarkets, which outweigh the limited incentives provided by duty or tax exemptions. Ayele concludes that "the implications are clear: assist enterprise development through developing infrastructure, expanding the education system (particularly technical education), by eliminating bureaucratic constraints and byproviding better access to enterprise sites" (p. 12). 3.64 "Formalizing" the Informal. Why do firms remain informal? This section summarizes the steps, and pros and cons, o f transitioning to the formal sector. There are several reasons firms might wish to remain outside the formal sector-perhaps the procedures or cost o f becoming licensed i s too onerous, or perhaps they feel unable to cope with the taxes and regulations that govern formal firms. With respect to the former, establishing a business inEthiopia requires 7 procedures, or an average o f about 32 days to complete, according to the World Bank's annually updated global Doing Business Database (Table 46). This represents a significant improvement over the situation a few years ago; infact, Ethiopia was one o f the "top ten reformers instartup procedures" inthe world from 2003 to 2004 (from a low base).33 In the ALMS, about a quarter o f unlicensed firms said that the main reason for not having a license was the cost or difficulty o f obtaining it, or because they did not know how to become licensed. 3.65 Especially as they grow, firms do perceive some benefits from having a license. These include freedom from fear o f their activities being curtailed, and access to services. Respondents to the MSE Development inEthiopia survey said they "would not have hid frondgive bribe to government officials" (47 percent), could "apply for credit" (17 percent), and could "apply for landhusiness premises" (10 percent) ifthey were licensed. 3.66 However, some businesses may remain informal to avoid taxes and regulations. According to the Addis Ababa Trade and Industry Bureau, "among the more than 2,200 businesses that returned their licenses during 2001/02, the major reason for doing so was tax burden and exorbitant rent" (World Bank and EDRI2004). The M S E Development inEthiopiasurvey found that "both accessing the formal sector.. and staying formal are . costly to the operator," meaning that many find the burden too great (Ageba and Amha 33All indicatorsavailable at www.doinabusiness.org. 96 ~~~~~. The m a ~ reasons given for not o b ~ a ~ n a~ license were %usiness is too small to n n g need a license" (51 percent), "no benefits to register in^," (20 percent), and "taxes too high" (13 pcrcen~~. i> ", " I f Ethiopia f Reglod ["`OECD iremcnts do not exist far f i m s nt for an^ operators. T~~ree-fou~hs they "didn't need a license" or Eotind it "easier to operate thou^ a From a policy s ~ a i ~ d ~ ~thisj nitn~ ~ ~ ~thatsencourag~n~ o i e f ` o m a l ~ ~ a t ~pernse is not the goal. Rather, pol~cyn~akersshould strive to change firms' o cost-~e~~efitd ~ c i ~ j coan~ c u l ~bys ~mpro~irrg relevance and p r o v ~ ~ o~fopublic goods the n (such as dispute resolution and pro~ectiono f property rights), $0that frmx consider it a net gain to join the formal e c o ~ ~ o ~ ~ y . 3.68 The i ~ f o m z asector sewcs a variety of critical functions in ~ ~ h ~ o p ieac' so ~ z o ~ ~ , ~ yet strate~icsfor raising p ~ o d u ~ ~and~ incomes in the sector r e ~ ~elusive.~ Most i ~ i ~ y a i ~ firms are small, ~ ~ ~ s i ofs onei nperson with limited e d ~ c a t ~and~ des~itethe ~ ~ o ~ ~ e ~ ~ r ~of`~the sectorj the ymajority s e e m to bc fairly s ~ ~ s ~ s t e n c c - @ rrather~ e ~ e n ~ ~ j c ~ for about 60 percent of the sector, md d ta be lower than th men. They arc s ~ ~ ~con ~ n ~ e ~ t have to balance their activitie 3.49 Mast workers da not seem to "choose" i al sector tvork in any meani sense of the ward$ although it i s not clear that they tvoutd ~ e c e s s ahave~ higher ~ ~ incomes or better ~ ~ o rc o~ ~i ~nj t~i in ~ o nthe fornzal sector jobs for which they would be he q~estjo~if f o ~ a l ~ is~o~f little~relevance for all but thc larger informal o ~ o n firms, ~ ~ hbegin to confront new obstacles-i.c. i c ~ ~ related to g a ~ ' c ~ m crn~t~ u ~ a t ~ o n s - ~ a s they grow, and the incentives to become licensed increase. The most ~ ~ p oself- ~ a ~ ~ t reported c ~ ~ ~ ~ rfaced a ~ Iby~ ~t ns f o ~firms across the board are related to preva~~ing a l market c o ~ d ~ t i o(highs e o ~ ~ p e t i ramong reIatjvely ~ ~ ~ ~ d ~ f f ~ r cgoods~ tandd ~ ~ ~ o ~ n t i c sewices, and thus l~miteddemand for a given firm's o f f e r ~ n ~and s ~ ~ i ~ ~accesse to ~ t d capital. 97 3.70 It seems that policy-related issues, and harassment by officials, are more significant problems for larger than smaller firms-which are also the firms, that have the most potential to contribute to growth. Different strategies may therefore be needed to support the smallest enterprises (e.g. to raise their human capital, and improve access to credit andworking places) and to provide a more conducive climate for larger enterprises (e.g. ensuring that firms are free from excessive regulatory and tax burdens, and arbitrary harassment). Regionally differentiated strategies may also be important. For example, facilitating access to goods and services insmaller urbanareas will require programs that fall well outside the scope o f labour market policies-for example with respect to building infrastructure and enabling higher volumes o f intraregional trade and market formation. InAddis Ababa, more action is needed to improve access to workplaces and the investment climate ingeneral. 3.71 More research on understanding informal norms and institutions would be useful, as this can be key to designing effective p o l i c i e s 4 r at least policies that do no harm. Hoddinott, Dercon, and Krishnan (2005) note: A misunderstanding o f the roles o f these [informal] networks can lead to policy changes that have unintended consequences on the functioning o f these networks, with potentially damaging effects on the capacity o f the poor to mitigate, and cope with, the effects o f shocks. At the same time, abetter understanding o f such networks can lead to the identification o f policies that complement existing networks that already serve the poor well, and to policies that can substitute for networks that simply are not reachingthe poor (p. 2). 3.72 It is also the case that some informal norms may not be entirely compatible with formal sector norms, complicating attempts to strengthen linkages between the two (Tripp 2001). 3.73 Overall, "an international best practice seems to be lacking" on improving productivity and incomes in the informal sector (Becker 2004). A somewhat experimental approach with respect to active programs such as those o f the Addis Ababa ReMSEDA is therefore warranted. An upcoming World Bank evaluation o f the employment impact o f the Integrated Housing and Employment Program in Addis, for example, will provide a more concrete basis on which tojudge whether this kindo f active intervention has a positive impact, and i s worth scaling up or replicating in other cities. An evidence-based approach would help ensure productive use o f scarce government resources. At the same time, as recognized by the Government's MSE Strategy, the private sector is ultimately the most important player in driving growth, even at the level o f small-scale enterprises. With this inmind, highpriorities for the Government vis-h-vis the informal sector include removing unnecessary impediments to firm success-by ensuring a level playing field for entrepreneurs, a predictable policy environment, and a minimalregulatory burden. 98 References Ageba, Gebrehiwot, and Wolday Amha. 2004. "Micro and Small Enterprises Development in Ethiopia: Survey Report." Addis Ababa: Ethiopian Development ResearchInstitute. Ayele, Seife. 2006. "The Industry and Location Impacts of `Investment Incentives on SMEs Start-upinEthiopia." Journal of InternationalDevelopment 18(1): 1-13. Barr, Abigail M. 1998. "Enterprise Performance and the Functional Diversity of Social Capital." WPS/98-1, Center for the Study of African Economies, Oxford. Becker, KristinaFlodman. 2004. "The Informal Economy." Stockholm: SIDA. Bigsten, Arne and Negatu Makonnen. 1999. "The Anatomy o f Income Distribution in UrbanEthiopia." African Development Review 11(1): 1-30. Chen, Martha Alter. 2005. "Rethinking the Informal Economy: Linkages with the FormalEconomy andthe FormalRegulatory Environment." Helsinki: WIDER. Cunningham,Wendy V. and William F. Maloney. 2001. "Heterogeneity among Mexico's Microenterprises: An Application of Factor and Cluster Analysis." Economic Development and Cultural Change 50 (1): 131-156. Doing Business Database. World Bank. http://w.doinEbusiness.ord (Ethiopia; accessed July 31,2006). EBDSN (Ethiopian Business Development Services Network). 2006. Marketing Problems. http://www. start-your-business.net/./business-services/marketing 1.html (accessed June 2,2006). Ersado, Lire. 2005. "Income diversification before and after economic shocks: evidence from urbanandrural Zimbabwe." Development SouthernAfrica 22 (1): 27-45. Gabre-Madhin, Eleni Z. 2001. "Market Institutions, Transaction Costs, and Social Capital inthe Ethiopian Grain Market." Washington, DC: IFPRI. Gobezie, Getaneh. 2005. "Regulating Microfinance in Ethiopia: Making in More Effective." College Park: IRIS. GOE (Government o f Ethiopia), 1997. "Micro and Small Enterprises Development Strategy." Addis Ababa. Hoddinott, John, Stefan Dercon, and Pramila Krishnan. 2005. "Networks and Informal Mutual Support in 15 Ethiopian Villages.'' http://www.economics.ox.ac.uk/members/stefan.dercon/hodd der kr.pdf 99 ILO. 2002. "Women and Meninthe Informal Economy: A Statistical Picture." Geneva: ILO. ILO and MTVWAD (Women's Affairs Department, Ethiopia Ministry of Trade & Industry). 2003. "Ethiopian Women Entrepreneurs:Going for Growth." Geneva: ILO. Maloney, William F. 2004. "Informality Revisited." World Development 32 (7): 1159- 1178. MOFED (Ethiopia Ministry of Finance and Economic Development). Forthcoming. "Ethiopia Participatory Poverty Assessment 2004-05," by Frank Ellis and Tassew Woldehanna. Consultant Report. Addis Ababa. Ravallion, Martin and Michael Lokshin. 2002. "Self-rated Economic Welfare in Russia." European Economic Review 46 (8): 1453-1473. Seboka, B. and A. Deressa. 2000. "Validating Fanners' Indigenous Social Networks for Local Seed Supply in Central Rift Valley o f Ethiopia." The Journal of Agricultural Education and Extension 6 (4): 245-254. Tripp, Aili Mari. 2001. "Non-formal institutions, informal economies, and the politics o f inclusion." DiscussionPaperNo. 2001/108, WIDER, Helsinki. UCW (Understanding Children's Work). 2006. "Child Labour and Youth Employment: EthiopiaCountry Study." Rome: UCW. UNECA. 2004. "Governance Profile of Ethiopia: Measuring and Monitoring Progress Towards Good GovernanceinAfrica." Addis Ababa: UNECA. WDI (World Development Indicators) Database. World Bank. www.worldbank.org/data/dataquery.html (Various countries; access July 31 2006). World Bank. 2005. "Ethiopia: Well-Being and Poverty in Ethiopia." Washington, DC: World Bank. World Bank and EDRI(Ethiopian Development ResearchInstitute). 2004. "Investment Climate Assessment Report." Draft, September 2004. World Bank. Zewde and Associates. 2002. "Preliminary Report: Women Entrepreneursin Ethiopia." Addis Ababa: ILO and SEED. 100 Annex 1: ALMS RegressionResults Table47: ProbitEstimations; MarginalEffects (1) How well-offfinancially doyou Right now, how satisfied with thinkyou are compared toyour your life doyou thinkyou are fi-iends, neighbors, and other compared toyourfi-iends, people you knowpersonally? neighbors, and other people you knowpersonally? Answered "Much better off' or Answered "Much more satisfied" "somewhat better off' or "somewhat more satisfied" Age -0.008 -0.030 (0.84) (3.21)** Age squared 0.000 0.000 (1-24) (3.59)** I Female 0.011 0.017 (0.35) (0.54) 0.014 -0.002 (0.83) (0.14) 0.001 0.001 (0.68) (1.20) Married 0.039 0.095 (1.96)* Government worker (including 0.177 0.270 12.64)** (3.88)** Private organizationemployee 0.114 0.099 (2.60)** (2.15)* NGO employee 0.240 0.337 (2.06) * Informal sector worker -0.054 -0.002 (Domestic, apprentice, cooperative, contract worker, (1.OS) (0.03) -0.178 -0.125 (3.46)** (2.29)* Self-emloved(formal) 0.332 0.336 (4.17)** (4.15)** Self-employed (informal) 0.002 0.024 (0.03) (0.38) professional 0.032 0.090 I (0.52) (1.35) Observations 1063 1068 Absolute value o f z statistics inparentheses. *significant at 5 percent; **significant at 1percent. 101 Annex 2: InformalSector Survey Cluster AnalysisandRegressionResults Each variableequals "I"ifthe statement below is "Upper Tier" "Lower Tier" true, "0" if false: Startup capital was borrowed frombank, individuals 0 0 on terms, or MFI ,1637 .1735 Keeps partial book o f accounts (rather than no 0 0 accounts) .1368 ,1111 Operates ina permanent buildingor temporary 0 0 structure outside the home 0.3158 ,2575 Operates 12monthdyear 1 1 ,4344 .4823 Operator has no other income-generating activity 1 1 .3617 .3918 Operator works more than 40 hourdweek 1 1 ,4714 .4936 Operator chose activity because "No alternative 1 1 source of income" or "Small investment needed" ,3355 ,3359 Operator chose activity because "Like activity" or I 0 0 "Activity brings high income" .3051 .3022 Operator intends to discontinue the current 0 0 activitylenterprise (`don't know' coded as NO) .3131 .3444 Operator's previous activity was inthe formal 0 0 sector A .3183 ,3173 Training 0 0 .2402 ,2056 Nondinary variables: Total number of workers 0 0 A394 ,6397 Average o f peak & slack month average sales (birr) 2850 215 1521.765 349.1615 Amount o f initial capital (non-linear; variables 1 1 coded from 1-10as inquestionnaire) 1.7957 1.5992 Years o f schooling 3 0 4.1799 3.9647 102 Table49: ProbitEstimation: LikelihoodofIntentionto DiscontinueCurrentActivity (marginaleffects) 103 (0.96) Lack ofmarket -0.001 (0.10) Lack of working place -0.005 (0.42) Healthproblem 0.038 (2.73)** Other problem 0.033 (1-72) Observations 13015 104 . Table 50: ProbitEstimation: OperatorsReportingeachProblemas among "ThreeMost DifficultProblems, (marginaleffects) 0 (10) z4z B !z I; 8 g% A e -0.006 0.000 -0.000 0.006 0.009 0.003 0.004 -0.005 (2.67)* (0.38) (0.18) (2.23)' (2.51)* (1.01) (2.72)* * (2.74)** 0.000 -0.000 -0.000 -0.000 -0.000 -0.000 0.000 (0.06) (0.18) (2.11)* (2.62)** (0.76) (0.32) - 0.000 0.001 -0.004 -0.002 0.006 -0.006 0.002 (1.18) (0.43) (0.92) (1.61) (0.45) (1.61) (2.47)' (0.84) * I -0.018 I -0.025 0.065 0.016 0.002 0.012 0.012 (4.06)* * (4.65)' * (4.04). (0.71) (0.12) (0.93) (1.09) i -0.012 -0.009 -0.010 0.013 (1.15) (1.56) 0.027 0.012 -0.002 0.009 (0.16) (0.92) -0.030 -0.004 -0.033 0.013 (1.50) (0.35) (2.84)' * (1.36) Literate I-0.007 0.005 -0.005 0.020 0.035 -0.020 -0.008 0.001 (0.84) (0.71) (0.95) (1.18) (0.74) (0.47) (0.06) w 0.001 0.002 0.010 -0.031 -0.026 -0.030 -0.015 -0.001 10.02) ..- (0.28) (1.04) (1.44) (0.66) (1.03) (0.57) (0.04) Mimated within last I-0.008 0.000 0.004 -0.024 0.048 0.005 -0.009 0.012 (0.67) (0.77) (1.34) =F KeeDssome accounts II -0.031 -0.028 0.002 for enterprise (0.83) (1.48) (0.84) (2.40)* (1.12) (0.78) (0.50) (0.06) F Average monthly 0.000 -0.000 0.000 0.000 0.000 0.000 0.000 0.000 -0.000 -0.000 I I I (0.66) I (1.07) I(0.84) I (0.44) I(2.04)* (0.40) (0.23) -0.007 I -0.000 I-0.017 I 0.006 I-0.015 0.042 0.021 -0.038 -0.001 (4.60)* * (0.09) 0.025 -0.012 (2.34)* (1.07) -0.000 -0.013 -0.026 0.405 0.089 -0.024 -0.079 (0.10) (2.15)* (1.01) (10.35). * (2.73)* * (138) (4.13)** -0.036 -0.072 ~ nts Sector (2.66)* (4.31)* (0.26) (13.28)* (3.57)* (2.73)' * (4.53)** 0 (10) 105 5 F `S Pb .6 1 $ 2 1 0 Ez" - E .-Y$2 -Bc L 32E 3 $ L =a een `E ? t7Y L - ff .; 0 2 % %i3 2i 0 L P v ) o 2 : -1 Firm intendsto 0.010 -0.002 0.001 -0.015 -0.062 0.014 -0.020 close/change services (`don'tknow'coded as NO) (0.71) (0.57) (0.18) (0.98) (2.44)' (0.68) (0.77) Firm establishedless 0.000 0.002 0.005 -0.028 -0.023 0.014 0.003 than 2 yrs ago (0.01) (0.46) (0.87) (1.78) (1.09) (0.59) (0.16) Operatorchose this -0.045 0.004 -0.003 0.013 0.056 -0.006 0.090 activity becauseN o alternativesourceof income'or `It is (3.00). ' (1.75) (0.55) (1.01) (2.24)' (0.39) (4.39)' ' Years since enterprise 0.000 -0.000 0.000 0.001 -0.000 0.000 -0.002 was started (0.08) (0.78) (0.05) (0.95) (0.38) (0.26) (1.74) Addis Ababa -0.032 -0.001 0.035 0.039 0.038 -0.067 -0.01 1 (1.35) (0.29) (2.80)' ' (1.31) (0.80) (2.95)* ' (0.20) 13015 13015 13015 13015 13017 13016 13015 I =FF 13016 13015 13015 parentheses Note; "Policy-related problem" includes "Barriers on free movement, "Government regulations," and "Too muchbureaucracy for license." *significant at 5 percent; **significant at 1percent. "Family-related problem" includes "Unable to cope with family responsibility," "Unable to cope with social responsibility," and "Credit to friendsh-elatives." 106 Female-ownedSole Proprietorship 0.052 (1.13) Male-ownedSole Proprietorship 0.049 Wholesale Trade -0.056 RetailTrade -0.082 Note: don't know' = NO, DSTS. Source:DSTS. Robust z statistics inparentheses. *significant at 5 percent; **significant at 1percent. 107 Table52: ProbitEstimation:OperatorsReportingeachProblemas one ofthe "Three Most DifficultProblemsAffecting the CurrentOperations 0(2)(3)(4)0(6)0(8) c -0.002 -0.058 -0.020 -0.049 0.054 -0.039 -0.015 0.016 (0.20) o* (0.42) (3.00)** (1.69) (1.13) (0.43) (1.14) 0.017 0.007 -0.059 -0.010 -0.031 -0.031 -0.026 -0.009 (2.06)* (0.27) (1.34) (0.68) (0.94) (0.88) (0.82) (0.68) -0.009 0.050 -0.043 0.021 -0.075 n.oox 0.03 1 0.009 I _... (0.76) (0.98) (0.56) (0.67) (1.12) (0.19) (0.59) (0.42) Retail Trade -0.010 0.108 0.048 -0.023 0.020 -0.031 0.080 -0.014 (0.66) - (0.63) (0.77) (0.33) (0.85) (1.34) (0.87) 0.01 1 0.038 -0.123 0.039 -0.045 -0.003 -0.019 -0.021 artial accounts (1.23) (1.81) (3.23)** (2.96)** (1.58) (0.14) (0.88) (3.29). * Whether firm 0.002 0.045 -0.058 0.014 0.043 -0.004 -0.011 0.003 intends to close/change services; 'don't know' coded as NO (0.29) (2.98)'. (2.40)' (1.38) (1.61) (0.27) (0.66) (0.57) Years since firm -0.001 0.003 -0.005 0.001 0.003 -0.004 -0.001 0.001 establishment (1.66) (4.12)" (3.00)** (2.38). (2.49)* (3.96)** (1.29) (6.85)** Total number o f 0.001 0.002 -0.019 0.001 -0.007 -0.007 0.002 -0.003 permanent employees (1.79) (0.99) (2.66)'. (0.76) (1.86) (2.12). (0.65) (1.30) t Total annual -0.000 -0.000 -0.000 0.000 0.000 -0.000 -0.000 -0.000 income (birr) (0.11) o* (0.25) (0.14) (0.00) (0.83) (1.56) (0.88) Maintenance and -0.005 0.002 0.190 -0.043 0.009 0.088 -0.063 -0.017 repair o fmotor vehicles (0.33) (0.04) (1.87) (2.61)** (0.10) (1.15) (1.31) (1.61) Sale of motor 0.013 -0.007 0.198 -0.039 0.169 -0.068 0.043 -0.014 vehicle parts and accessories (1.43) (0.20) (2.28)* (1.95) (2.43)* (2.43)* (0.90) (1.07) Retail sale o f food, 0.020 -0.071 0.120 -0.001 -0.060 0.007 -0.046 0.021 beveragesand tobacco in s cialized stores (1.11) 0' (1.98)* (0.03) (1.OO) (0.19) (0.94) (1.24) Retailsale innon- 0.007 -0.095 0.166 0.017 0.011 0.033 0.023 -0.013 specialized stores with food, beveragesor tobacco redominant (0.44) (2.89)'. (2.65)** (0.51) (0.17) (0.68) (0.41) (1.19) 108 (I) 1 (2) I (3) (4) I I I 0 (6) 1 (7) (8) I CI - 3Ly 23 2e 38 -I P E Retailsale of -0.015 -0.065 0.111 0.018 0.121 0.028 I -0.021 0.025 textiles, clothing, I footwearand leather goods (1.96) (1.99)* (1.74) (0.74) (2.63)* * (1.15) Retailsale of 0.007 -0.049 0.015 0.001 0.025 -0.007 household appliances,articles and equipment (0.29) (1.19) (0.19) (0.04) (0.37) (0.48) Other retail sale in -0.001 -0.060 0.110 0.031 -0.054 -0.024 -0.050 0.080 specializedstores (0.06) (1.69) (1.40) (0.84) (0.61) (2.13)* Repairof personal 0.040 -0.057 0.159 0.012 0.007 0.016 andhousehold (0.34) (0.10) (0.72) Kiosk 0.003 -0.063 0.194 0.041 -0.017 0.012 (1.63) (0.40) (0.91) I 1 Hotels; camping 0.053 0.161 0.133 0.059 0.132 0.002 sites and other provisionof short- accommodation (2.29)' (3.24)" (1.67) (1.60)0'1 (0.17) Restaurants. bars 0.020 0.052 0.219 -0.021 0.026 -0.010 0.015 0.004 and canteens (1.69) (1.44) (3.99)** (0.93) (0.53) (0.3 1) Hairdressingand 0.015 0.088 -0.036 0.002 0.043 -0.014 -0.037 0.012 other beau< treatment (0.07) (0.76) (0.65) Addis Ababa 0.001 0.071 -0.027 0.006 -0.009 -0.048 -0.018 3.15 ** (0.42) (0.30) 2.26 * 2.60 ** (3.24)** Observations 8561 8902 8902 8902 8902 8902 Source: Distributive and Service Trade Survey. Robust z statistics inparentheses. significant at 5 percent; **significant at 1percent. 109 4. TIME TO MOVE ON? INTERNAL MIGRATIONIN ETHIOPIAAND ITS LABOURMARKET IMPLICATIONS34 Highlights InEthiopiamigration hasbeenlow byinternationalstandardsthough there are signs that itmight be increasing. This paper is a first attempt at analysing the impact o f internal migration on urban labour markets, placing it inthe context of the urbanizationprocess inEthiopia, usingquantitative and qualitative evidence on the push and pull factors driving migration and drawing on internationalexperience. There i s a concern that, to the extent that urban labour markets are segmented, a large influx of migrants to urban areas would result inincreased competition inthe informal sector and inhigher unemployment. Available evidence suggests that from a labour market point o f view migration i s not a major concern. At the same time, enhanced policies are needed to ensure that internal migration can contribute to growth and poverty reduction. Migratory flows are highly heterogeneousbut overall the employment outcomes o f migrants are better than for non-migrants. Migrants tend to be more educated than non migrants and those who migrate for work relatedreasonshave a better employment performance than other migrants. Qualitative and quantitative evidence suggests that network effects play an important role inboth motivating migration and in facilitating access to jobs. The positive performance o f migrantsin the labour market is reflected also intheir wages. Migrants and non-migrants appear to face a different wage structure, with important gender differences. On average migrants have higher returns to skills than non-migrants, and the differences decline for higher levels of education. This mightreflect the correlation ofmigration with unobservable individual characteristics which also influence earnings. While migrants do not seem to "compete" directly with non-migrants and to drive down their wages, except for those with very low skills, they do seem to be in competition with each other. These findings suggest that there could be risks that increasing migratory pressures could exacerbatepoverty inurban areas. The PASDEP agenda, and inparticular the focus on stronger rural-urban linkages and urban development suggests that these challenges need to be addressed as part of an encompassing strategy. 34T h i s study draws on 2 background papers - Tadele, Pankhurst and Bevan (2005) and Blunch and Ruggeri Laderchi (2006). Copies o f these papers are available on request from cruggeriladerchi@worldba&.org, or through the "Job Creation, Core Labour Standards & Poverty Reduction In Africa" Trust Tund website searching at www.worldbank.org. 110 Introduction 4.1 Internal migration is central to the processes o f urbanization and structural transformation which cross-country evidence suggests are associated with higher development (Figures 23 and 24). It has beenobserved that, "as economies develop, the share o f the agricultural labour in total labour declines and converges to a level o f 2-3 percent" (Larson and Mundlak, 1997). While there is significant dispersion inthe degree to which countries have urbanized during their development process, a strong negative association can be found in cross-country data between income levels and share o f population inrural areas. Figure 23: The Rural Share of Population Figure 24: Agriculture's Share of Labour Declines at Higher Income Levels Declines as Countries Develop loo 1 0 5000 lw00 15wo 20000 250W 30000 55000 4w0045wo 54 prcapitl 1"Mrn. (2WOUSS) Source; World Development Indicators2005. 4.2 Migration can therefore play a major role in fostering growth and poverty reduction, by reallocating resources more efficiently both geographically and sectorally across the economy. China, offers a spectacular example o f the transforming role o f migration: an estimated o f 16 percent o f GDP growth over the period 1987-2005 has been attributed to migration (Box 12). Part o f the growth impact i s generated by internal remittances which, given the right conditions (an environment supportive to local business development, infrastructure to support communication and growth), can increase demand for agricultural produce and stimulate non-agricultural activities. 4.3 At the same time migration can contribute to new patterns o f exclusion. InIndia, decompositions o f the changes in poverty over 1983-1993 reveal that "rural to urban migration contributed to poverty reduction inrural areas by 2.6 percent," while resulting in an increase in urbanpoverty, though by a smaller amount. Similar patterns but less sizeable effects were found for the period 1994-1999 (Bhanumurthyand Mitra2003). 111 Box 12: China: Reapingthe Potentialof InternalMigrationfor SustainedGrowthand PovertyReduction InChina, the hukuo (registration) systemwas introducedin1958to prevent rural-urbanmigration by forbidding work and access to basic services outside the area of official residence. Reforms introduced in 1978 to provide greater autonomy on production decision in rural areas and development policies oriented to attracting FDI to the eastern parts of the country provided a more favorable environment for migration. As a result of those policies, the period between 1978 and 2004 has seen an increase in income per capita from US$153 (constant 2000) to US$1,162, while the share o f the rural population declined from 81 percent to 60 percent. In 2001 the Government startedreforming the hukuo introducingnew freedoms o fmovement from ruralareas modulated on the basis of the destination, i.e. decreasing from movements to small towns to megacities. Today unregistered migrants continue to play a large part in China's economy with estimates of those flows rangingfrom 50 to 120 millioninthe absence of official data. Internal migration flow are still male dominated owing to both cultural factors and demand in urban areas. In2000 37 percent of migrants were employed in manufacturing, with construction (14 percent), services (12 percent) and restaurant and commerce (12 percent) being the other major employers. Small and middle sized cities are significant migrant destination, with high percentagesof migrantsremaining within their province ofbirth. As inother East and South-East Asian countries, increases in autonomous female migration are registered because of its growing social acceptance and of specific demand in some sectors (domestic work for the uneducated and factory work for women with some education). In 2004 internal remittances have been estimated to provide about the same amount than agricultural earnings to rural incomes. Circular migration during the year, and high return migration are important factors in driving highremittances. Return migrants are more likely to invest inmachinery for agricultural production. Many migrants seem to contribute to the growth o f rural non-agricultural activities by returning to townships or small cities in their area rather than to farming. They are also likely to take leadership positions in their home communities, suggestingthat the skills andhowledgethey have acquired canhave externalities. Labour migration since the late 1990s has been recognized as one o f the national poverty reductionpolicies. Overall, considering all effects bothnegative (e.g. loss of agricultural labour), and positive ones (remittances) household income per capita inrural areas has been estimated to have increased between 14 and 30 percent due to the migration. There have been several policy initiatives to support the poverty reduction impact of migration. In2004, the Sunshine program which involves a partnership between the central and provincial governments inpoor areas and provinces with highrates of outmigration was launched. It aims to provide vocational training to 10 million rural labourers who plan to move out of agriculture or want to move to the cities. After ayear ofoperation 80percent ofthe 1.5 milliontrainees arereported to have foundjobs. New challenges are emerging. While "surplus labour" inrural areas still amounts to 150 million people, some manufacturing areas have started experiencing shortages o f manpower, and the increasing sophistication of products means emerging skills mismatches. Concerns on new vulnerable groups are also emerging: the poor and unskilled who continue facing barriers to migration and women and the elderly left behindin areas of intense outmigration in rural areas, and the landless and jobless migrants with no access to social services in urban areas. Further policy reform i s needed in continuing the gradual reform o f the hukuo system, making more flexible arrangements for land tenure and continuing efforts to increase education, for China to continue reaping the benefits o f rural-urban migration in the face of increasing inequality and emerging new forms ofpoverty., iources: Huang and Zhan (2005), Deshingkar (2005). 112 4.4 Ingeneral, the evidence points to great differences across countries inthe impact o f migration on poverty, depending on who migrates and the strength o f linkages with those left behind. Choosing to migrate i s not a random choice, and the empirical evidence suggests that those who can expect to benefit the most, such as younger and better educated individuals, are more likely to migrate. The extent to which migrants' gain from migration are spread widely i s affectedby whether they maintain linkages with source areas or not, whether their migration is temporary or permanent, and whether it i s part o f a seasonal pattern or a one-off, 4.5 Country specific factors affect the nature o f migratory flows, the success o f the migrants, and the linkages that can be kept with home communities thereby affecting the overall impact of migration on poverty reduction and growth. Relevant factors include institutions, geography, transport costs and information, as well as opportunities to integrate in receiving labour markets. For example in Mexico the poor migrate (internationally in this case) and this reduces national poverty (Stark and Taylor 1991). The geographic contiguity with a labour market characterized by large demand for low cost low skilled work (the US) i s clearly part o f this poverty reducing outcome. Inone study inWestern India, seasonal migration was found to be driven as well as amplifying rural inequality as it was directed towards urbanrural labour markets strongly segmented by ethnic lines, so that those without accessto recruiters' networks or kincan only access the worst paid forms o fjobs. Furthermore, recruitment from villages through networks decreased competition among employers (Mosse et al. 2002). Migration can also be associated with increasing inequality in urban areas, particularly as migratory processes tend to be cumulative, with successive waves o f migrants o f a certain type (location, ethnicity) attracting more o f the same - thereby generating competition and low earnings insectorswhere migrantsaretypically working (e.g. Borjas 2003). 4.6 The functioning o f the labour market plays a key role inmediating the impact o f internal migration on both growth and poverty reduction. Its structure and institutions affect: the incentives for migration. Wage differentials play a role in motivating the migration decision and the choice o f location. The degree o f labour market integration, the extent to which information on labour markets is shared and the efficiency and reliability o f information flows are crucial to ensure the economic success o fmigrants andtheir impact on the efficiency o f the economy. the labour market outcomes of migrants. While migration might or might not be motivated by labour market considerations, the extent to which migrants can find productive economic opportunities intheir new areas o fresidence and freely enter sectors inwhich there is demand for their labour depends from the functioning o f the labour market. Labour market flexibility, and ability to absorb new entrants, are the key characteristics that affect the labour market outcomes o f migrants. the impact of migrants on the labour market outcomes of the established population. While closely related to the point above, the way labour market functions, the degree o f segmentation and the returns to specific "local human capital" also affect how non-migrants are affected bymigration. 113 4.7 In Ethiopia overall migratory flows have been limited though there are signs the population (aged lo+) was living in urban areas, and o f those 18 percent had moved growing migratory pressures. According to the most recent national data, 14 percent o f there over the last 4 years up to 2005 (CSA 2006).35 While the lack o f recent census data does not allow an assessment o f current flows o f internal migration there is anecdotal evidence supporting the view that migration i s increasing. Further increases can be expected due to population pressures on land and environmental degradation in rural areas and the momentum o f the development process itself. Longer term determinants such as "regional imbalances in employment opportunity, improved communications, road and transport networks and changing aspirations o f the younger generation" (Deshingkar 2005, p. 25) are also likely to play a role. 4.8 The Government development agenda seems attuned to increases in internal migration and the PASDEP shows a greater awareness o fthe multiple roles played by the labour market, and its relation with the drivers o f migration and o f economic growth in urban areas. It includes reaping the benefits o f agglomeration in urban areas to create growth poles, increasing education and improving infrastructure. As the China experience shows, these flows are hard to contain, and once their potential to reinforce the pace o f economic growth and poverty reduction is realizedthe policy agenda needs to shift to strengthen the positive effects o f internal migration. 4.9 This chapter aims to assess the labour market implications o f current and future migration to urbanareas by trying to quantify migration, and explain its determinants and labour market impact. However, the labour market alone cannot explain fully either migration behavior or migrants' labour market success. For this reason this paper builds extensively on the insights provided by qualitative analysis undertaken by the WED project in four sites andby the recently completed ETPPA (MOFED 2005) to inform and integrate the quantitative analysis. 4.10 The analysis is organized as follows: section 2 briefly reviews how Ethiopia's urbanization and migration history shapes the institutional context for migration today, section 3 takes stock o f available evidence on the size o f migratory flows inEthiopia and places them in a comparative context; section 4 analyses the determinants o f migration and section 5, looks at different aspects o f the labour market performance o f migrants. Section 6 concludes by building on these findings to identify the policy agenda to maximize the impact o f migration on poverty reduction and economic growth. Data availability i s an obstacle to the detailed exploration o f these issues, but the use o f different qualitative and quantitative sources helpsproviding a roughpicture o f an area so far not sufficiently explored (see Box 13 on data sources). 35Note that while the overall share o f urban respondents appears relatively stable between the 1999 and the 2005 LFS (inthe former it was 13.5, 14.2 inthe latter), the share recent migrants inurban areas appears to have gone up: from 15% in 1999 to 18% in2005. 114 Box 13: Data Sources Usedfor this Study Labour Force Surveys (LFS). This study mostly relies on the 1999 LFS (see also Box 1, chapter 1 this volume) runby the Central Statistical Authority (CSA) though comparisons with the LFS 2005 are also provided to quantifL trends. Child Labour Force Survey (CLFS). A special LFS aimed at better understanding the determinants and patterns of child labour, run by CSA in 2001. The survey also contains a module covering the whole household therefore offering an updated picture of the LFS. The migration variable recordedinthis survey is rather imprecise as it refers to the whole household, raising doubts on how it has been understood by respondents. Mostly because of these differences in survey instruments, recorded levels of migration are about half of those recorded ~ by the 1999 LFS. Despite these measurement issues, the survey provided the only nationally representative information on wages by migrants, and was therefore used for that part of the analysis. Demographic Health Survey (DHS). The DHS surveys are a well established internationally comparable set of surveys which are run at regular intervals. At the time of writing, the latest ~ DHSsurveyavailablefor Ethiopiawas for 2000. Ethiopia Participatory Poverty Assessment. In 2005, the Ministry of Finance released the report of a large scale PPA exercise aimedto informthe ongoingprocessofrevision ofthe PRSP. The study surveyed 45 sites throughout the country and focused among other major themes on migration andits role inhouseholdlivelihoods inbothruralandurbanareas. Wellbeing in Developing Country study. This ongoing project is part o f a research program run with the University of Bath (UK) in several countries. The Ethiopia project, run in collabouration with the University of Addis Ababa, covers 12 sites - preliminary results from four of those sites (two rural and two urban) have been analyzed with the purpose of informing the presentwork. 2. The institutionalcontext for mipration today 4.11 Internal migration has always fueled urbanization in Ethiopia, with shortages o f landand low agricultural productivity inthe north as important "push factors". Already towards the turn o f the 19thcentury, sustained migration from the north accompanied the creation o f garrison towns in the south, southwest and east o f the country. Soldiers and peasants migrated from the north attracted by fertile land, and the possibility o f tribute and access to free labour at the south, southwest and east borders o fthe country. Among the "push" factors that contributed to this migration were the shortage of land, low agricultural productivity, high population densities and recurrent drought. The feudal institutions o f the north and insecurity in access to landmight have further compounded these factors. 4.12 The impetus was reinforced with the establishment o f Addis Ababa as the permanent capital with an influx o f labourers for construction o f roads and buildings as well as demands for goods and services. The improvement o fphysical infrastructure and development o f communications, notably the construction o f the Djibouti-Addis Ababa railway contributed to the emergence o f several towns along the way. The mushrooming o f towns, along with the establishment o f financial and public services such as banks, hospitals and schools also increased urban-rural linkages and the movement o f people from one area to another. The period up to the mid-1940s saw further consolidation o f urban infrastructure, and the emergence o f markets characterized by division o f labour, specialization and a cash economy. As the economic functions o f towns increased, even 115 ifmost remainedgarrisontowns, the urbaneconomic base increased and encouraged the migration o f non-agricultural `specialists' from rural areas such as artisans, traders, bar andrestaurant owners, shop-keepers andconstructionworkers. 4.13 In the post-war period urban growth, particularly the emergence o f small commercial towns, was reinforced by the developmental model adopted. Several factors contributed including the consolidation and centralization o f government structures, the renewed emphasis on road building, the setting up o f industrial enterprises and commercial centres, the designation o f industrial zones along the railroad lines, the `modernization' o f municipal services and the expansion o f urban handicrafts. Considerable rural-rural migration was observed in the 1950s and 1960s due to the development o f commercial agriculture and the eviction o f pastoralists for the development o f large-scale commercial farming. The combined effect o f infrastructure development, trade and the introduction o f mechanized farming strengthened the development o f towns where modern agricultural inputs and services were available. Urbancenters also served as stepping stones for evicted rural households to migrate to other provinces or work inthe towns as daily labourers. 4.14 The emergence o f commercial agricultural sites reinforced rural-rural migration. In 1976, for example, three-quarters o f the farm labourers engaged in 16 irrigation schemes inthe Awash valley were immigrants, mostly coming from areas o f considerable land pressure. The development o f coffee production in the south-west also attracted seasonal and permanent labour and the development o f new urban areas. There was considerable circular movement o f labour - in the early 1970s seasonal migrants to the coffee regions were estimated at 50,000. The majority o f these labourers reportedly migrated from areas o f environmental degradation and natural disasters. In general, the development o f the towns exhibited the migration o f various categories o f people who came to resettle as trade migrants, civil servants, soldiers, construction workers, domestic workers etc. Parallel to this voluntary migration, the Imperial Government also attempted to implement planned relocations with the aim o f reducing poverty, increasing access to land and protecting citizens from famine. However, until 1974 the scale was limited bythe private landtenure system andhighoperational costs. 4.15 Highurbanizationrates (albeit from a low base) andrural-urban migration came to a halt with the drastic political and economic reform introduced by the Derg, and most notably the confiscation o f rural and urban private lands o f 1974, the closure o f private mechanized agriculture, the de-prioritization o f urban development, and forced villagization and resettlement. Official registration requirements were introduced for both urban and rural inhabitants, with eligibility for obtaining rural landbeing limitedto permanent residents o f Peasant Associations, and the introduction o f a pass system and check-points along the main highways. 4.16 Some measures were explicitly aimed at restricting internal migration and preventing the consolidation o f land inlarge estates. For example, land would be subject to confiscation and redistribution if someone was absent for more than a year. The effects o f these regulations in rural areas were compounded by the requirements o f official pass letters to go to the cities, as well as o f official registration with the kebele for 116 urbanresidence. The climate o fpervasive violence and insecurity further diminished the attractiveness of towns. As a result, between 1975 and 1984 urban population growth was mostly due to natural population growth. Intensified conflict led however to the resumption of large in-flows into cities between 1984 and 1994. 4.17 Today's institutional framework characterized by controls o n voluntary migration and directing flows through resettlement (see Box 14). Despite the lack o f explicit migration policies, the SDPRP (Ethiopia's first PRSP, issued in 2002) reflected a negative attitude towards internal migration, by dwelling on its effects on urban poverty, HIV-AIDS transmissionand expansion o fcrime. Box 14: Understahdingthe Dynamics of ResettlementinEthiopia From 2003 to 2005 over half a million people were resettled in four national states: Oromia, Amhara, SNNP, and Tigray. Further resettlements are planned in two o f these regions. These resettlements, part of the plans of the New Coalition for Food Security established in2003 by the Government and its development partners, aim at avoiding the excesses and failures which had characterized earlier resettlement schemes. Detailed guidelines, pillars, principle and approaches have been laid out. A recent study by the Forumfor Social Studies offers insightson the barriers faced by settlers who proved less successfulthan others. Examples o f the problems that were encountered included: information on resettlement areas was not always accurate particularly with regard to the land size, number of oxen availability/quality of housing and availability/quality o f other supplies such as tools; some resettlements have been directed to areas close to rapidly dwindling forest reserves and without large availability o f underutilized land(rights of pastoralists inparticular appear not to have been taken into account); access to some sites was impossible duringthe rainy season making it impossible to receive food aid; access to water was not always safe, and maintenance systems for pumps proved non sustainable; in some sites consultation with local communities was not sufficient and some o f local people were displaced or deprived of access to naturalresources (e.g. forestry goods such as honey) important for their livelihoods; access to health and education services was problematic, for example with overcrowding o f existing facilities, sometimes with resettled population receiving services while the locals could not. Efforts had been made to avoid some of the pitfalls o f previous programs. Those include for example ensuring ethnic homogeneity in resettled areas to avoid conflict. Conflicts emerged, however, not o f ethnic nature but largely over use o f resources, particularly environmental resources. The need to adopt a more flexible approach, with space for experimentation and taking inlessonslearnt was one of the key messageso fthe study. Source: Adapted from Forumfor Social Studies 2005. 4.18 Specific elements of the regulatory framework continue to constrain migration. Despite progress towards increasing security o f tenure, the perception o f the risk o f land redistribution remains high and stifles diversification o f livelihoods in both rural non- agricultural activities and long term migration. Requirements for local registration and the need for an identity card provide means o f formally controlling internal movements. Perhaps equally important are perceptions on the role o f Peasant Associations and kebele officials in controlling internal movement (see Box 15). Arguably, the effects o f these institutional features have been heightened by the adoption o f a model o f ethnic-based 117 f e d e r a l i ~ m .This form o f federalism can be seen to compound the possible preference ~ ~ for migration to areas whose language and culture migrants already share particularly by affecting perceptions on the management and assignation o f housing and other assets, children's education and insecurity. Some discrimination by local authorities in finding jobs andproviding business licenses on language or ethnic ground has beenreported.37 Box 15: Perceptionon Constraintsto InternalMigrationinthe PPA The general case i s for people to be able to move freely, subject, as might be expected, to social constraints (as pertain to adult women in many places) and to carrying personal ID at all times. There was a sense in some reports that local administration kept a very close eye on people's movements and activities, although the degree to which this might inhibit individual initiative was difficult to assess. Inurban sites it was reported that resident registration with the kebele i s compulsory, and movement out of the kebele to go elsewhere requires an official leaving letter. Letters from the kebele are also required to get access to health services for free (or low fee). How uniformly these requirements are pursued intowns, and the penalties for failure to comply with them are not clear from information reported by respondents. Inrural areas it was reported that as people's movements are monitored quite closely by kebele administrations, it i s advisable in many circumstances to have cleared any planned travel with the relevant officials before setting out, to avoid trouble. Furthermore, it i s widely perceived that migration beyond a certain duration will result in forfeit of the land rights o f the person concerned by the kebele administration. Insome sites it was also reportedthat engagement innon-farm activities such as trading in consumer goods would result in forfeit o f a person's rights over land. Limits to migration, therefore, could be seen as part of a widespread perception in rural Ethiopia that engagement innon-farm activities as part o f a mixed livelihood strategy i s not permitted, so that people must choose in an absolute sense whether to stay wholly in agriculture or to engage wholly innon-farm business. burce: Adapted from MOFED(2005), ParticipatoryPoverty Assessment. 4.19 The desire to control internal migration - by hindering voluntary permanent movement and putting in place planned relocation schemes - has not been unique to Ethiopia. A recent survey o f PRSPs revealed that mobility is especially ill-represented (Black 2004), with 21 out o f 48 PRSPs examinednot mentioning migration at all, and the remaining ones referring to it in negative terms. Eight PRSPs expressed the need for internal migration to be actively controlled by the state; that i s for rural-urban migration to be curtailed or reduced. 4.20 Against this background, PASDEP provide a more liberal context for internal migration though a harmonization o f the overall policy environment i s required. For example, the draft PASDEP details as strategic priorities both the strengthening o f rural- urban linkages and a rebalancingo f the growth strategy with an emphasis on small towns development and growth poles and employment creation, particularly in urban areas. Even if the role o f internal migration in this context is not directly addressed, such strategic priorities seem to require a more favorable policy stance towards migration. 36The Migration, Gender and Health Survey conductedjointly by the Addis Ababa University and Brown University (2000) with a focus on the five most populatedregional states (Affar, Oromia, Amhara, SNNPR and Tigray), indicated that permanent intra-regional migration has become more frequent than permanent inter-regional migration with the exception o f temporary migration to Addis Ababa. 37Tadele et al, 2005. 118 3. The size and direction of migratory flows inEthiopia 4.2 1 Significant movements o f people have accompanied Ethiopia's development over the 19th and 20th century, though voluntary movements significantly slowed in the 1970s. As o f today, according to the participatory poverty assessment, in rural areas perceptions that "migration represents a failure o f some sort in relation to being able to exist adequately in one's own community" (p. 35) persist, and movement especially by some groups (women) are stigmatized. 4.22 According to the most recent national data an estimated 18 percent o f the Ethiopian population lived in a different area than where they were born, and 6 percent had moved over the last 4 years up to 2005 (LFS 2005). About 40 percent o f recent migrants were living inurban areas o f whom about 42 percent had moved from another urban areas. Among recent migrants living inrural areas intra-area migration was almost double at 79 percent. As the population is overwhelmingly found inrural areas, the most common type o f move was within rural areas (46 percent), while recent migrants represent a larger share o f the urbanthan o f the rural population (18 percent rather than 4 percent). Between 1999 and 2005 the pace o f migration grew significantly, albeit from a very low base (the share o f migrants who had spent one year or less in their current community almost doubled, going from 1.8 to 3.5). 4.23 These rates o f internal migration are quite limited in comparison with international evidence. Using a sample o f DHS surveys provides an opportunity to compare Ethiopian migration rates with other countries, by focusing on female migrants 15-49 inurban areas. This anal sis reveals that the prevalence o f migrants inurbanareas i s lower than the SSA average though the age profile o f migration i s similar to those ye observed inboth Sub-Saharan Africa andNorth Africa (Figures 25 and26). Figure25: Share of UrbanWomen of Figure 26: Share of Women who are Recent - Reproductive Age who are Recent Migrants, Migrants, by Age group and Region by Region 30 0% 25.0% 20.0% 15 0% 100% 5.0% 0.0% Source; National Research Council 2003 and own calculations for Ethiopia(from DHS 2000). 38 Sources are own calculations for Ethiopia and Mongtomery et al. (2003) for regional averages. Sub- Saharanaverages are based on 25 surveys and exclude Ethiopia. 119 4.24 The share o f migrants o f rural origin is significantly higher in Ethiopia, in line with lower urbanization rates. Among migrants o f reproductive age in urban areas, almost 70 percent are from rural origin, compared to less than 39 percent for the rest o f Sub-Saharan Africa. In contrast, only 11 percent o f female migrants are from larger cities against 33 percent inSub-Saharan Africa. 4.25 Significant variations inmigration rates are found across regions, with the lowest incidence o f migration in the Somali region, the highest in Gambella (Figure 27). With the exception o f Harari, Addis and Dire Dawa a large part o f the flows i s intra-regional (Figure 28), and qualitative evidence suggests quite short distance. Regional federalism might have contributed to these type o f flows by decreasing incentives for interregional migration. Reportedly, however, when the system was instituted it resulted in large migratory flows as some o f the resettled populations went back to their region o f origin (Forum for Social Studies 2005). Figure 27 :MigratoryFlows at the Figure 28 :Migration Destination by Regional Level Regionof Origin for 6 Regions Im y 0.14 ,lei I Addis Ababa ~Amhara 0 12 0 Osmyia 0 1 Ommyia ISomali O D 9 rn Benishaq ISNNP 003 Amham 0 Gambella OM 0 02 IAddisAt& 0 DireDava 0 0% 2w6 4wh 60% 80% 103% Source: LFS 1999, own calculations. Box 16: Methodological Difficultiesin Quantifying Migration Estimates o f the size o f migratory flows are difficult to obtain as estimates depend on the geographical unit and the time horizon considered. In the case o f the LFS, migrants are those who are currently residing in an area (in this case the town or rural part o f the wereda) which i s not the one where they were born. Imprecision is therefore inescapable: for example short distance moves who happen to entail crossing an administrative boundary would be recorded as migrations while comparable moves within the same areas would not. Further, it i s not clear how much these data are picking up seasonal mobility - as seasonalmigrants will appear as migrants to the extent that they cohabit with households established enough as to be part of the survey's sampling frame. Finally, other aspects of mobility such as intra-urban migration which i s an important element o f existing models o f urban development (withdifferenttypes o f urban settlers moving over time to areas where housing appropriate to their needs and ability to pay can be found) are unlikelyto be captured. Qualitative evidence underlinethe definitional difficulties since bothrural andurban livelihoods are characterized bymany types o fmobility outside anarrow definition o fmigration. The WED studyprovides examples o fthe variety o ftemporary movements inand out urbanandrural communities, fromtrips to other cities looking for work or cheaper livingconditions, to travel for 120 social activities -for exampleby urbanmigrants going back to their communitiesof origin, but alsobynon-migrant women going on visits -or as part o fthe business of small traders. These are clearly non-mutually exclusive categoriesas travelback to the homecommunity is an opportunity for exchangingandtrading different types of commodities. At the same time, as notedinthe PPA, "there isnot awell-developednon-farm economyandlabour market inrural Ethiopia" with the exceptionof areas near largetowns so that "there i s not a lot ofregular labour mobility associatedwith non-farmeconomic activities" (p. 35). Thus inone ofthe rural research sites close to urbanareasmore than one infive householdsinterviewedhadmembersaway (34 percent ofwhich to urbanareas inthe same wereda, and 18percentto Addis), while inamore remoterural researchsite, the percentagewas only 4 percent. 4.26 Smaller cities play an important role inmigration, a pattern common to Northern and Sub-Saharan Africa (see Figures 29 and 30), consistent with qualitative evidence that much o f migration, particularly within rural areas, is short distance. Difficulties o f disaggregating the urban category in the LFS and other major nationally representative household surveys prevent furthering exploring these issues. The PPA highlighted how migration to towns to the extent that it occurs, mainly comprises migration by young adult males in search o f employment opportunities, given their shrinking ability to participate in agriculture in their home communities. Inthis context, small cities can act as stepping stones for migrants allowing them to build up specific skills or gather information on their next move. Figure29: A Percentagesof Urban Figure30: PercentagesofUrbanWomen Women of ReproductiveAge who are of ReproductiveAge who are Recent RecentMigrants,by City PopulationSize MigrantsinEthiopia,by City Population andLocation, Restof the World Size andLocation 0.3 0 3 0.25 0 25 0.2 ff 2 0 2 0.15 0.15 0.1 O f 0.05 005 0 MrthAfrica SubSaharan Southeasth a South.Central. LainAmenca Afka Westhrn Source: National ResearchCouncil 2003 andown calculationsfor Ethiopia (from DHS 2000). 4. Who migratesandwhv? 4.27 The characteristics o f migrants affect their labour market performance, their sector o f occupation and their impact on the receiving areas, as illustratedbythe example o f China described above. Given that not all individuals are equally likely to migrate, understanding the determinants o f migration to urban areas helps explaining the characteristics o fmigratory flows and their impact on urbanlabour markets. 4.28 Migrants tend to be young and concentrated in the most productive age groups. About halfo f the migrants are aged less than 19, though the same age group represents a 121 much larger share o f the sedentary population (66 percent).39 Compared to those who have never moved, recent migrants tend to be concentrated in the most productive age groups such as the 15-19 years old (which represent 21 percent o f migrants, while the same age group represents 11percent o f those how never migrated), the 20-24 years old (16 percent versus 7 percent), andthe 25-29 years old (13 percent versus 6 percent). 4.29 Gender differences emerge in terms o f the reasons to migrate. Women migrants are more likely to be found inthe age group 15-25, while the opposite applies to migrants 25+ (Figure 31). These findings seem to be closely related to the main reasons for moving. Typically "female" reasons to migrate can be identified in Figure 32 as those related to marriage arrangement and dissolution (with women moving in order to get married, particularly in rural areas) and, though to a lesser degree, moving with family. Work related reasons such as searching for a job and having a job transfer are more prevalent for men, as are education and returning home. Shortages o f land are also more prevalent for men, and likely to be driven by customary gendered patterns o f land ownership. Reasons related to distress migration, such as displacement, war, drought and livingwith relatives (which qualitative evidence shows is a common coping strategy) are equally likely for men as for women. 39Migrationof children as young as 7 to work inweaving is reportedinthe WED study. 122 Figure 31: Gender Ratio (medwomen) by Figure 32: Reasonsfor Migration for Age and Migration Status Recent Migrants, by Gender 1, ... . . . . . .. 0.8 0.6 0.4 0.2 , n Source; LFS 1999, own calculations 4.30 Different reasons to migrate also characterize the choice o f migrating to either urbanor rural areas (Figures 33 and 34). Job related transfers are mostly towards urban areas (particularly so called "contracted moves" - i.e. moves part o f an agreement with an employer - though also "speculative moves" - i.e. searching for jobs - are slightly more directed towards urban areas). Not surprisingly also education related migration i s overwhelmingly directed towards urbanareas. Figure 33: Reasons to migrate, by areas Figure 34: Main reasonsfor move (single movers only 13+) BO 80 50 70 60 IBetterconditions 40 50 30 30 40 I 20 10 20 0 10 0 Men I Women Permanent migrants Tempwarymigrants Source; LFS 1999, own calculations and Migration Gender and Health Survey (1998). 4.31 Data from the Migration Gender and Health Survey show that men and women migrate temporarily mostly for job assignments or to look for better living conditions. Job assignments remain the main reason to migrate permanently for men, while for women permanent migration i s overwhelmingly linked to family and marriage (Figure 35). 4.32 While migrants are mostly unskilled, on average they are more educated than non migrants (Figure 36). In 1999, 58 percent of recent migrants had no education, significantly better than the 79 percent for those who have never migrated. Recent migrants are noticeably more likely to have achieved grades 5-8 and grades 9-12 than 123 those who had never migrated, partly o f course because requiring an education beyond basic primary might have required them to move. Educated migrants are more likely to bemen (Figure 37). Figure35: Educationaldistributionof Figure36: recentmigrants gender ratio migrantsandnon-migrants (medwomen) by educationalachie ment w I lW% 80 I 90% 70 80% BO mu w 63% 40 50% 40% 30 I I 32% 20 20% 10 10% 0 0% - No educallcn Graje 14 Grade 58 Grade 912Abmgrde 12 - Source: LFS 1999, own calculations 4.33 Migration can be seen as a cumulative process. The evidence available suggests that it is common for one household member to leave first and for others to join subsequently. For example, inthe DHS subsample for which more than one migrant is in the household, only 15 percent moved at the same time. Similarly, and o f great importance, established community o f migrants facilitate the sharing o f information and the settling o fnew migrants. 4.34 The importance o f the household and community history o f migration manifests itself indense social networks o f migrants. Migrants often have limited contacts with the population o f their destination area, and even after several years they might not have established strong social relationships. In contrast they rely heavily on contacts and relatives in their area o f destination, who often are the ones who helped them migrate in the first place. Migrants who follow "circular" patterns, by which they go with some regularity to their home communities often convince young men, unmarried women and children, to follow them to urban areas where they can help them find a job. Ethnic based iddirs burial associations and self-help groups are the key institutions migrants set up to provide mutual assistance. These organizations differ inthe way they operate and provide assistance for various purposes, including unexpected events (funerals, need to provide help to relatives in rural areas) or planned ones (organizing joint trips home for the Meskel holiday). Interestingly, migrants who keep contacts with their rural areas o f origin often continue contributing to iddirs also intheir home area, or contribute to efforts such as fundraising for local roads. 4.35 To assess the relative importance o f these factors, we ranreduced form models o f the determinants o f migration (Annex 1). While they cannot be interpreted as causal 124 relations, they provide a profile o f the characteristics most closely associated with migrantstatus.40 4.36 Probit models (Table 53 inthe Annex) show that education is a key characteristic associated with migration, with strong effects for the low and the high educated. The effects o f education differ for different segments o f the educational distribution. The likelihood o fbeing a recent migrant i s "bipolar": taking as base case having no education, individuals who have grades 1-12 though with differences appear less likely to be migrants. Those with more than 12 years o f education, incontrast appear just as likely to be migrants than those without any education. Non-recent migrants show a different pattern, with those with grades 1-4 showing a higher likelihood o f migration than the non-educated, while those with subsequent levels o f schooling display lower probabilities. Among the other individual characteristics, gender does not play a significant role.41 Age decreases the probability o f recent migration at an increasing pace. Life events such as marriage or marriage dissolution increase the chances o f migration, though effect is more muted for women. It is interesting to underscore the differences between these findings and those o f the descriptive analysis - for example, while marriage dissolution accounts for a highproportion o f female migration and a very low one o f men's, controlling for other factors such as age and education the increase in the probability o f migration following marriage dissolution is less for women than inthe case o fmen insimilar circumstances. 4.37 The cumulative nature o f migration i s evident with strong effects o f proxies for social networks. The share o f migrants livin in the community i s very significant in explaining the probability o f being a migrant. As for the composition o f the migrants $2 group it seems that migrants move to areas with a similar ethnic make up among migrants, though the opposite effect holds for those o f the same religious group. Similar findings apply to the effects o f the ethnic and religious composition o f the host community, with the likelihood o f being a migrant increasing with the share o f own- ethnic group inthe area, and opposite effects for the share o f own-religious group. While the effect o f the own ethnic group (either in the community or across migrants) is in line with expectations, those for religious groups are morepuzzling. The PPA offers a key o f interpretation as it finds that in rural areas communities with a more mixed religious make up are characterized by less strict application o f various religious bans and practices are less enforced. One can speculate that a similar effect applies to the migrant community, so that where it is very homogenous from the religious point o f view, migrants might find a higher degree o f social sanctions. 40 The household strategy element o f the decision to move is clearly missing from this analysis given the lack o f information on householdcharacteristics at the time o f migration. 41This finding is at odds with those o fprevious studies of migration inEthiopia which find that women are more likely to migrate than men. Our inability to distinguish permanent migrants from seasonal or circular migrants might help explain this difference (with women more likely to be permanent than seasonaUcircularmigrants). 42 Inthe census enumeration areas taken as definition o f local areas the mean share o f migrants in the community is 0.32 for Ethiopia as a whole and 0.26 and 0.67 for rural andurban areas, respectively. 125 4.38 Qualitative evidence helps to flesh out which are the mechanisms which lead to these characteristics of the migrants (see Box 17). An important finding o f the literature i s the role o f cultural and institutional factors, such as informal rights to land, agricultural practices, marriage practices, in affecting the opportunities o f different types o f individuals in migrating. Gender and ethnicity are important dimensions along which migration is segmented. Further, at variance with economic models who focus o n expected wage differentials as main drivers o f migration, migration is rarely based on a single motivation. Of the 25 or so reasons for migration put forward by rural ETPPA respondents "some (...) are individual (for example, being stood-down from the army), but most (. .) involve family interactions or events that place the individual in a weak . position for making a living inrural areas (divorce, loss o f land access, death o f a spouse andso on)" (MOFED2005, p.60-61). Box 17: Pushand PullFactors inRural and UrbanAreas QualitativeEvidence - For analytical purposes, reasons for migration are generally classifiedas pushfactors -i.e. factors insending areas that provide incentives for mobility-andpushfactors -Le. factors inreceiving areas that provide incentives for mobility. Qualitative evidence from the WED study and the Participatory Poverty Assessment provides insights on the motivations for mobility, and highlights the complex motivations driving migration. Such motivations go well beyond the simple differentials in expected wages at the centre of economic modeling of migration. A major difficulty ininvestigating these factors based on survey data i s that ideally one should explore them at the time o f migration, while in standard surveys migrants are observed only after the migration decision. A focus onrecent migrants goes some way towards addressing this problem. Unless otherwise specified what follows refers to recent migrants only. The rural push factors associated with male migration to urban areas are shortage of land, landlessness, destitution, the need to raise cash to pay the land tax and debts incurred inpaying for agricultural inputs.43 Urban areas are associated with the pull o f perceptions of better employment opportunities, particularly inthe informal sector and inconstruction, information on jobs from informal sources suggestingthat there i s demand for urban workers andbetter pay than inrural areas, andsocial supportbylongtermmigrantsareidentifiedas importantpullfactors.# Women are overwhelmingly reported to migrate to urban areas to join their spouses, though a distinct stream o f women looking for work can also be identified. The former groups participates inthe urban economy by working outside the house and undertakingpetty trading activities to a varying degree, depending on cultural factors linked to their ethnicity and origin. The latter group is constituted by young single women or divorced or widowed women moving to urban areas to look for work. Their migration i s related to institutional and cultural factors, such as the plough-based agriculture of the North which puts women at a disadvantage by not beingable to perform key agricultural tasks even when they have access to land. Further, the need to move away from the social norms typical ofrural areas or fromnegative perceptions o fwomen working 43 Inone o frural study site o f the WED study, decreases inout-migration were linkedbythe respondent to credit provision to the farmers, as well as lack o f success o f more recent migrants and ethnic based discrimination. 44 Inone urban study site o f the WED study, changes inthe migratory patterns o f migrants from rural areas were linked to the movingo f a second-hand clothes market from the market and the banning o f vending o n streets. 126 outside the house are closely associatedwith the choice o fmigrating to urban areas. Inrural areas, incontrast, significant mobility is found, though only inminimalpart it is long term and related to labour market behaviors. Much mobility to rural areas i s seasonal from both urban and rural areas. For example, daily labourers from urban areas who usually work inperi- urban areas in agricultural tasks travel seasonally to rural areas to take advantage of the various cropping seasons and to perform other agricultural tasks. Permanent migration in rural areas, both voluntary and involuntary, i s driven by land availability and the existence o f previous migrants. Women migrate a lot upon marriage but such migration tends to be short distance. Migration by women to work in towns i s seen as shameful, and risky, though perceptions o f female migrants (or commuters to nearby areas) by rural inhabitants can improve iftheir earnings allow them to support their families. An important element o f migration decisions to both urban and rural destinations i s the cost of migration, though we have only indirect evidence on its effects. The PPA notes that the cost o f travel i s relatively high in comparison to the ability of farm households to generate cash. Similarly, evidence from the WED study shows that location decision for migrants to urban areas are influenced by the closeness to home origin. Rising transport costs are also noted as factors hinderingcontacts with home communities, which are particularly important for those migrants who have left their households inrural areas and move back seasonally to continue to work their land. 5. Migrantsinthe UrbanLabourMarket 4.39 Given that migrants are significantly different from non-migrants in a number o f respects, such as being more likely to be educated, being younger than average etc .we would expect their employment performanceto be different from other urban dwellers. 4.40 Indeed, migrants tend to have higher employment rates than non-migrants. The employment rate o f recent migrant was 70 percent, against 65 percent for non-migrants. Such difference i s due to migrant men high employment rate (69 percent versus 54 percent for non migrants), while women's employment rate at 48 and 46 percent respectively i s very similar across groups. In general, migrants who moved for work related reasons exhibit the highest employment rates. Younger groups o f migrants (those who migrated because o f education, those who were sent to live with relatives and those who migrated along with their families) tend to exhibit the lowest employment rates. Groups who can be seen to have migrated as they have lost access to traditional sources o f livelihoods (men without land, women after marriage dissolution) have employment rates that compare favorably with the average. 4.41 The heterogeneity inmigrants' characteristics and reasons to move i s reflected in the differences intheir labour market performance (Figure 37). Breakingdown women's employment performance, it appears that recent migrants who moved for work related reasons have employment rates not very different from men's. 127 Figure37: EmploymentRatesfor RecentUrbanMigrantsby Gender andReason for Migration 1 male 1 ! 0.8 w female 0.6 0.4 0.2 0 T T Source: LFS 1999, own calculations. 4.42 Reduced form estimates (Table 54 in the Annex) confirm that urban migrants have higher probabilities o f being employed than non-migrants - among recent migrants the effect i s stronger for women, the opposite for non-recent migrants. Running the model separately by separate education groups shows that for recent migrants the effect o f migration on employment compared to non-migrants i s stronger for those with 1-4 years o f schooling and for those with more than 12 years o f schooling. The impact o f other determinants o f employment i s in line with other evidence for Ethiopia as a whole.45 4.43 Strong cumulative effects o f migration also emerge. The share o f migrants inthe local area, which is strongly correlated with the probability o f migration, does not affect the probability to work. In contrast, migrants' employment rate in the local area is strongly correlated with the probability o f individual migrants to be employed.46 One explanation for this finding i s that when the employment rate for migrants in an area is high,itjust means that overall that areadisplays highemployment rates. Ifthis were the case, however, the employment rate for non-migrants should be equally significant in 45The probability of being employed is lower for women and it increases with age at a decreasing rate. The highest probability of employment is for the non-educated and for those with more than 12 years o f schooling. Marriage i s associated with a higher probability o f being employed for men, a lower one for women. The share o f household members aged less than 1 decreases the probability for women to work, while the share o f older children increases it. 46Note, these variables referring to migrants in the local area exclude the individual observation in question. 128 affecting the probability for a migrant to work, while this i s not the case. Ina given area, there is little correlation between the employment rates for migrants and non-migrants. It seems more likely, therefore, that migrants employment rate is capturing something more specific: either the higher informational content that i s shared across social networks (the more migrants are employed, the more they have information about jobs) or the specific structure o f the local economy, and in particular the existence locally o f jobs that migrants can access (e.g. a growing construction sector). 4.44 Important gender differences on the probability o f employment are also found. Women migrants are less likely to work thanmen andthis effect is particularly strong for non-recent migrants. The positive effect o f age on employment is similarly more contained for women. For women, the highest probability o f working i s associated with no education and with more than grade 12 education (and the effect o f grade12t- is particularly high for non recent migrants), while for men it i s associated with less than 5 years o fprimary (also with more than 12 years for nonrecent migrants). Interms o f life- events, for women marriage i s associated with a lower probability o f working (while for men with a higher one). Marriage dissolution is associated with a higher probability o f working, with the exception o f men who have recently migrated for which it i s non- significant. For women who have not migrated recently, the negative effects o f marriage are less pronounced and those o f marriage dissolution are more positive. Of the network variables considered, the effect o f being o f the largest religious group in the community increases the probability o f employment for women - one can speculate that this might make it more acceptable for women to work. 4.45 The three largest sectors o f employment are trade, domestic work and hotels and restaurants, and migrants are slightly less likely than non-migrants to be found in self- employment. Evidence from the PPA hrther illustrates these findings as well as rural urbandifferences. Inurban areas migrantsmostly work in services and commerce -the low skilled moved to these activities mostly from farming. Inrural areas migrants are concentrated in farming, with migrants with primary education and above moving to this sector also from non-farming activities. Inboth urban and rural areas, the share o f self- employment for permanent migrants i s higher than for non-migrants, though it declines with education. Women are overwhelmingly reported to be in domestic work, though there are concerns on many o fthem working inbarshecoming prostitutes. 4.46 Qualitative data show that inurbanareas migrants, who are mostly unskilleddaily labourers, often are specialized by ethnic group (Box 18). In line with international evidence, specialization along ethnic lines i s found. While somejobs such as weaving or craft-work require skills which might be part o f some ethnic group traditional activities (e.g. the Gam0 in Kolfe, Addis Ababa) specialization in other sectors such as shoe- shiningor being foodpeddlers doesnot seemto bebased on specific skills. 129 Box 18: Networks of Migrants InformationFlows and Job Search Migrant networks are found to have a strongrole inmotivating migration decision, but the share o f migrants alone does not seem to be significant in explaining migrants' probability of employment. The role of networks in supporting migrants' livelihoods strategies should not be underestimated, however. Most migrants in the WED study mentioned that they obtained their first job though their informal networks (iddir, families, kin, ethnic members already established) -andthosearenotcapturedbysimplevariablessuchastheshareofmigrantsintheareawhich we can obtain from the LFS. Further, the literature on social capital highlights the importanceof the type of linkages that people have and the informational content that can transit through those linkages. Inthe WED study strong personal links with long-term inhabitants/previously settled migrants provide better employment opportunities. Contacts with migrants who are already working (more likely the higher the employment rate o f migrants in the area) might be able to convey higher value information content for those searching for jobs, both because working migrants might be able to provide direct employment opportunities, andbecause they are likely to Ihave information onjob opportunitiesintheir sector or type ofbusiness. An important consequence of the importance of network effects i s that it suggests that information is not freely available, with consequencesfor boththe efficiency of the matching and the sort of those who don't have access to that information. The ETPPA, for example, raised concernsabout the labourmarket chances ofyoung unskilledmigrants without connectionswhich exposethemto exploitation andexclusion. 4.47 The positive performance that migrants have in the labour market is partly reflected also in their wages (Table 55 in the annex), at least for men.47 Migration i s associated on average with a 14 percentage points premium for male migrants. Disaggregated analysis48 by educational group reveals that such premium applies to recent male migrantswho have higher education (more than 12 years o f schooling). 4.48 Closer analysis reveals that migrants and non-migrants face a different wage structure. Returns differ systematically for migrants and non-migrants. Gender differences are also very noticeable. On average migrants have higher returns to skills than non-migrants, though the differences decline for higher levels of education. This might reflect the correlation o f migration with unobservable individual characteristics which also influence earnings. The pattern differs for women and men. In the case o f female migrants, they enjoy higher returns to education than non-migrants at all education levels, and the gap decreases for higher educational levels. Inthe case o f male migrants, for lower educational levels they experience lower returns than non-migrants (for male migrants returns are not significantly different from those o f the non-educated up to 8 years o f education) and they experience slightly lower returns than non-migrants for levels o f education above grade 12. 41 Note that the analysis o f wages is conducted with the Child Labour Force Survey database, which collects information also on the overall labour market performance o f the whole family. Unfortunately in this survey it i s not possible to identify the individual as migrant, as the question i s only asked at the household level. Compared with data from the LFS o f 1999, the CLFS provides lower estimates o f migration. 48 See Blunch and Ruggeri Laderchi (2006) for more details. 130 4.49 The spatial clustering o f migrants is also related to lower wages, in turn suggesting competition between migrants inthe labour market. While this effect i s non- significant on average for the non-migrant population more disaggregated analysis reveals that the individualswith some education (grades 1-4) are those whose wages are negatively correlated with the concentration o fmigrants. 4.50 Does the competition between migrants extend to non-migrants as well? On average there i s a negative but non significant effect. Disaggregated analysis reveals a negative impact on the wages o f those without education. 6. Policv challenpes 4.51 From a labour market point o f view, migration i s not presently a major policy concern. Most o f migration is relatively short distance, with both step-migration and moves to urban areas directly from adjacent urban areas being quite common. Those who migrate for work related reasons have a better employment performance than the others, possibly because of the importance o f network effects in both motivating migration and in favoring access to jobs. While migrants do not seem to "compete" directly with non-migrants and to drive down their wages, they do seem to be in competition with each other. 4.52 There are concerns, however, around increased migratory pressures due to population growth and environmental factors inrural areas. Further, the PASDEP agenda poses a new emphasis on small town development and growth poles and on employment creation, particularly in urban areas. In this context increased migration i s likely to be W h e r spurred by urban growth, as well as being, to a certain extent, required for the success o f this development agenda. 4.53 A reformed approach to internal migration can contribute to growth and poverty reduction. The current regulatory system has de facto created two tiers o f citizens, and risko fa new form o fpoverty andexclusion inurban areas. Sustained flows o fmigration without reform in the residence requirements will result in the creation o f a new underclass o f citizens, with limited access to urban services and at risk o f losing their livelihoods inrural areas once their landis redistributed. As the experience of East Asian countries such as China and Viet Nam shows, deep pockets o f poverty and vulnerability can emerge in urban areas. Poverty and vulnerability are likely to pose the greatest policy challenges insmaller urban areas, which mightbe the least able to overcome these. 4.54 Specific policies to be considered with respect to migration include: Facilitatingflows of information andpeople. Strengthening rural-urban linkages through improved information flows on jobs and skills required, better transport facilities that can allow easier movement particularly for migrants that maintain links with their areas o f origin can have a strong impact on their development. Developing migrant-friendly financial services i s an important element in favoring the diffusion o f the benefits o f migration - this could be done by NGOs but also bythe private sector as intermediatingremittances canbeprofitable. 131 Deregulation. A short run priority is to address the exclusion o f migrants from goods and services that require registration - this could involve temporary identifications like those now issued in China. Temporary identification could also help reduce the need for a local guarantor required to access certain jobs (typically jobs that involve dealing with property), which hinders the employment prospects o f those migrants who do not possess strong linkages with the non- resident population. Women and children are particularly vulnerable to the denial o f service provision andprotection, particularly ifthey do not have strong support networks inthe host area. Civil society. The involvement o f NGOs in dealing with migrants has been successful elsewhere (e.g. state o f Rajastan in India) for example in setting up resource centers that provide information on job availability, wage rates and rights, or in providing skills and help in accessing better jobs. In another example, again from India, an NGO has worked closely with local governments o f source villages to develop an informal system o f identity cards for migrants, which gives them some protection against official harassment (Deshingkar 2005). Urban safety nets. This i s an area still to be explored, but to the extent that those who migrate under duress have no alternatives they will be present in cities and will require targeted measures, particularly if they have nowhere to return to and have no assets. 4.55 Migration issues could be better integrated into both the rural and urban development policy agendas. Inthe case o f rural development, migration needs to be seen in a context o f overall income diversification, favoring at least shorter range migration and connectivity with towns. Several policy elements currently pursued are in line with these objectives including strengthening the management, capacity and planning o f rural towns, which play a key role inproviding services and access to markets for the rural population; and scaling up the provision o f rural infrastructure such as roads and electricity. 4.56 The urban development agenda, faced with the prospect o f a growing population, needs to consider how to create an enabling environment which can turn the challenges posed by increased mobility into opportunities. This involves developing new alternatives for service delivery, providing urban infrastructure and considering access to land and credit by the newcomers. These issues are being picked up inupcoming World Bankanalysis o fthe Urbanagenda. 132 References Addis Ababa University and Brown University (2000) "Migration, Gender and Health Survey in Five regions of Ethiopia: 1998" Addis Ababa, United Nations Training and Research Project on the Interrelations o f Migration and Economic Change, Women's Status, Reproduction andHealth. BhanumurthyN.R.andArup Mitra (2003) DecliningPoverty inIndia: A Decomposition Analysis, Institute of Economic Growth, University of Delhi, Available at http://ieg.nic.in/workbhanu248.pdf. Black Richard (2004) Migration andPro-Poor Policy inAfrica, Sussex Centre for MigrationResearch, Working Paper C6, available at http://www.sarpn.org.za/docwnents/dOOO1833Nigration-Black-Nov2004.pdf. Blunch, Niels Hugo and Caterina Ruggeri Laderchi (2006) The winner takes all: migration and wages in Ethiopia, mimeo, the World Bank. Borjas, George (2003) "The Labour demand downward sloping: reexamining the impact of migration on the labour market" The Quarterly Journal of Economics, November 2003 pp 1335-1374. Central Statistical Authority (CSA) (1999) Statistical Report on the 1999 National Labour Force Survey, Statistical Bulletinn.225. Central Statistical Authority (CSA) (2006) Statistical Report on the 2005 National Labour Force Survey, Statistical Bulletinn. 365. Deshingkar, Priya (2005) "Maximizing the benefits of international migration for development" in IOM (2005) "Migration Development and Poverty Reduction inAsia", International Organization for Migration, Genev. Forum for Social Studies (2005) Understandingthe dynamics o f resettlement inEthiopia, PolicyBriefingn.4. Huang Ping and Zhan Shaohua (2005) "Internal migration in China: linking it to development" in IOM (2005) "Migration Development and Poverty Reduction in Asia", International Organization for Migration, Geneva. Larson, DonaldandYair Mundlak,(1997) "On the Intersectoral Migration o f Agricultural Labour" Economic Development and Cultural Change,Vol. 45, No. 2. pp. 295-319. Ministry of Finance and Economic Development (MOFED) (2005) Ethiopia Participatory Poverty Assessment 2004-05. Volume 2: Main Report. May 2005, Addis Ababa. Mosse D, Gupta S, Mehta MyShah V, Rees J, Team KP. 2002. Brokered Livelihoods: Debt, labour Migration and development in Tribal Western India. Journal of Development Studies 38: 59. National Research Council (2003) Cities transformed: Demographic change and its implications in the Developing World. Panel on urban population dynamics, MR 133 Montgomery, R. Stren, B. Cohen, and HE Reeds, (eds). Washington DC: The National Academies Press. Stark, Oded and J. EdwardTaylor (1991) "Migration Incentives, Migration Types: The Role of Relative Deprivation" The Economic Journal, Vol. 101, No. 408. pp. 1163-1178. Tadele, Feleke, Alula Pankhurst and Philippa Bevan (2005) "Migration, labour markets and the Informal Sector: an exploratory study in Ethiopia" , Wellbeing in Developing Countries Ethiopia Programme, ESRC WeD Research Programme, University of Bath, UK. 134 Table 53: Determinantsof Migration -RecentMigrants inUrbanAreas Only nic group incommunity o religious group incommunity of Observation 193952; regressions also control for experience and its square, technical and vocational training and non-formal education, and a dummy for female divorced or widowed. 135 Table 54: DeterminantsofEmployment RecentMigrantsinUrbanAreas Only - Obseivations 15029, Robust standard errors inbrackets *significant at 10percent; **significant At 5 percent; ***significant at 1percent. 136 Table 55: The determinantsof Wages by Migrant Statusand Gender Full sample Females Males I migrants I migrants 1migrants [0.178] [0.194] [0.3761 [0.2561 [0.233] [0.129] Amhara -0.041 -0.279*** -0.111 -0.351*** -0.02 -0.231 *** 1 [0.191] I [0.053] 1[0.263] I [0.094] I [0.262] 1[0.059] Gambella I 0.012 I 0.099 I0.161 I0.249** I -0.14 I 0.047 [0.158] [0.091] [0.2601 [0.106] [0.224] [0.122] Harari 0.055 -0.022 -0.11 0.034 0.089 -0.071 [0.201] [0.0391 [0.5 111 [0.0581 [0.2361 [0.055] DireDawa 0.064 0.131* -0.126 0.275** 0.03 0.02 [0.186] [0.077] [0.345] [O. 1121 [0.260] [0.075] Constant -2.562*** -2.082*** -3.606*** -2.502*** -1.988*** -1.943*** [0.461] [O. 1331 [0.802] [0.226] [0.703] [O. 1831 R2 0.45 0.44 I0.42 I0.39 I 0.37 I 0.36 I I I I I I N I 1, 147 19,267 349 3,879 798 5,388 Robust standard errors inbrackets *significant at 10 percent; **significant at 5 percent; ***significant at 1 percent 137 5. MOBILITYAND EARNINGS INETHIOPIA'S URBAN LABOURMARKET: 1994-200449 Highlights Ethiopia's economy has undergone a series o f major policy reforms since the early 1990s at the policy and structural levels. The structure o f the urban labour market has changed over the same period, and in this chapter we look at a set o f indicators that have a bearing on labour market segmentation andits relationto the rate and nature o f open unemployment inurban areas. Analysis of the five waves o f the urbanpanelreveals a large, persistent and residualpublic sector wage premium. A large proportion of the urban labour force i s also subject to long term unemployment. Taken together, this indicates that Ethiopia's urban labour market couldbe highly segmented, although we cannot provide formal evidence o f segmentation here. However, we do offer data on indicators that segmentation i s likely to have weakened over the period in question, if it does indeed explain any part o f the observed sectoral wage gaps. One such indicator i s that the rate of mobility between the two pairs o f sectors (public and private; and formal and informal, respectively) has increased since the late 1990s. Sample transitions rates have grown across survey waves, and state dependenceon sector choice has weakened. A second i s that the role o f comparative earnings in selection into the informal sector has increased in recent years. 1.Introduction 5.1 Ethiopia's economy has undergone a series of major reforms since the early 1990s at the policy and structural levels. Inthis chapter we investigate the extent to which the structure o f the urban labour market has changed over the same period. Specifically w e would like to assess changes that might have occurred in the mobility o f workers across sectors, in sectoral earnings gaps, and the sensitivity of mobility to the gaps. These have implication to the question o f whether or not the labour market i s segmented and to the nature and rate o f open unemployment. 5.2 Our analysis is based on data drawn from the Ethiopia Urban Household Socio Economic Survey o f the Addis Ababa University and the University o f Gothenburg, Sweden. The survey was started in 1994. There have been four subsequent waves, one in each o f the years 1995, 1997, 2000 and 2004. Some 9,000 to 10,000 individuals in 1,500 to 1,600 households were covered ineach of these waves. 5.3 A major strength o f the dataset generated by the survey is the sizeable panel component. More than 40 percent o f individuals covered in the first wave were tracked by all four subsequent waves. A much higher proportion were covered by at least three waves. A significant weakness i s that the survey samples were drawn exclusively from the country's seven largest urban centers. While these probably account for the bulk o f 49 This chapter is based on background papers by Arne Bigsten, Taye Mengistae and Abebe Shimeles. Copies of the papers are available upon request from tmengistae@worldbank.org. 138 the urban sector o f the economy, in the absence o f appropriate sample weighting schemes, our results may not apply to the broader (national) urban labour market. That said, we would be surprised if the more qualitative aspects o f our conclusions did not hold for the urban sector countrywide. 5.4 Segmentation i s a sign o f inefficiency in the labour market, because this implies that individuals cannot move to where they would be most productively employed. Segmentation i s indicated whenever there are persistent and uncompensated sector wage premiums that we cannot attribute to selectivity bias or to gaps in unmeasured ability between groups. Segmentation i s often hypothesized between public sector employment andthe private sector wage work, and betweenformal sector wage employment andown account workhnfonnal sector employment. As uncompensated differentials in pay are sustainable only if the mobility o f workers across sectors i s impeded, segmentation suggests that a proportion o f the unemployed have been rationed out o fjobs by non-price factors. 5.5 There are no widely accepted formal tests o f segmentation available. In this Chapter, we look at variables that directly relate to segmentation to see if these have changed over the period, an increase inworkers' mobility into highwage sectors over the survey period are taken to indicate a weakening o f segmentation. In order to see if this actually happens in the Ethiopian data we compare sample transition matrices across survey waves. More formally, we also estimate dynamic binary sector choice models for four states, namely, public sector employment, formal private sector wage employment, informal sector employment (including self employment), and open unemployment. 5.6 Another useful indicator would be changes inunexplained earnings gaps between sectors. Inparticular, a widening o f sector earnings gaps infavour o f a highwage sector could indicate a fall inthe rate o fmobility into the sector, and increased segmentation. A third useful indicator is the sensitivity o f sector choice to earnings. The less sensitive is sector choice to pay gaps the more likely that mobility into the high paying sector has been impeded. To look at these later indicators, we estimate sector earnings gaps based on estimated earnings equations for wage workers and informal sector workers. The sensitivity o f sector choice to earnings gaps i s estimated as a parameter o f the structural sector choice model. 5.7 We find large earnings gaps between the public and the private sectors, which have increased over the years. This could indicate a high degree o f segmentation o f the market. At the same time the sensitivity o f sector choice to earnings gaps seems to have risen in more recent waves, not only between the private and the public sectors of wage employment, but also between formal sector wage employment and informal employment. In particular, the role o f relative earnings in selection into the informal sector seems to have increased substantially suggesting that informal sector employment might be a choice driven by comparative advantage for a growing proportion o f those found inthe sector. 5.8 The overall results suggest that segmentation weakened over the survey period. Both the raw transition matrices computed from the survey sample and the estimation o f 139 dynamic sector choice models indicate similar trends. Raw transition matrices show that, although there was little mobility across sectors prior to the 1997 wave, it increased considerably between 1997 and 2000 to rates that have been sustained since then. This has been accompanied by a small but persistent decline in the sample rate o f open unemployment, which in large part was due to growth injob creation not so much inthe informal sector as the formal sector, and not so much inthe public sector as inthe private sector. 5.9 Our estimates o fthe dynamic sector choice models are also consistent with a story o f weakening segmentation. Although the state dependence (or persistence) parameter in the dynamic choice models is positive and statistically significant for both public and informal sector employment, it i s significantly smaller inmore recent survey waves. The state dependence parameter seems also to have been falling for the equation for the openly unemployed. 5.10 The rest o f the chapter proceeds as follows. Section 2 presents our framework for assessing the weakening or strengthening o f segmentation over time. Section 3 describes the structure o f Ethiopia's urban labour market in terms o f the demographic and educational profiles o f the various labour market states o f the survey sample. Insection 4 we provide a description o f patterns and rates o f mobility across states in terms o f raw transition matrices. We present results o f estimation o f dynamic sector choice models in Section 5 and those o f the estimation o f structural sector choice models and earnings equations inSection 6. We conclude in Section 7. 2. Empirical framework 5.11 One way o f investigating whether the labour market is getting more or less segmented between, say, formal sector wage employment and informal sector employment is to see if the probability that someone who was in the informal sector in year t -1 will have remained in the same sector in year t has increased, decreased, or remained the same over time. A fall in the probability indicates that segmentation has decreased over time, while an increase inthe same probability suggests an increase inthe degree o f segmentation. The more mobile are workers across the segmentation divide the poorer i s status at time t - 1 as a predictor o f status at timet. 5.12 Which are the factors that determine the probability inquestion? One option i s to take a sample o f individuals observed in the informal sector in yeart-1, and observe which o f those individuals will have left for the formal sector by yeart, and which ones will have remained, and to try to identify observable characteristics that discriminate between stayers and movers. The problem with this approach is that our estimate o f the influence o f covariates o f transition i s subject to selectivity bias if initial employment status in terms o f being in or out o f the informal sector is not exogenous, that is, if the sorting o f individuals between the formal and the informal sectors in year t-1 i s nor random and there are unobservables that influence both status at time t-1 and some o f or all o f the determinants o f subsequent transition out o f the informal sector. To allow for this possibility we will specify the determination o f employment status at time t as 140 Ii,=x;,p + +E, (1) where Ii,i s a dummy variable equal to unity if the individual i is observed in the informal sector in year t , xi, i s a set o f individual characteristics, p is a vector o f parameters, S is a constant, and ci, i s a zero mean iid random error term orthogonal to xi,. This formulation allows us to estimate the degree o f state dependence in informal sector status, which is now captured by the parameter S . Because we are controlling for initial status, we can now also interpret p as determinants o f the marginal effects o f xi, on the probability o f transition out o f the informal sector. However, whether or not we can estimate p and S consistently based on (1) depends on whether is exogenous. The latter will not be the case if si, includes permanent (unobservable) individual characteristics that influencing the sorting o f workers between the formal and the informal sectors in year. Estimates o f /?based solely on (1) would be selectivity biased inthat case. This problem ofpotentialbias is usually referredto as the 'initial condition problem' (Heckman 1981) and can be avoided by modeling Ii,simultaneously withIi,-l . We specify the reduced form o f the determination o f Ii,-l as Ii,-lz;y ui = + (2) where Zi i s a vector o f individual characteristics, y i s a vector o f parameters and 'i i s a zero mean iid error term orthogonal to 'i. Equation (1) and (2) can bothbe identified as long as not all elements o f ' iare in or vice versa. Ifwe assumed that 'i and E' are joint normal, we could estimate the two equations jointly as a particular restriction on the bivariate probit, but we do not do so here. Insteadwe use a two step procedure, whereby we first estimate equation (2) as a simple probit, andthen include the generalized residual thereof as a Heckman type selectivity term inthe estimation o f equation (1). Let be the component o f E, = ei +Vi, ei =au, +vi ' i tthat is correlated with 'isuch that Y and where vi i s orthogonal to ' i , and' i fis orthogonal to 'i.We canthen write (1) as I~,x;,p = + SI^,-^ +aui +vi +vi, (3) 5.13 In the second stage, we can estimate equation (3) as a random-effects probit, having replaced ui by the generalized residual from the estimation of equation (2). Letvar(ui and )= at =Os' Then . a=pa, lau andVar(Vi) =os'(1- p') .A test o f the statistical significance o f a i s consequently a test o f the null that p = ,that is, a test o f the hypothesis that employment status at time t-1 i s exogenous to current employment status'it . 5.14 The focus o f our analysis is the question o f whether state dependence in the sorting o f the labour force between labour market states increased or decreased over the 141 survey waves as measured by the parameter 6. Because we have a relatively short panel o f four waves o f the survey and a relatively small sample within each wave, we have chosen to analyze the data by taking different but consecutive two-wave cuts and seeing what happens to 6 as we move to later pairs o f waves, rather than considering a distributed lag structure o f labour market status in a dataset pooled into a single panel. This means we in effect estimate equation (3) in a cross-section and hence as a simple probit, rather than as a random effects probit. To underscore this we will write the equationwe estimate inthe second state as I,=Xi.?p 61it-l azi,+vi, + + (3b) where the hat symbol indicates that generalized residual from the estimation o f (2). 5.15 Changes in 6 that we observe over time tell us whether the labour market i s getting more or less segmented between the states in question. Decline in state dependence over time indicates that mobility between states is increasing just as an increase instate dependence should implydiminishingmobility over time. An alternative means o f gauging changes inthe degree of segmentation is to look at changes in sectoral eamings gaps between sectors o f employment. Wage gaps that we cannot attribute to selectivity bias or to differences in unmeasured ability between groups are sustainable only if some barrier to entry impedes mobility between sectors. The expansion o f uncompensated wage premiums over time should therefore be as good an indicator o f growing segmentation as would be decline in transition rates between sectors. One way o f formalizing this notion is to think in terms o f a structural sector choice equation underlying equation (3). Let Ylit be the earnings o f person i in the sector o f current employment, and Y2itthe person's potential earningsinthe alternative sector. Ifworkers respond to sector wage gaps in deciding where to work, then ylit -y2it should be one o f the determinants o f employment status, ' i t .We assume that the relationship betweenpay gaps and sector choice i s such that 5.16 where ' i t may now be defined as a dummy variable equal to 1 when i i s employed insector 1but is zero when i i s insector 2; A i s a parameter; ' i ti s a vector o f exogenous individual characteristics, n i s a vector o f parameters (including a constant term), and {it i s a zero mean iid error term orthogonal to pay gaps and to ' i t ,The proposition that workers' sector choice andmobility decisions depend on pay gaps can be tested by estimating equation (4). Moreover, assuming that there i s evidence o f segmentation in the data, one can get indications of whether it has been growing or diminishingover time by looking at changes in A over time. Ifsegmentation grows over time in the sense that workers have become less responsive to pay gaps this should be 142 reflected in A diminishingover time. Likewise an increasing sensitivity o f sector choice to pay gaps should be indicative o fweakening segmentation. 5.17 A problem in trying to estimate A on the basis o f (4) is that, although each worker might have a reasonable idea o f their earnings potentials in alternative fields of employment, we can observe only one o f these, which is the wages o f current employment. Inother words it i s only one o f ylitor y2it that we can hope to observe in ' any dataset. A commonly used solution to this problem is to replace the unobserved wage by a counterfactual pay rate implied by an estimated earnings equation. To obtain the estimates we assume that ylitand Y2itare linear ina set o f inparameters such that where xit is a vector o f exogenous covariates o f earnings, n1 and n2 are vectors o f parameters (including constant terms), and and are iid error terms orthogonal to 5.18 The counterfactuals based on the OLS estimation o f equations (5) and (6) would be biased if there are unobservable worker attributes that induce workers to self select into either sector and at the same time enhance or reduce their earnings in the same sector. In order to avoid this bias we estimate the two equations having added a Al(i,)=- #('it) selectivity term based on the estimation of equation (3b). Let @('it) where 4 is the pdf o f A {it,@the corresponding CDF and ' i f is predicted value o f ' i tthat one 4('it ) ~ 2 ( ' i t >= would obtain from parameter estimates o f equation (2) and l - Q ( ' i t I . We estimate the earnings equations by applying OLS to and Y21it=Xi'tn2 +0212('it +S2it (6b) where o1 and o2 are constants andSlit and S2it are iiderror terms orthogonal to all other righthandside variables intheir respective equations. 143 5.19 The structural sector choice equationwe actually estimate is Ii,= -j2J 2,;r q, + + (4b) where the hat symbol indicates estimates based on the estimation o f (5b) and (6b). 3. Results: profiles of the labour market sectors 5.20 Table 56 provides the breakdown o f the survey sample by age groups and labour market states. It implies a labour market participation rate o f about 54 percent ifwe limit the labour force to those in the 15-64 age group, and those who are currently self- employed, or are working for someone else for some form of payment. Table56: UHSESSampleAge Distributionofthe Sample. All waves andAge Groups All age groups Age =6-64 yrs. Age =11-64 yrs. Age =15-64 yrs. status Number Percent Number Percent Number Percent Number Percent self-employed 3,317 7.53 2,928 7.49 2,915 8.49 2,905 9.86 governmentworker 2,602 5.91 2,557 6.54 2,556 7.44 2,554 8.67 public enterpriseworker 1,004 2.28 986 2.52 986 2.87 986 3.35 formalprivatesector worker 1,721 3.91 1,668 4.27 1,663 4.84 1,656 5.62 otherprivatesector worker 2,990 6.79 2,821 7.22 2,811 8.19 2,735 9.29 unemployed 5,257 11.94 5,130 13.12 5,120 14.91 5,100 17.32 out of the labour force 27,140 61.64 23,009 58.85 18,292 53.26 13,514 45.89 Total 1 44,031 100 39,099 100 34,343 100 29,450 100 5.21 Table 57 shows the rate didnot change much over the decade, fluctuating within a 2 to 3 percentage point band. We also see that unemployment dropped significantly between 1997 and 2000 before somewhat stabilizing at just below 30 percent. No part o f this fall seems to have come about through the expansion o f the informal sector. If we define the latter to include all the self-employed, unpaid family workers, casual labourers and domestic workers, the informal sector employment rate fluctuates within the 35-36 percent range for the 1994, 1995, 1997 and 2000 waves, and drops slightly to 34 percent for the 2004 wave. The observed drop in the unemployment rate was therefore largely reflection o fthe expansion o f formal wage employment. 144 Table57: UHSESDistributionbyLabourMarketState andYear, Wave byWave ~ ~~ ~ ~~ 1994 1995 1997 2000 2004 Status Number Percent Number Percent Number Percent Number Percent Number Percent self-employed 609 10.47 636 9.73 530 9.8 556 10.05 574 9.32 government worker 523 8.99 561 8.58 469 8.67 453 8.19 54s 8.9 public enterprise worker 227 3.9 244 3.73 182 3.37 190 3.43 143 2.32 formalprivatesector worker 239 4.11 251 3.84 233 4.31 372 6.72 561 9.11 otherprivatesector worker 516 8.87 586 8.97 482 8.91 536 9.69 615 9.99 unemployed 1,075 18.48 1,140 17.44 938 17.35 926 16.73 1,021 16.58 out ofthe labourforce 2,627 45.17 3,118 47.71 2,573 47.59 2,501 45.19 2.695 43.77 Total 5,816 100 6,536 100 5,407 100 5,534 100 6,157 100 5.22 The share o f the formal sector inurbanemployment remained at about 31 percent inthe course ofthe 1994, 1995 and 1997waves before increasingto 33 percent by 2000, and further to 36 percent by 2004. The expansion of formal-wage employment came about through employment growth inthe public sector and even more so inprivate firms. Aside from showing a significant dip between the 1997 and the 2000 waves o f the survey, the public sector's share in urban employment seems to have remained at about 16 percent. The formal private sector's share also hovered around 14 percent up untilthe 1997 wave, but seems to have grown substantially since then, to 18.5 percent by the 2000 wave and to 20.4 percent by 2004.50 5.23 Within the public sector, the distribution o f employment between government agencies and state owned companies remained constant at least over period 1994-2000, with the former accounting for 75 percent. It then fell by about 50 percent over the next four years. 5.24 Gender and age profiles. Table 58 provides demographic and educational profiles o f the labour force by current market state. It is a multinomial logit fit to the sorting o f labour market participants in the sample into public sector workers, formal private sector workers, informal sector workers and the unemployed. There is a strong gender effect inparticipation inthat women are more likely to be found out o f the labour force (chiefly as homemakers) than in any o f the other four labour market states. Within the labour force the gender effect is strongest with respect to the formal sector, where women are less likely to be found than in informal employment or in open unemployment. If anything, these gender effects in public or formal sector employment seem to have grown over the years spanned by the survey. 50An important caveat is in order here. There is no employer size dimension to our definition of informality. People who have reported to be employees o f a private sector organization or enterprise have all been classified here as formal sector workers. I t i s possible that many o f these would be classified as informal businesses, inwhich case the figures o f the last paragraph could grossly understate the size o f the informal sector and overstate that o fthe private formal sector. 145 5.25 Table 58 also exhibits strong age effects in employment status. When in the labour force youth, aged 15-29 years, are more likely to be found inthe public sector or inopen unemployment than either inthe private sector or ininformal employment. The log odds ratio between being found inthe public sector and being out o f the labour force (or in school) is also statistically indistinguishable for a young person from that between being unemployed andbeingout o f labour force. 5.26 As is to be expected the 30-54 age group is the least likely to be found out o f the labour force. Within the labour force it is more likely to be found in the public sector, and more likely in the formal sector than in unemployment or in informal sector unemployment. A similar pattern applies to the 55-64 age group. The significant difference here i s that the log odds ratio between unemployment and retirement is also naturally higher for the 30-54 age group. 5.27 Age has become an even stronger predictor o f public sector employment over the decade. Like gender, however, its correlation with employment probabilities elsewhere, or with the probability o f unemployment, does not seem to have changed much. 5.28 Education Profiles. Labour market participation rates are highest among those without formal schooling (Table 58). Those without schooling are also more likely to be found in informal sector employment than in formal sector jobs or in open unemployment. In other words, the log odds ratio between informal sector employment andbeing out o fthe labour force is negative for all those who havehad formal schooling. 5.29 Among those with some schooling, those with tertiary education o f some kind are more likely to be found in the public sector than in the private sector and, within the private sector, least likely to be found in informal employment. Indeed they are more likely to be found in open unemployment than in informal sector employment. Those who have completed secondary schooling are also more likely to be found in public sector employment than in the private sector. They are also more likely to be unemployedthanto work inthe formal private sector. 5.30 Primary school completion does not seem to differ much from having no schooling at all in terms o f its prediction o f labour market status. However, those who have had some schooling but have not graduated from high school are more likely to be found informal private sector than anywhere else including the public sector. 5.31 Like gender and age, education effects on public sector employment have grown more pronounced inrecent waves. The association between probabilities o f employment inother sectors (or ofunemployment) andeducation haveremainedunchanged, however. 146 Table 58: MaximumLikelihood EstimatesofMultinomial Logit Models of Labour Market States, 1994-2004 Pooled 1994 2004 (1) (2) (3) (4) 5) (6) (7) (8) (9) (10) (11) (12) Public Formal Informal unemp- 'ublic Formal Informal unemp- Public Formal Informal unemp- sector private sector loyed iector private sector loyed sector private sector loyed worker sector worker worker sector worker worker sector worker employee employee employee Female -0.756 -1.116 -0.625 -0.466 0.592 -1.149 -0.650 -0.462 -0.951 -1.038 -0.604 -0.369 (14.53)" (23.01)" (17.72)" (12.67)' 5 10)" (9.88)" (8.16)" (5.67)" (6.87)" (11.21)" (8.02)" (4.65)" Married 0.138 0.191 -0.500 -1.132 1.125 0.264 -0.515 -1.215 0.116 0.059 -0.422 -0.880 (2.16)' (3.12)" (11.23)" (17.70)' 0 87) (1.81) (5.17)" (8.42)" (0.90) (0.50) (4.54)" (6.91)" Age group (reference=l5-28): 30-44yrs. 2.477 1.640 1.694 0.911 2.515 2.016 1.528 0.862 2.219 1.650 1.789 0.918 (36.63)" (29.25)" (35.29)" (16.88)' 16.45)" (13.05)" (13.96)- (6.91)" (16.42)" (13.92)" (17.61)" (6.20)" 45-54yrs. 2.091 1.506 1.284 0.033 2.205 1.869 1.358 0.014 1.677 0.867 1.030 -0.430 (22.84)" (17.84)" (20.99)" (0.33) 10.39)" (9 35)" (9.99)" (0.08) (9.51)" (5.36)" (8.03)" (2.01)' 55-64yrs. 0.372 0.192 0.687 -1.500 1.366 0.224 0.682 -1.502 -0.335 -0.229 0.178 -2.312 (2.73)" (1.69) (10.37)" (8.58)" 1.12) (0.78) (4.52)" (3.73)" (1.19) (1.07) (1.23) (5.01)" Education (referencewo formal schooling): grades 1to 10 -0.107 -0.005 -0.946 -0.164 1.502 0.112 -1.012 -0.289 -0.867 -0.272 -0.859 0.064 (1.24) (0.06) (22.25)" (2.48)' :2.20)' (0.61) (10.52)" (1.68) (5.27)" (2.01)' (9.58)" (0.49) preparatory school 2.164 1.577 -0.640 1.892 2.869 1.719 -0.975 1.891 1.100 0.934 -0.557 1.449 (24.38)" (19.52)" (10.73)" (27.64)' 112.11)" (8.52)" (6.92)" (11.96)' (7.06)" (6.53)" (4.84)" (10.57)' tertiaryeducation 2.636 1.739 -1.203 0.479 3.764 2.224 -1.032 0.840 1.617 0.835 -1.800 -0.338 (28.01)" (17.76)'' (11.17)" (4.51)" 13.98)" (9.02)" (4.01)'. (3.41)" (9.42)" (4.94)" (6.49)" (1.61) City(reference=Addls Ababa): Awasa 0.874 -0.810 -0,019 -0.154 1.842 -0.462 0.015 0.355 0.649 -1.057 -0.143 -0.637 (9.07)" (5.95)" (0.24) (1.82) 3.64)" (1.48) (0.08) (1.91) (3.40)" (4.31)" (0.95) (4.54)" Bahir Dar 0.440 -0.074 0.014 -0.305 1.754 0.387 0.433 0.124 0.049 -0.456 -0.064 -0.702 (4.22)" (0.73) (0.18) (3.49)" 3.29)" (1.73) (2.73)" (0.68) (0.23) (2.33)' (0.42) (3.67)" Dessie 0.520 -0.303 -0.089 0.046 1.551 -0.276 0.215 0.266 0.328 -0.554 -0.361 -0.710 (4.63)" (2.53)' (1.15) (0.54) 2.11)' (0.92) (1.26) (1.42) (1.46) (2.53)' (2.11)' (3.42)" Dire Dawa -0.299 -0,010 -0.127 0.324 0.137 0.365 -0.063 0.354 -0.448 -0.751 -0.125 0.262 (2.43)' (0.11) (1.84) (4.73)" 0.52) (1.87) (0.42) (2.35)' (1.73) (3.43)" (0.62) (1.78) Jimma 0.890 -0.644 0.028 -0.210 1.033 -0.324 0.218 0.171 3.685 -0.561 0.137 -0.387 (9.70)" (5.39)" (0.41) (2.57)' 4.83)" (1.14) (1.36) (0.93) (4.92)" (2.76)" (1.00) (2.36)' Mekele 0.273 -0.715 -0.025 -0.961 1.268 -1.013 0.429 -0.697 3.265 -0.655 -0,483 -1.635 (2.01)' (4.83)" (0.29) (7.56)" 0.79) (2.32)' (1.64) (2.34)' (0.79) (2.12)' (2.15)' (4.94)" Surveyyear (referenc~lSS4): 1995 0.057 0.322 -0.158 0.115 (0.43) (2.27)' (1.82) (1.13) 1997 0.062 0.332 -0.159 0.113 (0.46) (2.32)' (1.79) (1.10) 2000 0.021 0.364 0.052 0.086 (0.26) (4.98)" (0.98) (1.50) 2004 0.028 0.470 0.062 -0.008 (0.37) (6.67)" (1.20) (0.15) Constant -4.010 -2.476 -0.079 -1.315 4.634 -2.556 -0.036 -1.180 .3.001 -1.842 -0.515 -1.107 (14.93)" (11.02)" (0.48) (6.77)" 8.01)" (5.34)" (0.10) (2.84)" :9.93)" (6.44)" (2.48)' (4.67)" Ethnicity dummies Yes Yes Yes Yes les Yes Yes Yes Yes Yes Yes Yes Religiondummies Yes Yes Yes Yes Yes Yes Yes Yes Yes Yes Yes Yes Observations 30006 6123 6385 Log likelihood -34194 -6791 -7789 PseudoRsquared 0.17 0.19 0.14 147 5.32 Ethnicity, religion and regions. Though suppressed for reasons o f space, the specification reported in Table 58 includes dummies for first language groups (ethnicity) and religion. There do not seem to be significant ethnicity effects in labour market status ifwe define ethnicity narrowlyinterms oflanguage groups only. Religion does seemto matter at least in as far as Orthodox Christians are more likely to be found inthe public sector than elsewhere, while Muslims are less likely to be public or formal sector employees thanbeing inthe informal sector. 5.33 There are also significant location effects in sector profiles in that the public sector is a larger employer relative to either segment o f the private sector in five o f the six regional cities than it is in Addis Ababa, the exceptional regional town being Dire Dawa. The rate o f unemployment and the relative size o f the informal sector are also larger in Addis Ababa than inany o f the regional cities, again with the exception o f Dire Dawa. 4. Agerepate transition rates and patterns 5.34 As a prelude to more formal analysis o f the issue o f segmentation it is useful to describe the extent to which people have been mobile across labour market states, by comparing aggregate transition matrices across the three waves (see chapter annex for the transition matrices). 5.35 We beginbynoting four key trends inour panel, described inFigures 38 to 40. 0 The labour market participation rate fluctuated around 54 percent over the survey decade, dividing almost equally between formal sector employment, informal activities, andunemployment (Figures 38 and 39). 0 There was a small but statistically significant and persistent decline in the unemployment rate since 1997 (Figure 38). The unemployment decline can be traced to the formal rather than to the informal sector: the formal sector share in urban employment rose substantially (Figure 39). 0 The observed formal employment growth was in both the private and public sectors. The growth was significantly faster inthe private sector (Figure 40). 148 Figure38: UHSES ParticipationandUnemploymentRates(percent) 60 - 50- c a - v - A 40 - 30 - IC - I -- I 20 - 10- 0 I I I I I --eparticipation -m- unemployment Figure39: UHSESEmployment Ratesby Sector (percent) 38 - 36 - 34 - 32 - 30 28 I I , I , I 1994 1995 1997 2000 2004 +Formal-m- Informal Figure40: FormalEmploymentRatesby Sector (percent) 2ol1 15 I , I I I 1994 1995 1997 2000 2004 -e-Private-I-Pubic 149 5.36 Unemployment transitions. Tracking the unemployed o f the 1994 wave. The transition matrix between 1994 and 1995 does not indicate any movement between labour market states (not shown). But while the (marginal) probability o f someone reporting unemployment in the 1994 wave landing a job or shifting to self-employment within a year was zero, this increases as we expand the time horizon. 5.37 Table 68 in the Annex tracks the respondents o f the 1994 wave through to the 1997 wave. There i s some movement out o f unemployment, but at a very low rate: less than 7 percent o f those who were unemployed in the 1994 wave had some form o f employment by 1997. A little over a third o f these went into formal sector wage employment. 5.38 Things do change significantly over the nextthree years. By2000, more thanhalf o f the unemployed had changed status (Table 69). About one in eight left the labour force altogether, having stopped looking for work. About one in five went into own account work or into informal wage employment. Another 11percent hadjobs inprivate firms. Approximately 9 percent landed public sector jobs, about one third o f these being instate ownedcompanies. 5.39 Bythe 2004 wave, less than a thirdo fthe unemployed o fthe 1994wave were still available for work. What happened to the rest? Broadly, about 23 percent left the labour force; 19 percent joined the informal sector (in self-employment or in informal wage contracts); 17percent landedjobs inprivate firms, and 12percent got public sector jobs. 5.40 To assess changes in the intensity o f transition over time, we observe transition probabilities between three year intervals, Table 71 covers 1994-1997 and Table 75, 1997-2000. The results suggest that labour market mobility increased dramatically inthe second half o f the 1990s. The probability that someone reporting unemployment in 1997 had some form o f employment by 2000 was about 37 percent. This is more than five times the probability o f transition from unemployment to employment between the 1994 and 1997 waves. This segment suggests a significant increase in the rate o f gross job creationrate over the period. 5.41 Comparing the periods -- 1994-1997 and 1997-2000 -- increases in the rate o f absorption o f the unemployed into the formal sector increased from 2 percent to 18 percent; by the informal sector (both self-employment and wage contracts) from about 5 percent to 19 percent; by the public sector, from under 1percent to almost 8 percent; and, bythe formal private sector increased from under 2 percent to about 10percent. 5.42 The relatively high rates o f transition from unemployment during the period 1997-2000 were also sustained in 2004. The probability o f someone reporting unemployment in 2000 taking up some form o f employment by the time o f the 2004 survey was 41 percent. This is significantly higher than the corresponding transition rate for the period 1997-2000, but by an amount that could be explained entirely by the fact that the second interval is one year longer. About 24 percent of the unemployed o f 2000 landed formal sector jobs by 2004; another 17 percent became self-employed or entered 150 into informal job contracts; some 13 percent joined private firms; and about 11percent joined the public sector. 5.43 Where do the unemployed come from? Some 17 percent o f those who were out o f the labour force in2000 joined the unemployment pool by 2004 (Table 76). Giventhat the out-of-the-labour force are by far the largest segment o f survey respondents in any year, new entrants are the largest source o f entry into unemployment. 5.44 Private sector job losses were the second important source o f gross addition to employment over the period. The probability of a formal private sector job loss was 14 percent. The probability o f an informal sector job loss was even higher at 17 percent. This contrasts with 7-8 percent rate o f public sector job losses. Given that there were more private than public sector employees in the 2000 sample, a larger proportion of the unemployed of 2004 must have come from the private sector than did from the public ~ector:~' 5.45 The probabilities of losses o f various categories o f jobs-that is, public sector, private formal sector, andinformal sector-over the period 1997 -2000 (Table 10) are quite similar to those o f 2000-2004. However, job loss rates doubled in the state owned enterprise sector and nearly did so in government agencies. Job loss rates were also muchhigher for all sectors during 1997-2000 than for the period 1994-1997. Indeed, just as probabilities o f transitions out o f unemployment into any kind o f employment were negligible during 1994-1997, so were the probabilities o fjob losses in any sector-private or public, and formal and informal. A comparison of Table 76 with Table 71 shows that private sector formal job loss rates have remained similar between the 1997-2000 and 2000-2004. 5.46 Informal to formal sector transitions. Like rates of transitions out o f unemployment, probabilities o f transition between the formal and informal sectors o f employment seem negligible over one year (such as 1994-1995), and two-year intervals (Table 74). Less than one percent o f those who heldjobs or were self-employed in 1994 had moved either way by 1997. This seems to be a picture o f almost complete segregation o f workers betweenthe two sectors. 5.47 The probability o f transition from the informal sector to the formal sector does rise significantly when we expand the time horizon by three more years (Table 72). This is largely on account o f movement o f workers to the private formal sector. The probability that someone reporting self or informal employment inthe 1994 survey would have joined the public sector by the 2000 was only 5 percent. The probability that the same individual would have taken up ajob ina private firm was about 15 percent. These probabilities are similar when we expand the horizon further to 10 years (Table 73). The relatively highrates o f transition from informal sector employment to formal sector jobs for the 1997-2000 are sustained into 2000-2004 (Table 76). 5 1Interestingly, a significant proportiono f the unemployed in2004 were self-employed in2000. 151 5.48 To see if the rate o f mobility from informal sector employment to formal sector jobs has increased over time, we again compare table 71, which captures transitions over 1994-1997 with table 75, which relates to 1997-2000. As was the case with transitions from unemployment to employment, there was a drastic change here also: the probability o f transition from informal sector wage employment to formal sector jobs rose from less thanone per cent for 1994-1997 to 15percent for the period 1997-2000. 5.49 Like the contemporaneous drop in the rate o f unemployment, this must also reflect a higher formal sector job creation rate, especially in private firms. Again the larger shift occurs in the probability o f transition from informal sector employment to formal private sector jobs rather than from informal employment to public sector jobs. 5.50 Where do informal sector workers come from? By far the largest source o f new informal sector workers were those who were formerly unemployed or new labour force entrants. More than one inten o f people who were out o f the labour force in2000 would have become informal sector wage worker or self-employed by 2004. 5.51 Just under 20 percent o f those who were unemployed in 2000 would be self employed or in informal wage work by 2004. These rates are much larger than the probability o f transition from the formal sector. Because the combined share o f the out o f the labour force and the unemployed was 60 percent o f the full sample o f individuals in the 2000 survey, this means that by far the largest fraction o f those who joined the informal sector by 2004 could only have come from these sources. Infact the probability that anyone who had a formal sectorjob in2000 would be found inself employment or in informal paidjobs in2004 was quite low, standing at 6 percent for a government worker, at 8 percent for a public enterprise employee, and at 12 percent for an employee o f a private firm. 5.52 Transitions between the public and private sectors. There was a 48 percent chance that a civil servant 1994 would have left the public sector by 2004. This comprises (from table 73): 0 Those that retired or otherwise left the labour force (20.5 percent); 0 Those laid or pensioned offbut didnot leave the labour force (10 percent); and 0 Those that joined the private sector (17 percent). More than halfo f those who moved to the private sector worked for a formal sector firm. 5.53 The corresponding probabilities o f transition for a public enterprise worker were 35 percent o f leaving the labour force, 10 percent o f being unemployed, and 16 percent joining the private sector. Two-thirds o f those joining the private sector joined a formal sector firm. 5.54 Changes in transition rates reflect the higher transition probabilities in the later waves. There was practically no movement from the public sector between the first and the third waves (1994-97) (Table 71). Subsequently, there was a 12 percent chance that someone who was a government employee in 2000 would havejoined the private sector inself-employment (4 percent), or as an employee (8 percent) by 2004. The probability 152 o f a 2000 public enterprise employee moving to the private sector by 2004 was even higher, at 19percent. 5.55 Transitions to the Public Sector. Some 12 per cent o f those who worked for private firms in2000 hadjoined the public sector by 2004. Some 9.5 percent o f informal sector employees and 3 percent o f self-employees moved to the public sector over the same period. These rates are much higher than the negligible transition probabilities between 1994 and 1997, but are only slightly higherthan for the period 1997-2000. Thus the marked increase inmobility inthe late 1990s was sustained well into 2004. 5. State dependence insector choice 5.56 The dynamic binary choice models confirm that, while there has always been significant state dependence in all instances o f sector choice across survey waves, it has weakened considerably over time. This has been true o f transitions in and out o f the public sector, informal employment, and open unemployment. The estimates show that the probability o f being in any one o f these sectors during particular survey wave is higher for those who were in that state in the preceding wave, but this effect o f initial state on current choice has grown progressively weaker in more recent waves. We interpret this as evidence that inter-sectoral mobility has increased considerably over the years. 5.57 Details o f the finding are reported in Table 59 for public sector employment, in Table 60 for informal sector employment, and in Table 61 for unemployment, where we have estimated dynamic binary choice models (equation 3b above) using a two-step procedure for the 15-64 age group. The key estimation results are reported in the last four columns o f each o f these three tables. Taking the example o f table 59, the first two columns report results o f estimation o f the initial selection equation (equation 2 above). In the third column we estimate a selection equation for public sector employment in 1997 for those covered in the 1994 wave as well, the initial selection equation being estimated for 1994. The initial selection equation i s identified by including family background variables which we exclude from the current selection equation on the assumption that these influence current public sector employment status only through their effect on initial selection inor out o f the public sector. Incolumn 4 we estimate the selection equation for public sector employment in 2000 for those observed in the 1994 wave as well, again using column 1 as the initial selection equation. Likewise column 5 shows selection into public sector employment in 2004 for those observed in the 1994 wave. Incolumn 6 we re-estimate the equation for public sector employment in2004 for those observed in 2000, using the initial selection equation given in the second column. That state dependence in sector choice has diminished over time can be read from the first two rows o f each o f the three tables. 153 Initialselectionequation Dynamicselectionequation by year of observation by year of observation 1994 2000 1997 2000 2004 2004 Publiclppl 2.094 1.848 (12.11)" (13.38)" (9.47)" PublicBw 2.206 (8.34)" Genresiduaipublic,Bw 0.151 0.067 0.022 (0.60) (1.02) (0.34) Genresidpublih, -0.307 (1.99)' Female 0.129 0.036 -0.312 0.245 0.116 -0.161 (1.96)' (0.55) (137) (2.20)' (0.93) (1.68) Married,% 0.435 0.525 -0.018 -0.125 (6.01)" (2.01)' (0.14) (0.80) Marriednw 0.340 -0.095 (4.64)" (0.78) Age groups (reference group=Age 15-10]: Age 30-44 0.679 0.619 0.046 0.375 0.278 0.111 (9.22)" (8.36)" (0.17) (2.82)" (1.63) (0.91) Age 45-54 0.759 0.760 0.290 0.595 0.550 0.157 (7.35)" (7.36)" (0.83) (3.15)" (2.69)" (0.88) Age 55-64 0.191 0.173 -0.542 0.140 0.094 0.029 (1.15) (1.06) (1.09) (0.50) (0.32) (0.10) Educatlon (reference groupino formal schooling or primary Incomplete): Primaryschool completed 0.384 0.198 0.294 0.394 0.712 0.175 (4.49)" (2.32)' (0.96) (2.44)' (3.86)" (1.27) Secondaryschool completed 0.428 0.264 0.717 0.622 0.755 0.346 (4.80)" (3.00)" (2.26)' (3.84)" (4.11)" (2.52)' Some tertiary education 0.758 0.626 0.691 0.865 0.732 0.315 (6.12)" (5.24)" (1.63) (4.13)" (2.94)" (157) City dummies Yes Yes Yes Yes Yes Yes Ethnicityand religion dummies Yes Yes YeS Yes Yes YeS Paretnaleducationdummies Yes Yes No No No No Paretnaloccupationdummies Yes Yes No No No No Initialown occupationdummies Yes Yes No No No No Constant -2.509 -2.307 -3.087 -1.845 -2.003 -1.179 (15.47)" (15.03)" (5.61)" (6.92)" (6.09)" (5.33)" ObSeNatiOnS 3312 3144 1435 1233 760 1255 Log likelihood -1094 -1076 -0.86 -354 -287 -487 Pseudo Rsquared (0.38 0.32 lO.89 0.49 0.35 0.34 Absolute value of z statistic$ in parentheses * significantat 5%; ** significantat 1% 154 Table 60: MaximumLikelihoodEstimationofProbitModelof Selectionto Informal Sector Employment itial selectionequation Dynamicselectionequationbyyear of observation ,y year of Observation 394 2000 1997 2000 2004 2004 Informalrprw 13.767 1.792 2.004 (10.46)" (7.25)" (6.35)" Informall,, 2.825 (3.56)" Genresidualinfomllorw -0.102 -0.135 -0.395 (0.58) (1.07) (2.50)' Genresiduaiinformall ,, -0.953 (2.05)' Female 152 0.187 -0.107 0.172 0.063 -0.113 `.77)" (3.60)" (0.72) (137) (0.53) (1.10) Marriedrorw 024 -0.111 0.127 0.076 1.37) (0.63) (1.16) (0.55) Marriedarw 0.026 0.037 (0.41) (0.36) Age groups (reference group=Age 15-19): Age 3044 1.016 0.071 -0.035 -0.041 0.172 0.173 1.28) (1.17) (0.20) (0.39) (1.30) (1.74) Age 45-54 1.022 -0.064 -0.486 -0.221 -0.204 0.025 1.25) (0.73) (1.97)' (1.47) (1.14) (0.19) Age 55-64 430 0.458 -0.266 0.195 -0.033 0.200 1.37)" (3.60)" (0.66) (0.93) (0.14) (1.OO) Education(referencegroup=no formal schooling or prlmary Incomplete): Primaryschoolcompleted 1.887 -0.741 -0.093 -0.329 -0.371 0.025 3.75)" (11.79)" (0.45) (2.35)' (2.07)' (0.10) Secondaryschool completed ,665 -1.308 -0.955 -0.367 -0.263 0.177 !5.19)" (19.31)" (3.26)" (2.02)' (1.12) (0.45) Some tertialyeducation ,955 -1.521 -1.047 -0.444 -0.507 -0.060 6.11)" (14.07)" (2.24)' (1.97)' (1.67) (0.13) City dummies es Yes Yes Yes Yes Yes Ethnlflty and religiondummies es Yes Yes Yes Yes Yes Paretnaleducationdummies es YeS No No No No Paretnaloccupationdummies es Yes No No No No Constant 744 0.463 -1,112 -0.993 -1.224 -1.695 i.76)" (4.12)" (2.58)" (3.42)" (3.12)" (3.28)" ObSeNatlOnS 312 3144 11435 1233 760 1257 Loglikelihood -1575 -1707 -165 -552 -349 -594 PseudoRsquared 0.28 0.18 0.6 0.32 0.3 0.25 Absolutevalueof z statisticsIn parentheses * significantat 5%; ** significantat 1% 155 Initialselectionequation Dynamic selectionequationbyyear of observation by year of observation 1994 2000 1997 2000 2004 2004 Unemployed,, 3.195 1.345 0.672 (8.53)" (5.14)" (2.01)' Unemployedzm 1.349 (4.06)" Genresidualunemployed ,= 0.112 -0.029 -0.039 (0.50) (0.18) (0.17) Genresidualunemployedd,, -0.429 (2.13)' Female -0.039 -0.071 0.106 -0.207 -0.022 0.125 (0.73) (1.33) (0.78) (2.13)' (0.18) (1.37) Mamed,, -0.761 -0.338 -0.282 -0.381 (10.47)" (1.61) (1.88) (2.06)' Marned,, -0.650 -0.380 (8.86)" (2.54)' pee groups (referencegroup=Age 15-19): Age 30-44 -0.599 -0.547 -0.180 -0.238 -0.502 -0.354 (9.50)" (8.71)" (1.OO) (2.14)' (3.75)" (3.46)" Age 4554 -0.759 -0.678 0.262 -0.276 -0.737 -0.729 (6.64)" (6.08)" (1.OO) (1.52) (3.37)" (4.01)" Age 55-64 -0.915 -0.837 -0.000 -0.414 -0.728 -0.828 (4.87)" (4.53)" (0.00) (1.44) (2.23)' (2.56)' Educatlon (referencegroup=no formal schooling or primaryIncomplete): Primaryschwl completed 0.709 -0.265 0.062 0.138 0.048 -0.358 (9.60)" (2.18)' (0.28) (0.94) (0.25) (1.61) Secondaryschoolcompleted 1.187 -0.246 0.572 0.120 0.118 -0.137 (16.60)" (1.93) (2.26)' (0.70) (0.54) (0.46) Some tertiary education 0.123 0.145 0.318 -0.842 0.037 -1.014 (1.04) (1.03) (1.03) (2.90)" (0.14) (2.88)" Citydummies Yes Yes YeS Yes Yes Yes Ethnicityand regligiondummies Yes Yes Yes Yes Yes Yes Paretnaleducationdummies Yes Yes No No No No Paretnaloccupationdummies Yes Yes No No No No Constant -0.692 -0.577 -2.304 -1.197 -0.418 -0.980 (5.21)" (4.63)" (6.10)" (4.60)" (1.30) (3.74)" Observations 3312 3138 1430 1228 726 1253 Loglikelihood -1540 -1561 -218 -474 -301 -536 PseudoRsquared 10.26 0.18 10.76 0.29 0.18 0.21 Absolutevalue of z statistics Inparentheses `significant at 5%; **significantat 1% 5.58 Further highlights from these binarychoice model estimates include: e The selectivity terms used to test for the endogeneity o f initial sector choice -- shown in the third and fourth rows -- reveal clear evidence o f the endogeneity o f employment status in 2000 to employment status in 2004. However, there i s no evidence o f the endogeneity o f employment status in 1994 to subsequent employment status inthe same sector. e The effect o f schooling on transition probabilities into the public sector and into informal employment, while still strong, has nonetheless weakened. e There are no statistically significant gender or ethnicity effects on transitions, across all survey waves, which i s consistent with a less segmented labour market. 156 0 There are no age effects either in any o f the transitions except for transition to unemployment, where there have always been strong and growing negative age effects. 5.59 Results from the reduced form multinomial transition models (Annex Tables 77 to 78) are basically consistent with the foregoing. It should be noted, though, that the effects inthis caseare onprobabilitiesconditionalonspecifiedinitiallabour market states. 6. EarninPsgaps and sector choice 5.60 We now turn to our second approach to gauging changes the degree o f segmentation, which is testing for uncompensated earnings gaps between sectors and the sensitivity o f sector choice to these gaps. A structural model o f the sorting o f workers between the public andprivate sectors wage employment on the assumption that equation 4b correctly specifies the model are reportedinTables 62-64. 5.61 The model is estimated by maximum likelihood on the formal sector employee sub-sample o f the 2004 wave that was in the 2000 wave (column 4; Table 62). The earnings gap that enters the structural equation i s obtained from sector specific earnings equations that we report in the first (public sector) and the second (private sector) columns o f the same table. These are obtained by applying OLS to the specifications given by equations 5b and 6b respectively. The selectivity terms o f these columns are obtained from a binary dynamic public sector choice model as given by equation 3b, on data points of wage workers only. The selection equation i s identified inthe same way as the selection equations o f Tables 4 and 5, that is, by including family background variables in the initial selection equations but excluding them, in this case, from the earnings equations. Inthe third column we estimate a single earnings equation for both private and public sector workers so as to obtain a point estimate o f the sector wage gap, by including a public sector dummy. Since the dummy is assumed to be endogenous in this context, we include the generalized residual o f equation3b as a selectivity correction term inthis column. 5.62 We estimated this model for three waves - 1997, 2000 and 2004 (results available on request) - and find a sizeable and growing public sector wage premiumthroughout the survey decade. The implied public sector premium (at the mean o f the wage distributions) was 49 percent for 2004 as compared to 40 percent for the year 2000 wave, and 32 percent for 1997. Moreover, the coefficients o f the public sector wage premium inthe last columns suggest that the premiumhas become an increasingly more powerfbl driver o f selection into public sector employment. This premium is computed over and above possible sector differences inrates o f return to schooling and to market experience. 5.63 Next we turn to the estimation o f a structural model o f the sorting o f workers between formal sector wage employment and informal employment, based on the assumed structure given by equation 4b. Since there are practically no earnings data for informal sector wage workers we confine the analysis to data points on formal sector wage workers and own account workers reporting sales revenue which i s a proxy for earnings from self employment. The "self-employment earnings premium" i s the log 157 difference between annual business revenue and the counterfactual annual wage, which we assume to be monotonically increasing in the observed gap between earnings from self-employment and wages. The sector specific earnings equations are identified and estimated as before, as are the underlyingreduced form selection equations. Table 62: Earnings and SelectionintoPrivateandPublicSectors: 2004 (1) (2) (3) (4) Public Private Log Selection sector log sector log monthly into public monthly monthly earnings sector earnings earnings in 2004 in 2004 in 2004 in 2004 (Probit) Female 1-0.214 6.388 -0.302 -1.891 (4.49)" (5.60)" (7.21)" Marned 6.020 (0.12) Logage 4.085 3.961 -9.877 (1.89) (2.34)' (1.80) Logage squared 6.485 -0.468 1.621 (1.58) (1.95) (2.08)' Education(reference=noformal schooling): grades 1 to 10 0.157 0.254 -1.746 (1.21) (2.74)" (5.66)" preparatoryschool 0.281 0.457 -4.095 (2.01)' (4.61)" (7.49)" some tertiaryeducation 0.769 0.886 -3.021 (4.31)" (7.93)" (6.38)" Own occupation (reference-unskilled workers): Professionalor technical 0.822 0.466 6.654 (4.95)" (5.27)" (10.39)" Admin or clencal 0.334 0.286 0.967 (2.63)" (3.52)" (5.15)" Skilledproductionworker 0.482 0.426 1.777 (4.29)" (5,55)" (7.82)" Public, 0.492 (4.64)" publicsectorwage premium,, 12.174 (9.72)" Genresidualpublic,, -0.168 (2.44)' Selectivityterm 1 Selectivityterm 2 0.088 (0.62) Constant -2.532 -2.630 18.126 (0.68) (0.89) (1.85) city dummies Yes Yes Yes ethnicitydummies Yes Yes Yes religiondummies Yes Yes Yes Observations 312 603 603 R-squared 0.45 0.51 Loglikelihood -286 PseudoRsquared 0.31 158 Table 63: EarningsandSelectionintothe WageEmploymentandSelfEmployment:2004 (1) (2) (3) (4) (5) Selectioninto Selectioninto Log of annual Log annual Selectioninto self employment self employment sales revenue wages Self employment in 2000 (probit) in 2004 (probit) in 2004 in 2004 in 2004 (probit) Female 0,182 0.081 -0.129 -0.184 -0.009 (2.77)" (0.50) (0.22) (1.87) (0.05) Marned 0.163 0.060 -0.146 (2.30)' (0.34) (0.83) Log age -20.522 9.216 76.826 (1.29) (3.27)" (12.44)" Log age squared 2.781 -1.171 -10.063 (1.30) (2.96)" (12.15)" Age group (reference=l5-29): 30-44yrs. 0.204 0.106 (2.77)" (0.55) 45-54yrs. 0.261 0.417 (2.67)" (1.85) 55-64 yrs. 0.666 0.431 (5.17)" (1.42) Education (reference=noformal schooling): grades 1to 10 -0.337 -0.292 1.412 0.319 -2.853 (4.54)" (1.62) (1.93) (2.28)' (10.07)" preparatoryschool -0.566 0.049 1.871 0.508 -3.440 (6.14)" (0.22) (2.21)' (3.14)" (10.01)" some tertiary education -1.106 -0.243 2.906 0.897 -5.526 (8,31)" (0.71) (2.58)' (4.15)" (10.71)" Self-employedsw 2.501 (3.87)" Genresidselfeemw -0.188 (0.49) Self-employment earningspremium,, 2.734 (13.38)" Selectivityterm 1 0.060 (0.46) Selectivityterm 2 -0.630 (1.33) Constant -0.303 -1.395 47.048 -9.869 -147.524 (2.09)' (3.36)" (1.60) (1.99)' (12.71)" city dummies Yes Yes Yes Yes Yes ethnicltydummies yes Yes Yes Yes Yes Fatheh education Yes No No No No Mother%education Yes No No No No Fathehoccupation(referencegroup= Yes No No No No Observations 2206 509 115 212 507 R-squared 0.19 0.30 Log likelihood -1150 -191 -183 Pseudo Rsquared 10.11 0.45 0.47 Absolutevalue of z statisticsin parentheses * significantat 5%: **significant at 1% 159 5.64 A key finding that emerges from the estimation o f sector choice and earnings equations for 1997, 2000 and 2004 (only 2004 presented, other results available on request) i s that the informal sector coefficient o f relative earning i s always positive and statistically significant. This suggests that anticipated earnings gains could be a driver o f selection into the informal sector employment. This result could be concealed inpractice by the fact that schooling is negatively correlated with selection into self-employment, Noteably, however, while the less educated have greater propensity for self-employment, the more educated are more successful among the self-employed. 7. Conclusion 5.65 Based on the data from the Ethiopia Urban Household Socio Economic Survey over the decade through 2004, this chapter has assessedthe extent to which the structure o f the urban labour market has changed. Specifically we assessed changes in rates o f mobility o f workers across sectors and sectoral earnings gaps and the relationship between the two. A key motivation for the analysis was to explore whether the urban labour market has become more or less segmented over time. 5.66 Our analysis shows large andpersistent public sector wage premium, which could reflect a high degree o f segmentation. More generally, however, there are also strong indications o f weakening segmentation. First, the rate o f mobility has increased between the two pairs o f sectors since the late 1990s. Sample transitions rates have grown over across survey waves. As a result state dependence on sector choice has weakened considerably over the decade. What this means is, roughly, that the fact that we observe a randomly chosen individual in a given labour market state today is becoming a less and less powerfbl predictor o f the probability that the individual will be found in the same state at a future date. Secondly, the sensitivity o f sector choice to the premiums has increased inrecent years. Inparticular the role o f comparative earnings in selection into the informal sector has increased inrecent years. 160 References: Heckman, J. (1981). `The incidental parameters problem and the problem of initial conditions in estimating a discrete time-discrete data stochastic process,' in C.F. Manski and D. McFadden (eds.), Structural Analysis of Discrete Data with Econometric Applications, Cambridge, Mass.: MIT Press. 161 Annex Table 64: Transitions ProbabilitiesBetween1994 and 1997, age group 15-64 Mainactivity I 1 2 3 4 5 6 7 Total 1. self-employed 94.74 0.21 0.42 0.21 0.63 1.26 2.53 100 2. government worker 0.68 97.29 0 0.68 0 0.45 0.9 100 3. public enterprise worker 1.18 0 92.94 1.76 0 0 4.12 100 4. formal private sector worker 0 0.51 1.02 94.42 0 1.52 2.54 100 5. other private sector worker 0.76 0.25 0.51 0 95.67 2.8 0 100 6. unemployed 1.47 0.56 0.34 1.69 2.37 93.22 0.34 100 7. out of the labour force 1.36 0.05 0.09 0.42 0.8 1.78 95.51 100 Total 10.64 9.36 3.6 4.62 8.87 18.83 44.07 100 Table 65: Transition ProbabilitiesBetween1994 and 2000, Age 15-64 Main activity5 I 1 2 3 4 5 6 7 Total 1. self-employed 59.39 2.39 0.34 4.44 9.56 4.44 19.45 100 2. government worker 2.01 58.63 14.86 6.02 0.8 4.82 12.85 100 3. public enterprise worker 8.06 23.39 29.84 7.26 2.42 4.84 24.19 100 4. formal private sector worker 11.11 0.85 7.69 46.15 11.97 8.55 13.68 100 5. other private sector worker 11.35 2.13 2.13 14.89 40.43 14.18 14.89 100 6. unemployed 8.65 5.91 2.53 10.97 11.6 48.73 11.6 100 7. out ofthe labour force 7.89 3.44 1.93 4.45 5.12 19.04 58.14 100 Total I 13.63 9.85 4.71 8.38 8.49 20.04 34.9 I 100 Main activity5 1 2 3 4 5 6 7 Total 1. self-employed 49.26 2.96 1.48 6.3 8.15 7.78 24.07 100 2. government worker 3.49 45.35 6.98 9.69 3.49 10.47 20.54 100 3. public enterprise worker 3.15 21.26 17.32 10.24 3.15 9.45 35.43 100 4. formal private sector worker 8.91 8.91 3.96 37.62 7.92 14.85 17.82 100 5. other private sector worker 15.7 4.65 1.16 12.21 31.98 9.3 25 100 6. unemployed 8.08 9.39 2.4 16.59 10.48 30.35 22.71 100 7. out ofthe labour force 8.46 6.08 2.03 8.99 6.7 20.44 47.31 100 Total 12.5 11.15 3.33 11.58 8.81 18.33 34.31 100 162 Main activity 1 2 3 4 5 6 7 Total 1. self-employed 94.74 0.21 0.42 0.21 0.63 1.26 2.53 100 2. government worker 0.68 97.29 0 0.68 0 0.45 0.9 100 3. public enterprise worker 1.18 0 92.94 1.76 0 0 4.12 100 4. formal private sector worker 0 0.51 1.02 94.42 0 1.52 2.54 100 5. other private sector worker 0.76 0.25 0.51 0 95.67 2.8 0 100 6. unemployed 1.47 0.56 0.34 1.69 2.37 93.22 0.34 100 7. out o fthe labour force 1.36 0.05 0.09 0.42 0.8 1.78 95.51 100 Total 10.64 9.36 3.6 4.62 8.87 18.83 44.07 100 Table 68: Transition ProbabilitiesBetween1994 and 2000, Age 15-64 Main activity I 1 2 3 4 5 6 7 Total 1. self-employed 59.39 2.39 0.34 4.44 9.56 4.44 19.45 100 2. government worker 2.01 58.63 14.86 6.02 0.8 4.82 12.85 100 3. public enterprise worker 8.06 23.39 29.84 7.26 2.42 4.84 24.19 100 4. formal private sector worker 11.11 0.85 7.69 46.15 11.97 8.55 13.68 100 5. other private sector worker 11.35 2.13 2.13 14.89 40.43 14.18 14.89 100 6. unemployed 8.65 5.91 2.53 10.97 11.6 48.73 11.6 100 7. out ofthe labour force 7.89 3.44 1.93 4.45 5.12 19.04 58.14 100 Total 13.63 9.85 4.71 8.38 8.49 20.04 34.9 100 Main activity 1 2 3 4 5 6 7 Total I 1. self-employed 49.26 2.96 1.48 6.3 8.15 7.78 24.07 100 2. government worker 3.49 45.35 6.98 9.69 3.49 10.47 20.54 100 3. public enterprise worker 3.15 21.26 17.32 10.24 3.15 9.45 35.43 100 4. formal private sector worker 8.91 8.91 3.96 37.62 7.92 14.85 17.82 100 5. other private sector worker 15.7 4.65 1.16 12.21 31.98 9.3 25 100 6. unemployed 8.08 9.39 2.4 16.59 10.48 30.35 22.71 100 7. out ofthe labour force 8.46 6.08 2.03 8.99 6.7 20.44 47.31 100 Total I 12.5 11.15 3.33 11.58 8.81 18.33 34.31 I 100 163 Main activity5 1 2 3 4 5 6 7 Total 1. self-employed 95.03 0.2 0.4 0.2 0.6 1.19 2.39 100 2. government worker 0.63 97.05 0 0.84 0 0.63 0.84 100 3. public enterprise worker 1.09 0 93.44 1.64 0 0 3.83 100 4. formal private sector worker 0 0.47 0.95 94.79 0 1.42 2.37 100 5. other private sector worker 0.66 0.22 0.44 0 96.02 2.65 0 100 6. unemployed 1.39 0.53 0.32 1.6 2.35 93.48 0.32 100 7. out o fthe labour force 1.14 0.04 0.08 0.35 0.79 1.53 96.07 100 Total I 9.96 8.84 3.43 4.37 9.03 17.69 46.67 I 100 Main activity 1 2 3 4 5 6 7 Total 1. self-employed 58.28 1.03 0.69 6.21 9.31 5.52 18.97 100 2. government worker 2.41 59.44 14.86 6.83 0.8 4.42 11.24 100 3. public enterprise worker 3.77 27.36 33.02 9.43 4.72 4.72 16.98 100 4. formal private sector worker 9.76 0 8.94 44.72 10.57 13.01 13.01 100 5. otherprivate sectorworker 9.94 3.11 1.86 12.42 40.37 15.53 16.77 100 6. unemployed 7.78 5.56 2 10.22 11.56 51.33 11.56 100 7. out of the labour force 7.21 3.19 1.19 4.01 4.68 16.94 62.78 100 Total 12.44 9.28 4.15 8.07 8.33 19.52 38.2 100 Table 72: Transition Probabilitiesof 2000 RespondentsBetween2000 and 2004 age group 15-64 Main activity5 1 2 3 4 5 6 7 Total 1. self-employed 50 2 1.33 3.67 6.67 9.33 27 100 2. government worker 3.63 55.65 11.29 6.05 2.42 7.26 13.71 100 3. public enterprise worker 2.73 37.27 21.82 10 5.45 8.18 14.55 100 4. formal private sector worker 4.97 7.73 3.87 45.86 7.18 13.81 16.57 100 5. other private sector worker 8.65 8.17 1.44 14.9 26.92 16.35 23.56 100 6. unemployed 6.82 8.47 2.89 12.6 10.54 35.74 22.93 100 7. out of the labour force 5.85 3.95 1.11 6.32 4.5 16.9 61.37 100 Total 10.58 10.98 3.36 10.44 7.47 17.91 39.26 100 164 Table 73: MultinomialLogitModelof TransitionOut PublicSector Employment: 1994- 2004, BaseOutcomeis beingout ofthe LabourForce, Age group 15-65 Beginningof period 2ooo-04 1887-2W 1897-2004 1984.2004 valued fordmevarylnp (1) (2) (3) (4) (5) (6) (7) (8) (8) (10) (11) (12) (13) (14) (15) (16) explanatoryvanables Public Formal Informal unemp Public Formal Informal unemp Public Formal Informal unempPublic Formal Informal unemp sector pnvate sector loyed sector pnvate sector loyed sector pnvate sector loyed sector pnvate sector loyed Iwn e r sector w h e r sector wrker wrker sector wrker Iwrker sector wrker employee Iwrker employee employee employee (0.87) (0.85) (1.12) (0.22) (1.10) (1.02) (0.41) (0.63) (0.71) (0.67) (0.68) (0.83) (0.04) (0.52) (1.61) (0.27) Mamed 0.551 0.472 0.456 0.607 -0.015 0.558 -0.100 -0.861 -0.407 0.187 -0.290 -1.183 -0.577 -0,087 -0.688 -1.382 (1.52) (0.83) (0.78) (1.07) (0.03) (0.75) (0.14) (1.11) (1.28) (0.38) (0.51) (2.28). (1.88) (0.21) (1.24) (2.74)" Ape proups (reference pmup=Ape 1519): Age 3044 0.720 0,808 -0.480 -0,280 1.276 -0.518 2.528 0.OW 0,958 0.183 0.829 -0.077 0,514 0.380 0.682 -0.565 (1.74) (1.14) (0.79) (0.44) (2.08)' (0.62) (2.05)' (0.07) (2.43)' (0.32) (1.09) (0.15) (1.38) (0.64) (1.02) (1.18) Age 45-54 -0.337 0.256 -2.175 -2.310 -0.488 -1.823 0.838 -1.565 -0.894 -1.227 -0.586 -2.525 -1.362 -0.718 -0.813 -2.605 (0.70) (0.31) (2.55)' (2.44)' (0.80) (2.11)' (0.61) (1.42) (1.48) (1.82) (0.63) (3.05)"(3.03)" (1.00) (0.85) (3.28)" Age 55-64 -2.148 -34.031 -2.852 4.480 -3.132 -2.514 0.915 -38.113 -1.204 -0.733 -0.633 -35.81!-1.4@5 0.052 -0.783 -31.788 (2.34). (0.00) (1.90) (1.77) (3.28)" (2.28)' (0.63) (0.00) (1.87) (0.73) (0.47) (0.00) (1.81) (0.05) (0.60) (0.00) schooling or primary Incomplete): Primaryscllwl mmpleted -1.487 -2.072 -1.530 .1.742 0.528 0.630 .0.128 1.728 -0.368 -1.898 -1.318 -1.582 .0.182 -1.237 -0.780 -0.830 (2.75)" (2.50)' (2.02)' (2.21)' (1.01) (0.83) (0.11) (1.46) (0.82) (2.00)' (1.53) (2.24)'(0.47) (1.51) (1.00) (1.25) Semcdarysdool mmpleted -1.057 -0.891 -1.487 -1.447 0.482 0.567 1.341 1,018 -0.830 -0.162 -0.788 -1.877 -0.311 0.343 -0.424 -1.250 (1.83) (1.34) (1.94) (1.87) (0.91) (0.74) (1.38) (0.82) (1.38) (0.23) (1.04) (2.64)*'(0.74) (0.52) (0.58) (1.83) Ternary education -0.504 -0.876 -1.647 -0.870 0,340 0.028 0.878 .33.018 -0.388 0.218 -0.161 -2.081 -0.128 0.841 0 . W -1.388 (0.88) (1.07) (1.88) (1.18) (0.84) (0.03) (0.87) (0.00) (0.83) (0.30) (0.21) (2.67)'(0.28) (0.88) (0.00) (1.87) City Addis Ababa -0.204 0.758 -0.263 0.789 0.528 0.823 1.888 -0.223 0.150 1.118 -0.082 1.840 0.158 1.025 0.410 1.380 (0.80) (1.25) (0.48) (1.31) (1.36) (1.63) (2.21)' (0.33) (0.52) (2.03)' (0.16) (2.71)"(0.58) (2.10)' (0.81) (2.51). Ethnic or llnpuistl) groups (reference-groups 2 and 5): Ethnlcgmup 1 0.183 0.038 -0.205 1.274 0.276 -0.338 0.652 -0,118 (0.43) (0.05) (0.30) (1.45) (0.74) (0.65) (0.84) (0.21) Ethnicgmup3 -0.523 0.339 .0.326 0.888 0.237 -0.332 0.534 -0.778 (1.07) (0.45) (0.43) (0.94) (0.54) (0.51) (0.66) (1.03) Ethnicgmup4 0,265 0,281 0.071 0.830 0.152 -0.184 1.803 -1.339 (0.34) (0.21) (0.M) (0.57) (0.28) (0.21) (2.06)' (1.13) Reliplon proups: Ortrodox Chnstian -0.050 -0.182 1.307 -0.084 -0.521 -0.577 .0.004 -0.662 (0.10) (0.21) (1.15) (0.12) (0.86) (0.60) (0.00) (0.87) Moslem 0.585 41.488 2.301 1.387 0.328 0.124 2.408 -0.803 (0.65) (0.00) (1.81) (1.05) (0.38) (0.11) (1.71) (0.56) Canstant 2.M8 -0.747 0.278 .O.W 0.682 .0.537 -4.525 -0.491 0.753 -1.578 -0,821 0.188 1.222 .1.187 .1.785 1.103 (2.82)" (0.58) (0.20) (0.65) (0.96) (0.55) (2.77)" (0.36) (1.38) (1.73) (0.65) (0.22) (1.58) (1.03) (1.13) (0.95) Lcg likelihwd -383 -282 -435 -476 Pseudo Rwuared 0.08 0.15 0.10 0.12 observstions 376 344 368 404 165 Table 74: MultinomialLogitModelofTransition Out ofInformalSector Employment: 1994-2004 Beginningof penod 20002004 1997-2000 1997-2004 1884-2004 valued fortlmevawng vana (1) (2) (3) (4) (5) (8) (7) (8) (8) (10) (11) (12) (13) (14) (15) (18) Public Formal Informal unemp Public Formal Informal unemp Public Formal Informal unemp Public Formal Informal unemp sector pnvate secior byed sector private seclor byed sector private sector loyed sector pnvate sector byed I worker sector worker worker sector worker worker sector worker worker sector worker employee employee empbyee employee Female -1.681 -2.187 .1.122 -0.707 -1.177 .3.091 -1.598 -2.290 .0.407 -2.173 -0.802 -0.875 -0.405 -1.302 -0.288 -0.455 (3.33)" (4.78)- (4.74)`. (2.08). (1.74) (5.74)- (5.34)" (4.92)- (0.78) (4.a3)- (3.94)" (2.28)' (0.47) (2.07)' (0.38) (0.82) Mamed 0.094 -0.250 0.387 .i.zm -0.788 -1.293 -0.194 -1.204 -0.444 0.041 0.539 -0.757 .0.831 0.882 1.383 1.088 (0.19) (0.57) (1.57) (2.88)" (1.08) (2.53). (0.71) (2.48). (0.70) (0.08) (2.25). (1.82) (0.68) (1.05) (1.38) (1.07) Age groups (reference group-Age 15-19): Age 30-44 1.433 0.888 1.088 0.279 0.088 0.752 0.688 0.181 -0.057 .0.575 0.432 -0.214 2.417 0.882 .0.365 m a 3 (2.90)" (1.98)' (4.08)- (0.74) (0.09) (1.44) (2.19)' (0.33) (0.10) (1.22) (1.88) (0.48) (2.87)" (1.34) (0.41) (0.28) Ape 45-54 -0.831 0.164 0.885 .0.300 0.587 1.087 0 . ~ 9 2 a i 7 4 0.264 -0.528 0.180 -0.378 -3i.wa -0.358 -3.448 -1.108 (0.55) (0.27) (2.79)" (0.49) ( O M ) (1.62) (0.80) (0.27) (0.34) (0.85) (0.80) (0.80) (0.00) (0.38) (2.14)' (0.91) Ape 55-64 0.883 0.418 0.464 -32.881 -33.971 -0.736 0,108 -34.308 -32.431 -0.559 0 . 1 ~ -0.482 -32.841 -2.310 ~ -35.489 a.018 (0.87) (0.61) (1.29) (0.00) (0.00) (0.64) (0.27) (0.00) (0.00) (0.77) (0.50) (0.71) (0.00) (1.42) (0.00) (0.00) Primaryschwlwmpleted 0.495 0.858 0.254 0.564 o m a 1.283 -0.381 0.041 1.388 0.8w 0 . m 0.423 2.258 1.185 -0.126 0.838 (0.84) (2.09)' (0.98) (1.48) (0.85) (2.83)- (1.24) (0.09) (2.28)' (1.88) (0.52) (1.03) (1.58) (1.32) (0.12) (0.52) Secondary school wmpletec 2.157 1.202 0.483 1.301 1.545 0.543 -0.254 0.073 2.537 1.817 0.547 1.503 0,801 -0.272 -1.800 0.301 (3.88)" (2.28)' (1.33) (2.98Y (1.87) * (om) (0.48) (0.10) (3.28)" (2.58)" (1.15) (2.50). (0.43) (0.31) (1.54) (0.28) Tertiaryeducation 1.418 2.224 1 127 0.890 -0.387 -0.880 33.569 -1.232 2.148 0.875 -0.500 -33.537 0.873 1.029 -0.687 -0.097 (1.05) (2.28)' (1.28) (0.75) (0.00) (0.00) (0.00) (0.00) (1.81) (0.52) (0.51) (0.00) (0.48) (1.00) (0.54) (0.06) CW Addk Ababa -0.332 -0.705 .0.925 -0.188 0.053 0.115 0.182 0.542 0.081 .o.080 -0.a5i 0.266 .os48 -0.489 1.831 0.352 (0.82) (1.58) (3.71)** (0.48) (0.08) (0.25) (0.74) (1.16) (0.13) (0.18) (2.88)'. (0.81) (0.54) (0.83) (1.42) (0.34) Ethnlc or llngulml)groups (reference=groups 2 and 5): Ethnlcgmup 1 .0.829 0.088 .0.391 .0.472 -0.151 0.237 0.797 -0.380 (1.18) (0.17) (1.44) (1.12) (0.18) (0.31) (0.79) (0.43) Ethnlcgmup3 0.424 1.238 0.088 0.492 -2.098 0.127 0.351 -0.340 Beginningof penod 2000.2004 1997-2000 1997-2004 1994-2004 valuedfortimevaryinp vana(1) (2) (3) (4) (5) (a) (7) (8) (8) (10) (11) (12) (13) (14) (15) (16) Public Formal Informal unemp Public Formal Informal unemp- Publlc Formal Informal unemp Public Formal Informal unemp. seclor pnvate sector byed seclor pnvate sector byed sector pnvate sector loyed seclor pnvate sector byed I worker sector worker worker sector worker worker secior worker worker sector worker employee employee employee employee Ethnicpmup 4 -1.311 0.383 -0.252 -0.443 -1.208 -0.419 1.380 -0.758 (1.15) (0.50) (0.83) (0.85) (0.73) (0.38) (1.03) (0.53) Rellglongroups: OrthodoxChnstlan 1.113 -0.533 1.247 2.075 -2.284 0.251 0.408 -0.285 (1.27) (0.92) (2.64)" (1.82) (1.58) (0.21) (0.28) (0.24) Moslem -0.177 .1.195 0.553 1.352 .1.784 -0.474 -0.378 -0.008 (0.17) (1.84) (1.07) (1.18) (0.85) (0.28) (0.18) (0.01) Constant 2.643 .0.583 .0.081 -2.350 -1.405 .0.078 1.887 0.533 -2.882 -0.858 0.921 -1.001 0.858 0.120 -2.142 .OS28 (2.28)' (0.72) (0.11) (1.94) (1.50) (0.12) (4.74)" (0.88) (3.30)" (1.25) (3.03)" (1.87) (0.27) (0.07) (0.90) (0.25) Log likelihood .a71 -487 -817 -144 Pseudo Rsquared 0.14 0.12 0.10 0.18 Observations 588 498 557 i i a 166 6. A SIMULATION PERSPECTIVE ON ETHIOPIA'S LABOURMARKET UPTO 201552 I Inorderto developaquantitativesense ofthe benefitsofreducinglabourmarketrestrictions aset of labour-focusedsimulations has been designed with MAMS (Maquette for h4DG Simulations) for the period 2005-2015. M A M S i s a dynamic CGE (Computable General Equilibrium)model andrelative to other CGE models it offers a detailedtreatment of government services and of the tax system and other sources of government receipts. The labour market is disaggregatedinto three segments on the basisof educationalachievement. The simulations show that the elimination of barriers to the efficient functioning of the labour market, all else equal, could lead in 2015 to national output that i s 2.4 percent higher than inthe base case, with an increaseinaverage welfare of about 2 percent. A conservative estimate of the poverty impact i s a 3 percentdeclinewithrespectto the base. Removing suchbarriers would also acceleratethe transition out of agriculture, with the urban economy becoming more dynamic and the exDansion ofall the non-governmenturban sectors. 1. Introduction 6.1 As noted inthe preceding chapters, Ethiopian labour markets are characterized by a strong segmentation between rural and urban labour markets (with legal and defacto restrictions on rural-urban migration) and, within urban areas, an institutional setup that undulyrestricts labour movements between different employments. Inqualitative terms, these characteristics lead to an inefficient allocation o f resources, reducing overall welfare. 6.2 From a short-run perspective, the effects o f this situation relative to a labour market where workers are free to move between regions and production activities are obvious: labour does not leave agriculture, the predominant rural sector, raising its employment and production but depressing the wages o f its workers. In urban areas, wages and prices are higher, while production i s reduced in those sectors, primarily private services that would have absorbed the bulk o f additional unskilled labour migrants from rural areas. Interms o f distribution, limited rural-urban migration harms those who are trapped inrural areas while benefiting those in urban areas who would see their wages decline ifmigration restrictions were lifted. Via upstream (supply-side) and 52This chapter has beencontributedbyHans LofgrenandRahimaisaAbdulabothwith the WorldBank. 167 downstream (demand-side) linkages, other parts o f the economy are further affected by these restrictions. In urban areas, restrictions on labour movements have similar consequences, segmenting the labour market and generating wage differentials that are larger than those that would prevail iflabour was allowed to move freely, with a negative impact on overall welfare. 6.3 Inorder to determine whether it is apolicypriorityto addressthese restrictions to mobility, it is necessary to develop a sense o f whether the effects discussed in the preceding paragraph are large or small. The above discussion was also looking at the labour market from a relatively short-run perspective - it i s quite possible that, from a medium- to long-run perspective, everyone (also the workers who today are shielded from inflows o f more labour from rural areas or other sectors) would gain as the dynamic effects o f a more integrated and flexible labour market have a positive impact growth and incomes throughout the economy. 2. The simulations 6.4 In order to develop a quantitative, dynamic sense of the importance of reduced labour market restrictions and how they interact with complementary policies, we designed a set o f labour-focused simulations with M A M S (Maquette for MDG Simulations) for the period 2005-2015. M A M S i s a dynamic CGE (Computable General Equilibrium) model designed for economywide analysis o f the impact o f policies and external shocks on poverty and human development (HD) in developing countries (See Appendix for a description o fMAMS).53 Inthis Ethiopian version o fMAMS, production i s divided into 11 activities. The Government is divided according to function into eight activities - four in education (first primary, second primary, secondary, and tertiary), health, water-sanitation, other public infrastructure, and other government. The rest o f the economy is dividedinto three production activities (agriculture, industry, and private services). This classification reflects data limitations and the fact that the key purpose behind the model design was to analyze the impact o f government policy and other economic developments on the evolution o f selected MDGindicators. The labour market i s disaggregated into three types on the basis o f educational achievement - workers with less than completed secondary education (the bulk o f the labour force), workers with secondary education but not completed tertiary, and workers with completed tertiary education. The labour market i s linked to the educational system. The part o f the population that is enrolled in school is not in the labour market. Among those not enrolled in schools, a fixed share enters the labour market at the time when their cohort would have entered the second primary cycle. This group consists o f those who either never attended school or had left school before starting the second primarycycle. Others leave the educational system at a higher level, with their educational achievement at the time o f their exit determining their labour type. 53 For a technical documentation, see Lofgren and Diaz-Bonilla (2006a). Inan appendix to this report, we provide a brief, less technical model description. For other applications o f MAMSto Ethiopia, see Lofgren and Diaz-Bonilla (2006b) and Sundberg and Lofgren (2006). 168 6.5 Inthe labour market, the default assumption (from which we will inpart deviate; see below) is that the workers are free to move between the different production sectors. For any given labour type, the model captures observed relative wage differences between employments in different production activities for any given labour type. For example, workers at the lowest educational level (referred to as unskilled) in agriculture earn considerably less than those o f this type who work in other activities. For each labour type, a flexible economywide wage clears the market in each time period, scaling wages in the different production activities up or down in a manner that retains relative wage difference^.'^ For each labour type, the unemployment rate i s fixed (a fixed share o f those in the labour force are not employed). Technically, it is possible to endogenize the unemployment rate. However, given that our understanding o f the determinants o f unemployment is limited, we opted for the simpler solution o f keeping these rates fixed. These means that we are implicitly assuming that, in the context o f the current set o f simulations, the unemployment rates do not change. 6.6 The different simulations are defined inTable 75. The results are summarized in Tables 76 (macro effects and MDG indicators) and 77 (wages, wage gaps and employment). Note that in Table 80, the effects on monetary poverty are to be considered as conservative estimates as they are based on poverty elasticity to changes in average consumption. As such they are likely to underestimate changes inpoverty when policies have a pro-poor poverty impact, affecting mostly those at the bottom o f the distribution. The simulations were designed with an eye toward enhancing our understanding the effects o f a more efficient allocation o f labour across the economy and how those would affect other policies. Table75: SimulationDefinitions 6.7 As shown in Table 76, the BASE simulation is our benchmark for comparisons. It assumes growth in most macro aggregates around 5 percent, i.e. close to historical trends. It shows moderate to significant improvement in the different MDGs, but not 54Technically, the wage paid to a worker o f typef inproduction activity a, WFAaf,, i s defined as follows: WFAfa= wfdisq,. WF, where wfdisq, is a fixed wage distortion factor and WF, is the market-clearing economywide wage variable. 169 enough to reach them by 2015. Employment growth in different sectors i s such that the 2005 wage gaps between different activities (due to segmentation) are retained. 6.8 The simulation MIG5O addresses some o f the segmentation in urban labour markets, as it simulates the effects o f reducing the wage gap for the labour type with the least education (e secondary) between agriculture and private services by 50 percent (depending on the simulation), gradually during the three-year period 2006-2008. 55 Given data limitations (no full disaggregation o f production into rural and urban), agriculture (which dominates the rural economy) i s our proxy for rural whereas the other (private) service sectors (all private services except for health) i s our proxy for the urban informal sector. MIGDESSO simulates a more liberal setting where the cut in 50 percent inwage gaps is generalized to all labour types and all sectors -it therefore simulates the effects o f removing also other sources o f segmentation inurban areas. MIGDES100 i s in a similar vein, with a generalized cut inwage gaps for all labour types and all sectors by 100percent, i.e. everybody with a given educational achievement receives the same wage irrespective o f sector o f employment. The purpose o f this simulation is to explore the upper limit on gains from more efficient labour allocation. 6.9 The interactions between labour market segmentation and other policies are explored inthe remaining three simulations. More specifically they explore the effects o f the educational expansion (here assumed to be a doubling o f growth in all educational sectors) which i s an important policy priority and the functioning o f the labour market. As inthe previous simulations direct tax changes are the balancing item. Tables 76 and 77 summarize the results o fthese different simulations. Table 76: Simulatia Simuiatlons' Indicator 2005 base mig50 migdes50 mlgdesIO0 mig5Oe mlgdes5Oe migderlooe bn birr %annual growth 2006-2015 Absorption 104.60 4.73 4.78 4.84 4.91 4.68 4.73 4.80 Privateconsumption 67.71 4.85 4.90 4.96 5.01 4.31 4.37 4.42 PrivateInvestment 14.49 5.00 5.13 5.28 5.49 4.29 4.44 4.66 Governmentconsumptlon 12.04 5.00 5.00 5.00 5.00 6.41 6.41 6.41 Governmentinvestment 10.36 3.11 3.11 3.11 3.11 5.38 5.38 5.38 Exports 23.14 5.46 5.65 5.73 5.95 5.38 5.46 5.65 imports 45.88 4.68 4.79 4.83 4.95 4.65 4.68 4.79 Realexchange rate 100.00 0.06 0.08 0.22 0.29 0.11 0.25 0.32 EV" 28.96 29.57 30.26 30.90 22.41 23.08 23.69 Real GDP at total 75.54 4.99 5.07 5.16 5.29 4.94 5.03 5.15 factor cost agriculture 38.09 4.39 4.10 3.99 3.55 3.80 3.69 3.24 Industry 7.05 5.10 5.25 5.68 6.18 5.08 5.48 6.00 privatesewlces 25.96 5.80 6.33 6.59 7.28 6.02 6.28 6.95 governmentsewices 4.45 5.00 5.00 5.00 5.00 7.29 7.29 7.29 rate- rate- In 2015 MDG I(headcountpoverty 0.360 0.256 0.254 0.251 0.249 0.279 0.277 0.275 indicators 2 (1st cycle primarycompletion) 0.335 0.622 0.624 0.617 0.614 0.883 0.879 0.877 4 (Under4 mortality) 0.127 0,112 0.112 0.111 0.110 0,114 0,113 0.111 5 (Maternalmortality) 0.871 0.695 0.692 0.680 0.663 0.712 0.701 0.685 7a (Improvedwater access) 0.400 0.430 0.430 0.430 0.430 0.430 0.430 0.430 7b(improved sanitation access) 0.120 0.166 0.166 0.166 0.166 0.164 0.164 0.164 55A technical point: if there were no endogenous wage change, the cut inthe wage gap would be exactly 50%. However, this i s not exactly the case when wages change. 170 6.10 MIG50: Increased mobility o f low skilled from agriculture into urban services leads to very small (but positive) changes at the macro level while the main changes are in the labour market and at the sector level. As employment grows more slowly in agriculture than in BASE, GDP growth in agriculture declines from 4.4 percent to 4.1 percent. At the same time there is higher growth in other, non-government sectors, especially private services, where the relative wage declines. Improvements intotal GDP could be higher if there are productivity o f labour differentials between agriculture and the urban sectors, and if there are dynamic effects, like increased household incomes, savings, investment, and capital stock growth. The Equivalent Variation in 2015 i s 0.61 percent higher than for base, i.e. a 0.61 percent improvement inwelfare at end o f period relative to BASE. As already mentioned, the impact on the headcount is likely to be underestimated given that the simulated change inrelative wages i s infavor o f those with the lowest wage. 6.11 MIGDESSO: Removing more sources o f segmentation results in still small but more positive (roughly twice as large) changes at the macro level compared to MIGSO. Also the improvement in the Equivalent Variation are double those in MIGSO, as aggregate welfare is 1.3 percent higher in 2015 than in 2015 for BASE. On the top o f reduction in agricultural growth due to the reallocation o f labour already present in MIGSO, employment growth in agriculture is lowered further as the urban economy becomes more dynamic with expanded employment inall non-government urban sectors. Inparallel with slower employment growth inagriculture, also agricultural GDP growth declines to 4 percent. Interms o f employment composition, the Government switches to more skilled labour (since reducing wage differentials implies raising the wages o f the least skilled and lowering those o f the more skilled). There i s also an overall decline in government employment since the more skilled are more productive. 6.12 MIGDESlOO: Removing entirely wage differentials leads to larger positive changes both in terms o f growth and aggregate welfare (1.9 percent higher in 2015 than in BASE). Also in terms o f employment growth the trends for MIGDESSO are intensified, including the negative impact on employment growth for agriculture. The reduction in segmentation therefore increases the contribution o f the urbanprivate sector to production and employment. The improved efficiency inlabour allocation also leads to a higher rate o f real GDP growth because o f higher private consumption and investment. Private consumption increases due to the higher wages in the agricultural sector and due to increased employment in the urban sector. Therefore, the gains o f both the rural and urban sectors contribute to improvements in welfare, total productivity and poverty reduction. 6.13 MIGSOe, MIGDESSOe and MIGDESlOOe: The outcomes are dominated by the large public investment in education which results in shifts o f final demand from non- government sectors to government consumption and investment. This leads to a minimal decline in real GDP growth over the period up to 2015. Increased final government demands i s at the expense o f private final demands, especially spending on investment, the payoff from which appears with a considerable lag. As a result o f less private consumption, the Equivalent Variation in 2015 i s lower than for BASE and monetary poverty worsens (non-monetary dimensions o f poverty such as education would 171 improve). The only condition under which this large public investment could end up not EV, consumption, and poverty is if foreign grants provided the financing (permitting the country to import more, export less, and maintain reasonable growth not only for government but also for private final demands). Note that across the simulations, as the efficiency gains are larger, the negative effects on the Equivalent Variation and poverty are more contained. However, only slightly: for the most efficient simulation, MIGDESlOOe, EV welfare in2015 i s 5.3 percent lower than for BASE in2015. Interms o f employment growth, the patterns are similar to MIG5O but with one major change - growth in government service employment is much stronger, leaving less room for employment growth elsewhere. Still, outside the Government, agriculture declines relative to other non-government sectors whereas private services gain. The relatively lower growth in the production and employment o f the private sector observed in these scenarios vis-&vis the previous ones (in which education is not scaled up) explain the relatively lower overall productivity and welfare gains. The expansion in the Government sector crowds out to a certain extent the private sector and limits the gains from reduced segmentation. Furthermore, increasing educational services attracts potential labour participants into the educational system, reducing the total supply o f labour and output. These observations however reflect the short-run dimension o f the policy change. The eventual gains from human capital development, i.e. higher skilled and more productive labour force and therefore higher overall productivity and welfare, are likely to be reaped inthe longer run. 3. Conclusions 6.14 The elimination o f barriers to the efficient functioning o f the labour market, all else equal, could lead in2015 to national output that i s 2.4 percent higher than inthe base case, with an increase in average welfare o f about 2 percent. A conservative estimate o f the poverty impact is a 3 percent decline with respect to the base. Removing such barriers would also accelerate the transition out o f agriculture, with the urban economy becoming more dynamic andthe expansion o f all the non-government urban sectors. 6.15 An interesting aspect o f these simulations is that removing barriers in the labour market would result in increasing returns to secondary and tertiary education outside o f agriculture. This i s good news in the medium term given plans o f accelerating investments in education, suggesting that new cohorts o f better educated workers would face more favorable labour market prospects. 6.16 The efficiency gains from labour market reform could help generate resources to face some o f the development challenges o f Ethiopia, such as education expansion. The labour market effects o f improved human capital outcomes would require additional modeling work, as they are likely to affect the labour markets inthe medium to the longer run,beyondthe 2015 timehorizon adopted inthis analysis. 172 Table 77: Simulation results:Wages, Wage Gaps, andEmployment Slmulatl 2005 base mlg5O mlgdes5O mlgdes100 000 blrr Kannual growl Realwage agrlculture 1.006 3.95 4.56 4.78 5.70 4.79 4.99 5.90 c secondary Industry 2.262 3.95 3.77 1.14 -2.51 3.96 1.34 -2.32 prlvatesewIces 1.321 3.95 3.17 3.36 2.66 3.36 3.57 3.08 governmentservices 0.919 3.95 3.77 5.31 6.66 3.96 5.51 6.69 total 1.171 3.92 3.96 4.04 4.13 4.17 4.25 4.33 Realwage agrlculture 1.754 3.69 3.67 5.35 6.99 3.32 5.02 6.68 secondary Industry 3.689 3.69 3.67 1.39 -0.68 3.32 1.06 -0.97 prlvateservices 3.226 3.71 3.66 2.04 0.66 3.32 1.72 0.37 governmentsewlces 4.471 3.51 3.50 0.40 -2.56 4.56 0.96 -2.65 total 2.699 4.15 4.15 3.25 2.46 4.72 3.39 2.18 Realwage agrlculture 10.443 6.83 6.67 6.61 7.12 6.92 6.67 7.19 tertlary Industry ' 20.593 6.83 6.67 3.73 0.09 6.92 3.79 0.15 prlvatesewlces 6.916 6.92 6.93 8.03 8.62 6.94 8.07 8.69 governmentsewlces 12.770 7.02 7.06 5.65 4.99 7.75 6.24 5.06 total 11.054 7.15 7.17 6.77 6.51 7.70 7.06 6.58 Ratlo toavg Ratioto averagewage for la type (byedu) In 2015 Wage gap agriculture 0.861 0.66 0.91 0.92 1.oo 0.91 0.92 1.oo c secondary Industry 1.932 1.937 1.896 1.456 1.000 1.sa 1.46 1.oo privateservices 1.128 1.131 1.045 1.057 1.000 1.05 1.06 1.oo governmentsewlces 0.765 0.767 0.770 0.666 1.000 0.77 0.89 1.oo Wage gap agrlculture 0.650 0.622 0.620 0.795 1.000 0.57 0.76 1.oo secondary Industry 1.367 1.306 1.304 1.140 1.000 1.19 1.os 1.oo privatesewlces 1.196 1.147 1.142 1.063 1.000 1.05 1.02 1.oo governmentservlces 1.656 1.556 1.555 1.253 1.000 1.63 1.31 1.oo Wage gap agrlculture 0.945 0.917 0.919 0.946 1.000 0.68 0.93 1.oo tertlary Industry 1.663 1.609 1.612 1.396 1.000 1.73 1.37 1.oo prlvatesewlces 0.607 0.790 0.789 0.906 1.000 0.75 0.69 1.oo governmentservlces 1.155 1.142 1.145 1.060 1.000 1.16 1.07 1.oo mn. workers '%annual grow 006.2015 Employment agrlculture 15.619 1.01 0.45 0.34 -0.41 -0.03 -0.13 -0.06 c secondary Industry 1.464 0.95 1.17 2.34 4.067 0.92 2.06 3.62 I prlvateservices 8.865 1.66 2.46 2.61 3.41 2.10 2.23 3.02 governmentservlces 1.525 4.84 4.90 4.27 3.85 5.20 4.56 4.17 ItOtal 27.474 1.47 1.46 1.47 1.48 1.os 1.10 1.10 Employment agriculture 1.337 1.16 1.06 -0.04 -1.25 0.96 -0.16 -1.38 secondary Industry 0.134 1.13 1.24 2.16 2.74 1.37 2.27 2.62 prlvatesewlces 0.561 1.84 2.14 3.51 4.92 2.15 3.49 4.87 governmentservlces 0.471 4.69 4.65 5.32 5.64 6.76 7.17 7.47 total 2.5032.12 2.12 2.10 2.09 2.57 2.54 2.52 Employment agrlculture 0.064 -0.90 -1.07 -0.99 -1.33 -1.43 -1.37 -1.71 tertlary Industry 0.007 -0.96 -0.69 0.55 2.19 -1.03 0.39 2.02 prlvatesewlces 0.071 -0.23 0.01 -0.71 -1.oo -0.27 -1.01 -1.32 I governmentsewlces 0.073 3.33 3.26 3.65 3.94 5.12 5.47 5.73 0.216 0.92 0.92 0.94 0.95 1.57 1.59 1.61 ltohl 17.020 1.02 0.49 0.31 -0.47 0.04 -0.14 -0.92 1.605 0.96 1.17 2.32 3.97 0.95 2.09 3.73 9.497 1.66 2.45 2.64 3.46 2.09 2.29 3.11 2.070 4.60 4.64 4.50 4.26 5.57 5.26 5.06 30.192 1.52 1.52 1.52 1.52 1.22 1.23 1.23 *The simulations are defined in Table i Apoendix 1 Descriotionof MAMS - MAMS is an economy-wide simulation model that the World Bank has developed to analyze development strategies in different countries with an emphasis on the determination o f MDG outcomes. Many o f the policies and foreign aid flows targeting MDGshave strong effects throughout the economy that feed back on the MDGindicators 173 through markets for labour, goods, services and foreign exchange. Therefore, economy wide analysis o f MDGstrategies i s a necessary complement to sectoral studies. MAMS integrates a standard (recursive) dynamic general equilibrium (GE) model with an additional MDG module that links specific MDG-related interventions to MDG achievements. Output and factor employment decisions are drivenby profit-maximizing producer behavior. Flexible wages or rents clear the markets for the different production factors (including labour and capital except for government capital). In most applications, the model has a relatively detailed treatment o f government activities, which are classified by function into: education (typically disaggregated into three-four cycles), one or more health services, water-sanitation, (other) infrastructure, and other government. Like other production activities, these government activities use production factors, and intermediate inputs to produce an activity-specific output (in the case o f the Government, different types o f services). The factors o f production typically include three types o f labour (those with less than secondary school, secondary school, and more than secondary school education), public capital stocks by government activity (function), and a private capital stock. The Government finances its activities from domestic taxes, domestic borrowing, and foreign aid (borrowing and grants). Provision o f education, health, and water-sanitation services contribute directly to the MDGs. Growth in the stock o f public infrastructure capital (including roads, energy and irrigation) contribute to overall growth by adding to the productivity o f other production activities. In health, the model accounts for the fact that households, NGOs and the private sector cover part o f spending, service provision and investments. The rest o f the economy (representing agriculture, industry, and other private services) may be very aggregate or highly disaggregated, depending on the analytical context and data availability. Its output i s exported and sold domestically, competing with imports. Apart from the Government, the institutions o f the economy include one or more households, the rest o fthe world and, optionally, NGOs. The model i s intended to capture key interactions between the pursuit o f the MDGs and economic evolution. To keep it relatively simple, it does not cover all MDGs. It focuses on the ones with the greatest cost and the greatest interaction with the rest o f the economy: universal primary school completion (MDG 2; measured by the net primary completion rate), reduced under-five and maternal mortality rates (MDGs 4 and 5), halting and reducing the incidence o f HIV/AIDS (part o f MDG 6), and increased access to improved water sources and sanitation (part o f MDG 7). We also address achievements interms o fpoverty reduction (MDG 1). These different MDGs are covered in an additional set o f functions that link the level o f each MDG indicator to a set o f determinants. The determinants include the delivery o f relevant services (in education, health, and water-sanitation) and other indicators, also allowing for the presence o f synergies between MDGs, i.e. the fact that achievements in terms o f one MDG can have an impact on other MDGs. Outside education, service delivery i s expressed relative to the size o f the population. Ineducation, the model tracks base-year stocks o f students and new entrants through the different cycles. Ineach year, students will successfully complete their grade, repeat it, or drop out o f their cycle. 174 Student performance depends on educational quality (quantity o f services per student), household welfare (measure by per-capita household consumption), the level o f public infrastructure, wage incentives (expressed as the ratio betweenthe wages for labour at the next higher and current levels o f education for the student in question; an indicator o f payoff from continued education), and health status (proxied by MDG 4). The achievement o fMDG2 requires that (very close to) all students inthe relevant age cohort enter the primary cycle and successfully complete each year within this cycle. The functions for education and the other MDGs have been calibrated to assure that, under base-year conditions, base-year performance is replicated and that, under a set of other conditions identifiedby sector studies, the target i s fully achieved. 175 7. MONITORINGLABOURMARKETDEVELOPMENTS: PRIORITIES TO OPERATIONALIZEAN EFFECTIVE LABOURMARKET INFORMATION SYSTEM56 I Hiphlivhts I Policy actions and developments inthe labour market, particularly inurban areas for which data collection i s more frequent, need to be part o f the PASDEP monitoring and evaluation framework. Indeed, including employment creation as a strategic priority has been an important innovation and an opportunity to enhance the value and policy impact o f labour market information. Monitoring and reporting on PASDEP should not substitute but build upon an integrated Labour Market Information System(LMIS). This chapter provides a discussion of both the role o f an LMIS within the overall monitoring of PASDEP and at the operational steps of setting up an LMIS. A central theme i s the need for collabouration among all relevant stakeholders to build on the strength o f existing efforts and systematize them within the framework offered byPASDEP monitoringmatrix. From an operational point o f view, an incremental approach based on prioritizing informational needs i s suggested. LMIS-Ethiopia Icould cover labour market information needed for policy design, implementation, and monitoring - while LMIS-Ethiopia I1 could cover information neededfor the design of active labour market interventions, such asjob-matching activities. This division responds to the different users and indicators on which they would focus, as well as the different nature o f the indicators to be monitored which might require different frequency of updating, different dissemination strategies, possibly even different hosts for different parts o f the monitoring system etc. 1.Introduction 7.1 A labour market information system (LMIS) like the other sectoral monitoring systems i s an integral part of monitoring PASDEP, particularly given the new emphasis on labour markets as a strategic priority. Timely information about labour market developments i s necessary to respond to ongoing trends as well as to develop and implementnew strategies. 56 This chapter draws o n the COW1 (2005) report "LMIS-Ethiopia. Towards a new Labour Market Information System" produced by Peter Madsen and B o Rosendhal, as well as o n discussions at the WB- ILO Labour Market workshop held in December 8 and 9, 2005 at the Sheraton Addis. More details, particularly o n the operationalization o f the L M I S are to be found in the background paper, available on request. 176 7.2 This chapter looks at the informational needs an LMIS system needs to fulfill within the context o f monitoring PASDEP, in light o f the dialogue which has accompanied the process o f producing this study. It then provides a quick overview o f the more operational side o f setting up an effective LMIS system, and its possible developments over time. The two issues are clearly linked as monitoring and reporting on labour market developments for PASDEP should not substitute but build upon an integrated Labour Market Information System. These themes will however be discussed separately for ease o f exposition. 7.3 A common thread linking these themes is that collabouration o f all the stakeholders interested in labour market policy both within the Government and outside will be needed to rationalize the significant efforts in terms o f data collection and reporting that Ethiopia undertakes, and to maximize the gains in feeding into effective policy making. Such collabouration will have to take place at various levels all along the spectrum from data production to data analysis and reporting and evaluation o f specific program outcomes. 2. Labour market monitoringas a sectoralinputto PASDEPmonitoring 7.4 Highlighting employment creation as a strategic priority for PASDEP is an important innovation and an opportunity to enhance the value andpolicy impact o f labour market information. Policy actions and developments inthe labour markets, particularly inurbanareas for which data collection is more frequent,needto bepart ofthe PASDEP monitoring and evaluation framework. 7.5 The design and updating of PASDEP's monitoring matrix and the identification o f target indicators provide the starting point. It i s important that the indicators chosen for the monitoring are clear and o f easy interpretation, and that they can be disaggregated by age group and gender to focus on priority groups such as youth and women. Indicators should also be prioritized, in light o f the international experience pointing to the need to keep the number o f indicators manageable. 7.6 The consultation process around this study has highlighted a variety o f opinions on appropriate indicators for monitoring the labour market policy side o f PASDEP. Internationalevidence suggests two useful criteria to consider inidentifying indicators: 0 The indicators chosen should offer a clear sense o f progress. Current practices o f monitoring employment for the age group 10+ might not fulfill this criterion, as enrolment is increasing. Complementing this indicator with an indicator for the age group 15+ would help disentangling the effects o f increased enrollments from those o f labour market developments such asjob creation.57 0 The indicators chosen should be clearly linked to policy. For example, while unemployment figures can steal the headlines, rising inactivity may be an equally important policy concern that risks neglect. 57Monitoring separately child labour (as an indicator to reduce) and employment rates will help in distinguishing these different policy outcomes. 177 7.7 Important indicators that fulfill these criteria are the employment rates for the individuals 15 and over and the wage levels o f the employed. Reporting regularly on these indicators, in total and disaggregating by gender, age group (focusing inparticular on youth, likely to display idiosyncratic behaviour), skills, andregion (including min and max) would provide a good summary o fhow the country is usingits productive potential. At the same time, giventhe limitedspread o fwage work, further efforts at understanding and measuringbetter the earnings o f the self-employed and what drives them should also be undertaken. 7.8 Reporting on even such a minimal set o f indicators could prove to be a challenge. As part o f ongoing efforts at the CSA to develop a strategic plan for its survey activities and to linkthem more closely with the PASDEP, monitoring issues o f data coverage and schedule o f collection would need to be addressed. Now that the CSA i s planning to undertake an annual UBEUS type o f survey, urbanareas could be regularly monitored. T he challenge is to strengthen the quality o f wage data (whose inclusion inthe survey has been a major step forward), o f considering urban domains capable o f providing more details on different types o f urban areas (larger cities, smaller, secondary cities and for rural towns), and o f exploring ways o f capturing earnings inthe informal sector. 7.9 Data coverage for regular monitoring in rural areas remains more problematic. Different alternatives ranging from the inclusion o f an all purpose nationally representative survey in the statistical survey plan to the customization o f existing surveys (e.g. agricultural sample survey) to fully include labour market aspects in additional modules/questions might be considered. Learning opportunities such as the ongoing Rural I C A could offer a platform to start a discussion on the specificities o f rural labour markets, on the type o f instruments which might better capture them and on the appropriate tools for analysis. 7.10 Similarly, the ongoing work for the new I C A could offer an opportunity for discussion, collabouration and capacity building on firm level surveys, building on international experience and on the specificities o f the Ethiopian statistical program and its multiple firm level surveys. 7.11 Overall, it i s important to ensure that all the various elements o f labour market monitoring aim to reflect closely the informational and analytical needs raised by PASDEP. This should be responsive to emerging areas for which demands for more detailed information are arising, such as those related to the scaling up o f existing government programs either directly or indirectly supporting firm development (through the provision o f infrastructure, for example), and those relatedto the new urbanagenda. 7.12 For labour market indicators to fulfill its role in informing the monitoring o f PASDEP as well as providing timely and appropriate information to all the actors interested in labour market developments a functional Labour Market Information System needs to be inplace. At present, such a system does not exist inEthiopia, so that different actors find ad hoc solutions for their own informational needs. 178 3. Stakeholders andtheir informationalneeds 7.13 An effective LMIS system needs to be shaped by the informational needs o f the various users, and a clear prioritization of these needs. The main actors involved with using andproducing labour market information are: 0 MOLSA - Planning and Programming department, which normally represents MOLSA in interdepartmental committees and working groups and has the responsibility to ensure that labour market issues are taken appropriately into account. MOLSA - Department o f Labour Manpower Research and Statistics Team. This - team which is an office within the Department o f Labour - is the main analytical unit inMOLSA. It uses labour market data provided by the CSA, and it collects and uses data on unemployment, vacancies etc. from the 11 BOLSAS and also some from the 49 public employment service (PES) offices. The data and analyses are then disseminated to other users - primarily via the annual labour statistics bulletin inhard copy. This bulletin focuses on the urban labour market, although to some extent it addresses the differences across rural-urban labour markets. 0 BOLSA -- the 11 bureaus o f labour and social affairs are primarily concerned with the use of information for designing and implementing active labour market policies such as vocational guidance to ensure job-matching etc. To that effect, they need timely information about demand and supply patterns. They are also the centre for information about work accidents, work permits to work abroad, andthe number o fdisputes betweenemployers andemployees. Regional PES offices. The 49 public employment service offices have similar functions to the BOLSAs by which they are guided to a large extent. Their needs are mostly for information to ensure efficient job matching. A joint plan by MOLSA and the ILO aims to improve their hnctions and include their coverage o fvocational training. 0 The Ethiopian Employers Federation which aims to promote the interests o f employers, and developing and maintaining good relations among social partners as well as to promote social progress by creating new jobs. It presently lacks information about labour market developments - in particular specific demand- side information such as vacancies, but also more general analysis, though it uses some of the data produced by CSA and by MOLSA. It plans to produce a monthly bulletinfor its members. The Confederation o f Ethiopian Trade Unions which represents around 350,000 workers - 85 percent o f which work inagriculture, and o f the remaining 2/3 work inpublic enterprises and 1/3 inprivate enterprises. CETU gathers some statistics from its members (9 federal unions and 450 trade unions). At present it i s not a major user of labour market information though they consult LFS results and the 179 labour statistics bulletins. The information is mainly disseminated to member organizations at the general assembly held every second year. Information on wages i s a major need. MOFED,MOEandMOTI. These ministries, which are represented inthe Labour Market Technical Committee, are users, albeit not very active, o f labour market data. The MOE needs information on the supply o f and demand for skilled and educated people and i s planning new activities in this area. With the new high policy profile o f labour market developments, it is expected that MOFED will become a more intensive user o f labour market information, particularly as its Welfare MonitoringUnit i s responsible for the monitoring o f PASDEP. Other actors. Labour market data are currently usedby development partners such as the ILO and the World Bank intheir policy dialogue. Over time, it i s expected that a broader group o f actors will be using the information system, as the availability o f statistics and analysis on labour market developments on a regular basis can foster broader participation and interest by governmental and non- governmental actors. 4. Availabilitv and accessibilitvof data 7.14 CSA i s the principal data provider for survey data in Ethiopia. As such, it i s essential that its statistical activity plans and its strategic priorities are discussed in the context o f the labour market (and other) sectoral information system as well as for the overall monitoring o fPASDEP. 7.15 Regular monitoring o f types o f indicators discussed above can build on the data sources which CSA is already collecting, integrating those information with administrative data and with others, possibly even collected ad hoc. The main primary sources available for the LMIS are: i. Labour force survey data. The existing labour market surveys (the quinquennial LFS and the UBEUS, planned to be held on a yearly basis) comprise a strong data set to inform LMIS- I,particularly now that both surveys allow the monitoring o f wages. The picture is not completely rosy, however, as the current schedule o f data collection is not working as smoothly as it would appear - e.g.preparations for the large scale LFS disrupt the fielding o f the UBEUS, and the questionnaires o f the two surveys are not completely comparable. More work i s also needed to understand the differences in the assessment provided by the two types o f surveys. Significant efforts have gone into improving the LFS questionnaire, ensuring that the UBEUS and the LFS methodologies and questionnaires are fully comparable (and that data collection is timed so as to avoid capturing seasonal variation together with structural trends) is now the priority to be addressed to further strengthen these primary data sources. Making LMIS-I operational and the dialogue among stakeholders surrounding it will contribute to address these difficulties and improve the data collection 180 process. Further, the work done within the CSA to establish a user-friendly databank will significantly increase the accessibility o f these data sources and ease the updating o f LMIS-Ethiopia I. ii. Registerofunemployed, vacanciesetc. Accordingtoagenerallawondata provision, these data are provided by the 11 BoLSA to MOLSA. The arrangement includes in turn that MoLSA in exchange provides training courses for BoLSA staff. It appears that several o f the BoLSA and the PES regional offices have capacity problems both human and technical - which - impedes the transfer o f quality data in useful formats. In this context, the provided key measures o f labour supply and labour demand will only represent the labour market partially. This issue should be addressed and strengthened as part o f the practical implementation o f the two labour market information systems. iii.Otherprimary data sources include: data collected by other Government institutions - such as the Ministry o f Education, which has an ongoing dialogue with MOLSA (eg on human resource planning) likely to lead to improved availability, accessibility and quality o f these data; o f the Federal Civil Service Commission - and other survey data collected either by CSA (e.g. firm level data: informal sector survey, small and medium enterprise survey etc.) or by other actors (e.g. Addis University Urban panel). While the coverage o f these surveys might not ensure the national representativeness o f the analysis, these data can integrate other information available, and technical dialogue with those involved in the collection and analysis o f these data can help shape the implementation o f the data collection agenda. 7.16 Secondary sources on labour market appear limited, though the Labour Statistics Bulletin (produced by the Manpower Research and Statistics Team within MOLSA to disseminate labour market data and analyses) and employer and employee information available through the Employer Federation or CETU might be o f help. Data quality and accessibilitymightbe however, concerns. 7.17 Additional data collection activities have also been suggested, such as a new enterprise survey to be conducted jointly with the Ethiopian Employer Federation, an in depth analysis o f wages to be conducted with CETU, and a survey to be conducted by BOLSAs to better capture regional realities, particularly on mismatches between demand and supply. It seems likely, however, that most o f the issues which such additional surveys would tackle can be already accommodated in the existing CSA statistical program, if efforts at improving existing surveys continue (e.g. the introduction o f wage questions in the LFS offers the possibility o f in depth wage analysis; another area on which work is needed i s for example the development of firm level surveys spanning both the formal and informal sector to analyze issues such as earnings across these sectors etc.). 181 5. Labour market indicators 7.18 Discussions with the different stakeholders have led to the compilation o f lists o f indicators that should ideally be monitored by the LMIS (see Annex l), and possibly linked with targets within the context o f the PASDEP objectives. In the Ethiopian context standard labour market indicators are most relevant in urbanareas. As discussed above, further work i s needed to develop relevant indicators for rural areas. These lists can be the basis for further technical discussion and agreement, based also on the data collection plans o f the CSA. It i s worth noting, however, that the elements o f these lists span several types o f indicators in order to provide a full picture o f labour market developments andhow they can be linked to growth and poverty reduction. 7.19 The main types o findicators included fall under the followingheadings: Poverty indicators. As the overarching objective o f Ethiopian policies is poverty reduction, poverty indicators provide the backdrop and motivation for labour market analysis. Further, poverty i s correlated with low human capital outcomes (low health, low education) which affect the quality o f labour,supply, as well as the returns that poor people can have from their labour (earnings o f the self- employed, wages). 0 Labour supply. Detailed breakdowns o f activity and inactivity rates by age, gender, area and skills provide insights into the evolution o f labour supply. This knowledge i s central for designing policies to reduce barriers to entry on the labour market (e.g. for young people making a transition from school to work) and to help in retaining people who already participate. Data on internal migration provide an understanding o f how labour supply i s reallocating itself following pushandpull incentives. Employment. This is the largest group o f indicators and it aims at revealing how well the economy i s usingits resources andwhich are the groups which are facing greatest problems in accessing jobs. Disaggregation by sector and type o f employer (public and private) also help tracing the process o f structural transformation o f the economy. Equity. Equityissues mightnot be apart o fthe first version o fLMIS-Ethiopia I- in particular because they should await the actual specifications o f targets in strategies or policies. Nevertheless, equality issues should be addressed as they comprise central elements o f a well-functioning labour market. 0 Humanresource development. This includes the educational attainment o f those in the labour force, of those on their way into the labour force, and the characteristics o f those who because o f illiteracy might find it particularly hard to upgrade their skills. 0 Working conditions. Good working conditions are fundamental for securing long-term productivity increases and thus economic development. This group o f 182 indicators might be included over time as capacity for data collection at the BOLSA level is put inplace. 7.20 Additional indicators could be monitored if public employment services were scaled up to become a major active labour market instrument at the regional level. The type o f indicators suggested would be administrative data relative to the hnctioning o f the public employment system and the efficiency o f the matching process and would include: 0 Supply o f labour. Information about job seeking through the employment services as well as through other means" provides information on the demand for such services. 0 Demand for labour. Similarly to the supply o f labour, information about registered as well as non-registered vacancies i s central for getting a picture o f the regional demand for labour. This information would be central for the design o f the active labour market policies such as vocational training and guidance. 0 Imbalances between supply and demand. Finally, a combination o f the supply and demand pictures gives rise to specific indicators for imbalances on the regional labour markets. The two suggested indicators are in particular suitable for assessing cross-regional differences and thus highlighting the option o f favoring relocation efforts byjob-seekers. 7.21 Whichever the final list o f indicators arrived at through fuller consultation as the basis for the development o f the LMIS, the system should be sufficiently flexible to incorporate new indicators (and remove obsolete ones) in response to new emerging needs and/or when new or improved data sources become available. The need for continuity and the benefits o f having longer series o f indicators need however also be considered inthis process o f upgrading. 6. Organization of the LMIS 7.22 Inthe present system, MoLSA has formal responsibility of collecting, analysing and disseminating data with respect to their specific policy areas. Inpractice much o f the production and dissemination o f labour market information centers around the producer role o f the CSA. supplemented at present with Canadian experts. Other labour market data - e.g. detailed The CSA is well-organised, with qualified and motivated staff - data from the regional BoLSA and PES offices - are collected, collated and hosted by MoLSA Computer Department. The staff members - although low innumber, e.g. really only one - i.e. the Manpower Research and Statistics Team in collabouration with the computer/programming expert - appear also well-qualified and capable o f hosting LMIS- 58It is suggested that the information about registered job-seekers is supplemented by estimates of non- registeredjob-seekers -e.g. via a survey carried out by BOLSA or PES (guided by the manpower research and statistics team inMOLSA) - to get a better regional picture o f the labour supply situation. 183 Ethiopia in the future. There i s still a need however, both at MoLSA and CSA, to increase the analytical capacity and the ability to provide targeted dissemination. 7.23 There is at present time not a single access point for labour market information and data. While (most of) the data at the CSA are stored electronically, part o f it is only available to the users in paper version. However, CSA has launched a new website which will make much o f the data directly accessible to the users. Inthe first place, the statistics will be made available via files inpdf-format. The next step inthe plani s then to establish an actual databank - i.e. storing the data indatabase format and thus enabling the generation o f special data extracts requested by the users. A fbture step is then to make website facilities which allow the users - interactively by themselves -to extract specially-designed data sets. Incontrast, the data from regional BoLSA and PES offices, as well as from other data providers to the Labour Statistics Bulletin, do not come in electronic format - so that many resources are spent on generating the actual input to the databases. 7.24 At the moment, the main ways o f disseminating labour market data and analyses are MOLSA's Labour Statistics Bulletin (disseminated mostly as a paper version though a homepage will be launched as well), and the CSA reports o f its LFS type o f surveys. Smaller and more user-targeted publication with key labour market indicators are also being considered. 7.25 The establishment and the proper working o f the organisation for LMIS-Ethiopia will take time. It will involve at least the following practical steps: First, physical establishment o f LMIS-Ethiopia unit. Office space and some computer equipment has already been allocated by MoLSA to this effect. Clear statutes for the unit need to be established, laying down the responsibilities and powers o f the unit, and the man-days to be spent by each unit member on LMIS-Ethiopia tasks, second, upgrading o f skills o f the LMIS-Ethiopia unit staff members on a variety o f topics ranging from technical issues to labour market analysis and communication and collabouration skills, reviving the role o f the Labour Market Technical Committee as a forum to holddiscussion and take decisions on the development o f the system, and announce formally the existence o f the LMIS unit, establishment o f collabouration agreements with data providers. At first such agreements should be pursued informally, with the LMIS unit developing forms and formats, or actual Excel-files to be filled in by the different data providers, to ensure a common understanding o f the data need and to ease the data transfer and thus the inclusion in an LMIS-Ethiopia databank. More formal data delivery agreements, including on modalities and timing o f delivery o f data etc, can become institutionalized over time initiating activities by producing an initial publication for web distribution, to generate interest in the output o fthe units, and obtain initial feedback from potential users. 7.26 Other steps would include the selection o f hardware and the development o f a prioritized procurement plan; the selection o f software with a view to increase the skills o f the stuff and the comprehensiveness o f the analysis over time; the development o f guidelines to the programmers on how to make the data accessible and available given capacity constraints on the part o f some o f the users; and identifying an appropriate dissemination strategy (in the case o f the LMIS-I, an annual frequency o f data dissemination, supplemented with more thorough analysis seems desirable), 184 Annex Table 78: PossibleList of Labour Market Indicators Definition Rationale Primary data sources Poverty indicators Share of population Percentage of population Quantifylng poverty - Householdincome below poverty line below national and relevant both because labour and consumption international (US$l and 2 a i s the main asset o f the poor survey. day) consumption poverty and because i s correlated with Possibly lines. low human capital outcomes LFSLJBEUSto which affect labour supply. have proxy measuresbasedon earnings Wages level by Average annual (or monthly) Understanding evolution of LFSLJBEUS - occupation or wages by occupationfsector. sources o f income and returns earnings frompaid sector to labour. Also a measure for employment. identifying benefits from Work neededon changing occupation or developing sector. measures of earnings of the self- employedfamily businesses Labour supply Labour force Number of persons in the Obtaining a general overview LFS andUBEUS 3articipation (LFP) labour force as percentage of and understanding o f the -ateby age, gender, working-age population Ethiopian labour force indarea (WAP) in total, by five-year age groups, male and female, and an urbanfrural division. Jrban labour force Number of persons in the Monitoring detailed urban LFS and UBEUS iarticipation rate urban labour force as labour market participation. ~ :LFP) by additional percentage of working-age :haracteristics population (WAP) by e.g. employment status, no of jobs, skill levels, city size etc. 185 Definition I Primary data sources ~~ ~ ~~~~~~ ~ ~~ ~~ Inactivity rate by Proportion age, gender, and outside the labour force - by some o f population Understanding reasons why LFS and UBEUS people are not school attainment five-year age groups, male participating in the labour and female, and LFS force (retired, caring for education classifier. attending school etc.) - and so family members, disabled, knowledge of potential barriers to overcome to avoid non-participation. Internalmigration Influx to urban areas as Urbanisation is still low - but LFS percentage o f labour force - internal migration i s expected by skill/education to play an increasing role. Employment Employment-to- Share of urban working-age The ratio provides LFS, UBEUS, and population (ETP) population in formal information on the ability of the Population and ratio by working employment and informal the Ethiopian urban economy HousingCensus status employment (self-employed, to createjobs. A highratio i s family workers etc.) good as it implies low unemployment. Employment by Employment (in urban areas) Monitoring the development LFS and UBEUS sector and by skills at least divided into three in employment by sector and broad groups (agriculture, by skills i s central for industry and service), and by planning education activities educational level (low and overall active labour mediumand high) market policy priorities and for analyzing how well the urban economy i s using the resourcesavailable. Unemployment Share of urban persons in the Most widely-used and LFS and UBEUS rate by age and labour force without work, monitoredindicator for labour gender actively seeking work - and people's problems in getting available for work, market imbalances - i.e. by five-year age groups and male jobs. and female. Unemployment Share of urban persons in the Information to education and LFS, UBEUS rate by educational labour force without work, re-training providers on the ittainment available actively seeking work - by skills - for design of future for work, and possible extent of obsolete LFSeducation classifier. education and retraining programmes. Youth Rate of unemployment among Special attention and thus LFS, UBEUS, and inemployment rate urban persons aged 15-24 monitoring o f developments - Child Labour (alternatively, share o f young as high youth unemployment Survey (only aged 186 Definition Rationale Primary data sources unemployed in total can have serious under 17) unemployment). consequences for future labour market participation. Share of long-term Number of unemployed While short periods of LFS andUBEUS unemployed by divided into groups according joblessness are of less duration and to the length of the concern - long periods often gender unemployment spell (less imply serious loss of income than 1months, 1-3 months, 3- and employability. Hence, 6 months, 6-12 months, above special measures are needed 12 months) - by male and to alleviate such female. consequences. Extent of Share of employed working Underemployment reflects LFS and UBEUS underemployment less than full working hours underutilisation of the (hours of work by gender (and willing to work more) - by less than 10 percent employed population - often underemployment) productive capacity o f the and indicator of underemployment, 10-25 as a result of an inflexible percent, 25-50 percent, and labour market. Hence, an above 50 percent increasing underemployment underemployment. The LFS problem might call for also offers information about changes to the labour relation workers who self report laws. underemployment. Size of informal Share of employees working The measure provides partly LFS and UBEUS sector by gender in the informal sector using an indication of the potential - the adopted definition of resource base for a informal employment (which developing formal sector. might change over time) by - male and female. Number of Number of vacancies The development in the Aggregated vacancies by sector registered by the regional number of vacancies is a information and area BoLSA and PES offices - prime indicator for deciding received from possibly supplemented by on the targeting of active BoLSA estimates of nonregistered labour market policies and vacancies. educational priorities. Ratio of Number o f unemployed per Overall measure of LFS and UBEUS, unemployed to registered vacancy (possibly imbalances on the (regional) and aggregated vacancies by area including estimates o f non- labour market - although information from registered vacancies). somewhat inaccurate due to BoLSA or new differences in data sources to BOLSA survey be used. Number o f Number of enterprises sector hdicator for developments in Enterprise surveys enterprises by grouped according to number labour demand. sector, size and o f employees (less than 10, area 10-25, 25-50, 50-100, above loo), branch and location of 187 ~~ _ _ _ _ ~ Definition Rationale Primary data sources head office. Demand for Recruitment plans three Indicator for developments in Enterprise surveys employees by skill months and one year aheadby labour demand. level skill level. Gender pay gap by Absolute (or relative) Monitoring the efficiency o f LFS/UBEUS - sector difference in monthly pay the labour market by focusing earnings from paid between male and female by on the incentives provided to employment. sector (for the samejob). menand women Work needed on developing measures of earnings o f the self- employedfamily businesses Gender Difference in rate o f Monitoring the efficiency o f LFS/UBEUS - unemployment gap unemployment between male the labour market by focusing earnings from paid in total and by and female in total and by on the incentives provided to employment. skills skill levels. menandwomen Work needed on developing measures of earnings o f the self- employedfamily businesses Regional Difference in unemployment Monitoring o f regional LFS and UBEUS unemployment gap between Addis Ababa and disparities. total and for youth other urban areas for two main age groupings (total and for youth, 15-25) Human resource development Education Education according to LFS Indication o f the human LFS, UBEUS and attainment of the education classifier by five- capital available for the possibly Ministry labour force by age year age groups, and by male labour market. of Education and gender and female. ~~ ~~ Degree of IlliteracyShare of population that Indication of constraints to LFS, UBEUS and in total and for cannot read and write. upgrading o f qualifications possibly Ministry youth via e.g. active labour market of Education measures. Number of school Educational attainment o f Indication of the extent of LFS, UBEUS and leavers by level of school leavers according to new entries into the labour possibly Ministry 188 Definition Rationale Primary data sources education and LFS education classifier - by market -useful for both o f Education gender male and female. enterprises and labour market institutions. Number o f work Number o f persons involved Increases in the number o f Aggregated accidents by sector in a work accident - by type work accidents might require information o f accident and sector. actions to avoid limitations to received from production and thus economic BoLSA developments. ~~~ _______ ______ ~~~ ~ Number o f disputes Number o f registered disputes Increases in the number o f Aggregated at work by sector between employers and disputes at work might information employees by sector. require actions to avoid received from limitations to production BoLSA developments. Definition Rationale Primary data source Supply of labour Number o f Number o f persons in a given Indication o f the part o f the BoLSA and PES registered job- period (quarter or year) regional labour supply seeking seekers by area, visiting BoLSA or PES jobs and thus putting pressure and skills or requesting assistance to find a on service education job - by urban areas (PES employment resources spent on job- regions) and preferably by matching. LFS education classifier. Estimates o f non- Estimates o f persons - BoLSA and PES registered job- registered at BoLSA or PES - not Together indicator - an indication o f the with the above estimates seekers taking upon a job - by urban resources immediately available for immediately available labour for regional areas (PES regions) and employers. preferably by LFS education classifier. Demandfor labour Number o f Number o f vacancies Indication the need for job- BoLSA and PES registered registered in a given period matching actions by the vacancies by area (quarter or year) by BoLSA or and sector PES - by urban areas (PES employment service. regions) and sector. I ~~ Estimates o f the Estimates o f the number o f Indication o f the extent to BoLSA and PES I 1 189 Rationale Primary data source number o f non- vacancies which are not which vacancies are filled estimates registered registered at BoLSA or PES. vacancies by area and sector Demand for skills Recruitment plans three Indicator for expectations to Enterprise by enterprises months and one year ahead by the number o f future survey I enterprises. skill level - by regional vacancies. Registered Number of registered job- Indicator o f the failure in LFS and unemployment rate seekers compared with the ensuring job matching. UBEUS, and by area number o f persons inthe urban Information labour force - by PES region. of regional BoLSA and PES differences might promote effort to favour the mobility o f job-seekers. Ratio Bottleneck indicator pointing BoLSA and PES vacancies - i.e. indication o f to difficulties in filling o f non-registered job-seekers skillshortages. 190