Report No. 47648-ET Ethiopia The Employment Creation Effects of the Addis Ababa Integrated Housing Program March 2009 Poverty Reduction and Economic Management Unit (AFTP2) Water and Urban Development Unit (AFTU1) Africa Region Document of the World Bank CURRENCY EQUIVALENTS (as of February2007) CURRENCY UNIT = ETHIOPIANBIRR(ETB) US$1 = BR8.9 FISCALYEAR July 8 - July7 WEIGHTSAND MEASURES Metric System ABBREVIATIONS AND ACRONYMS AACES Addis Ababa ConstructionEnterprise Survey IHDP Integrated Housing Development Program CWIQ Core Welfare Indicator Questionnaire HDPO Housing Development Project Office HICES Household Income Consumption and Expenditure Survey IHDP Integrated Housing Development Programme LFS Labour Force Survey MDG Millennium Development Goal MFIs Micro-Finance Institutions MOFED Ministryo fFinance andEconomic Development MSEs Micro and Small Enterprises PASDEP Plan for Accelerated and Sustained Development to EndPoverty PRSP Poverty Reduction Strategy Paper TVET Technical and Vocational Education and Training Vice President : Obiageli KatrynEzekwesili Country Director : Kenishi Ohashi, Ishac Diwan Sector Director : Sudhir Shetty Sector Manager : Kathie Krumm Task Team Leaders : Caterina Ruggeri Laderchi and Rumana Huque .. 11 ACKNOWLEDGEMENTS We would like to thank our primary counterparts, Dr Wubshet Berhanu, City Manager, Addis Ababa City Government, and Woz. Tsedale Mamo, Manager, Addis Ababa Housing Development Project Office, AACG. Staff at the Addis Ababa Housing Development Project Office has been most helpful inaccessing the program documents and inupdating the available information. We would like to thank in particular At0 Solomon Lemma and Ato Shimelis Alemayehu for their patience and collaboration. Others inthe City Administration to whom we are particularly indebted are Ato Mulugeta Assefa and At0 Yohannes Solomon. During the course o f this study we benefited greatly from insightful comments from Dr Abraham Tekeste o f the Ministryo f Works and UrbanDevelopment. We also benefited from consultation with our GTZ colleagues, and inparticular Gerhard Mai, Andreas Koening and Michael Meiwald. Generous funding from the Poverty and Social Impact Analysis Trust fund and from the German Trust Fund program on "Job creation, core labour standards and poverty reduction" i s gratefully acknowledged. The team was composed by Caterina Ruggeri Laderchi and Rumana Huque (task team leaders), Bob Rijkers, Emily Kallaur and Abebaw Alemayehu. The Ethiopian Economic Association runthe data collection activities and provided invaluable support inthe design o f the survey instrument. We would like to thank in particular Ato Kibre Moges, Ato Assefa Ademassie, and Ato Daniel Aklilu. Laila Nuru, Samuel Abate and Sandu Cojocaru provided research support. Impeccable administrative support was provided by Marjorie Kingston and Dora Harris. The team benefited from the overall guidance o f Ishac Diwan, Kathie Krumm and Jeni Klugman. Dr Francis Teal o f Oxford University provided invaluable guidance inthe analysis o f the survey. We also benefited greatly from informal conversations with various colleagues and inparticular we would like to thank Louise Fox, and Ato Membere Taye . Special thanks go also to Marianne Fay, Dileep Wagle and Magdi Amin who acted as peer reviewers. Their contribution has been important to help frame the findings o f this study in the broader context ofEthiopia's current development challenges. ... 111 TABLE OF CONTENTS ABBREVIATIONSAND ACRONYMS ............................................................................................ i1 ACKNOWLEDGEMENTS .............................................................................................................. i11 EXECUTIVE SUMMARY .............................................................................................................. vi1 1. INTRODUCTION ...................................................................................................................... 1 The Challenge of Unemployment inEthiopia................................................................................... 1 Employment Creation through MSE Support in PASDEP: The IHDP............................................. 2 Aims andOrganization ofthis Study ................................................................................................. 8 2. THE ADDIS ABABA INTEGRATEDHOUSINGDEVELOPMENT PROGRAM:A DESCRIPTION .................................................................................................................................. 10 Program Objectives and Targets ...................................................................................................... Program Administration................................................................................................................... 10 12 Program Participation: Procedures.,................................................................................................. The Low Cost Technology Adopted Bythe M D P .......................................................................... 13 14 Program Participation: Benefits ....................................................................................................... 16 Program Costs.................................................................................................................................. 18 Achievements to Date ...................................................................................................................... 20 3. METHODOLOGY AND DATA SOURCES .......................................................................... 22 Key Hypotheses and Analytical Tools............................................................................................. 22 24 The Addis Ababa ConstructionEnterpriseSurvey.......................................................................... Limitations of the Analysis.............................................................................................................. 25 Sampling Strategy-Firms ............................................................................................................. -25 Sampling Strategy-Workers ........................................................................................................... -26 Survey Instruments .......................................................................................................................... 27 4. THE EMPLOYMENT CREATIONEFFECTSOF THE MDP ......................................... 28 28 Selection into the Program............................................................................................................... A Profile of ProjectBeneficiaries:Workers.................................................................................... 31 FirmSelection.................................................................................................................................. 35 Project Beneficiaries: Firms............................................................................................................. 33 The Job Creation Impact o fthe Program ......................................................................................... 36 Net Employment Creation Effects: Differences inTechnology -37 Net Employment Effects: Differences in Factor Proportions .......................................................... ..................................................... The Impact ofthe IHDP on Labour Demand: Dynamic Considerations ......................................... 39 41 41 Program Participation and Growth .................................................................................................. FirmCharacteristics at Start-up ....................................................................................................... 42 Annex 1: DescriptivesofBeneficiaries........................................................................................... 44 Annex 2: DerivingEstimates ofFullTime EquivalentWorkers from the AACES ....................... 45 5. THE IMPACT OF SUPPORT ONPROGRAMFIRMS ...................................................... 47 47 Program Support and FirmPerformance ......................................................................................... Constraints and Support to ProgramFirms' Operation.................................................................... 50 Broader Impacts ofthe IHDP .......................................................................................................... 52 iv Entry and Exit.................................................................................................................................. 52 Perceptions of Program Impact ....................................................................................................... 53 6 . THE BENEFITSOFPROGRAMPARTICIPATION: INCREASEDEARNINGS .........56 A Descriptionof Earnings................................................................................................................ Simulated Distributional Impact...................................................................................................... 56 58 7 . CONCLUSION ......................................................................................................................... 59 APPENDICES ..................................................................................................................................... 63 Appendix A: Sample Scheme ......................................................................................................... 64 Appendix B: Construction of KeyVariables .................................................................................. References....................................................................................................................................... 67 -70 LIST OF BOXES Box 4.1: Qualitative Evidence on Hiringand Wages................................................................................ 29 33 Box 4.3: Qualitative Evidence on the Decisionto Participate inthe IHDP .............................................. Box 4.2: Definition of FirmTypes ............................................................................................................ 35 46 Box 5.1: Anecdotal Evidence on Firms' Key Constraints ........................................................................ Box 4.4: Estimating full time equivalent workers ...................................................................................... Box 5.2: Firms' Perceptions of the Effectiveness of Program Support..................................................... 47 51 LISTOFFIGURES Figure 1.1: A Schematic Representation o fthe Rationale of the IHDP ...................................................... Figure 4.1: Capital Intensity...................................................................................................................... 3 40 Figure4.2: InputDensity........................................................................................................................... 41 49 Figure 5.2: Rankings o f Benefits of Program Participants ........................................................................ Figure 5.1: Types of Support Received by (All) Program Firms.............................................................. 50 Figure 6.1: Median Wages by Occupation, Firm Size, and ProgramParticipation Status........................ 57 Figure 6.2: Simulated Earnings Distributions ........................................................................................... 58 V LISTOF TABLES 12 Table 2.2: A Comparisonof ConstructionProcesses: Private Sector and IHDP....................................... Table 2.1: Quantifiable and Non-Quantifiable ProgramTargets .............................................................. 14 19 Table 2.4: ProgramTargets and Results as of October 2006 .................................................................... Table 2.3: Program Costs ....................................................................................................................... 20 Table 3.1: Composition of the Sample ...................................................................................................... Table 4.1: OccupationalBreakdown by Gender........................................................................................ 25 30 Table 4.2: Average Characteristicsof Workers by FirmParticipation and............................................... 30 Table 4.3: Mean Proportion of Workers with DifferentTypes of Work Experience, ............................... 31 Table 4.4: Characteristicsof Firms by Size andProgramParticipation Status ......................................... 34 Table 4.5: Median Inputand Output Pricesby FirmType........................................................................ Table 4.6: ProductionFunction: Deflated Revenue.................................................................................. 34 Table 4.7: Human CapitalAugmented ProductionFunction: Deflated Output........................................ 38 39 Table 4.8: Activities of Programvs.Non-ProgramFirms (percent) ......................................................... 44 Table 4.9: Age ofFirms (percent and cumulative) by firm type ............................................................... Table 4.10: Labour Absorption in2006 by Firms by Size and ProgramParticipation Status...................44 46 48 Table 5.2: Distribution of Types of Support by Types of Firms (percent)................................................ Table 5.1: Use of credit by firm type (percentage).................................................................................... 49 Table 5.3: Percentof Firms Reporting Problemswith Utilities andMarket Access ................................. 50 Table 6.1: Median Monthly Wages (inBirr)............................................................................................. 56 Table 6.2: The Impact of ProgramParticipation on Workers at DifferentQuantiles ofthe Earnings Distribution................................................................................................................................................. 58 Table A 1: Non-Contractors...................................................................................................................... 65 Table A 2: Contractors.............................................................................................................................. 66 vi EXECUTIVESUMMARY 1. Ethiopia's second Poverty Reduction Strategy, the Plan for Accelerated and Sustained Development to EndPoverty (PASDEP), outlines a continued planfor increasing agricultural productivity but also incorporates a renewedemphasis on tackling urbandevelopment issues. Two o f the key urban challenges facing Ethiopia are inadequate infrastructure, including housing, and insufficient creation o f quality employment opportunities (as evidenced by a large cohort o f working poor and significant levels o f open unemployment). To address these challenges, in 2004, the city o f Addis Ababa introduced the Integrated Housing Development Program (IHDP), an innovative program designed to provide low-cost and affordable housing while also generating employment and building human capital and entrepreneurship in the construction sector. The ultimate objective o f the program was to create better quality employment opportunities and a competitive market structure in the construction sector. To have a sense o f the scale o f the program, consider that the program aims to create jobs for the equivalent o f nearly one-fourth o f the active population inits main beneficiary age group (Le. 80,000 jobs over 4 years, relative to an active population o f 350,000 25-34 year olds). As part o f its urban development strategy, PASDEP envisions launching a national program based on the Addis Ababa initiative, and incorporating lessons learned from the experience with the IHDP. 2. It is therefore timely to conduct an assessment of the impact o f the IHDP thus far. However, the aims o f this study are more modest-it focuses exclusively on the employment creation side o f the rogram, which has not been evaluated to date, taking as a given the demandfor housing. The study relies primarilyon a purposely-designed quantitative survey P o f program beneficiaries and their counterparts outside the program, but also draws on administrative data and qualitative evidence. Although it i s not a full impact evaluation, given the reliance on a survey conducted at a single point intime and the fact that program beneficiaries are not selected randomly, the study represents a first attempt to quantify the extent to which the assumptions behindthe program design seem to hold. 3. The central question o f the study i s whether the IHDP has generated more jobs than would have been created if it had only been a housingprogram (and relied on hiringexisting construction firms) rather than a housing plus employment creation program. The employment creation side o f the program includes several major elements. First, the program actively stimulates the creation o f Micro and Small Enterprises (MSEs) in the construction sector by screening qualified workers (via a skills test), teaching them how to form legal business enterprises, and allowing them to group themselves into new firms. Second, the program contracts these new MSEs to work on the housing projects, which incorporate innovative low-cost and labour-intensive technologies (most notably, pre-cast ' Other aspects of the program, such as its dynamic role in fostering entrepreneurship and the distributional aspects that such entrepreneurship might have are dealt with only indirectly, by looking at the profitabilityand sustainabilityofthe MSEs createdby the program. vii beams and prefabricated hollow blocks) in order to build affordable housing. Third, the program provides wide-ranging support to the MSEs, including access to land, access to credit, input provision (e.g. re-bars, cement, and iron) on credit, machinery leases at favourable conditions, and skills training (though not all firms receive all types o f support). This type o f MSEs support i s extended by the Municipality to other priority sectors as well, though the construction sector is the only one to benefit also from a source o f guaranteed demandthrough the Housingprogram. 4. The program intends to accelerate employment creation inthe construction sector by supporting the formation and development o f MSEs, as described. The effectiveness o f this approach i s therefore predicated on two basic assumptions: 1) MSEs in the construction sector, and particularly MSEs in the IHDP, are more labour-intensive than larger firms; and 2) the creation and support offered by IHDP to MSEs in the program is needed to overcome market failures. The first assumption explains the program's focus on supporting MSEs rather than buildingthe capacity o f large contracting firms, while the second summarizes the rationale for an activist approach to growing the MSE sector. This study tests each o f these assumptions to determine whether they reflect the reality in Addis, with an eye toward informing the scale-up o f the program to other cities and towns. Giventhe program's long term objective o f improving "the standard o f living o f citizens, especially low-income residents o f the city through the creation o f employment opportunities and the provision of decent and affordable housing'' (IHDP 2006), this study also looks at the distributional effects o f the program-i.e. whether the beneficiaries of the program are disproportionately drawn from relatively poor or vulnerable segments o f the labour force. One limitation o f our analysis i s that, by focusing on the MSEs segment o f the construction sector, we do not explicitly assess the jobs created by large-scale firms working as contractors on the structural works.2 5. The survey designed to enable this analysis was run in December 2006. It is representative o f the construction sector in Addis, and the data allow comparisons o f firms and workers inside and outside the program, as well as between large contractors and other firms in the sector, which are mostly MSEs. The survey enables a detailed look at questions o f technology use, which are at the heart o f the IHDP's design. It also allows us to consider a number o f additional issues, such as the nature o f constraints experienced by firms and the effectiveness o f the type o f support provided. The timing o f the survey, which as mentioned intended to provide elements to inform the roll-out o f the program to other cities, might imply that some o f the findings reflect the specific experience o f the first years of operation o f the program. These years have been characterized by delays in implementation and by shortages o f key inputs, particularly inthe latter part o f 2006. The influence o f these transient factors i s noted inthe discussion of the findings. Note that one can assume that such additional demand for labour would be expressed by large contractors even if the program were designed as a housing program only, rather than as a housing and employment creation program. While our inability to control for employment creation by the large scale contractors means that we cannot assess the overall employment creation effects o f the program, it does not seem to affect our ability to judge on whether the employment creation emphasis o f the program has resulted in additional jobs than would have otherwise been created. ... Vlll 6. The analysis supports the notion that MSEs are more labour-intensive than large contractors, although among MSEs there does not appear to be a difference in labour intensity between firms that participate in the IHDP and those that do not. In other words, there i s no systematic difference in the amount o f capital per worker between program and non-program MSEs. However, at a fixed point in time (e.g. when the survey was fielded), program firms did have more workers than non-program firms-a median o f 13 workers, versus 5 workers for non-participants. This does not necessarily mean that firms hire more workers as a result o f being inthe program, but the results do indicate that program firms are larger (interms o f number o f workers) at start-up. This seems to be associated with the fact that program firms have better access to capital than their non-program counterparts. Once established, program firms do not grow more quickly than non-program firms overall-but program firms do grow faster than non-program firms that started at the same size. 7. Turning to the second assumption, there do appear to be significant market failures, particularly in terms o f access to land and credit, as reported elsewhere in the literature. Moreover, although not all firms in the program receive support (85 percent receive some support, but very few receive all types o f support), those that do receive support from the program perform better than those that do not, but not better than firms not working for the program. Access to credit seems to be the most effective type of support and i s associated with significantly higher production. Also, the provision of a space to work is associated with significantly higher profitability. 8. A central element o f the IHDP is providing a market for the MSEs that have been created through the program, and this demanddoes seem to be critical to their survival, since most are not working for non-IHDP clients. However, given shortages o f key inputs (notably cement) inthe construction sector as a whole, the IHDP's contracting system (defacto fixed bid with some ad hoc revisions o f the terms of the contracts ) puts some firms at risk-firms commit to delivering outputs at a given price but are not compensated for rising input prices, which threatens their profitability and sustainability (except when inputs are directly provided by the program). The Housing Development Programme Office has realized the importance o f this issue and i s currently considering how to put inplace a system for a more systematic review o f contractual terms. 9. Interms o fthe targeting o fthe program, the survey data do not reveal that workers in program firms come disproportionately from the ranks o f the unemployed, or that women disproportionately benefit from jobs with program firms, despite the program's intention to target them. While this does not rule out a significant contribution o fthe program to reducing unemployment, it underscores that such effects are rather indirect, i.e. through a general increase in labour demand or by providing incumbent MSEs with a way o f overcoming market failures. From a distributional point o f view, program firms do pay workers on the lower end o f the educational distribution a higher wage than non-program firms, so that the program does seem to have a positive distributional effect+ven ifthe poorest workers are not its main beneficiaries. Further, the program provides opportunities for casual workers to become permanently employed. 10. Looking at the business environment more broadly, about three-fourths o f firms believethat competition inthe construction sector has increased inthe past 3 years, and about ix a third o f those attribute the increase to the IHDP. Program firms are most concerned about competition from non-program firms, which raises concerns about their ability to compete in the larger market. As one of the ideas underpinning the program is the idea of fostering a more competitive market structure this can be seen as a success, though it also raises the question of how to ensure the growth of a good number ofthese newly born firms. 11. Note, however, that to the extent that the program has exerted upward pressure on input markets, there also may be some crowding out o f non-program firms that are not able to secure key inputs. The direct impact o f the program on labour markets (both in terms o f absorption and driving up wages) seems to have been limited, even though the program offers the benefits o f higher wages to lower skilledworkers. 12. Overall, the study finds that the program has been successful in several respects; it has generated jobs and the support provided to MSEs seems to have enabled firms to grow more quickly than non-program firms o f the same size. Yet, the study finds that program MSEs do not appear to be more labour intensive than non-program MSEs. This suggests that alternative modalities o f delivering housing could be considered without affecting the employment creation potential of the program. The scaling up o f the program to other urban areas offers the possibility o f experimenting with alternative MSE-based arrangements for delivery which might offer advantages in terms o f a simplified administration and coordination o f the program, while retaining the dynamic benefits o f the program itself (e.g. in terms of strengthening the skills of workers in the sector and introducing new technologies). 13. The scale-up o f the program could also potentially benefit from addressing some o f the issues that emerged from the analysis. The program supports firms growth mostly by addressing the technology and skills constraints. Possibly because those are addressed across the board, some other elements o f support, such as facilitating access to credit and land, seem to be quite important to firms' success, and the program could be strengthened by expanding coverage to more firms. Current plans to introduce systematic mechanisms for reviewing contracts may also be critical to ensuring that firms continue to be interested inworking for the program without the risks associated with fixed-prices. Further, policy reform could strengthenthe program's ability to reach its own goals by addressing the constraints infactor markets and input provision (e.g. design, credit, materials) which affect the construction sector as a whole (and more generally urbandevelopment). For example, the development o f a more robust mortgage market could allow the extension o f commercial financing to a greater segment o f the housing market. With a smaller supply gap in the provision o f housing, public resources could be more effectively targeted to the provision o f housing affordable for very low-income households. These possible developments seem in line with PASDEP vision o f a national Integrated Housing Development Program that "integrates public and private sector investment with MSE development and the provision o f basic services" (p. 163). Similarly, reform options for the way land i s allocated could be considered. These are discussed indetails inWorld Bank 2007, b. 14. The study also highlights the importance o f strengthening the program's monitoring and evaluation system to guide further implementation and scale-up. Indicators being tracked (most importantly, number o f job opportunities created) could be better defined to X allow for more precise measurements o f program outputs. At the same time, measuring only the number o f job opportunities (in full-time equivalents, for example) does not take into account the sustainability or profitability o f the firms in question. Since existing firms appeared to have high levels o f unused capacity, targets for firm creation may need to be adjusted. Capacity underutilization also reflects underlying issues (such as input shortages) that may need to be addressed through different kinds o f interventions outside the purview o f the IHDP and o f labour marketpolicies. 15. Finally, since the study reveals that some o f the main groups o f low-income residents targetedby the program (e.g. women) are not the main beneficiaries o f the jobs createdY3this study draws attention to the need for additional programs to support poverty reduction in urban areas. Related to the IHDP's key concerns is the ongoing TVET reform, intended to make TVET more responsive to market demand andthe needs o f employers and improvejob prospects for new graduates. A well functioning TVET system can contribute to reducing the educational disadvantage o f the most marginal groups. Further, more specific interventions might be needed to reach women, for example by addressing constraints in the legal and institutional environment faced by female entrepreneurs and facilitating their firms productivity growth, as highlighted for example inthe recent Investment Climate Assessment findings. Furthermore, the international evidence on the effectiveness o f active labour market policies, however limited with relation to developing countries, suggests that often they play the role o f a safety net rather than providing sustainable employment creation. A closer look at the nature o f vulnerability in urban areas and the specific challenges that marginal groups face might help inform the formulation o f an urban safety net strategy appropriate to the needs o f Ethiopia. 3 Note however that vulnerable groups such as female headed households have been given priority in the allocation o f the apartments - in two recent rounds o f allocation o f houses more than 50 percent o f the beneficiaries were female headedhouseholds, a group particularly prevalent inurban Ethiopia. xi 1. INTRODUCTION 1.1 Ethiopia's second Poverty Reduction Strategy, the Plan for Accelerated and Sustained Development to End Poverty (PASDEP) outlines a strategy o f complementing a continued strong focus on increasing agricultural productivity with an increased emphasis on urban development. Inthis context it highlightsthe importance o f facilitating accelerated employment generation to address the issue o f high levels o f urban unemployment. The Addis Ababa ' experience with the Integrated Housing Development Program (IHDP) i s singled out as an effective tool for employment creation and for addressing housing shortages inurbanareas. The Challengeof UnemploymentinEthiopia 1.2 Recent analysis (World Bank 2007) estimated urban unemployment rates at around 14 per~ent,~ also identified great heterogeneity in its concentration both geographically and but across groups o f people. Larger towns, including Addis, are characterized by lower employment rates and higher ~nemployment.~ Further, women and young people face particular challenges in terms of labour market success. Despite some positive trends,6 which are at least partly related to increases in secondary school enrolment, Ethiopian unemployment levels remain high relative to international comparators. 1.3 The challenge o f reducing unemployment in Ethiopia today is characterized by a number o f relatively new developments. These include: (i)a rising skills profile o f the urban workforce-by 2005, three-quarters o f youth had at least 4 years o f education, while this was only true for half o f the older cohort; (ii)changes in the educational composition o f employment, with lower education levels accounting for a smaller share o f employment; (iii) the overall growth in the labor supply in urban areas compounded by internal migration (which appears to be growing); and (iv) high levels o f unemployment, which particularly affects the growing cohort o f young people. According to evidence from an urban panel ~urvey,~ new entrants to the labour force constitute the main source o f the newly unemployed with 17 percent o f the inactive population (this would include students) in 2000 being unemployed in 2004. 4 All indicators are for population 15+. Note that estimates of unemployment vary significantly depending on definition-while the standard international definition of unemployment requires active job search, Ethiopia (and some other countrieswhere very longunemploymentduration is common)relaxesthis criterion. At 21percent, the populationage lo+). official unemployment rate i s therefore higher than the one reported here (the official rate is also based on the In Addis the employment rate (15+) is 49 percent (40 for women) and the unemployment rate is 24 (34 for women). Comparisons of the 1999 and 2005 Labour Force Survey (LFS) reveal a reduction in unemployment over the period from 14.8percent to 13.5 percent driven by decreases for youth and older males. The Government relaxes the "active search" criterion when measuring unemployment, on the grounds that long unemployment durations leadto large numbers of discouragedworkers (Le. those who would like ajob, but are not actively lookingfor one because they believe a search would be unsuccessful). Between 1999 and 2005, passive unemployment declined even more than active unemployment (the share of population in passive unemployment declined by 38 percent, versus a decline of 9 percent inactive unemployment), in favour of inactivity. The shift inpassiveunemployment was particularly visible among youth, which is consistent with higher school enrolment. Other positive developmentsover the periodinclude the large reductioninthe median duration o f unemployment from morethan 1.5 years to less than 1year over the period, and increaseddynamism of the labour markets. For more details see WorldBank (2007). 7The EthiopiaUrbanHouseholdSocio-economicSurvey, collectedin 1995, 1997,2000, and2004. 1 These new developments imply that while on the one hand a more educated labour force has increased expectations in terms o f jobs and remuneration, a significant segment o f the population faces reduced economic opportunities due to their lack o f skills. EmploymentCreationthroughMSESupport inPASDEP: The IHDP 1.4 To tackle the challenge o f unemployment, PASDEP's strategic emphasis i s on the growth of labour-intensive sectors, and on facilitating the growth o f Micro and Small Enterprises (MSEs). In particular the effort o f employment creation through the growth o f the MSE sector is seen to require integrationo f efforts to increase educational attainment, both via general education and TVET skills training, with the provision o f capital for the unemployed (within a well-functioning financial system), and with "specialized programs to promote opportunities for self-employment" (MOFED 2007, p 54). 1.5 Inthis context PASDEP emphasizes that "the recent experience of the Addis Ababa City Administration insmall and medium scale enterprise development linked with TVET and a low cost housing program is going to be scaled up and rolled out to other towns inthe country" (ibid.)as it contains the key elements o f the government's strategy to fight unemployment. The Addis Ababa Integrated Housing Program (IHDP), launched in 2003, integrates housing constructionto address the housing shortage (see Box 1.1) with MSEsupport to createjobs. Box 1.1: The HousingShortage inAddis Ethiopia's urban areas are characterizedby an acute housing shortage, and Addis is no exception. Estimatesof the housing shortage in Addis vary between 250,000 and 300,000 housing units, and the shortage i s increasing by approximately 40,000 units each year. In addition, at least a third of the estimated total housing stock of 640,000 units is of very poor quality. For example, about 80 percent of the 150,000 housing units administeredby the city administration are built with mudand straw and are older thantheir estimatedlifetime of 30 years. Housing demand has been increasing in Addis in the recent past as a result of population growth, migration to urban areas and the dilapidation of the existing housing stock due to poor maintenance. Other drivers of the demand for housing include progressively increasingdiasporademand for housing, a lack of alternative investment opportunitiesand speculation. Recent reforms in the areas of customs, businessregulation and registration have helped stimulate housing supply by relaxing financing constraints, alleviating the burden of bureaucratic procedures, and marginally increasingthe availability of land. Despite the improvements,however, high and variable land prices, and difficultiesin obtaining land, continue to pose a challenge to would-be developers. Several other key challenges also persist, such as obtaining finance, poor regulation, the absence of quality insurance, lack of technological know-how and adequate equipment, unpredictability oftax liabilities, andthe vulnerability ofbiddingandtender proceduresto corruption, `ources: Addis Ababa City Administration Housing Development Project Office/HDPO, "Brief Project Profile," p 4., September 2004; "Addis Ababa Integrated Housing Development Program," (no author), March 2004; "Low Cost Housing: Would this be the Solution?" in Construction Ahead: Bi-monthly interface with the construction industry, " p 32., Butterfly Publishing, September-October 2005; World Bank (2004); World Bank 2007 b (the CEM); the "Integrated Value Chain Analysis for the Housing Construction Sector in Ethiopia" by Global DevelopmentSolutions, 8 June 2006. 1.6 The IHDP's long-term objective i s "to improve the standard o f living o f citizens, especially low-income residents o f the city, which are the majority, through the creation o f 2 employment opportunities and the provision o f decent and affordable housing."8 From the employment creation point o f view, the IHDP aims to stimulate labour demand by fostering the adoption o f input-saving technologies by newly created MSEs, which the program itself supports and to which it provides a guaranteed market. Figure 1.1 below provides a schematic representation o f the rationale o f the program in addressing employment generation and constructiono f low-cost housing (more details on the program are provided inChapter 2). Figure 1.1: A SchematicRepresentationof the Rationaleof the MDP Low cost housing Improvedurban Subsidies on available for poor livelihoods construction construction particularly for inputs low income Subsidieson residents, loans through Introduction employment of new tl opportunities technologies A andprovisionof Skills Creation of decent and development MSEswhich creation use input Skills affordable housing. Coordination saving development of various technology Viable MSEs activities created supportto 1.7 While in the rest o f the analysis we will focus specifically on the employment generation effects o f the program through its MSE support component, it i s worth drawing attentionto other defining features o fthe IHDPwhich make it quite a unique program: 0 Direct provisionof housing.Incurrent debates, the emphasis on state-built housing for sale or rent which characterized the period o f the 1950~-1970s,has been replaced by a focus on the state as an enabler o f low-income housing construction by ensuring favourable land legislation, efficient markets, and provision o f urban infrastructure and services. Both the Ethiopian and international experience provide examples o f alternative interventions for housing delivery (Box. 1.2). 0 Labour intensive methods. Labour intensive methods for public works have become quite common, though the literature provides examples mostly focused on their application to roads construction and maintenance (with labour intensive methods for maintenance o f federal roads used in Ethiopia, among other countries, I L O 2004) or on irrigation. For roads, absorption o f up to 5 times as much labour, and cost savings in the range o f 25-30 percent, have been found (Keddeman 1998). Yet, as in the case o f other active labour market policies, the effects have been found to be mostly short term and 8Presentation by Ato Salomon Lemma, IHDP, at World Bank's planning meeting on this study, November 27, 2006. 3 related to the provision o f some elements o f a safety net. Longer term poverty reduction impacts appear more difficult to identify. Building capacity in the housing construction sector. An important element of the program i s its emphasis on building capacity inthe use o f low-cost building technologies which were not widely available in the country. Yet a recent value chain study (GDS 2006) highlighted that several other challenges stifle the growth o f the sector and its potential to increase the supply of housing. These have been identified as ranging from problems o f design (due for example to the unavailability o f adequate standards and norms, and lack o f technology know-how and training), procurement (due for example to the lack o f standard contract documents) and difficulties in accessing inputs (material inputs, infrastructure, and land) and equipment. While the program addresses some of these constraints, such as those on design and know-how, by adopting standardized designs and linking with the TVET systemYga broader set o f interventions could strengthenthe sector as a whole andreduce the need for direct housingprovision. 0 Joint pursuit of different goals. The most salient feature o f the IHDP is its joint pursuit o f a variety o f goals, including housing development, MSE support, sector development and employment creation. These are clearly all important goals for addressing urban development and poverty reduction. Yet, the choice o f pursuingthese goals jointly implies costs of coordination between different sets o f activities, which, though hard to quantify (see also chapter 2) needto result inincreased efficiency or strong dynamic benefits. Box 1.2: Strategies to Improve Access to Low-Income Housing The severe housing shortage in Addis Ababa has prompted the municipal government to adopt a range of strategies for improving the condition of existing housing, and facilitating new construction. Of these, the IHDP's constructionof new condominiums is the most prominent.But initiativesalso include: 1) infrastructure upgradingin kebeles; 2) a kebele "renewal" program that aims to move residents into the new condominiums, thus freeing up land for higher-value commercial use; 3) regularization of some illegal settlements (those in conformity with city regulations) and demolition of the remainder; 4) provision o f incentives (e.g. free or subsidized land) for real estate developers, especially if they intend to build low-cost housing; and 5) implementing sites and services schemes to deliver serviced plots to housing cooperatives and individuals. There is some initial evidence that the different types of programs pursued may reach very populations, underscoring the importance of making affordability concerns central to the design of any housing program. This might have limited the IHDP ability to reachthe poorestbeneficiaries as compared, for example, to slum upgradingprograms.(R. Fein, unpublisheddata; see also LohnertandFein2006). Additional insights into successful housing policies and programs can be gleaned from international experience.loPrograms to stimulate housing demand by addressing liquidity issues, through housing microfinance programs for example, have recently gained traction. Since many of the poorest cannot access commercialfinance, and mortgage finance is fairly uncommonin countries like Ethiopia, short-term loans can help families gradually improvetheir housing-which is consistentwith the incrementalprocess through which informalhousingtends to be upgradedand enlargedover time. Mibanco of Peru, a well-regardedmicrofinance Contactswith the EngineeringCapacity BuildingProgramare currentlybeingpursued, to explorepossible collaborationsonmattersrelatedto temporarytrainingonthejob, internships, the introductionofmore advanced technologiesandalternativeconstructionmaterials(e.g."agrostone" partitionwalls, sustainablescaffolding systems etc.) andthe provisionof advancedjob opportunitiesfor newly graduatedengineers. loThis draws onFay and Wellenstein, 2005. 4 agency, provides an example of successfully applyingmicrofinance techniques to the housingsector. In2000, it beganthe Micasa Program, which provides home improvement loans with lower interest rates, longer terms, and in larger amounts than typical microfinance loans for entrepreneurs. The program became profitable-boding well for sustainability-within the first year, accumulating 3,000 clients and an outstanding portfolio o f $2.6 million. Taking a housing program to scale requires cost-effective techniques. Slum upgradingprograms (elements of which are underway in Addis) seem to be among the most sustainable in this respect, as they tend to be about one-tenthas expensive as programs aimed at demolishing slums and relocatingtheir inhabitants. A large payoff can also result from regularizing tenure; S I in government spending on average leads to about S7 in private investment (SIGUS 2001, cited in Fay and Wellenstein 2005). In Ethiopia, this could suggest significant potential benefits from accelerating efforts to regularize informal settlements and privatize kebele housing, which could lead to better living conditions for some of the city's most vulnerable residents. Other possibilities include promotion of low-cost rental housingthrough the private sector and cooperatives, or self-help housing programs (such as those undertaken in Botswana and Mexico) that provide standardized materials, credit, and technical assistance, but leaveorganizationand labour to communities. Source : Fay, Marianne, and Anna Wellenstein. 2005. "Keeping a Roofover One's Head: ImprovingAccess to Safe and Decent Shelter." InThe UrbanPoor in Latin America, ed. Marianne Fay, 91-124. Washington, DC: World Bank. Evidence on the Effectiveness o f MSE Support and Employment Generation 1.8 Since 2006 the IHDP has been integrated into a "national integrated housing program" based on the lessons learnt from the Addis experience (ibid, p. 163). While PASDEP refers to the lessons learnt in implementing the Addis Ababa IHDP, no systematic assessment has been conducted on the MSE support component o fthe program and its effectiveness increatingjobs. 1.9 Some initial assessment has focused on the pilot site o f Bole-Gerji (see Box 1.3 for their main findings). While the evidence that the housing component o f the program has not been pro-poor has been disputed, it is worth noting that the different methods o f assigning housing which have been introduced inthe IHDP after the pilot experience are unlikely to have improved this distributional outcome, as they consist in a lottery, with a 10 percent o f units reserved for people who have been displaced or otherwise affected by the program itself. Efforts to give priority to female headed households in the allocation, might have been more effective in addressing distributional concern though no systematic evidence on the pro-poor nature o f the allocation has been collected as yet. Based on this initial evidence it can be expected that to date the mainpro-poor effects o f the programs have come from its employment creation effects. Box 1.3: ExistingAssessmentsof the IHDPPilot In March 2006, researchers from the University of Bayreuth conducted a survey of residents of IHDP condominiums in Bole-Gerji, the site of the IHDP pilot project (Lohnert and Fein 2006). The survey asked residents about their family situation, income, andliving conditions, and found on the whole that residents(most of whom were owners) were generally satisfied with their experiences thus far and felt that living conditions were better than in their previous homes. Studio and mid-size apartments were renting for 600-800 Birr per month, though a limited number of large or furnishedapartments had monthly rents of 1,000 Birr or more. A follow-up study inMay 2007 revealed that rents had risento 800-1300 Birr per month for studio and mid-sizedapartments, and 1500 Birr (or much higher) for large or furnished apartments; additionally, the percentage of renter-occupied units had increasedfrom 22 percent to 43 percent (R. Fein, unpublisheddata). 5 The survey revealed that most residents could be consideredmiddle-income, and most apartments were occupied by families. However, there was a high level of heterogeneity, with some lower- and upper-incomehouseholds present, and a wide range of household types. Although some apartments were inhabited by youth receiving financial support from their families, all other households had one or more members with a steadyjob (usually a civil servant), while informalsector work was generally found only as a secondarysource of income (andtypically inbighouseholds). While the methodologyofthis study has beendisputed, these findings raise concernsemphasizedin a WorldBank study of housing in Addis Ababa-namely that the IHDP condominiums would be unaffordable for the poor (whichalso meansthat they mightnot effectively draw from those currently livinginthe worst housingconditions) (World Bank 2005). According to the study, a one-bedroomunit would be unaffordable for 85 percent of the Addis Ababa population. As a result, costs would be expected either to be borne by the government-requiring an enormous public outlay-r the units would have to be sold mainly to the well-off. (Lohnert and Fein found that most beneficiarieswere civil servants). As mentionedinthe text, following the pilot, the programhas changed some of its allocation mechanisms, though no systematicmonitoringofthe pro-poornature of the allocationshas beenundertakenas yet. Sources: Loehnert and Fein, "Survey in the CondominiumApartment Houses o f the Pilot Project in Bole-Gerji, FirstFindings, 2006; WorldBank2005. 1.10 The international literature provides some pointers on the types o f impacts that the IHDP can be expected to have through its MSE support/employment creation component and the importance o f program design features. A literature survey by Betcherman et al. (2004) finds that programs that provide support on a variety of fronts (mentoring and business counseling, financial aid, etc.,) are more effective than those that focus on only one aspect. Training for MSES" can foster higher rates o f capacity utilization and quality practices, as well as increased productivity growth, though these effects appeared to vary over time. Finally, MSEshelf-employment assistance programs can provide effective support for the small minorityo funemployedworkers who are interested instartingtheir ownbusiness" (ibid. p. 51). 1.11 An extensive literature exists (Biggs 2002; Halberg 2001) onthe technology and factor proportion choices o f MSEs, though mostly for firms in the manufacturing sector. The main rationale for promoting MSEs is that they are more labour intensive, i.e. they hire more workers per unit o f investment than larger firms.12 Evidence from the manufacturing sector in developing countries shows that this is indeed typically the case (Teal 2007, Bigsten and Soderbom 2006), though in many cases this pattern i s violated. Inparticular it has been shown that the more disaggregated the data are in terms o f productive activities, the less this pattern holds (Little and al. 1987). 1.12 Further arguments brought in defense o f supporting MSEs include dynamic considerations as MSEs are thought to grow faster than large firms. The empirical evidence for this proposition is thin. For example, in the case o f the manufacturing sector in Africa, Biggs and Shah (1998) find that small firms were not the main source o f net job creation in countries I 1E.g. Hong Tan, and Gladys Lopez Acevedo (2005) Evaluating Training Programs for Small and Medium Enterprises.Lessonsfrom Mexico. WorldBankPolicyResearchWorkingPaper 3760. '*Arguably there are trade offs in terms of benefits as more capital intensive and larger firms can offer higher wages. In contrast, however, it could be argued that fostering a plurality of specialized MSEs can offer a more diversified economic basis and thereby support the creation of a middle-class. This argument does not appear to havebeenformallytested inthe empiricalliterature. 6 where netjob additions took place. Only in Zambia, which experienced an overall netjob loss, did small firms create morejobs thanlarge firms.13 1.13 While the economic literature raises at least some doubts on whether supporting MSEs necessarily results in greater employment creation than relying on other types of firms, it i s worth underscoring that while the scaling up o f the IHDP is predicated on static arguments on employment creation, the program pursues a much broader set of objectives, including building capacity in the construction sector. These effects, though hard to quantify could result in increased productivity and greater employment creation in the sector as a whole. Furthermore, even if MSE in the construction sector in Ethiopia did not adopt more labour intensive technologies than larger firms, by introducing new input saving technologies the program could be helpingto release input shortages (see Box 1.4) and could result in greater employment per unitofexpenditure bythe housingprogram. Box 1.4: The ConstructionBoom and Input Shortages Over the last ten years construction has accounted on average for 9.3 percent o f non-agricultural GDP, and in the last five years the sector has been growing at 10 percent. Between 1999 and 2005 employment in the sector has been growing at 95 percent (based on MOFED national accounts data and National Labour Force Survey, Central Statistical Authority). Both private and public sector investment are driving construction activity, which is also enabled by favourable lending conditions allowing investors that meet collateral requirementsto borrow as much as 70 percent o f project costs. Over 2006 particularly severe shortages o f key construction inputs have been registered, triggering price escalations and delays in the delivery o f buildings.As an example, the demand for cement was almost 60 percent higher than the available supply. High international fuel prices exacerbated the cement price increases, as fuel accounts for about half o f the price o f cement. The limited local supply o f cement and the arrangements for its marketing and distribution have led to a segmented market-in June 2006, prices in the informal (secondary) market were about double the price offered by one o f the two Ethiopian factories. These price increases have had significant repercussions on the costs o f building, as cement represents about 23 percent o f the total cost o f building materials. The price hike for inputs, while significant, appears to have been exacerbated by temporary factors. Starting in 2007 increased imports in fact have ledto decreases inconstruction input prices. 1.14 Furthermore, the emphasis given to the IHDP within PASDEP also emphasizes its potential for contributing to the broader poverty reduction agenda through its progressive distributional impact. Indeed, recent findings point to the crucial role played by MSEs in absorbing particularly vulnerable categories of workers, such as the unemployed and the young.14 The evidence points to small firms being a last resort employment option, characterized by low dynamism and facing significant constraints to expansion due to the perceiveduncertainty o f the regulatory environment and difficulties inaccessing factor markets, 13 Other advantages o f promoting MSEs discussed in the literature include their contributions to competition, entrepreneurshipand innovation, creation o fproducts which are more suitable for the poor, as well as political and social dividends. l4 For example, evidence from the urban panelreveals that self-employment is a residual or last resort option inthat flows are largely from the ranks o fthe unemployed and new entrants. The 2005 LFS reveals that over 80 percent o f employed youth work in the informal sector. Note that the defmition o f informal sector in the LFS is substantially overlapping with the one o f MSE, as the former is defined as a fm with less than 10 workers or not having a license or not keepingbooks o f accounts-where the latter two requirements are simplifiedfor small firms. 7 particularly land and credit (World Bank 2007). Supporting this sector can therefore contribute to strengthening the livelihoods o f workers who would otherwise find themselves with only dead end labour market options, offering inadequate remuneration to guarantee them a life out o f poverty. Aims and Organization of this Study 1.15 Against this background, this study aims to provide an assessment o f the effects o f the Addis Ababa Integrated Housing Development Program on employment creation. The novelty and complexity o f program design suggest that much can be learnt from assessing the Addis experience, which can then contribute to both its scaling up to other cities and the setting up o f appropriate monitoring and evaluation systems. 1.16 The approach o f the study i s very focused; we take as a given the housing construction element of the program and consider exclusively the employment creation effects o f the programthrough its Micro and Small Enterprise component. The main question that we aim to address i s whether the program has created more jobs than would have been created relying on existing firms. 1.17 As will be further discussed in Chapter 3, to perform such an analysis we test the two main assumptions at the heart o f the economic rationale for the program: (i)MSEsintheconstruction sector, andparticularly MSEsintheprogram, aremore labour intensive than larger firms. (ii)The creation and support offeredbyIHDPto MSEsinthe program is neededto overcome market failures. 1.18 Other aspects which can contribute to strengthening the impact o f the program will also be analyzed, including the profile o f the program beneficiaries as compared with the stated objectives o f the program in terms o f targeting. Together with an analysis o f the distributional consequences o f the job creation, this helps inplacing the program withinthe broader context o f a poverty reduction strategy in urban areas. Furthermore, analyzing the distributive impact o f the program helps in better understanding the trade-off between efficiency and equity that characterizes the program. This is particularly relevant since existing assessments o f the program show that the housing component per se has not had so far a pro-poor impact, so it i s the employment creation componentthat is expected to perform this role.I6 1.19 This report is organized as follows. The next chapter provides more detailed background information on the program's operation and goals. Chapter 3 focuses on the methodology for this study, by articulating the key hypotheses tested in our analysis and identifying its limitations. The chapter also describes the main source o f information for this l5The lower-cost technologies adopted by MSEs in the program should allow greater labour intensity per unit o f investment than the technologies used by f m s outside the program, though this is not explicitly addressed in the programdocuments. l6Existing studies of the distributional impact of the housing componentwere based on the Bole-Gerji pilot. While changes have been made to the mechanismadopted for the allocation of housing, it i s not clear whether the current lottery mechanismeffectively improves such distributional impact. 8 study, a purposively collected survey o f firms and workers inthe construction sector. Chapter 4 addresses the central issue o f this report by providing a snapshot o f the beneficiaries o f the program, focusing both on firms and workers and reviewing the effects o f the program interms o f employment creation, both in static and in dynamic terms. Chapter 5 provides evidence on the effectiveness o f the support offered by the program, and provides some evidence on the general equilibrium effects o f the program. Finally, Chapter 6 provides an assessment o f the distributional effects o f the program through its employment creation effects. Chapter 7 reviews the main conclusions emerging from the study and draws the policy implications for the program, particularly inlight o f its ongoing scaling up. 9 2. THE ADDIS ABABA INTEGRATED HOUSING DEVELOPMENT PROGRAM:A DESCRIPTION1' 2.1 This chapter provides a brief overview o f the main elements o f the Integrated HousingDevelopment Program (IHDP) as currently implementedinAddis Ababa, based on existin program documents and consultations with the IHDP office in Addis on earlier drafts.l$ PROGRAMOBJECTIVESAND TARGETS 2.2 The IHDP was launched inlight o f the massive shortage o f housinginAddis Ababa (estimated at 300,000 in 2004) and the relatively poor housing conditions across the board (70 percent of the urban population in Ethiopia live in slumshb-standard living conditions, according to the 2004 Millennium Development Goals (MDG) Needs Assessment). The current program, launched in2005, built on the experience o f the L o w Cost Housing Project, a pilot project conducted by GTZ, GTZ-IS and the Housing Development Program Office (HDPO) in 2004-05, in the Bole Gerji area.lg The IHDP vision i s to improve the living standards o f Addis residents, especially low-income citizens, through the creation o f employment opportunities and the provision o f affordable housing. This vision translates into a number o f more specific objectives: i)regenerating the slum areas o f the city; ii) increasing the land delivery amount in the inner city as a process of densification; iii) promoting micro and small-scale enterprises, which can absorb more labour force and operate at a lower overhead cost; iv) promoting cost efficient housing construction technology; v) empowering citizens o f the city through ownership o f houses and tenure security; and vi) changing the image o f the city. 2.3 The need for an integrated program emerges from the diagnosis that "[tlhe market cannot deliver low-cost housing at the required quantity and reasonable price" (Project Profile, pl). On this basis, the program aims to promote low-cost and low-skill intensive technologies which can be deployed in a short period of time, by involving micro and small- scale enterprises (MSEs). As previously mentioned, this analysis focuses exclusively on the MSE support and employment generation aspects o fthe IHDP. 2.4 The program has set itself specific targets summarized in Table 2.1 below (the targets are revised periodically depending on circumstances-the table below shows the "OurspecialthanksgotothededicatedstaffoftheIHDPcentralofficeforhelpinginthecompilationofthe information in this chapter and in clarifying discrepancies between existing program documents and newer sources o f information. l8These interactions have revealedhow thinkingonprogram design has evolved during implementation, particularly after the completion o fthe Bole Gerji pilot. For example there is now a greater emphasis on "affordable housing" for the lower middle classesthan inearlier phases when the program was mostly characterized as based on "low cost" technology. 19Note that when inthe text we refer to the IHDP as "low-cost housing" (as this is one o f the aims o f the project) we do not aim to refer specifically to the GTZ and GTZ-IS pilot. 10 original targets, and the targets as currently set). Over the period 2006/7-2009/10 the IHDP aims to construct 192,500 houses (in addition to the 32,000 units already underway from 2005/6), generate 80,000 job opportunities, support 1,300 existing MSEs and create another 1,000 new ones. In addition, the program aims to reduce slum areas by 50 percent and to strengthen the construction sector by developing 1,200 ha of land, promoting low-cost technologiesY2'changing training systems, ensuring minimum construction standards and developing the institutionalcapacity requiredto construct 50,000 houses each year. 2.5 It is commendable that such specific targets have been specified, as this allows regular monitoring of progress. Sub-cities are in charge of the monitoring by collecting data on each building site, and on participating MSEs. They report monthly to the central IHDP office on the number of MSEs created; the number of jobs created; the amount of productionhervice each MSE produces; and the contracts awarded and their recipient. The reports from the sub-cities are reviewedperiodically by the central IHDP office. 2.6 It is worth noting, however, that not all these indicators are easy to measure. Particularly challenging i s the "job opportunity" indicator mentioned in the program documents as such an indicator requires specifying and monitoring the duration of the job opportunity in question as well as its sustainability. The program has not yet identified specific indicators that would take into account boththese dimensions (e.g. establishing how long ajob should last)*l or determined how these should be monitored (e.g. if a beneficiary works ontwo differenthousing sites, how to avoid double-counting this as twojobs). 2oSee GTZ & Ministry of FederalAffairs, "Low-Cost Housing:TechnicalManual" Addis Ababa, 2005, p. 7 for *'a Notethat moredetaileddescriptionofthese cost-savingtechnologies. eveninternationalstatisticalconventionsoffer little guidanceonthe minimumrequirementsfor work durationto be consideredajob (they establishthat workingfor at leastan hour a week is a minimum requirement,butthat countries shouldrefine sucha conceptbasedon localcircumstances).InAnnex 2, chapter 4, we discuss the ideaof estimatingfull time equivalents as a measureofjob creation.This mightinvolve estimatingthe totalnumber ofmonthsworkedby workers anddivide it by 12. 11 Typeof goal Original Targets2004-2008 Revised Targets2006-2010 Housing construction Construction o f 150-200,000 Construction of 192,500 housing units in houses 4 years (32,000 units already underway 2,000 MSEs 1 Increasethe number ofMSEs to 2300 Slum upgrading II II Reduce slum .- Reduce slum and decaying areas o f Land development 1I the . - and decaying areas ofthe . - city by 50% Prepare and develop 1,200 ha of I Prepare I city by 50% and develop 1,200 haof land landrequired for housing required for housing development and development and relatedlocal related local development works development works InstitutionalCapacity Build institutional capacity capable Build institutional capacity capable of o f building 50,000 houses per year building 50,000 houses per year NON-QUANTIFIABLEPROGRAMTARGETS 1 Changethe current training systems inthe construction industry 1 Ensure the filfilment of minimum construction standards 1 DeveloDandwidelv utilise low-cost construction PROGRAM ADMINISTRATION 2.7 The IHDP is managedby a Housing Development Project Office (HDPO) incharge of implementing and coordinating the program. The Project Office's responsibilities entail preparing land and ensuring the supply of infrastructure facilities, executing design works, procurement and distribution of major construction inputs, MSE support, human resource management, financial management, expansion of construction capacity, as well as co- ordination and supervision of the activities of the various institutions involved in project work. A number of these tasks are delegatedto Sub-City Project Offices23which are set up to coordinate the program at Sub-City level. Administering construction sites, collecting and delivering construction materials and supervising their use, selecting and organizing MSEs, selectingand preparing landare tasks ofthe Sub-City Project Offices. 2.8 The Project Office works closely with other government offices such as the Department of Trade and Industry, the Ministry of Urban Works and Development and the City Administration MSE department, which in charge of MSE support to all the priority sectors (textile and garment, food and beverage, metal and wood work, construction and municipal services). It also cooperates with GTZ, which has been involved inthe design of the housingprogram, for theprovisionoftechnical support andas an implementingpartner. 22This table relies on two separate sources: the housing construction, employment creation andMSEpromotion targets are from a presentation by IHDP to the World Bank in October 2006, while the targets for slum upgrading, land development and the non-quantifiable targets are taken from "Addis Ababa Integrated Housing Development Program" (no author), Addis Ababa, March 2004. 23Addis Ababa is divided into 10 Sub-City Administrations, with delegated responsibility for service delivery. The Sub-City Project Office refers to staff o f the IHDP in charge o f IHDP implementation who are located at the sub-city administration offices. 12 THELOWCOSTTECHNOLOGY ADOPTEDBYTHE IHDP 2.9 The IHDP has developed a production process that deviates from the one conventionally used inthe private sector. The low-cost aspect arises from homogeneous type o f housing, using novel construction technologies, cheaper inputs, fixed-price contracts24and a standardized production procedure permitting greater specialization. The main features o f this process are: 0 Standardization. The condominiums constructed by the program are fairly standard 4 to h t o r y apartment buildings (Ground-plus-3 or Ground-plus-4 levels). IHDP clients can only choose between a studio or a one-, two-, or three-bedroom a~artment.~' 0 Quality of housing provided. The condominiums constructed by the program are less luxurious thanhousing units constructed by the private sector. 0 Introduction of new technologies which call for different inputs. The two most prominent new inputs are pre-cast beams and ribslabs (prefabricated hollow blocks), which reduce material inputs and the needfor formwork. 0 Almost all inputs used in IHDP construction are produced by firms selected to participate in the program who sell their inputs at a fixed price below the market price. 0 Differences in terms of process (Table 2.2) which allow savings on designcosts and require support from sub-city staff for project supervision and procurement. Itis not possible, on the basis ofproject documents, to ascertain which elements account for the largest contribution to cost reduction. In particular, it i s not possible to capture fully the importance o f administrative costs (see below). 24Note that for specific activities ad hoc revisions o f prices have been undertaken (e.g. since prices for the sanitary and installation activities have been revised 3 times). The HDPO is aware of the need to build the institutional capacity for the systematic revisions o f contracts, and is getting organized for this purpose. 25GTZ has been making an effort to increase diversification inthe design o f condominium construction. Table 2.2: A Comparisonof ConstructionProcesses:Private Sector and M D P Stage of Construction Private Sector (LargeProjects) IHDP Process Design Consultant responsible for i)proposing rn Design is uniform and therefore a design ii)preparing the bill o f cheaper than in the conventional quantities (document detailing all the production process inputs required for construction, Program engineers check quality thereby enabling the estimation of construction costs); and iii) supervision, coordination o f construction activities and quality control NB design cost is approximately 4 percent o f total construction cost in Ethiopia Procurement Tender document based on the bill rnProcurement i s largely taken care o f quantities o f by the sub-city project offices Bids for the tender are collected based on participating firms. ftom contractors and a winner is Fixedprice system selected to sign a contract with the Profit margins lower than in the client, making him responsible for private sector the construction o fthe building. Purchaseo f inputs Purchase o f material inputs is typically B Sub-city project office organizes the responsibility ofthe contractor. the supply and distribution' of almost all outputs (from production MSEs inthe program) and purchases inputs such as cement, reinforcement bars and iron. (Key) inputs can usually be obtained on credit. Construction The contractor may either conduct each rnSuper-structure: by contractors i)sub-structure activity himself or subcontract some or with license grade 5 and lower, (excavation, foundational all activities to thirdparties. rnProgram MSEs construct the sub- works, etc) structure and do the finishing ii)super-structure(walls, where possible; otherwise, done roofing, etc), by non-program MSEsISMEs iii)finishing(sanitary& andor small contractors electrical installation, Supervisory activities are carpentry, etc) conducted by program engineers Analysis for the Housing Construction PROGRAMPARTICIPATION: PROCEDURES 2.10 The IHDP uses MSEs to produce construction inputs and build low cost housing. To this end, the program aims to support the development of MSEs capable of adopting low cost, labour and low skill intensive technologies (described above), to award them contracts andto provide themwith a wide array of support interventions. 14 2.11 The MSEs participating inthe program engage inboth construction and production activities. Construction activities include block work or wall construction, electrical installation, sanitary installation, finishing (ceramics, tile laying, painting), and site works. The production activities include pre-cast beam production, hollow block production, metalwork, woodwork, stone work and gravel production (the latter known as aggregate production). 2.12 Prior to the introduction o f the program, pre-cast beam production andprefabricated slab production had not yet been used in Ethiopia, nor were there MSEs which specialized exclusively in electrical and sanitary installation. Support to MSEs to specialize in these activities i s in line with the program's aim to strengthen the construction sector by fostering specialization and stimulating diversification. Creation of MSEs 2.13 The process o f creating an MSEoccurs inthree steps discussed below inturn. This i s a general process which affects all MSE support activities and is therefore not exclusive to IHDPparticipatingMSEsonly. (i) Registration of interested and eligible individuals. Kebele (the smallest unit within the city administration) MSEs promotion offices in each sub-city announce that individuals interested in participating in the program can register to apply. Announcements for registration can be repeated as necessary and are not restricted to the IHDP-firms can register in any one o f the growth sectors that are promoted under the city-wide MSE support program, which includes food processing, textiles, leather, etc., in addition to the construction industry. In order to be eligible to register, an individualhas to have a valid identity card and have either graduated from a TVET college26 or show proof o f experience. For the IHDP, experience in the construction sector (formal or informal) i s required, thus building on the capacity o f the informal sector. According to the guidelines, men and women should be treated equally. Thus far, the total number o f registered candidates for IHDP figures over 21,000.~~ (ii)Testing. All eligible applicants are tested to verify that their skills meet the minimum standards. The test has a theoretical component which accounts for 20-30 percent o f the score and a practical component accounting for 70-80 percent, thus placing a premiumon practical skills. The pass rates vary between 50 percent and 68 percent. The tests are organized by the sub-city SME Development Office in collaboration with the Project Office. Test centers are TVET schools or other technical training 26 Note that while the TVET system was founded long before the program started and is independentfiom the IHDP, there are a number of links betweenthe two. The IHDPrelies on the skills of TVET graduates, whilst providinga learning opportunity for bothTVET students and instructors, as well as employment opportunities for TVET graduates. Inaddition, the designofthe TVET curriculum is adjustedpartlyinresponseto the needs o f the IHDP. Furthermore, the number o f TVET graduates trained in construction-related activities is adjusted to meetthe projecteddemands ofthe constructionsector at large, andthe IHDP inparticular. "AccordingtotheProjectOffice, registrationwasdiscontinuedinearly2006whenitwasdeterminedthat there were enoughemployees/fmsto carry out the full program. 15 centers which provide relevant skills. The costs o f the test are covered by the Project Office. Individuals who fail the test are allowed to take the test again at a later date. They may attempt to upgrade their skills by joining successful candidates on the job as assistants. (iii) Enterpriseorganization(MSEs).Candidates who have passedthe test are provided with a one-day orientation course which teaches them how to form a legal business enterprise, how contracts work, what the IHDP objectives are, and what i s expected from IHDP program enterprises. Candidates then make groupings on their own, and identify the type o f organization they would like to establish. They can choose between three different organizational structures: a cooperative, a trade association (share company, Plc or partnership) or sole ownership. A cooperative comprises a group o f at least 10 people who jointly own the firm and share its profits. Most candidates choose to become a member o f a cooperative, as this allows them to combine their skills, take on larger contracts, share the financial risk, and offers the advantage o f exemption from profit tax (despite the potential disadvantage o f free- ridingthat can arise because o fprofit sharing). Awarding of Contracts 2.14 MSEs they register at the Kebele Micro and Small Scale Enterprise Promotion Office one-stop shops), and secure certificates to operate as business entities. Such certification allows the MSE to submit a job request to their respective sub-city project office. Depending on when they registered and the availability o f work intheir area, MSEs are awarded contracts and work under the supervision o f engineers and foremen from the Project Office. MSEs registered ina given sub-city will have priority for constructionworks inthat sub-city over others outside the sub-city. Inthe future, the IHDP also intendsto make the offering o fjobs dependent onpast performance. 2.15 Only MSEs formed through this process are awarded IHDP jobs. However, when newly organized MSEs are unwilling or unable to complete certain works, pre-existing licensed MSEs are invited to take up the job. Anecdotal evidence suggests it i s not very common for pre-existing SMEs to re-register in order to obtain IHDP work. It i s only for foundation and structural works, which are generally beyond the capacity o f MSEs that large contractors are usually hired. PROGRAMPARTICIPATION: BENEFITS 2.16 The IHDP provides wide-ranging support to MSEs by providing, and in certain cases subsidizing, a place to work, facilitating access to credit, providing training and access to inputs (on credit) and subsidizing machinery for firms producing re-bars (reinforcement bars) or hollow blocks, more specifically: 0 Land Grant. By enabling access to land and subsidizing certain types of land and sheds, the IHDP tries to ensure MSEs have a place to work. The program provides certain sites, such as TVET compounds, for free if the working shed i s built with wood (80 percent of cases), while it charges full rent if the shed on the site was 16 constructed with metal (approximately 20 percent o f all cases). In contrast, plots provided by the Bureau o f Trade & Industry are not subsidized. 0 Access to credit without collateral (through a joint bank account). The IHDP does not extend credit itself, but rather connects program firms with existing Micro- Finance Institutions (MFIs). The collateral requirements for program firms and non- program firms are different as firms inthe program can open a bank-account together with the IHDP without providing the collateral which is normally required. The program's office signature provides the required collateral, even though the project office i s not responsible for repayment in the case o f defaulte2*In all other respects (interest rates, repayment periods and grace-period) program firms face the same lending conditions. 0 Inputs on Credit. The IHDP provides re-bars, iron and cement (when not available inthe market) on credit. The costs o f inputs are deducted from future payments for the outputs. 0 Subsidized Machinery-Lease. Program firms producing re-bars and/or hollow blocks can purchase machinery with leases below market rates; the IHDP communicates to the MFIs which machineries to buy and to which MSEs to lease them out; upon completion of all payments, the program firms own the machine. Again, payments are not made upfront but deducted from each payment made to the MSEby IHDP. Training. Firms which engage inpre-cast beams and/or hollow block production are trained before deployment. The program also intends to provide training for firms which engage in other activities, but to date, no systematic training system has been set up for that purpose. In the future, the IHDP also wants to organize courses to teach business skills. 0 Demand. Awardingwork to new firms and shielding them from competition by non- program firms is perhaps the most important support the IHDP provides to program MSEs. Firms which have beencreated by the program are free to carry out jobs for clients other thanthe IHDP. 2.17 While the IHDP is aimed at sustaining program MSEs, other firms can also benefit from its contracts, most notably contractors. Large contractors (Grade 6 and above)29 are hired for foundational and structural works, on a fixed price basis, at least in the past - a anecdotal evidence suggests that such fixed price system has been disincentive for large 28 However, the program MSEneedsto obtainpermissionfi-omthe IHDP every time it withdraws money from the shared bank account in which the loan has been deposited. The IHDP and the MFIs have a general agreement that the IHDP automatically repaysthe MFI once a fm completesa contract, whichreducesthe risk of default for the MFI. The IHDP deducts the repaymentsfrom the amount it pays out to the firm. Supervision by quality engineers is supposed to reduce the risk of default for the IHDP by ensuring that program firms successfullycompletethe contractsthey have beenawarded. 29Contractorsare gradedbased on capitalrequirementsand skills. Contractors of grade 7 and above have more than 500,000 Birr incapital andmeet certain skillrequirements. 17 firms and that many have been reluctant to register for these contract^.^' The selection of contractors i s inthe order inwhich they registered with the Project Office, unless the number o f registered contractors exceeds the number o f available jobs (inwhich case a draw is held). Contractors who participate inthe program do not enjoy any special benefits or support. PROGRAM COSTS 2.18 It is difficult to obtain an accurate estimate of the full costs of the program due to price fluctuations, incomplete cost accounting, decentralized administration and delays in program implementation, despite the best efforts o f the Project Office to collect this data. Giventhe ambition o f the program, it is to be expected that its costs are large. However, it is believedthat available estimates o f costs are biased downward. 2.19 Some importantelements emerge from Table 2.3. a The cumulative administrative and MSE support cost incurred by 2006 are equivalent to 6,756 Birr per MSE created and 288 Birr per housing unit currently under con~truction.~' a The development o f infrastructure has been the biggest expense in the provision o f MSE support, while the provision o f quality controls has been the highest administrative expense. a Costs related to technology, such as testing and training are relatively modest. This partly reflectsthe limitedimplementationofthe training component. a The costs o f site development have also been fairly modest thus far though this largely reflects the low administrative price of land. Lease prices for program firms are lower than for non-program firms, a subsidy value which i s not reflected inthese cost estimates. 2.20 Program costs appear to exceed the initial estimates, probably due to the shortages o f buildingmaterials (especially cement) and hightransportation costs (rising fuel prices). In 2004, the Project Office estimated the price o f IHDP housing at 850-860 Birr per m232or 30,000 Birr for an average unit sized 35m2 (excluding the cost o f land and infrastructure development, which would add another 2,500 Birr per housing unit).33According to more recent data provided by the IHDP Project office, these estimates were too optimistic. At the end of2006, the actual total cost o f construction (excluding only landusage) was 42,645 Birr per 35 m2 apartment: Le. 31 percent higher than the original estimate. The cost increase seems driven by increasing input costs, most notably the costs o f cement and metal, but also by rising wages. It should be noted that a cumulative cost increase o f 27 percent in three 30 Contractors of lower grade (7-10) and SMEsMSEs are not eligible for foundational and structural work because they do not posses the required constructionmachinery, equipmentandor skills. 31The total numberof MSEs createdthus far is 1,386; the total number ofhousing units is 32,495. 32"Addis Ababa IntegratedHousing Development Program" (no author), Addis Ababa, March2004. 33According to estimates made in2004, about 50m2 of land per house needs to be preparedat an average cost of 50 Birrper m2, so that the average cost of land preparationper house amountsto 2,500 Birr. 18 years i s broadly in line with aggregate inflation. 34 Thus at current prices the costs would be close to 9 billion birr (about US$1 billion), excluding the cost o f administering the program. Table 2.3: ProgramCosts Support & Administrative Costs Description Years Cumulative Total 2003l2004 2004l2005 2005/2006 (1996 E.C) (1997 EX) (1998 E.C) Administrative Costs Salaryofpersonneldirectly 25,584.00 315,198.00 440,100.00 780,882.00 involvedto promote SMEs in IHDP Management. 13,950.00 561,888.00 739,154.00 1,314,992.00 Quality control 191,537.00 1,294,475.00 1,486,012.00 Stationeries 625.00 33,682.00 33,483 .OO 67,790.00 Fuel 2,400.00 71,647.00 76,024.00 150,071.00 Utilities 1,410.00 33,914.00 55,359.00 90,683.00 DepreciationofVehicles 59,536.00 71,912.00 131,448.00 Others (maintain) 1,120.00 4,480.00 2,240.00 7,840.00 Sub-Total 45,089.00 1,271,882.00 2,7 12,747.00 4,029,718.00 support costs Trade testing 1,200.00 162,550.00 9,056.00 172,806.00 Training 84,657.00 22,235.00 106,892.00 BuildingofProductionSheds 796,549.96 123,876.00 920,425.96 TransportationCost for 99,588.00 6,813.00 106,401.OO productionmaterials Infrastructure 3,356,203 -62 385,304.00 3,74 1,507.62 Installationofproduction 700.00 700.00 materials Sitedevelopment 280,180.00 5,000.00 285,180.00 Sub-Total 1,200.00 4,779,728.58 552,984.00 5,333,912.58 GrandTotal 46.289.00 6,051,610.58 3,265,731.00 9.363.630.58 , , Source: Project documents 2.21 The program is financed mostly through the sale of housing units on a lease basis with repayment periods o f up to 20 years. To meet its immediate financial obligations the IHDPrequires down payments varying from 7.5 percent for a studio to 30 percent for a three bedroomapartment. Reportedly, however initially beneficiaries have been chosen based on their ability to pay the entire price upfront, rather thanrelyingon a downpayment system.A lottery is now usedto allocate apartments, with a 10 percent o f units reserved for those who have been displaced by the program. The Project Office sells the commercial facilities built inthe context ofthe condominium's development on a marketbasis. 2.22 Rising costs, delays in project implementation due to input shortages, and the events following the 2005 elections appear to have resulted in some financial strain for the 34The consumerprice index inEthiopiahas beengrowingrelativelyfast over recentyears: 8.6 percent in2004, 6.80 percentin2005 and 12.3 percentin2006. 19 program. The original price structure for example, contained an element of cross- subsidization between larger and smaller units, in order to provide lower cost housing to the poorest. But this was affected by the rise in prices as subsidies have de facto been extended also to larger units. ACHIEVEMENTSTO DATE 2.23 Despite the challenges provided by input shortages (particularly cement) and the events following the 2005 elections, the IHDP has delivered significant results (Table 2.4). However, the indicator currently used for monitoring the job opportunity objective could be more accurately refined to account for different durations o fjobs and avoid double counting. A measure o f "days-worked per year," would provide a more accurate description of what is currently possible to monitor. Further, a distinction between "jobs" for casual workers and those for permanent workers would be useful. Table 2.4: Program Targets and Resultsas of October 2006 Quantifiable Program Targets: Typeof Goal Target Accomplishmentsthusfar Housing Construction of 192,500 32,495 housing units construction housing units in4 years Employment Create 80,000 job opportunities 52,600 job opportunities created Creation MSE promotion 1 Strengthenthe existing 1,300 1,386 MSEs developed MSEs 1 292 inconstructiontrades 1 Increase the number of MSEs 1 1071inproduction trades to 2,300 Slum upgrading Reduce slum and decayingareas ofthe city by 50 percent Land development Prepare and develop 1,200 ha of 173.8 ha of land developed land required for housing development andrelated local development works Institutional Buildinstitutional capacity 1 20,000 applicants tested Capacity capableofbuilding 50,000 1 3,500 workers trained innew houses per year technologies 1 800 workers trained on thejob Sources: The housing construction, employment creation and SME promotion targets are IHDP (2006) and ftom the Project Description (No author, 2004, p. 3). Accomplishments are from IHDP (2006) 2.24 Other elements not reported in the table need to be considered in a technical evaluation o fthe IHDP effectiveness ingenerating employment. These include: The possibility o f crowding out o f existing firms and projects in an input supply- constrained market sojob creation could be lower than expected; The need for more careful targeting o f program support to enhance the likelihood o f MSE survival (EDRI2003, also note that evidence from other countries suggests low survival rates); 20 Possible trade-offs between low costs due to high standardization of inputs (which favors efficiency and might be better exploited by larger firms) and employment generationvia MSEs; The possible trade-offs between speed of implementation and quality standards, as well as betweenhighlevel ofcoordinationby the programand low cost; and The possibly distortionary effects o f a fixed price system, which may result in inefficient production. 21 3. METHODOLOGYAND DATA SOURCES 3.1 The IHDP is a complex and innovative program and inthis report we only assess its employment creation effects, taking as a given the demand for housing from the program. The question is whether the creation o f support of MSES which use input saving technologies that are novel in Ethiopia creates more jobs than if the program had taken a different approach to constructing housing. 3.2 The identification o f an appropriate counterfactual i s key for this analysis. The typical practices outside the program constitute the natural counterfactual against which to compare the IHDP. Our analysis looks at the whole spectrum o f firms in the construction sector inorder to assess this co~nterfactual.~~ 3.3 The complexity o f the program, its size as a significant player in input markets and its potential dynamic effects limit our assessment on the employment creation effects based on cross-sectional data. Yet, our assessment can provide very useful insights into whether the program i s having the desired impact-at least ina static sense-and on the channels through which it has this impact. Further, this type o f analysis can highlight potential discrepancies betweenexpected results and what is occurring inpractice, thereby identifying areas where corrective actions may be called for ifthe original goals are to be reached. 3.4 The next section reviews the analytical framework, both interms o f key hypotheses and analytical tools used to test them, in order to provide the methodological background for the analysis presented in the next chapters. This provides the background for the design of the survey discussed inthe following sections. KEY HYPOTHESES AND ANALYTICALTOOLS 3.5 There seem to be are two main hypotheses which are relevant to the economic rationale for the program: (i)MSEsinthe constructionsector, and articularly MSEsinthe program, are more labour-intensive than larger firms.r36 3.6 Testing this hypothesis requires examining in detail two related but theoretically separate issues. One is the technology37used per se, and the other i s testing for differences in 35 Our initial focus was on large contractors and MSEs. As discussed below, it has not beenpossibleto sample non-contractors based on size. However, the vast majority of firms who are not contractors appear to be relatively small. For this reason, we refer to them more generally as MSEs. 36 The lower-cost technologies adopted by MSEs inthe program should allow greater labour intensity per unit o f investmentthan the technologies usedby f m s outside the program, though this is not explicitly addressedin theprogramdocuments. 37 Note that "technology" is used here in the economic sense. This concept refers to the way inputs are transformed into outputs when looked at interms ofvalue, rather than to the specific material technology used. 22 factor proportions3* given the technology used, since technology and factor proportions jointly determine labour intensity. These two issues are explored through a production function framework, whereby the output o f each firm i s modeled as a function o f the inputs as well as o f firm characteristics (e.g., being a large contractor, being an MSE, participating inthe programor not), andby direct modeling o f factor proportions. 3.7 An important caveat is that differences in factor proportions and technologies do not tell us what would have happened ifthe program hadnot been implemented.The current choices made by non-program firms provide some indication as to factor proportion and technology choices o f incumbents and potential entrants, had the program not been introduced. Yet, they are not perfect indications since the implementation o f the program itself may have affected (the evolution of) (i)the prices and availability o f inputs; (ii) the prices o f and demand for outputs; and (iii) available technologies. the (ii) creationandsupportofferedbyIHDPtoMSEsintheprogramisneededto The overcome market failures 3.8 Ideally this would be tested with longitudinal data (information collected at different points in time for the same firms) and for a randomized assignment o f program support. While our study, based on data collected at one point o f time only, does not allow a full impact evaluation o f the support provided by the program, it can provide some useful insightsinto effectiveness. 3.9 We compare whether support i s effective in an expanded production function framework, contrasting firms which receive support with firms that do not. An assessment i s also conducted o f the impacts o f program support on employment growth. Finally, we complement the analysis with firm managers' perceptions o f the value o f the support received, and their expectations o f what would happen if program support were withdrawn. The combination o fthese findings provides detailed evidence on channels o fimpact. 3.10 Inaddition to testing the two mainhypotheses, we also analyze the effectiveness of program targeting, and its distributional consequences based on a profile of beneficiaries and an analysis o f the benefits they receive from the program. This seems particularly relevant inassessingwhether the program is an effective way o f targeting the urbanpoor, and its role in the context o f an effective poverty reduction strategy. Furthermore, analyzing the distributive impact o f the program helps in identifying trade-offs between efficiency and equity that characterizes this program. This is particularly important because existing assessments o fthe program suggest that the housing component per se has not hada pro-poor impact, meaning that it i s the employment creation component that i s expected to perform this role. 38Note that a given technology encompasses all the combinations of capital, labour and inputs which give a certain outputs. Given desired output levels, different f m s could therefore decide to adopt different combinations of inputs (Le. factor proportions) to attainthat productionlevel. 23 Limitationsof the Analysis 3.11 The most serious limitation o f this study is our reliance on cross-sectional data collected when the program had already been operating for at least two years. Any exceptional circumstances at the time o f the survey would affect our results. Infact, late 2006 was the most acute phase o f the input shortage which hampered the construction sector. To the extent that firms inand outside of the program were affected differently by the shortage, this could indeed cloud our comparisons. Given the emphasis of the program on facilitating access to inputs, it i s unlikely that their access was worst than for non-program firms. This particular circumstance should not therefore bias negatively our assessment o f effectiveness in supporting firms' performance. Furthermore, the indicators o f firms' performance are based on recall o f the last 12 months, a factor which should attenuate the effects o f the peak o f the shortage (as well as affect similarly both program and non-program firms). It i s possible, however, that our assessments o f capacity at which firms were operating might have been affected by such transitory factors. 3.12 It is not possible to assess the full impact of the support provided since the absence o f randomization o f program treatment implies that there may be something specific about the firms who received (or who did not receive) program support which drives both the performance o fthe firm, and whether support was received3' cannot be eliminated. 3.13 The lack o f longitudinal data means that we cannot account for the effects o f the program on non-participants, though complementary information could shed some light on the issue. This applies inparticular to assessinghow the program has affected the likelihood o f entry o f non-participating firms into the construction sector and the evolution of input prices. 3.14 Our analysis cannot account for the dynamic effects o f the program. Nevertheless we can estimate the relative importance o f static and dynamic factors in driving the employment creation effects o f the program, which can have important implications for the design o f its scaling up. 3.15 Finally, we recognize that this assessment glosses over the specificities o f the e program. For example, when looking at technology and factor proportion we focus on elements such as the capital/labour ratio rather than the specific nature o f the equipment used inthe different trades. This allows us to conduct an overall analysis which `focuses on the main economic drivers o f performance, but specific challenges faced in one particular activity or by a certain set o fworkers are not be addressed. 39For example, we cannot rule out the possibilitythat programsupporthas beengivento firms whichotherwise would havebeenmost likely to fail. Ifthat is the case, anestimateofthe effectivenessofthe programwould not be able to distinguish betweenthe (say) negative effects of beinga fm likely to fail as opposedto the positive effects ofreceivingsupport from the program. 24 THEADDIS ABABACONSTRUCTIONENTERPRISE SURVEY 3.16 In order to test the key hypotheses above a purposely-designed survey was conducted by the World Bank in collaboration with the Ethiopian Economic Association. The Addis Ababa Construction Enterprise Survey (AACES) included quantitative information on firms and their workers, and was complemented by qualitative evidence collected by means o f in-depth interviews with respondents. The reminder of this chapter details the approach adopted in the development o f the survey and sampling, and the challenges it posed. 3.17 The survey matches firms and workers in the construction sector in Addis Ababa, and covered 240 firms and 971 workers, 241 o f whom were casual workers.40 The sample is equally divided between program- and non-program firms, containing data on 120 in each category. A program firm fulfils at least one o f three criteria: (i) having been created by the program; (ii)having received support from the program; or (iii)having worked for the program. Inaddition, a distinction can be drawn between MSEs and large contractors, where contractors are defined as firms which have a contracting license grade between 1 and 6. MSEs are all other firms, including small contractors, i.e. firms with a contracting license grade between 7 and 10, as well as some non-contracting firms which are relatively large. Yet, the majority o f MSEs are small non-contracting firms. The differences between MSEs and large contractors will be shown to be considerable; large contractors tend to hire more workers, be more capital-intensive, use more inputs and have higher labour productivity. The sample distribution is shown inTable 3.1. Table 3.1: Compositionof the Sample Type of Firm Number of Firms Non-ProgramMSE 109 ProgramMSE 103 Non-programLargeContractor 11 I ProgramLargeContractor 17 SamplingStrategy-Firms 3.18 The main emphasis o f this study i s on construction firms, and in particular the MSEs.A stratum o f contractors was added to the sample, inorder to capture the employment creation effects o f the structural works in the condominium construction, as well as to provide a counterfactual on an alternative mode o f delivery o f public housing. Specifics to note include: Given the heterogeneous nature o f activities in the housing construction sector, the sub-samples o f program and non-program firms had to be both comparable and representative. Program firms were sampled on the basis o f their activities, with the 40The sample we ended up with differs slightly from the sample we intendedto have, because it turned out to be impossible to trace the exact intended number of f m s in each (program) activity. Moreover, f m s can engage indifferent activities, a fact whichwas nottakeninto account inthe samplingdesign. 25 sample frequency corresponding roughly to the population frequency as derived from the IHDP list, with a slight oversampling o f firms in activities whose population frequency i s low, and excluding marginal activities. 0 The population o f non-program firms engaging in comparable activities was identified. Again the sample frequency of firms engaging in different types o f activities roughly corresponds to the observed population frequency. 0 A distinction was made between contractors and non-contractors. Contractors with a license grade between 1 and 6 are considered large, as they are allowed to execute structural works. Large contractors were oversampled since there are relatively few o f them. The samplingfor non-contractors didnot take the size o f firms into account.41 3.19 A major challenge o f this study was the identification o f the construction sector universe from which to draw a sample. The Addis Ababa City Administration keeps a registry of all firms which have been licensed in the city. Firms categories were drawn mostly from the construction sector, but included also the manufacturing sub-sectors included inthe program (e.g. hollow block construction andpre-cast beams). 3.20 The listing has several shortfalls for our purposes. Inparticular tracing firms proved to be very difficult; contact details were often conflicting, outdated, or missing altogether, and in half of the cases firms which could not be contacted had to be replaced with others from the same stratum. 3.21 The listings provided by the City Administration were integratedwith those kept by the IHDPprogram administration. Fromthe program listingaplannedhalfofthe sample was drawn. Tracing firms posed difficulties, as some o f the firms on the list had exited and had to be replaced, while others had moved. 42 From the population registries offered by the various available listings, weights have been obtained to make the survey representative o f the whole sector. SamplingStrategy-Workers 3.22 A maximum o f up to 4 workers (including the firm manager) were interviewed, stratified by occupational category. Women, particularly those not employed as casual workers, were oversampled and enumerators were instructed to attempt to sample one casual worker per firm.43 The sample o f workers contains disproportionately more workers from higher occupational echelons.44 41Giventhe information at our disposal, it was not possibleto sample f m s onthe basis of size. 42On the 1406 firms in the IHDP listing we attempted to contact by phone the 721 firms over which the program office had doubts whether they were hlly in operation. O f those 721, 506 could not be contacted and ofthe remaining firms, only 66 were operating. 43Additional efforts were made to collect information fiom casualworkers looking for work at usual gathering places. The quality o f these data turned out to be very poor and therefore these data have been mostly disregardedfor the purpose of analysis. 44Note that inthe worker sample the definition is based on workers' own categorization; it could be that some workers categorizethemselves as skilled labourers,while being categorisedas unskilledby the fm. 26 3.23 One potential caveat i s that the approval o f the manager was needed in order to interview workers and thus the sampling of workers may not be entirely random.45 In addition, response bias cannot be ruled out. Yet, most o f our questions are factual in nature and most managers were cooperative. Our impression was that the assignment o f workers for interviews was more due to chance, i.e., proximity at the time o f our interview, than to strategic selection on the manager's part. Survey Instruments 3.24 Two questionnaires were developed-one for firms and one for workers, with adjustments to differentiate between questionnaires for contractors and non-contractors and for casual and permanent workers. The survey was piloted on 10 firms and 39 workers and slight revisions were subsequently made. 0 The firm-level questionnaire covers a rich set o f characteristics, including their activities, age, size, capital stock, inputs, outputs, expenditure, revenues, organizational and occupational structure, program participation and support, access to finance, inputs and skilled personnel, constraints, expectations, the number and type o f workers they employ, the wages paid and employment dynamics. Data on the volume and total costs o f inputs and outputs, as well as on expenditure and revenues were collected; though appear to be quite noisy. Only 35 percent o f firms in the Sam le keep complete books o f account, while 32 percent do not keep any books at all:' While contractors and large firms are typically better at keeping books of account, they were less capable o f providing accurate information on the precise amounts o f different inputs used. 0 The worker-level questionnaire gathers detailed information on earnings, employment history, experience, skills, educational background, program participation, job satisfaction, motivation for choosing their current activity and on a number o f socio-demographic characteristics including household characteristics, parental background and household assets. Casual and permanent workers were administered slightly different questionnaires. 45The decisionto oversamplewomen was taken after pilotingofthe survey as the piloting revealedthat women were the least likely to be assigned to give an interview. 46Programf m s keptmuchbetterbookso f accounts thannon-programf m s . 27 4. THE EMPLOYMENTCREATIONEFFECTS OF THE IHDP 4.1 This chapter draws on the AACES to present a fuller characterization o f how the program operates. It starts by providing a profile of the beneficiary workers and o f program firms. The next sections addresses the central issue o f this report, i.e. the effectiveness of the program in generating jobs, particularly for the poor, first in the static, then in terms o f dynamic effects. A Profile of Project Beneficiaries: Workers47 4.2 To better understand who the program i s reaching the profile o f beneficiaries, both workers and firms, was analyzed. Highlights include: 0 According to the AACES, workers in the construction sector are mostly men (72 percent), and about 30 years old. Many (47 percent) are migrants, and three quarters migrated as adults.48 Overall education levels are high for Ethiopia, with an average education inthe sector equal to 9.5 years o f schooling. 41 percent o f all workers inthe housing construction sector in Addis are employed by a firm that works for the program (47 percent o f all permanent workers and 40 percent o f all casual workers). 0 Permanent workers are older and better educated than casual workers, work more days permonth andearn more. When asked why they choose to become casual workers, 94 percent o f the casual workers answered they lacked alternative opportunities. Contrary to popular wisdom but in line with the argument made by Garrett (2003), there is no evidence that casual status i s a product o f income diversification; only 6 percent o f casual workers had another income generating activity, compared to 12 percent o f permanentworkers. A much larger proportion of casual employees had an unemployment spell in the past, andfewer had an experience working as a government employee. The labour market position o f women in the construction sector is particularly disadvantaged. While the average years o f schooling are very similar, men earn more than twice as much as women on average (839 Birr vs. 439 Birr per month). This is partly explained by women being much less likely to occupy higher level positions (Table 4.1); they are much less likely to be employed as a manager, engineer, foremen or skilled labourer, but much more likely to be employed as casual workers. Yet, pay discrimination also occurs within occupations; when asked whether they rewarded 47We refer here to workers who are not currentlystudentsand only workers inregisteredenterprises. 48 This is consistent with the argument made in World Bank (2007) on migrants being concentrated in sectors, suchas construction, with low barriersto entry. 28 men and women differently for the same job, 40 percent of managers claimed they paid men more, while only 4 percent indicated paying women more. Thirty-eight percent of firm managers claimed men work harder than women, while 18 percent believe the opposite. Cultural factors thus seem to play an important role in determining pay differences. Some findings on hiring and wage setting from the supplementaryqualitativeinterviewswith firm managersandworkers are discussed in Box 4.1 Box 4.1: Qualitative Evidenceon Hiring and Wages Inthe process o f conducting the AACES, supplementary qualitative evidence was gathered by asking some survey respondents additional questions covering a broad range o f issues. These in-depth interviews were conducted with 3 contractors and 6 MSEs, and included both program and non-program firms. In 4 o f the cases, more than 1 interview was conducted (e.g. workers were interviewed, in addition to the manager). The interviewed f m s are not necessarily representative o f the total population, and we must therefore be careful not to generalize based on their responses-the number o f interviews was quite limited, and may reflect some selectivity bias, since certain types o f firms (particularly large contractors) were reluctant to be interviewed. However, when used to complement the findings from the quantitative survey, the interviews are useful inproviding a richer characterization o f firmbehaviour and perceptions o f the program. Some qualitative findings related to hiringpractices and wage setting are reported here: Social connections play a pivotal role in hiring. Typically, casual workers are recommended by peers andor selected on the basis o f their physical characteristics. Before being hired, both skilled workers and unskilled workers typically have to go through a tryout period. Workers who participate in the IHDP and set up a fm together have typically known each other for an extended period for time and in a context different from the IHDP alone. While some f m s complain o f difficulty in hiring skilled workers, others find it easy as long as the firm can afford to pay the going rate. Skilled workers for installation works, masonry and carpentry seem to be particularly sought after. Women in the construction sector suffer several disadvantages in comparison with men. it is rare for women to occupy a managerial or a professional position, while the majority o f daily labourers are women. Gender pay differentials do not seem to exist for professionals, yet female casual labourers arguably face pay discrimination. Of course, the pay differential may be due to productivity differentials or because women have different tasks than men, yet this does not seem to be the case-as illustrated by the answer o f a firm manager to the question why women get paid less than men: `Women get only 10 birr [per day, as opposed to 12 birrfor men], but I prefer women; infact, 75percentof my daily labourers are women; they work hard and they do what I want them to do, and they don't complain. Infact, for the work they do, I would be willing topay them 15 birr and men 10birr.' wce: Interviews with workers and fmmanagers, Dec. 2006. 29 Table 4.1: Occupational Breakdownby Gender Women Men Total % employedas Manager 4 10 8 %of all managers 14 86 %employedas Engineer 0 9 6 %of all engineers 0 100 %employedas Foreman 0 2 1 % of all foremen 4 96 % employedas SkilledLabourer 26 52 44 % of all skilledlabourers 18 82 %employedas UnskilledLabourer 70 26 40 %ofunskilledlabourers 53 47 Workers in program and non-program firms have somewhat different profiles, as summarized in Table 4.2. Permanent workers inprogram firms are slightly older than in non-program firms, and a larger share has a TVET degree, which illustrates the linkages betweenthe program and the TVET educational system. Interestingly, permanent workers in non-program firms are far more likely to be female, though they still only account for about a third o f all workers. Less than a fifth o f permanent workers in program firms are women-although women account for a larger share o f casual workers in program firms relative to non-participating firms. Casual workers in program firms work more days than their non-program counterparts. Table 4.2: Average Characteristicsof Workers by FirmParticipationand EmploymentStatus Workers Non-program Program Characteristics Permanent Casual Permanent Casual Percent female 35 13 19 23 Age 30 27 32 26 Daysworkedper month 23 17 24 20 Years of schooling49 9.8 9.1 6.7 7.4 Percent with TVET 18 4 26 9 degree Percentprior activity is 29 33 27 22 unemployment 0 Workers hired by program firms differ from workers innon-program firms, at least in some respects (Table 4.3). Looking at their employment history, it seems that program firms do not draw disproportionately on the unemployed, students and inactive workers, although they are more likely to employ those whose prior activity was 49 Notethat permanentworkers innon-programf m s appear better educated thanpermanentworkers inprogram firms. This result is not consistentwith the informationonthe schooling of workers fi-omthe fm questionnaire, which suggests that the average educational attainment of workers in program f m s is 9.4 yrs, while that o f workers innon-programf m s is 8.1. 30 casual labor. Finally, when comparing casual workers alone, the differences between program participants and non-program participants are much less notable. Table 4.3: Mean Proportion of Workers with Different Types of W o r k Experience, by firm type (percent) Non-program Program Permanent Casual Permanent Casual Prior Activity Unemployed 29 33 27 22 Private sector firm 27 14 25 19 Self employed 4 4 4 2 Casualworker 19 13 26 29 Student 17 17 14 16 Government 3 3 2 3 employee Inactive 0 1 0 0 Employment History Government 9 10 17 9 employee Private sector fm 36 37 42 46 Domestic employee 7 13 9 9 Self-employed 5 4 3 6 Cooperative 1 1 12 7 Unemployed 42 70 71 84 SELECTION INTO THE PROGRAM 4.3 The initial insights offered by the description o f program participants have been enriched by modeling the determinants o f passing the IHDP test, and the determinants o f working for a program firm. We focus in particular on the role o f human capital and employment history variables. 4.4 O f permanent workers, 28 percent took the test. The probability o f taking the test in itself i s associated with their sub-city and to their employment history. Those who have experienced unemployment spells inthe past are more likely to take the test, while those who have experience as employees were less likely to take the test. 4.5 Just over half-54 percent-f the workers who took the test passed. Modeling the determinants o f passing the test shows that educational attainment i s the best predictor; completing grade 10 and completing a TVET are both strongly positively correlated with passing the test. Having completed an apprenticeship in the past does not enhance the likelihood o f passing the test. Gender, age, employment history do not affect the probability o f passing the test. Inlight o f the program's goal o f women representing at least one third o f beneficiaries it is notable that the test does not appear to discriminate against women, at least once other characteristics are taken into account (note however that women are less educated thanmen). 31 4.6 Running a similar kind of model for the probability o f being hired by a program firms0shows that: a Predicted household expenditure5' i s not a significant correlate of being hired by a program firm. This suggests that the workers employed by program firms are not on average poorer than workers hiredby firms not working for the program. a More educated workers are much more likely to be employed in program firms, especially ifhaving a TVET degree. a Having completed an apprenticeship i s significant, though this may reflect the fact that those who do not pass the test canjoin the firms as apprentices. a Age and gender are not correlated with the probability o fbeing hired. a There i s strong geographical variation in the probability of participating in the program, as evidenced by the significance o f some o f the subcity dummies. 4.7 When controlling for the workers' prior activities to test whether program firms draw disproportionately on unemployed workers, casual workers and workers in otherwise marginaljobs suggest that this is not the case - these variables are not statistically significant. Workers inprogram firms are not more likely to have been unemployed, to have worked as a casual labourer or to have been active in an otherwise marginal activity immediately prior to obtaining their current jobs. Unemployment spells become however significant when considering having experienced a significant one (i.e. longer than three months) at some point in the past in the past. Other important elements of workers past employment history are: having experience working in a cooperative and having experience as a domestic employee.52 Workers who have experience being employed as a casual worker are less likely to be employed inprogram firms.53 50 The determinantsfor participating inthe program do not seem to differ betweencasualworkers and permanent workers. This may partly be due to the fact that the determinants of program participation for casual workers can be estimatedonly very imprecisely. " Predictedhouseholdexpenditurewas constructedby using informationon asset holdings collectedbythe AACES, using the methodology developedby the CWIQ surveys. In practice data from the 1999/2000 HICES survey were regressed on asset holdings for the subsample of Addis residents. The resulting parameter estimates were used to predict household expenditure using the information on asset holdings in the AACES. While this variable represented our best attempt at capturing monetary poverty, the variable appeared to suffer from a number of shortfalls, including the low predictive power of the HICES regression, the significant weights of outliers, the nature of the assets available (note that ownership of a radio and a TV are lumped into a single category and that the price of goods, particularly electronics, has changed over the years) and conceptual doubts on whether asset ownership was an appropriatepredictor of welfare levels in this context, particularly given the high levels of migrants. In what follows, while testing for the significance of the variable, we rely on other long term indicators associatedwith long term welfare levels (such as education) to capturethis dimension. 52 There is anecdotal evidence that women inparticularly are finding it more profitable to work on construction sites than as domestic workers. 53 These effects are all significant at the 5 percent level. 32 PROJECTBENEFICIARIES: FIRMS 4.8 To create a profile o f firms inthe program, a two way classification i s used to reflect both the nature o f production and our interest in the employment creation effects. A distinction i s drawn between program and non-program firms, and between large contractors, defined as firms which have a license grade between 1 and 6, and MSEs. The latter encompasses both small contractors, and non-contracting firms, few o f which are relatively large. Box 4.2 summarizes our definitions o f the four categories o f firms. 4.9 Comparing large contractors and non-contractors provides insight into alternative modalities for delivering housing, such as awarding contracts to existing private sector firms. This distinction is also relevant as the program employment creation effects operate differently across the two groups. While the MSEs are the main vehicle for employment creation in the IHDP, the program's demand for structural works also creates job opportunities for large contractors (though they also engage inactivities outside the program). The mainfocus o fthe analysis is on MSEs, since these benefit most from the program. Box 4.2: Definitionof Firm Types Program Large Contractor: a contracting fm with a license grade between 1 and 6 which works for the program. Typically, these f m s are hiredto execute the structural works for the IHDP condominiums. Non-Program Large Contractor: a contracting firm with a license grade between 1and 6 which does not work for the IHDP. Program MSE: any firm that works for the program that i s not a large contractor. Note that this category also includes some relatively large firms which are not contractors. Yet, the bulk o f program MSEs are made up by cooperatives, which tend to be relatively small. Non-Program MSE: any fm that i s not a large contractor but does not work for the program. This category also includes small contractors, those with license grade 7-10, as well as some relatively large f m s which are not contractors. According to the AACES, almost one third o f construction firms are inthe program (Table 4.4). Program firms engage in fewer activities at the same time, suggesting that they mightbemore specialized thannon-program firms. The differences in median and mean firm age between program and non-program firms are substantial (Annex 1, Table 4.9). The average age o f a program MSE i s 2.3 years, while the average age o f a non-program MSE i s 3.8 years.s4 Large contractors are typically much older than non-contractors. In addition, 91 percent o f the program MSEs were created inthe last three years, i.e. the period over which the program has beenactive. Overall, firms o f 10 or fewer workers account for about 60 percent o f employment, while firms o f 20 or more workers account for only about 15 percent and only 1.5 percent o f all firms employ more than 100 workers. Program firms employ significantly more workers than non-program firms-a median of 13 workers, compared to 5 workers for non- program MSE. Because firms participating in the program are generally larger than non- 54Incontrast, program contractors are slightly younger thannon-program contractors (with an averageage of6.2 years vs.7.8 years). Note that the age o fthe fm is strongly related to fm-size; fm (p=0.24). 33 program firms, they account for a disproportionate share o f employment (43 percent o f all workers in the construction sector in Addis). Some key descriptives are shown in Table 4.8. 0 While the construction sector is booming, the AACES reveals that its MSE component is not operating at full capacity. On average firms operate 9 months per year and at 49 percent o f capacity. Only 3 percent o f firms report operating at full capacity. Program firms operate at similar capacity than non-program firms (48 percent vs. 49 percent, respectively). As already mentioned in chapter 3, this low level might reflect the acute shortage o f inputs which was registered at the end o f 2006, though firms were asked about their capacity utilization over the last 12months. Table 4.4: Characteristicsof Firms by Size and Program Participation Status Non-program Program Non-program ProgramLarge SME SME Large Contractor Contractor Share o f Employment in the construction sector 33% 38% 25% 4% accounted for Share o f f m s inthe 59.7% 31.4% 7.6% 1.2% construction sector Average number o f 9.0 7.1 8.9 9.6 months operated Capacity utilization 49% 49% 55% 65% Size at the 10" percentile 3 6 4 10 Size at the 25" percentile 4 10 8 12 Median Size 5 13 15 21 Size at 75" percentile 8 21 58 40 Size at the 90" percentile 15 32 102 122 0 Firm specific price indexes for both outputs and inputs55summarize the relative prices faced by different types o f firms. Program and non-program firms are found to face similar prices for inputs, but different prices for outputs.56These indexes will be used in chapter 7 to compare the productivity o fprogram and non-program firms. Table 4.5: Median Input and Output Prices by Firm Type Firm Type MedianOutputPrice Median InputPrice Non-Program M S E 1.08 1.oo Program M S E .97 1.oo Large Contractor (Non-Program) 1.19 .74 Large contractor (~rogram)'~ 1.50 .90 55Such indexes are based on the deviation o f the prices faced by each fm for each good (output or input) from the median price for that product. They are then aggregated based on the importance in the revenue and cost structure, respectively, for each fm. 56Note that this finding is based on the respondents' answers to the survey, despite the fact that the program subsidizes several inputs. As the price indexes take into account the cost structure o f the firms, the finding that inputprices are similar for program andnon-program f m s is likely to reflect the weight o f other non-subsidised inputs or of inputs(such as cement) which were of limited availability through the program. 57 Note that the prices for contractors are not very reliable. The reason i s that only a select number o f large contractors responded to the questions on inputs and outputs, which were used to construct fm-specific price indicators. 34 FIRMSELECTION 4.10 To obtain a richer characterization o f program firms' characteristics such as factors o f production used and the initial sources o f financing, and o f its manager, some simple models o f program participation have beenrun. Key findings were: a Firms in the program were significantly larger in terms of employment when they started, especially for firms created inthe last 3 years. This i s consistent with what we know about the majority o f MSEs in the program being organized as cooperatives, which require a minimumo f 10 workers. a Firmsinthe program, particularly those created inthe last 3 years are also more likely to have a better educated workforce. 0 The age o f the workforce is not correlated with program participation, other than for a negative correlation between age o fthe workforce and participation for older firms. a When controlling for firm size and amount of capital per worker, the level of education o f the workforce i s even more correlated with program participation. a The age o f the firm is negatively correlated with participation, suggesting that firms established before the program launch are less likely to join. a Firms inthe program are more likely to have received a formal loan. To complement the quantitative analysis of the factors correlated with program participation, Box 4.3 presents some o f the findings from the qualitative interviews. 4.11 These estimates provide a sense o f multiple correlations between characteristics, rather than ~ausality.~'To understand what makes firms join the program, we also limitedthe analysis to firms who were in the program but did not receive support. In this way, one can eliminate the effects o f the support o f the program itself from the factors which might be driving the associations observed. The main findings on the size and average educational attainment o fthe labour force remain, though access to credit i s not longer significant. Box 4.3: Qualitative Evidence on the Decision to Participate in the IHDP MSEs who joined the IHDP can be grouped into those founded prior to the inception of the programme and those who were newly created in response to the program. The former group typically responded to invitations by the project administration to join the programme and were often motivated by the opportunity to receive support and have additional demand for their work. Most of them report to have been operating below full capacity when the programme started. For some, receiving training and being exposed to new technologies were additional motivatingfactors. Newly-created MSEs are typically composed of either workers with previous experience in the construction sector or of recent TVET graduates. Many respondents indicated that by working together they could accomplish more than by working as individuals; participating in the program is the least difficult way of 58 For example, we are not claiming that firms hire more workers as a result of being inthe program, only that f m s currently in the program have more workers. It would therefore be wrong to see the associationbetween size and programparticipation as causal. 35 setting up one's own company. Many respondents appear to be motivated by the opportunity to be their own boss and the belief that the exposurethey would get by being a member or manager of a program fm would be more beneficial than beingan employee in construction f m s not partaking inprogram. As expressed by one MSE manager when he was asked why hejoined the program: `it is better to be your own boss than to workfor someone else; the pay is better [than being an employee outside of the IHDP]. We get experience. also, we do not get any supportfrom any other source.' However, MSEworkers did express concerns about the sustainability oftheir f m s . ExistingMSEsthat decidednot to attempt to participate inthe program typically believedthat they would not benefit fiom being in the program, either because the payment was too low or because it would involve cooperatingwith less productive individuals. According to an MSE manager of a metal-working firm, `Ihave much more experience than people in the program. rf I had to co-operate with people whojust graduated from TVET, the division of labour would be unequal.' Contractors' decisions to join the programme are overwhelmingly driven by a profit motive; contractors who didjoin the programmewere motivated by the opportunity to be employed at a time when demand was slack, while contractors who did not join the programme were discouraged by the low price and the fixed price system. One contractor regretted the decision to join, saying that: `Iam losing at least 15percent insteadOJ making a small profit of 5 percent. But what can I do now? I have signed a contract and I must complete the building. But then I will stop; I have invested all my own money and now it is wasted.' As explained by a contractor who regarded the IHDP favourably, `Iam not in the program. They offered me to be in it, but I refused it, Themainproblem is that theprice isfuced it is too low. ' ource: Interviews with workers and fmmanagers, Dec. 2006 (see Box 4.1 for more backgroundinformation). 4.12 This profile both confirms and challenges what is commonly believed about the program. While the program aims to absorb young workers, participating firms are not on average different in this respect, even when taking into account the current size o f operation and the type o f technology used. That the educational structure o f the workforce i s strongly associated with program participation presents a policy challenge. While the program i s explicitly targeted at TVET graduates and other skilled workers (to keep high quality standards and operate the new machinery that the program has introduced), poorer workers have typically lower levels o f educational attainment. Since support offered by the program includes credit, it is not surprising that having received a formal loan i s significantly associated with program participation by firms. THEJOBCREATION IMPACTOFTHE PROGRAM 4.13 The main target o f the IHDP interms o f employment creation is the number ofjobs created. As the program i s large, a large number o f workers are found in program firms. In October 2006 the number o f jobs created by the program was estimated at 52,600 (IHDP 2006).59 4.14 The official estimate, however, is problematic. I s it enough to do any work (even if only for a day) to consider that 1job has been created? Or should the measure incorporate some sense o f duration (for example in terms o f full-time equivalent) to be considered a job? Other complexities arise because any worker working for a firm contracted by the program i s considered as a job created. However, program firms might have exited the market (in our 59To give a sense ofthe scale ofthis achievementconsider that in2005 accordingto the LFSthere were 349,000 active individuals inthe 25-34 age cohort (medianage inthe program i s 30). The program would therefore be reachingthe equivalentof one insix workers inthis age group. 36 estimates this occurred to 22 percent o f participating firms). Further, to the extent that casual workers move between different sites, there i s the possibility o f double counting. Finally to the extent that programfirms also work inpart for customers outside o fthe program, it is not clear to what extent all the jobs inprogram MSEs should be attributed to the program itself. 4.15 There i s no unique answer to how jobs should be measured, but it i s important to have clarity on what exactly different indicators capture. If exposure to the program even if for a very limited amount o f time i s the key indicator of interest (perhaps because o f the impact on the workers' skills or productivity) then a very broad definition o f jobs might be appropriate. Ifone i s concerned about effective labour demand, then a more narrow definition i s more appropriate. Annex 1 presents a rough estimate based on the AACES on the numbers o f jobs in the program in 2006 which illustrates how all the different considerations made above can helpmaking an estimate o f the number o fjobs created. Net EmploymentCreationEffects:DifferencesinTechnology 4.16 The true test o f the IHDP's effectiveness in creating jobs is whether it has created more jobs than would otherwise have been created if, for example, the construction o f condominiums had been tendered to existing firms (either MSEs or large contractors). We test whether MSEs are more labour intensive than larger firms, as well as whether MSEs in the program are indeed more labour intensive than non-program MSEs. If these hypotheses are disproven the employment benefits o f the program must be found elsewhere-for example in terms o f their dynamic effects. This issue i s addressed via production function estimation, thenfactor proportion.60 4.17 Table 4.6 presents estimates o f the production function used by construction firms. We use revenue, deflated by firm-specific price deflators, as our measure o f output. Inputs are not deflated since input prices faced by different types o f firms are very similar. Column 1 offers a basic model6' where revenues are modeled as a function o f capital, the number o f workers and inputs.62Since large contractors might use a different technology we interact the different parameters o f the production function with being a large contractor. With respect to contractors, our analysis for large contractors i s not conclusive because the sample becomes very small (10). The results however support the commonly held view that large contractors use a differenttechnology than MSEs. 60 Note that the production function approach implies a number o f assumptions, including full employment o f resources which, given intermediate input constraints and the resulting excessive capacity in the construction sector inAddis at the time o f the survey mightnot be appropriate. While there is some support for this concern in the high elasticity for intermediate inputs and the small or insignificant coefficient o f capital, our estimates for the coefficient on capital are broadly in line with others found in the literature. We do not feel that these concerns, therefore, are such as to invalidate our analysis. "Alternativeregressionsaddressingtheissueofpotentialendogeneityofinputandoutputchoiceshavebeenrun also, but given the instruments available (input prices, output prices, initial capital and employment at startup) the hypothesis o f endogeneity has been rejected. Note that due to the difficulties o f collecting price and input data these regressions have been run on the subsample for which data were available, comprising 120 f m s . 37 Table 4.6: ProductionFunction: DeflatedRevenue All MSEs Program Non-program MSEs MSEs coeflsd coeflsd coeflsd coeflsd Labour(1og) 0.488*** 0.667*** 0.425** 0.667* ** (0.135) (0.214) (0.193) (0.225) Capital(1og) 0.099* 0.167** 0.007 0.167** (0.055) (0.071) (0.083) (0.074) Inputs(1og) 0.496*** 0.404* ** 0.569*** 0.404** * (0.059) (0.093) (0.071) (0.097) Largecontractor 3.940 (4.199) Largecontractor Inputs * -0.867 (0.786) Largecontractor*Capital 0.655 (0.518) Largecontractor *Workers 0.077 (0.361) Program*Labour -0.242 (0.294) Program*Inputs 0.165 (0.119) Program*Capital -0.161 (0.112) Program-dummy 0.229 (1.387) Constant 4.117*** 4.148*** 4.377*** 4.148*** (0.656) (0.959) (0.963) (1.007) R2 0.714 0.630 0.634 0.574 Adjusted R2 0.696 0.604 0.615 0.544 Number of observations 120 110 62 48 Note: significant at the 10 percent level, * ** significant at the 5 percent level,*** significantat the 1percent level. 4.18 ,In the other models we focus only on MSEs and find that ingeneral program MSEs and non-program MSEs utilize rather similar technologies. None o f the interactions between being inthe program and the parameters o f the production function i s statistically significant (column 2). This i s confirmed in Column 3 and 4 for the sub-sample o f program and non- program MSEs respectively. It should be noted that the models presented in Table 4.6 seem to describe the data well, asjudged by the R2.' 4.19 We also explored whether the apparent similarity intechnology betweenfirms inand out o f the program is in fact driven by differences in the labour they use. Table 4.7 presents human capital augmented production functions. The model suggests that given the overall 38 amounts o f capital and inputs, the education o f the workforce does not affect output significantly. The average age o fthe workforce i s also found to be in~ignificant.~~ Table 4.7: Human CapitalAugmented ProductionFunction: DeflatedOutput All SMEs All SMEs coeflsd coeflsd Labour(log) 0.672*** 0.974** (0.217) (0.388) Capital(1og) 0.165** 0.156 (0.072) (0.103) Inputs(1og) 0.407*** 0.454** (0.094) (0,190) Program*Labour(log) -0.242 -0.422 (0.295) (0.463) Program*Capital(log) 0.163 0.078 (0.120) (0.217) Program*Inputs(log) -0.156 -0.179 (0.115) (0.165) Program-dummie 0.176 1.618 (1.413) (2.446) Average Education Workforce 0.007 (0.030) Average Age Workforce 0.022 (0,019) Constant 4.087*** 2.415 (0.999) (2.05 1) R2 0.630 0.684 Adjusted R2 0.601 0.628 Number of observations 110 54 Note: significant at the 10 percent level, *** * ** significant at the 5 percent level, significant at the 1percentlevel. NetEmployment Effects: Differences inFactor Proportions 4.20 Plotting the amount o f capital per worker against the total number o f workers for different types o f firms (Figure 4.1) shows that capital intensity (defined as the amount of capital per worker) does not differ betweenprogram and non-program firms and does not vary systematically with firm size. MSEs thus do not seem to be less capital intensive than larger firms. 63These results are robust to estimating this specification on subsamples of contractors and non-contractorsand using a measure o f value-added as dependent variable. Note also that differences inproductivity are not due to differences inthe occupational structure. 39 Figure4.1: CapitalIntensity Capital Intensity 0 0 0 2 4 6 a Workers (log) Non-Program MSEs 0 Program MSEs Non-Program large contractors Program large contractors 4.21 Likewise, program firms are not systematically different in their use o f capital per worker than non-program firms when capital intensity has been regressed on number o f workers, program participation, and being a large contractor program, participation i s not significant. A result which holds across alternative specifications, even when controlling for the activities firms engage in.64 4.22 Interestingly, large contractors are found to be significantly more capital-intensive than other firms, but their capital intensity o f large contractors does not vary with their size. The higher capital intensity o fcontractors is at least partly due to the fact that these firms have been operating for longer thereby accumulating more capital, and to the activities they engage in. 4.23 A similar analysis has been conducted for input intensity, defined as the total value o f material inputs per worker. Figure 5.2 shows that input intensity does not seem to depend on program participation or on firm size. Regression analysis confirms the intuition from the graphs, that size variables and program participation have no impact on input inten~ity.~' Capital intensity, incontrast, i s significantly correlated with input intensity.66 64Firms in different activities differ in their capital requirements, with electrical installation requiring the least capital and firms engaging in structural works and wall construction possessing the most capital. In addition there is evidence that f m s which have been around for longer have been able to accumulate more capital. 65Note that the results presented are based on inputs deflated by firm specific price indexes. The pattern o f results remains the same if the value of input i s not deflated. 66Model 5 which controls for different activities finds that f m s engaging ingravel production or woodwork use significantly less input per workers than other f m s . 40 Figure 4.2: Input Density . Input Intensity e.: I I I I I 0 2 4 6 a Workers (log) Non-Program MSEs Program MSEs Non-Program large contractors Program large contractors 4.24 Insum, technology andfactor proportion do not seem to vary with firm size nor with program participation. We do find differences, however, betweenlarge contractors and other firms. These findings suggest that the program has not createda more labour intensive supply chain for housingthan would have otherwise existed. THEIMPACTOFTHE IHDPONLABOUR DEMAND:DYNAMICCONSIDERATIONS 4.25 The program might have significant dynamic effects on employment creation since program firms typically employ more workers than non-IHDP firms and typically employ more workers at startup. The question i s whether the program MSEs are larger now because they startedlarger or because they grew faster. 4.26 To explore this hypothesis, inthis section we compare survival and growth patterns of program and non-program firms. We examine this issue in the context of the AACES cross-sectional data which provides insights on the different channels of impacts of the program. These effects are particularly relevant given the IHDP's goal to create sustainable construction capacity. First we discuss the determinants of size and labour-intensity at startup, andthen examine growth patterns. FirmCharacteristics at Start-up 4.27 Program firms on average employ more workers than non-program firms and this effect is strongest for firms younger than 3 years, suggesting that the IHDP creates firms which are on average larger thannon-program firms. Once capital stock at startup is modeled as a function of characteristics of the manager and access to capital, program firms appear to start out larger becausethey tendto have better access to loans from micro-finance institutions (MFIs). Programparticipationis associatedwith a significantly higher size ofthe capital stock 41 even after conditioning on characteristics of the manager, yet the program participation becomes insignificant once obtaining credit from microfinance institutions is accounted for, an effect which holds both for young and older firms. Only one non-program firm in our sample relied on M F I credit to establish a firm, whereas such credit was the most important source of credit for over 30 percent of program non-contractors. 4.28 The capital labour ratio of program and non-program firms on average does not differ at startup, a conclusion valid for both young and older firms. Yet, disaggregating the analysis, program firms which obtained an MFI loan at startup are much more capital intensive than non-program firms, while program articipants which did not receive credit from MFIs tendedto start at lower capital intensity,' This suggeststhat the IHDP contributes to creating more capital intensive firms, though the effect is visible only for firms that receive access to credit. ProgramParticipationand Growth 4.29 We investigated whether program firms grow at a different pace than non-program firms and found that they do not. This finding on program firms' growth i s surprising since they start up larger than other firms. Yet, program firms grow faster than non-program firms of the same size. The coefficient onthe initial size of the firm is negative, consistent with our expectations.68In the longer run, program firms program firms will stay larger than non- program firms as they converge to a higher long term level of emplo ment,69a finding which suggests a beneficial impact ofprogram participation infirm growth. 4 4.30 Another aspect o f firms' growth is the growth of the capital We find that program firms on average do not expand their capital stock faster than non-program firms. While the initial capital stock is negatively associatedwith subsequent average annual growth of the capital stock, program participation has a significantly positive effect once capital stock at startup i s controlled for. This provides further evidence against the hypothesis of 61 This may partly reflect the fact that firms which tend to receive support from an MFIto start tendto engage in more capital intensive activities such as pre-cast beam production, hollow-block production and wall construction. Trying to control for this by includingcontrols for firm-activities only strengthens this effect. 68 The negative coefficient on initial size can be explained in a number o f ways. To start with, it may reflect survivor bias. Suppose for example that ability is the key determinant o f survival and suppose furthermore that it is orthogonal to start-up size. Also assume that f m s which entered large take a longer time to exit, for example becausethey shrink before quitting.Taken together, these assumption implythat a negative coefficient on the log o f initial start-up size. Alternatively, one could interpret this coefficient as evidence o f a tendency for firms to evolve towards the average size (regression to the mean), but the data do not support the idea that all firms converge to the same size in the long run; once the initial size at start-up is included, the program dummy becomes positively significant. 69 Since this regression uses information on two points in time only it is a bit exaggerated to use the concept o f long run. Yet, if the parameters are stable over time, program f m s will converge to a larger size than non- porogramf m s . Note that in the literature larger f m s are found to grow more slowly, so program participation appears to contrast this effect. 71 Growth in capital stock was estimated by comparing capital stock at startup to current capital stock. The survey question on initial capital was a multiple-choice question with 15 choices (ranges o f capital, in birr), while the question on current capital stock was an open-ended question. In order to compare the two amounts, the dataon current capital stock were recodedperthe 15 categories. 42 convergence to one unique capital stock level. Further, it seems that program firms end up with a larger capital stock as, while not expanding their capital stock faster or slower than non-program firms, they start larger in terms o f employment and maintain similar factor proportions. 4.3 1 Insum, the investigation o f potential dynamic effects o fthe IHDP on employment suggests that firms created by the program start with more capital and more workers, yet are not on average more labour-intensive than non-program firms when they start. Program firms which obtain credit from an MFI start at a higher capital-intensity than non-program firms, because they start out with a much larger capital stock. Program firms which did not obtain such credit started out operating at lower capital intensity. These differences in size persist over time. Program firms exhibit growth rates which are similar to those o f non-program firms. Compared to non-program firms which entered at a similar size, however, program firms have grown faster. 43 ANNEX1: DESCRIPTIVES BENEFICIARIES OF Table 4.8: Activities of Program vs. Nowprogram Firms (percent) Activity Share ofProgram firms Share of Non-program engagingin this activity Firms engaging in the activity Pre-Cast Beam Production 1 26 Hollow/Concrete Block Production 29 43 WoodMetal Works 21 44 Gravel production 3 8 Wall construction 14 14 Structural Works 9 4 Electrical installation 9 9 Sanitary works 11 9 Finishing 12 7 Site Works 4 5 Production o f Construction Inputs 4 38 Other 4 2 Note that the columns do not add to 100as f m s engage inmultipleactivities. Table 4.9: Age of Firms(percent and cumulative) by firm type Non-program Non-program Program Large SME Program SME large Contractor Contractor Years YO cum YO Cum YO cum % cum 0 14.77 14.77 3.37 3.37 1.75 1.75 10.65 10.65 1 27.27 42.04 71.91 75.28 7.14 8.89 21.43 32.09 2 19.22 61.26 13.48 88.76 1.75 10.64 11.36 43.44 3 14.48 75.74 2.25 91.01 14.29 24.93 7.14 50.59 4 andmore 24.26 100 8.99 100 75.07 100 49.41 100 44 ANNEX2: DERIVING ESTIMATES FULLTIME EQUIVALENT OF WORKERS FROMTHE AACES 4.32 As already mentioned, quantifyingthe exact number o fjobs created by the program, while important for monitoring performance, is not a good indicator o f the effectiveness of the program at creating jobs. Nevertheless, we use the AACES to provide an order o f magnitude on thejobs inthe program and outside in2006. Two characteristics o f these estimates needto be underscored: one is that we see them as estimates o f labour absorption over a given time period, rather than estimates o fjobs created, as there i s no satisfactory way o f accounting for exit o f program firms (and there are also doubts on whether having had briefly a job with a program firm has intrinsic benefits), even though we try.72.The second one i s that they should be taken as broad orders o f magnitude as they are obtained from sample estimates. As there should be no inherent bias inhow these calculations account for labour absorption indifferent type of firms, however, we believe that they are comparable across firm types. 4.33 The approach we adopt starts by estimating the work that is performed by firms o f different types. This total amount of work is then allocated to full time equivalents which control for the fact that not all firms inthe sample work year-round. This approach guarantees us comparability across different types o f firms. At the same time, by construction, this approach focuses on the amount o f work rather than on the number o f beneficiaries o f the program-a one year full time equivalent o f work could be performed by different people (for example 4 casual workers working 4 months each).73 Box 5.1 provides further details on the methods usedto estimate the number o f full-time equivalent workers. 4.34 As Table 4.10 illustrates, the program appears to have absorbed over 2006 almost 20,000 jobs. This i s a major achievement considering that the program was certainly not operating at full capacity duringthe period. Furthermore, program firms appear to account for approximately 40 percent o f full time equivalent jobs inthe construction sector. And MSEs in the program cover some 8 percent more o f full-time equivalentjobs thannon-program MSEs. Yet, it is worth stressing that any type of public works and any form of construction will create employment. To analyze the effectiveness o fthe program at creatingjobs it i s therefore not sufficient to show that it provides employment for a very large share o f workers ina given sector. 72 The reason why we cannot estimate exit rates satisfactorily i s that exit rates can only be estimated for non- contractorsandnot for contractors;we correct for exit of non-contractors,but not for exit of contractors. 73 An analysis more oriented to the welfare effects of the program might need to focus more heavily on the number of beneficiaries, yet it would have to specify how the benefits of participation vary depending on the intensity of labour market attachment(e.g. benefitsof working for one day vs. benefits of working for one year). 45 Box 4.4: Estimatingfull time equivalentworkers The first step i s the estimate of total number of workers employed by f m s of each type from our sample. This estimate takes into account the number of workers in each individual fm and how many workers in the population they represent (Le. the weights of the sample design). Note that a similar sample estimate can be obtained for the total number of workers inthe constructionsector as awhole. The total number of workers is then corrected to account for the share of the output that is bought by the program. A direct link between output and employment can be drawn because of constant returns to scale technology adoptedby the firms and the findings that factor proportion do not vary by size. This step gives us an estimate ofthe "true number" of workers who can be accounted for by the program. A fixther correction i s introduced to account for the fact that not all firms work 12 months a year. This leads us to a full-time equivalent estimate which is comparable across f m s type. Note that we do not specifically introducefurther corrections for days andhours worked as variation across fmtypes is on averageminimal. Inthe case ofprogram firms, to arriveat our final estimatesof fulltime equivalentworkers in the program we consider that 43 percent of all workers in the formal housing construction sector in Addis work in a firm which is participating inthe program,74and that on average program f m s sell 86 percent of their output to the IHDP. This leads us to an estimate ofjob opportunities provided by the program of 22,438 workers over the past year. Factoring in that firms were not operating all year we arrive at an estimate of labour absorption by the programof 17,138 full-time equivalentjobs for 2005/6.75 Table 4.10: Labour Absorptionin 2006 by Firms by Size and Program Participation Status Non-program ProgramSME Non-programLarge ProgramLarge SME Contractor Contractor Survey estimate of 18149 19322 15918 3116 Numder ofWorkers Surveybasedestimate 13280 14378 14054 2760 of full-time equivalent worker^'^ Memo items: Average number of 22.5 23.4 23.5 23.4 days eachemployee works each month Average number of 9.0 7.1 8.9 9.6 months operated ~ ~ 74 Note that the total value of output produced for the IHDP is lower because of the fixed price system. This percentage was calculatedby dividing the share of output produced for the IHDP as a percentageof total output, deflatedby the relative price ofprogram outputs. 75 Note that variation in average number of days and hours worked turned out to be rather low across f m s and was consequentlynot correctedfor. 76 Full-time equivalent obtained as (number of workers*months operated)/l2. Note that the number of days worked per month does not differ drastically betweendifferent types of firms, on average, which i s why we leave it out ofthe calculation. 46 5. THE IMPACT OF SUPPORT ON PROGRAM FIRMS 5.1 This chapter reviews how the support the program provides affects the performance o f firms, complementing the findings on the impact o f support already presented. A description o f program support inalleviating firms' constraints and tackling market failures i s followed by an analysis o f the impact o f program support on productivity and profitability trying to unpack the effect o f individualtypes of support. Inthe last section we expand the analysis to discuss the possible effects o f the program on non-beneficiaries. CONSTRAINTSAND SUPPORT TO PROGRAM FIRMS'OPERATION 5.2 The AACES suggests that construction MSE components are far from operating at full capacity.77This suggests that the sector is already very competitive, one of the objectives that the program tries to address. Input constraints are the most frequently cited reason for capacity underutilization (see Box 5.1 for qualitative evidence on firms' constraints). Indeed, 75 percent o f non-program MSEs and 86 percent o f program MSEs report facing difficulties accessing inputs.All large contractors claim that they face input constraints, and44 percent o f all firms indicate they have had to refuse contracts because they lacked inputs. Cement i s the scarcest input. Surprisingly, a larger proportion o f program firms reports facing input constraints than non-program firms. Box 5.1: Anecdotal Evidenceon Firms' Key Constraints At the time o f the survey, virtually all f m s , with the exception o f MSEs engaged in electrical and sanitary installation, complained about the lack o f inputs and the lack o f demand, two issues which are closely related. Small firms tend to see fmance as their biggest challenge, while large f m s claim the lack o f inputs as their major problem. The shortage o f cement i s one o f the most serious issues facing the construction sector as a whole and frequently causes construction delays. Some illustrative comments from interviews with managers and workers include the following: `We have stopped workingfor 20 days now. Since starting theproject 6 months ago, we have not been able to work 2 months because of cementshortage. `We arefortunate that we are constructing a buildingfor the government; that makes it easier to get cement. yet, even they [the government entity that ispayingfor the buildind have dfjculty getting cement.' `Theprice of cement used to be 60 birr, now it is I20 birr per quintile. On the black market theprice is 250 birr per quintile. ' Fortunately, it seems that the government has recently been taking important steps to alleviate input constraints, for instance by allowing the import o f cement. Source: Interviews with workers and firm managers, Dec. 2006 (see Box 4.1 for more background information). "Onaveragefmsoperate9monthsperyearandat49percentoftheircapacity. Only3percentoffirmsreport operating at 100 percent o f their capacity. 47 5.3 Credit is the most important obstacle to operating successfully on a day-to-day basis according to the majority o f firm managers. Interestingly,program firms are as likely as non- program firms to cite credit constraints as their ma'or problem, even though a much larger proportion o f them managed to obtain formal credit." Of non-program MSEs, 35 percent took out a loan from any source, while 77 percent o f program MSEs took out a loan. The difference can be attributed to the better access program firms have to formal credit (65 percent vs. 13 percent). While interest rates for formal loans are similar, program firms are less likely to have to put up collateral for credit. Assuring access to credit seems a major achievement o f the program. It is worth noting, however, that 53 percent o f program firms that got a loan from a formal credit institution are currently in default on their loans, against only 11 percent o f non-program firms (Table 5.1). This could suggest that the program is allocating credit inefficiently - yet, at the time o f the survey the program hadjust started and input constraints were severe, which mightexplain why so manyprogram firms were indefault. Table 5.1: Use of credit by firm type (percentage) Type of Firm Proportion out of Defaulting (as a Recipient of Collateral firms that ever percentageof all firms a Formal required for this took a loan taking out a loan) Loan formal loan Non-program MSE 35 11 14 78 Program MSE 77 53 65 43 LargeContractor 54 18 54 100 (Non-Program) Large contractor 55 12 39 78 (Program) 5.4 Comparing sources o f start-up capital confirms the importance o f access to credit for firm operation. For 32 percent o f program MSEs, an MFI constituted the major source of initial capital, whereas none o f the non-program MSEs in the sample reported this. Own saving was the most important source o f initial capital for 81 percent o f the firms outside o f the program and for 50 percent o f firms inthe program. The importance of access to credit is further evidenced by the fact that the 75 percent o f the program firms that exited claimed it was becausethey could not obtain accessto credit. 5.5 Not all program firms received all types o f direct support. O f MSEs inthe program, 85 percent received some support, but only 6 firms received access to a place to work, a building, access to credit, machinery and training. Firms engaging in pre-cast beam production and/or hollow block production are supported the most. O f all firms which received support (MSEs or large contractors), 68 percent received a piece of land, 36 percent received buildings on their land, 31 percent received machinery and 50 percent received access to credit via the program. 78It is awell-known limitation ofthis type ofdatathat the interpretation ofself-reported credit constraints are difficult to interpret as many o f the respondents could be "bad risks" rather than sufferingfrom a market failure. The difficultiesof obtaining credit without collateral suggest however that credit constraints are likelyto affect a significant number o f MSEs. 48 '' Table 5.2: Distributionof Types of Support by Types ofFirms (percent) Number of ty es Program MSEsPCB PCB HCB of support MSEs and HCB production 0 22 33 0 15 1 27 33 4 23 2 28 20 25 13 3 20 10 38 29 4 8 3 21 14 5 5 2 13 7 5.6 In general, different types of support are only loosely correlated, though there are some notable exceptions. For instance, firms which get access to a place to work typically also get access to credit (p=.35). Inaddition, firms which receive machinery are likely also to receive training (p=. 14) and credit (p=. 13). Incontrast, training is negatively correlated with receiving a place to work (p=-.20). Inother words, support seems to be packaged in different ways. The packaging of support i s important since the literature on active labour market programs suggests that programs that provide assistance on multiple fronts produce better results than support programs that tackle single constraints. More analysis i s needed to identify the most effective way o f bundlingdifferent interventions inthe case o f Ethiopia. Figure5.1: Types of Support Receivedby (All) ProgramFirms Land Training Credit Building Machinery 0 0.1 0 2 0.3 0.4 0 5 0.6 0 7 0.8 I 79Notethat other support is excluded. 49 Figure5.2: Rankingsof Benefits ofProgramParticipants prousion of land training guaranteed demand access to inputs access to credit prowsion of building prowsion of machinery 0 0.1 0.2 0.3 0.4 0.5 5.7 It is interesting to compare the types of support received with the perceptions of the main benefits o f program participation by the managers o fparticipating firms. Land i s seen as the most valuable benefit by 42 percent o f them, while access to inputs, demand and training are considered to be so by 12percent, 12percent and 16 percent o f them, respectively.'' 5.8 Note that the finding on the importance o f land i s in line with results from other firm-level research. For example, focus groups conducted by the World Bank in 2005 as a follow-up to the 2002 Investment Climate Survey found that firms consider access to land for construction to be the most severe investment climate constraint they face. Note however that access to land o f good quality can remain a problem also for program firms: 42 percent o f firms complain about problems with utilities and 45 percent about problems with market access (Table 5.3). For a fuller treatment o f the issue o f landregulation inurban areas see also World Bank 2007 b. Table 5.3: PercentofFirms ReportingProblemswith Utilitiesand MarketAccess Non-program ProgramMSE Non-program ProgramLarge MSE LargeContractor Contractor Problems with 43 39 44 47 Utilities Problemswith 45 52 47 40 market access PROGRAM SUPPORT AND FIRMPERFORMANCE 5.9 Inthe previous section we ascertained that not all firms receive support from the program and that not all those that receive support receive the same type o f help from the program. It i s interesting to analyze whether firms' perceptions o f the types of support that they receive (see Box 5.2 for qualitative evidence on this) appear to be confirmed by the *'Note that the support provided by the program in terms of land is visible also in the low mean payments for landby firms; 3253 Birr for non-programMSEs versus 2644 Birr for program MSEs. 50 analysis o f the impact o f support on firm performance, both interms o f effect on productivity and o f profitability, and then aiming to identify which elements o f support appear to be drivingthose effects. Box 5.2: Firms' Perceptions of the Effectiveness of Program Support Program firms received support in terms of land provision and subsidization, training, access to credit, inputs and (subsidized) machinery on credit. Perhaps the most important form of support is the opportunity to have a job. Not all f m s received each type of su port. The commontheme is that the support that was providedby the IHDPis appreciated, yet not sufficient' (althoughthe programenvisionsthat inthe longrunSMEs oughtto P be self-sustaining). Land. Land is one of the types of support that is oftennot providedby the program, yet when it i s provided, it i s appreciated. Land i s often highly subsidized. Some firms reported paying as little as 2 birr per m2, while marketprices o f over 2000birr per m2 are not exceptional.Moreover, obtaininga pieceof landinthe restricted land market is otherwise difficult. Unfortunately, many of the plots used by construction f m s , both in and outsideof the program, are locatedat the outskirts of the city and lack essentialinfiastructuralfacilities such as access to water, electricity and all-season roads, leadingto underutilizationof capacity and hightransportation and transaction costs. As expressedby one of the respondents: `here is nothing, only eIectriciQ, not even a road; in the rainy seasonyou won't see a single car here. ' Nevertheless,the value of having a piece o f land shouldnot beunderestimated.82 Training. Not many respondents receivedtraining and those who did receive training had varying opinions about its usehlness. Some claimed it was the most important benefit providedby the program, while others claimed the training was practically irrelevant for their activities, for example by stating: it [the training they had received] was very theoretical and there was nofollow up; since it was notpractical it was of little use to us. ' Access to credit. Despite facilitating access to credit for a number o f f m s , many firms remain credit constrained. They have either not managed to obtain a loan, or would want to borrow a greater amount of money.Furthermore,many ofthe firms which are providedinputsandmachinery on credit claimto have hadto repay their loansmuchsooner thanthe agreedrepaymentarrangementspecified, causingthem difficulties. Demand. Virtually all f m s in the construction sector, with the exception of those in electrical installation, complained about not having enough work. Unsurprisingly therefore, the opportunity to obtain work through the programmeis highly appreciatedby all programparticipants. Yet, virtually all MSEs complainabout a lack of demand. Many IHDP f m s are out of work, includingmany f m s which were set up with support o f the IHDPyet have notyet hadthe opportunityto work for the IHDP. The problemof not gettingprogramcontracts is most serious for f m s producing pre-cast beams, because there are limited alternative customers. The fact that many programme f m s are struggling to survive on their own suggests that the support from the programme is critical and that sustainability may be a real issue. Furthermore, firms that could in theory completeassignments for clients other thanthe IHDPhave complainedabout regulationspreventingthem from producing for such clients. Such rules, however, if they indeed exist, are incongruous with the program guidelineswhich stipulatethat f m s are allowedto serve clients other thanthe IHDP. Source: Interviewswith workers and fmmanagers,Dec.2006 (see Box 4.1 for morebackgroundinformation). 5.10 In order to study the impact of program support on performance, the production functions already presented in Chapter 4 (see Table 4.6), have been re-run drawing a distinction between firms participating receiving support and those that did not receive 81 Of course, one should be aware that it is not inthe interest of respondentsto claim the support they received was fully adequate; perhaps the knowledge that one was being interviewed by the World Bank made provided firms with a strategic incentiveto try to ask for moresupport. ''Itis perhaps telling that it was not uncommonfor respondents to answer "yes" to both the questions"is your locationgood" and"would you like to move ifyoucould?" 51 support, and exploring the effects o f different types o f support. This analysis reveals that from the point o f view o f productivity, access to credit i s the most effective type o f support as it is associated with significant1 higher production. Access to a workplace andprovisionof a buildingappear also important. B 5.11 A similar analysis has been conducted for and profitabilityYg4running regression of profits per worker on capital per worker and the number o f workers. Program firms are on average much less profitable than non-program firms. However, this effect i s driven by program participants who did not receive any type o f support, as those firms are significantly less profitable than all other. Firms which receive a building from the program are significantly more profitable than program firms which do not receive such support, while other types of support do not appear to be significant. 5.12 Inshort, these findings suggestthat when looking at the impact of program support on firm performance, credit i s strongly related to productivity, while receipt o f a building to work ini s strongly related to profitability. BroaderImpactsof the IHDP 5.13 While we cannot draw conclusions on the overall impact o f the entire IHDP, we can draw some inferences on the impact o f the IHDP on the construction sector at large. .Giventhe scale o f the program and input constraints inthe sector, it might have affected non-beneficiary firms by leading them to shed jobs, or stunting their growth. These effects, o f course, would lower the employment creation effects o f the program itself. We explore these issues by presentingdescriptive statistics on entry and exit inthe construction sector, and by reviewing the subjective assessments o f the impact o f the program made by managers o f firms in the housing construction sector. ENTRY AND EXIT 5.14 Analysis o f entry and exit dynamics explores whether crowding out of non-program firms by the new entrants created by the program has occurred. Inthe AACES firm turnover i s high, as reflected by the age distribution of firms; the average firm is in our sample is 3.2 years old, while the median age i s 2 years. However, most firms that exit do so inthe first few years o f operation; 44 percent o f firms that have exited have done so after less than a year o f existing, while fewer than 10 percent o f firms that exited had been operating for more than 3 years. This is consistent with the pattern o f entry and exit in other countries and sectors (see e.g., Scarpetta et al., 2004). 83 Howeverthese dummies are not significant at conventionallevels. 84 Note that our profit measure is defined as revenues minus expenditure on material inputs and wages. It considers, therefore, only direct costs, while it is likely that program firms have substantially lower indirect costs. For instance, for some program firms capital and the place where they work are subsidized. In addition, because the majority of program f m s are cooperativesthey pay significantly less profit tax. Alternative profit measures, basedonrevenuesand expendituresand self-reportedprofitshave also beentried. Yet, these measures o f profit do not overturn the pattern of results (while doubts can be raised on the quality o f the indirect costs informationcollected). Inaddition, programMSEspayhigher wages. 52 5.15 Inadditionto the AACES a short phone interview was administered to a sub-sample of firms from the Addis Ababa City registry. O f firms in our sample, 21 percent started in 2006, suggesting that net entry rates are positive and substantial. 5.16 The evidence suggests that the program has had a limited impact on entry, thoughit mighthave exerted indirect pressures on some of the incumbents. More than three quarters of entrants into the construction sector over the past three years have not participated in the program. In addition, gross entry o f non-participants in 2005 and 2006, the years for which the data are most reliable, are almost identical, despite variation in gross entry of program filTllS. 5.17 Estimates o f exit rates from these data are much more difficult as there are strong suspicions that the might be biased by the deletion o f firms who do not renew their license from the registry. r-5Our estimate is o f an exit rate of 14 percent o f firms in the registry in 2006.86 5.18 An interesting element emerging from these findings is that for program firms gettingwork from the IHDP is the key determinant of survival: 95 percent o fprogram firms in the sample have ever completed a program assignment, while only 13 percent of program firms that perished claimed they were working for the program. Intotal, at least 22 percent o f the firms supported by the IHDP have exited since the start of the program. The average exit rates o f program firms are thus marginally lower than the typical exit rates o f non-program firms. Perceptionsof ProgramImpact 5.19 The AACES includes a module with questions on perceptions by firm managers on the performance o f the IHDP. These data offer some insights on the elements o f the program which seem to be effective, and on its perceived impact on the sector as a whole. The main findings are: 54 percent o f the managers claimed they did not join the program because they were not aware o f it, 21 percent because participation would not have been beneficial (the rest either had no opportunity to register, or did not fulfill the criteria). 0 With respect to the overall business environment, 77 percent of firms believe competition has increased over the past 3 years because o f the program, while 11 percent believe it has decreased. O f those that believe competition increased, 35 percent (or 27 percent o f all firms) believe it was because o f the IHDP. Firms working 85 Note that prior informationfrom the AA trade and Licensing office suggested that f m s who fail to renew their license are not deleted from the registry itself. Inspections of the data reveal that this is very unlikely; thoughit was notpossibleto identify with whichregularity the registry is updated. 86 Retrospectivecomputation of survivalrates o f f m s in previous years, suggest these have dropped from 96 percent in2003 to 86 percent in2006, hintingthat the program may have contributedto higher exit rates. Note however that the size of the bias inunderestimatingexit rates tends to increase as time passes, since eventually all f m s whichhave exitedin any givenyear are removed. 53 for the IHDP are much more likely to indicate that competition increased because o f the IHDPthannon-participants (intotal 15percent vs. 38 percent). Only 10 percent o f non-participating firms believe termination o f the IHDP would improve their business, while 18 percent o f IHDP participants indicate that termination o f the IHDP would cause their company to go bankrupt, and another 33 percent fear their businesswould slow down should the program be terminated. The IHDP, on average, purchases 86 percent o f the output produced by firms participating inthe program, and only 56 percent o f IHDP firms have ever completed contracts for clients other than the IHDP. For IHDP firms, support i s considered very important. Firms are more concerned about competition from non-program firms than about competition from IHDP firms. 36 percent o f all firms indicate that they experience competition from IHDP MSEs, while 48 percent o f all firms report experiencing competition from non-program MSEs. Program contractors, who work for the program on a fixed-cost basis, are vulnerable to rising input prices. As program contracts are not indexed some firms claim they are struggling to adjust, which may lessen their willingness to participate inthe program inthe future. Regarding the impact o f the IHDP on the labour market, 46 percent o f managers believe that the program has not caused changes, or do not know about the program. 13 percent claimed the program had contributed to the shortage o f skilled workers, while 9 percent contended that the program had sparked a general wage rise. 8 percent o f employers took a more positive view and replied that the program had created employment and income. When asked about the impact o f the IHDP on input markets, only 15 percent o f managers claimed either that the program had not had an impact, or that they did not know o f the program; 49 percent claimed that the program had contributed to increasing shortages o f raw materials; and 23 percent argued along similar lines that the programhadledto risinginputprices. When asked about the impact o f the IHDP on output, 54 percent o f firm managers responded that the IHDP had not affected their level o f output, while 12 percent claimed that the shortage o f inputs resulting from the IHDP had forced them to delay certainprojects. 5.20 In sum, the program seems to have contributed to competition in the housing construction sector and has exerted upward pressure on input markets, although this finding may be at least in part driven by the fact that the survey was conducted at a time when input shortageswere particularly acute. The impact o f the program on labour markets seems to have been limited, even though many o f the interviewed managers complained about price escalations inrecent years. 54 5.21 The fact that inputs used by the program are no longer available for non-program firms suggests that the program might have crowded out non-program firms. This would suggest that the net creation effect o f the program would be lower than what discussed in Chapter 4. Yet, given market failures in both land and capital markets the extent to which firms outside the program might have really thrived i s debatable. 55 6. THE BENEFITS OFPROGRAM PARTICIPATION: INCREASED EARNINGS 6.1 A crucial element of the pro-poor nature of the IHDP is the extent to whichworkers benefit from program participation. This chapter investigates this issue by comparing how muchworkers inprogram firms would have earned hadthey beenworking for a non-program firm.'7 This chapters starts by presenting an overview of the wages earned by workers in different occupations in different types of firms, before proceeding to analyze the impact of the program using anearnings functions framework." A DESCRIPTION OF EARNINGS 6.2 The median wage in the construction sector i s 450 Birr a month, but there i s significant variation (Table 6.1). Engineers, the best educated workers, earn the most with a median wage o f 2000 Birr per month, while unskilled labourers, the least educated workers, are the worst off with a median monthly income of 300 Birr. Large firms pay higher wages thansmall firms for almost all occupations. Table 6.1: Median Monthly Wages (in Birr) Labour Labour Manager Engineer Foreman (skilled) (unskilled) Apprentice Non-ProgramSME 650 1800 720 400 273 150 ProgramSME 600 1000 550 500 312 200 Non-ProgramLarge contractors 1200 2500 1400 780 320 150 ProgramLarge Contractors 2000 1400 900 675 360 250 6.3 Program firms pay lower wages than non-program firms for more specialized jobs (foreman, engineer manager) but they do pay more than non program firms for skilled and unskilledlabourers and for apprentices. Contractors pay more than SMEs for alljobs (except apprentices) and hire more skilled workers than SMEs outside the program. Figure 6.1 shows medianmonthly wages, inbirr, by occupation, program participation, and size of firm. 87 Clearly, other counterfactuals, such as being employedor working in another sector are possible. The one we choose for this analysis is however consistentwith our overall methodological approachwhich takes the goal of housingconstructionas a given and our focus on the IHDP. 88As already mentionedin chapter 3, ideally analyzingthis issue would require ideally the availability of panel data to analyze how program participation has influenced earnings and to consider potential dynamic benefits on the workers. 56 Figure6.1: MedianWages by Occupation,FirmSize, and ProgramParticipationStatus manager 0engineer Oforeman 3000 I mlabour skilled labour unskilled Sapprentice I- 2500 n 2000 1500 1000 500 0 Non-Program Program Non-Program Program Large SME SME Large Contractors contractors 6.4 Regressions o f earnings as a function o f worker and firm specific characteristics have been run.The main findings are: Returns to schooling are higher for higher educational level, and progressively increasing with the level o f educational attainment. In addition, having completed an apprenticeship i s associated with higher earnings. Being a casual worker i s related to lower earnings." Women are paid about 30 percent less than men-this finding remains even when controlling for occupational variables, pointin therefore to unequal pay for the same types o f occupations between women andmen. s Program participation has a positive and significant effect on earnings suggestingthat workers in program firms earn more than comparable worker in non-program firms. This programpremium exists for workers inprogram MSEsonly. The program premium seems to be related to the size at which firms operate, and, to the extent that differences inremunerationreflect differences inproductivity, and the higherinput intensity o ftheir technology. Estimating the models for individuals in different ranges o f the wage distribution (quantile regressions, with key coefficients summarized in Table 6.2.) show that the 89 Interactions between education dummies and being a casual worker are not significant, suggesting that the returns to education are similar for both casual and permanent workers-these results are not shown. 90Women's disadvantage is compounded by the fact that they are less likely to complete their schooling - which indirectly further affects their earnings potential. 57 effect o f program participation on earnings i s strongest for those at the bottom o f the earnings distribution. Table 6.2: The Impact of ProgramParticipationonWorkers at Different Quantiles of the EarningsDistribution 10% 25% 50% 75% Individual Controls 0.303 *** 0.160** 0.0867 0.0688 only (4.16) (2.79) (1.28) (1.29) Individual Controls 0.150 0.0152 -0.0452 -0.0220 andFirm Controls (0.64) (0.58) (-1.16) (-0.26) Note: significant at the 10 percent level, * ** significant at the 5 percent level, *** significant at the 1 percentlevel. 6.5 An important caveat o f these results is that the size and direction of the program premium identified could be influenced by non-observable differences between workers in and out o f the program. Several econometric techniques failed to identify this effect with the data available. SimulatedDistributionalImpact 6.6 To provide an estimate o f the aggregate effect o f the program on monthly earnings, we simulated earnings distributions. Figure 6.2 contrasts the current distribution o f earnings with the earnings predicted in the absence o f the program. The figure suggests that the program has shifted the earnings distribution to the right. In addition, the earnings distribution without the program has thicker tails, suggesting that the program has contributed to a more egalitarian wage distributionhas ledto wage compression. Yet, though the program premiumis largest for those at the bottom o f the distribution, as relatively fewer of them are inthe program (e.g. 21 percent inthe bottom quintile vs. 67 percent inthe second lower and 52 in the third lowest), the shift in the distribution i s most noticeable for those who are in higher deciles o fthe distribution. Figure6.2: Simulated EarningsDistributions Simulated Earnings Distributions A 4 I 5 log 6of monthly earnings 7 8 9 ___ Earningsdistributionwithouttheprogram ~ Earningsdistribution with the program (current) 58 7. CONCLUSION 7.1 Inthis report we have provideda first look at the.employment creation effect of the IHDP focusing on the central issue of whether the program has created more jobs than would have been created relying on existing firms. The centrality of this question i s that ifthis is not the case, then investing in coordinating the M S E support with the housing program is only justified if there are other benefits to doing so, such as job creation for the poor, or strengtheningcapacity in the construction sector as a whole. This study also addresses some of these other potential benefits, although a complete assessment would require a comprehensiveimpact evaluation study. 7.2 The limitations of this analysis should be made clear. The employment creation impact of the program has been assessed conditional on the housing demand exerted by the IHDP-in other words, the analysis does not compare the employment effect of the program versus the absence of the program. The use of cross-sectional data does not permit us to assess adequately the dynamic benefits of the program. Last, the survey was undertakenin December 2006, at a time when there was a particularly severe shortage of material inputs such as cement and re-bars inthe market. This situation may have biased some of the results (these effects are highlighted inthe text). 7.3 It is also worth underscoring that our emphasis in this analysis is different from program documents which measure the success of the program in terms of the absolute numbers of jobs created. Nevertheless, some of the insights from this study point to the difficulties of calculating what is a job; particularly when sustainability and capacity utilization inthe sector might be a concern. 7.4 Two main assumptions at the heart of the design of the program have beenidentified and tested: (i) the MSEs inthe construction sector, and particularly MSEs inthe program, that are more labour intensive than larger firms; and (ii) the support offered by IHDP to MSEs that inthe program is neededto overcome the market failures which would otherwise hinder their establishmentand sustainability. 7.5 With respect to the first assumption, our analysis shows that MSEs are indeedmore labor intensivethan large contractors, an effect which does not dependon the size of the firms bur rather of the activities they engage in. However, when comparing MSEs, labour demand i s found to be similar among program and non-program MSEs. These findings suggest that the program has not createda more labour intensivesupply chainfor housingthan would have beenthe case using existing MSEs, at least instatic terms. 7.6 We also tested whether the employment creation effects of the program occur through dynamic effects by determining whether program firms grow faster than non-program firms. IHDP firms are found to employ typically more workers than non IHDP firms. Our analysis finds that the growth in employment in program firms i s not faster than in non- program firms. Rather, they are larger becausethey started larger, employingmore workers at startup. The finding i s interesting as there is a large body of evidence showing that large firms tend to grow more slowly than smaller ones. The fact that program firms manageto stay 59 larger suggests that program participation helps firms grow faster than non-program firms which startedwith a similar size. 7.7 For what concerns the second hypothesis on the need o f program support to overcome market failures, we find that MSEs report significant constraints particularly in terms o f access to land and credit. Even if these constraints are self-reported, and as such not necessarily indicative o f market failures per se, the existing literature confirms that the regulatory environment as well information asymmetry can seriously limit access to factor markets. Our findings show that not all firms in the program receive support, but those that receive support perform significantly better than those that do not. Access to credit seems to be the most effective type of support and is associated with significantly higher production. In addition, the provision o f a space to work i s associated with significantly higher profitability (suggesting that renting space might be expensive, and the provision o f a working space significantly reduces costs). Both these elements o f the program seem therefore to be having a positive impact, and it might be worth considering whether greater efforts need to be made to ensure that all firms inthe program have access to them. 7.8 While ensuring greater coverage o f support is within the program remit, it is worth underscoring that other broad rangingpolicy reforms could strengthenthe program ability to reach its own goals. The development o f a more robust mortgage market, for example, could allow the extension o f commercial financing to a greater segment o f the housing market. With a smaller supply gap in the provision o f housing, public resources could be more effectively targeted to the provision o f low income housing. These possible developments seem in line with PASDEP vision o f a national IntegratedHousing Development Program that "integrates public and private sector investment with MSE development and the provision o f basic services" (p. 163). Similarly, reform options for the way land is allocated could be considered and prove to be beneficial to the development of the construction sector (and more generally o f urban development). These are discussed indetails inWorld Bank 2007, b. 7.9 One o f the crucial elements o f program support appears to be the guaranteed demand for the firms' products that it provides. However, there are signs that fixed price system de facto in vigor resulted in lower profits (as input prices are similar across program and non- program firms, but output prices are lower). Consistent with our diagnosis, options to introduce a system which would provide for a more systematic review o f contractual terms, are currently beingconsidered. 7.10 While this points to the positive impact o f program interventions, the findings also raise some concerns about the sustainability o f the jobs created. Despite signs o f a construction boom, capacity utilization inthe sector appears far from full. This i s explained by IHDP staff as a result o f the inability o f the IHDP to build the total number o f housing units according to the planned schedule. Nevertheless, this illustrates that MSEs created by the program might be unable to adjust to changing demand conditions. This highlights the possible lack o f competitiveness o f program firms in the wider market. This seems to be supported by the fact that when asked about their competitors, the majority o f firms indicate that their greatest competition i s fiom firms outside the program. A related concern i s that the program might be encouraging an excessive specialization in products which would not be 60 demanded in the present extent (e.g. pre-cast beams) by construction systems different from the one currently adopted bythe IHDP. 7.11 It is worth underscoring that the overall effect o fthe program on employment partly depends on how the program has affected non program firms. From the perspective o f non- program firms, the program has increased the pressure on limited input markets, making it more difficult for non-program firms to compete. The direct impact o f the program on labour markets (both interms o f absorption and drivingup wages) seems to have been limited. 7.12 A further hypothesis that has been explored is whether program firms create employment which i s more pro-poor. This aspect i s particularly relevant because program documents emphasize its role in improving the livelihoods o f low-income residents, while available evidence shows that such role i s not fulfilled by the housing component o f the program. Further, the program sets itself goals in terms o f reaching particular groups such as women. Our analysis o f the beneficiaries highlights that program firms do not significantly draw upon the low skilled workers or the unemployed, nor do they employ significantly more women than non-program firms, though it offers a possibility for casual workers to become permanently employed. These finding does not challenge the usefulness o f creating jobs in urban areas. However, alternative programs or interventions may be necessary to specifically target the poorest andless skilled populationinurbanareas. 7.13 While the program i s not particularly effective at reaching particularly vulnerable groups, program firms remunerate workers more, particularly those at the bottom o f the wage distribution. Overall, therefore, the program seems to have a positive distributional effect, even though such effect i s bounded by the program limited coverage o f workers inthe bottom decile - a finding which i s not surprising given that participation in the program appears to require better skills than working innon-program MSEs. Further, while the program does not seem to reach particularly some o f the groups it was meant to (e.g. women), and benefits others, for example migrants (a large proportion o f workers in the sector appear to have migrated as adults), it might have contributed to reduce unemployment in other ways such as by increasing labour demand and removing marketfailures which would have preventedMSE developments. 7.14 The main implications o fthis study as the IHDP continues to be implementedinAddis . Ababa and as it is being rolled out to other regions, can be summed up as follows: While the large absorption o f labour by the program is unquestionable, the issue o f whether the program i s the most efficient way o f creatingjobs needs to be considered. We find evidence o f possible significant dynamic benefits o f the program (e.g. in terms o f firm growth), which suggests that from the point o f view o f the employment creation effects o f the program, elements such as the introduction o f new technology to the sector, building skills which might make firms sustainable in the future, and . building skills for specialized workers deserve to be strengthened. It is not clear that the complexity of coordinating the housing and the MSE support side o f the program pays off interms o f benefits. Note that while the program aims to address mostly the constraints in terms o f skills and technology, de facto its greatest 61 benefit in a very competitive environment such as the construction sector i s the provision o f demand. The program appears to have fostered the entry o f a great number of firms, but at least in the short run the entry of these firms and their employment creation impact do not appear to have increased demand for labour more than if alternative MSE-based housing delivery options had been pursued. Inlight o f the costs o f coordinating the program, and in particular the supply of inputs and the construction side o f this MSE based model, it i s worth considering whether alternative MSE-based housing delivery options could free resources to focus on the more successful aspects o f the program, such as for example by strengthening skills . development. The high degree o f capacity underutilization o f MSEs deserves closer scrutiny to ascertain whether the targets interms o f firm creation are realistic. Setting appropriate targets for firm creation will be particularly important as the program i s rolled out to other regions. Further, the extent to which underutilization might reflect excessive . specializationinproducts used mostly by the IHDP should be assessed. Measures should be taken to strengthen and increase access to factors o f production that are found to be critical for firm survival and growth - namely, the provision o f . credit and working spaces. The fixed price system can be detrimental to the survival o f the firm if the prices are not regularly reviewed to take into account market prices o f inputs not directly provided by the program. Current efforts to address this challenge are therefore very welcome. 7.15 The monitoring system o f the program needs to be strengthened by clarifying the type o f indicators to be collected by the program and their purpose. If the program's employment targets are to be measured primarily by the absolute number o fjobs created, this needs to be better defined taking into account issues o f capacity utilization and sustainability. Inaddition, as the program is rolled out to different urbanareas and faces challenges specific to local conditions, it will be important to experiment with alternative design features and learn from those experiences, inthe spirit o f PASDEPs' emphasis on "seeking results through learning" (p. 224). 7.16 To increase the overall efficiency o f the program, it might also be worth considering implementing the incentive system for rewarding good performance with repeated contracts, as originally planned by the program. Such system could contribute to the competitiveness o f firms and thus their sustainability; and help improving the targeting o f support. 7.17 Finally, it i s worth noting that while the program appears to have an overall progressive impact, its employment generation component is not necessarily reaching some o f the intended target groups (women) or the poorest. To the extent that these objectives are an important motivation for the program, other specific interventions with a strong pro-poor impact might also be needed. 62 APPENDICES 63 APPENDIXA: SAMPLE SCHEME A.1 Samplingof Firms Insamplingfirms a distinction was madebetweenprogramandnon-program firms, as well as betweencontractors andnon-contractors. Non-Contractors Table A.1. presents an overview o f the proportion o f firms in the underlying population o f interest, sorted by activity, and the corresponding proportion o f firms in our sample. To ensure our sample was representative, we made sure that the proportion o f firms inour sample rou hly corresponds to the proportion o f firms inthe population engaging ina particular activity.8 Ifonly very few firms engaged in a particular activity, such as carpentry, 1 that activity was typically not included in the sample, while firms in other relatively minor activities were marginally oversampled, at the expense o f slightly undersampling firms engaging inpopular activities. One problem was that firms can engage indifferent activities at the same time. Since it was difficult to establish which firms engaged in multiple activities, this was not taken into account when sampling firms. It turned out that 15 program non- contractors and 32 non-program non-contractors engaged inmultiple activities. One difficulty in sampling firms was that many were registered under the rather general headings o f "Production & Sales o f UnfabricatedBuildingMaterials" and "Production & Sales o f Fabricated Building Materials"; while clearly relevant for the purposes o f comparison, it was difficult to guess a priori to what extent these firms engaged in activities comparable to program firms. Contractors Contractors in Addis fall in two types: building contractors, which only focus on the construction and general contractors, firms which engage in a multitude o f activities which may include building construction but also road construction and infrastructure development. The sample was confined to building contractors, since we wished to avoid comparing the incomparable. Contracting firms were sampled on the basis o f their license grade, rather than on the basis o f their activities. As can be seen in Table A.2 below, large contractors were oversampled to ensure they were represented in our sample. The table also reveals that the sampling frame provided by the list o f program firms was not fully accurate as a number o f firms had a license grade which was either higher or lower than the license grade listed in the list o fprogram firms and the general registry o f firms. 91Also note that for pre-cast beam production, and to a lesser extent for hollow block fabrications, there are no suitable counterparts for program firms, since frms engaging in these production activities use methods not previously applied inEthiopia. In addition, electrical and sanitary installation was not practiced separately prior to the introduction o f the program. Consequently, it was difficult to trace firms engaging in such activities amongst non-program participants. 64 Table A 1: Non-Contractors Program Non-program # in # in # in # in population sample population sample 1418 89 2395 97 Activities engagedin by different firms9` Pre-CastBeams 186 24 1 1 Hollow Blocks 440 39 297 32 Wood andmetalWork 408 41 461 25 Gravel 92 9 2 4 Wall 162 13 0 6 ElectricalInstallation 80 8 3 5 Sanitary 32 8 0 7 Installation Site work andFinishing 23 11 0 9 Other 0 2 61 5 Fabricated93Building 0 0 1450 0 Materials UnfabricatedBuilding 0 0 120 0 Materials 92Note that firms can engage indifferent activities. 93The categories "fabricated buildingmaterials" and "unfabricated buildingmaterials" comprise a multitude o f activities which are comparableto the activities program f m s engage in, which is why firms inthese categories were selectedfor the controlgroup. 65 Table A 2: Contractors Program Non-program # in # in # in # in population sample population sample License-grade 1 License-grade 2 License-grade 3 License-grade 4 4 5 60 7 License-grade 5 7 7 141 4 License-grade 6 57 11 107 4 License-grade 7 3 82 4 License-grade 8 140 3 License-grade 9 40 1 License-grade 10 4 2 Total 70 31 610 270 A.2 Samplingof Workers For each firm, up to 4 workers were interviewed. Enumerators were instructed to interview workers in different occupations, if possible. In addition, they were instructed to attempt to interview one casual worker and at least one woman per firm. In addition, they were instructed to attempt to sample women not engaging inunskilled tasks, ifpossible. High- skilledworkers, casual workers andwomen are consequently overrepresented inour sample.94 Unfortunately, the manager's permission was required to be able to interview workers. As a result the sample o f workers may not be entirely random, yet whether one was sampled does not seem to be determined by strategic considerations on the manager's part, but rather on random factors, such as being around at the time o f the interview. 94Inthe descriptives this is corrected for by using sample weights; the regressions are typically unweighted, unless otherwise stated. 66 APPENDIX3: CONSTRUCTION KEYVARIABLESOF Capital The capital stock variable measures the value o f all capital owned by the firm, including buildings, machineries, vehicles, tools, and other assets, at replacement cost. The measure does not account for rental capitaLg5Only 7 firms in our sample indicated to rent capital. Moreover, imputingthe value o f such rental capital and usingthe sum o f rental capital and owned capital did not affect the results. Inputs Data on total inputs were obtained by adding up the expenses for individual inputs. The inputsvariable measures the total value o f all inputs,not the physical quantity o f inputs. Constructionof the Welfare-indicator The predicted welfare indicator i s constructed by using information on asset holdings to predict household expenditure, using the CWIQ methodology. The model was calibrated usingdata from the 1999/2000 HECIS; household expenditure inthat survey was regressed on asset holdings for the subsample o f Addis residents. The resulting parameter estimates were used to predict household expenditure usingthe information on asset holdings inthe AACES. Estimates o f the predicted welfare expenditure turned out to be rather high, perhaps because ownership o f a radio and a TV were lumped into a single category, or because the price o f goods, particularly electronics, has changed over the years. Imputingthe Average EducationalAttainmentof the Workforce Taking a weighted average o f the typical educational attainments of workers in different occupational categories (from the firm-level questionnaire) yields an estimate of the average educational attainment of the workforce. Firm managers' responses to questions about the educational attainment o f workers in different occupational categories (managers, engineers, foremen, skilled labourer, unskilled labourers and apprentices), were used to impute the educational attainment o f workers in different occupations, while the weights correspond to the relative proportion o f all employees of the firm within a certain occupation. Unfortunately, firm managers only had to provide a coarse estimate o f workers' educational attainment, forcing us to impute more specific numbers. The following imputations were used: 1) Never been to school-0 years o f schooling; 2) Primary school incomplete-3 years o f schooling; 95Imputingthe value of such capital was difficult since the replacement cost o f these items was not documented inthe questionnaire; instead items were valued at the median value for each item, multiplied by the percentage deviation o fthe rental price from the median rental price. 67 3) Primary School complete-6 years of schooling; 4) Grade 10 complete-1 0 years o f schooling; 5) Grade 12 complete-12 years of schooling; 6) College diploma-1 4 years o f schooling; and 7) College degree-1 6 years o f schooling. Such imputations are inherently arbitrary and the measure o f educational attainment constructed is noisy. Ideally, one would have observed the educational attainment o f all workers inthe firm, yet this was impossible since only 4 workers per firm were interviewed, which is why we rely on the managers' responses instead. The Average Age o f the Workforce The procedure for constructing a variable representing the average age o f the workforce i s similar to that for imputingthe average educational attainment o f the workforce; a weighted average o f the manager's responses to questions about the typical age o f workers in different occupations, with weights proportional to the.share of the workforce within any given occupation, were used to construct to construct a measure o f the average age o f the workforce. Again, coarseness inthe response options for managers forces us to make arbitrary choices. The following imputations were used: 1) 15-20 years-17.5 years; 2) 20-25 years-22.5 years; 3) 26-35 years-30 years; 4) 36-50 y e a r s 4 3 years; and 5) Above 50 years-56 years. Price-deflators The data on the amount and costs of different inputsused and the amount and prices of different outputs produced allowed us to construct firm-specific input- and output prices. Collecting these input- and output data was arduous because firms, particularly large ones, simply do not keep track o fthe precise amounts and exact costs o f (all of) the different inputs they use. The data on inputs and outputs consequently seem to contain a lot of measurement error and some severe outliers, which were dealt with by scaling down reported prices to appropriate orders o f magnitude that is to the same order o f magnitude as the median price. It should be noted that input and output data were missing for 30 contractors and for 49 firms in total. To enable a price-comparison between firms using different inputs, input- and output price deflators were created as follows. Firstly, firm-specific inputs and output-prices were created for each in-and output by dividing quantity o f input (output) used by the total amount o f money paid (received) for the particular input (output). Secondly, these prices were divided by the median price for that input (output) for all firms, to arrive at a relative price. The input price-deflator is then constructed by taking the weighted average o f these relative prices, with weights corresponding to the relative cost-share of the respective input inthe total 68 costs of inputs (outputs). Similar deflators were constructed for program and non-program inputs and outputs separately, enabling us to compare the cost of program inputs with the costs of non-program inputs. The resulting price-deflators are reasonably well-behaved as they are symmetrically distributed around 1, although there are some implausible outliers, which are excludedby trimming values greaterthan2 and smaller than 5. 69 REFERENCES "Addis Ababa Integrated Housing Development Program" (no author), Addis Ababa, March 2004. "Low Cost Housing: Would this be the Solution?" (no author) in Construction Ahead: Bi- monthly interface with the construction industry ", p. 32, Butterfly Publishing, September- October 2005. "Mugher Cement Enterprise Increases Cement Price" (no author) in Addis Fortune, 26 December 2006. AccessedJanuary 5 2007, via All Africa Global Mediavia COMTEX: http://www.docenp;ines.com/ArticleFile.asp?Instance= 105&ID=27255 19135DA48A6B3 1793 33B3196C72. Addis Ababa City Administration Housing Development Project Offce/HDPO, 2004. "Brief Project Profile", September2004. Betcheman, Gordon, Karina Olivas and Amit Dar, 2004. Impacts of Active Labor Market Programs: New Evidence from Evaluations with Particular Attention to Developing and TransitionCountries World Bank Social ProtectionDiscussion Paper Series n. 0402. Biesebroeck, van. J., 2005. "Firm Size Matters: Growth and Productivity Growth inAfrican Manufacturing", Economic Development and Cultural Change, 53 (2005), 545-583. Biggs, T. (2002), Is Small Beautiful and Worthy of Subsidy? International Finance Corporation, Washington. Bigsten, Arne, and Mans Soderbom, 2006. "What Have We Learned from a Decade of Manufacturing Enterprise Surveys in Africa?," World Bank Research Observer, Oxford University Press, vol. 21(2), pages 241-265. Fay, Marianne, and Anna Wellenstein. 2005. "Keeping a Roof over One's Head: Improving Access to Safe and Decent Shelter." In The Urban Poor in Latin America, ed. Marianne Fay, 91-124. Washington, DC: World Bank. GDS Global Development Solutions, 2006. "Integrated Value Chain Analysis for the Housing Construction Sector inEthiopia." 8 June 2006. GTZ, 2004. "Grand Addis Ababa Housing Program. Considerations on SME, training, standards, stakeholder involvement & organizational structures (draft)." Addis Ababa, March 2004. GTZ & Ministry o f Federal Affairs, 2004 "Low-Cost Housing: Technical Manual" Addis Ababa. 70 Halberg, K. 2001 "A market-oriented strategy for small and medium-scale enterprises." DiscussionPaperNo. 40. International FinanceCorporation IHDP, 2006. Presentation of the IHDP Program Office at the World Bank Office. Addis Ababa, November. ILO (2004) ASIST. Advisory Support Information Services and Training for Employment Intensive Infrastructure, Bulletinn. 17. March2004 Keddeman, Willem "Of NETS and ASSETS. Effects and impacts of employment creation programmes - A review of ILO experience". Socio-Economic Technical Papers (SETP) n.1. ILO Geneva Little, I."Small Manufacturing Enterprises in Developing Countries" World Bank Economic Review 1(2) pp. 203-235. (1987) Little, I.M.D., Mazumdar D. and J.M. Page, Jr. Small Manufacturing Enterprises: A Comparative Analysis o f India and Other Economies" Oxford University Press, New York, N.Y., (1987). Lohnert, Prof. Dr. Beate, and Regina Fein. ,2006. "Socio-economic Survey on Acceptance of New Living Patterns in the Bole/Gerji Pilot Condominium Apartment Construction Site." In Integrated Housing Development Program Vol. I: Management Manual-How to Plan and Implement Integrated Housing Development Programs, p. 120-122. Addis Ababa: GTZ, MWUD,Addis Ababa City Government, andIHDP. Teal, F. "A Note on Using Firm Surveys for Policy Analysis in Low Income Countries `I, unpublishedmanuscript (2007) World Bank, 2004. "Land and Housing inAddis." Unpublisheddraft World Bank, 2005. "Well-Being and Poverty in Ethiopia: The Role of Agriculture and Agency." Washington, DC: World Bank. World Bank, 2007 a. "Urban Labour Markets in Ethiopia: Challenges and Prospects." Washington, DC: World Bank. World Bank, 2007 b. "The Challenge of Urbanization in Ethiopia. Implications for Growth andPoverty Alleviation." Washington, DC: World Bank. 71