TANZANIA MAINLAND POVERTY ASSESSMENT Structural Transformation and Firms Performance PART 2 STANDARD DISCLAIMER This volume is a product of the staff of the International Bank for Reconstruction and Development/The World Bank. The findings, interpretations, and conclusions expressed in this paper do not necessarily reflect the views of the Executive Directors of The World Bank or the governments they represent. The World Bank does not guarantee the accuracy of the data included in this work. The boundaries, colors, denominations, and other information shown on any map in this work do not imply any judgment on the part of The World Bank concerning the legal status of any territory or the endorsement or acceptance of such boundaries. COPYRIGHT STATEMENT The material in this publication is copyrighted. Copying and/or transmitting portions or all of this work without permission may be a violation of applicable law. The International Bank for Reconstruction and Development/ The World Bank encourages dissemination of its work and will normally grant permission to reproduce portions of the work promptly. For permission to photocopy or reprint any part of this work, please send a request with complete information to the Copyright Clearance Center, Inc., 222 Rosewood Drive, Danvers, MA 01923, USA, ­ telephone 978-750-8400, fax 978-750-4470, http://www.copyright.com/. All other queries on rights and licenses, including subsidiary rights, should be addressed to the Office of the Publisher, The World Bank, 1818 H Street NW, Washington, DC 20433, USA, fax 202-522-2422, e-mail pubrights@worldbank.org. TANZANIA MAINLAND POVERTY ASSESSMENT Structural Transformation and Firms Performance PART 2 Table of Content Key Findings.............................................................................................................................................................................................................vii Chapter 1: Structural Transformation...................................................................................................................................... 1 I. Economic Growth and Structural Transformation..................................................................................................................2 II. Labor Market Transformation..................................................................................................................................................7 III. Productivity Gaps...................................................................................................................................................................13 IV. Occupation Decision .............................................................................................................................................................17 Chapter 2: Firm Profiles and Structural Transformation...................................................................................................19 I. Profile of Tanzanian Firms .....................................................................................................................................................20 II. Which Firms Create the Most Jobs.......................................................................................................................................31 Chapter 3: Access to Finance ............................................................................................................................................... 37 I. Extent and Determinants of Individuals’ Financial Inclusion..............................................................................................39 II. The Effects of Access to Finance on the Growth of Tanzanian Firms ................................................................................45 Chapter 4: Economic Choices by Socially Embedded Individuals: recent advances in economics research and their potential implications for informal finance in Tanzania.................................................................................49 Occupational choices and structural change are central issues within poverty and development discourses..............50 I. What do We Know About Occupational Choices?..............................................................................................................51 II. Structural Change in Tanzania: The Role of Household Enterprises..................................................................................56 Appendixes.................................................................................................................................................................................61 Appendix A: Measuring Structural Transformation.....................................................................................................................61 Appendix B: Data and Descriptive Statistics of the 2013 Enterprise Survey.............................................................................68 Appendix C: Determinants of Job Creation.................................................................................................................................71 Appendix D: Determinants of Financial Inclusion .......................................................................................................................75 Appendix E: Access to Finance and Tanzania’s Firms Growth ...................................................................................................79 Appendix F: A Summary of Theoretical Contributions on ROSCAs...........................................................................................84 References.................................................................................................................................................................................. 87 List of Boxes Box 1.1: Measuring Labor Productivity..........................................................................................................................................................................................................14 Box 1.2: Biases of Productivity Measures in Tanzania..................................................................................................................................................................................15 List of Figures Figure 1.1: GDP Growth, 1999–2017, Percent.................................................................................................................................................................................................2 Figure 1.2: Contribution of Economic Sectors to GDP, 1998–2017, Percent................................................................................................................................................3 Figure 1.3: Contribution of Supply-side Factors to GDP Growth, 1998–2017, Percent and Percentage Points.......................................................................................4 Figure 1.4: Demand-side Contributions to GDP Growth, 1998–2017, Percentage Points..........................................................................................................................5 Figure 1.5: Growth Accounting Decomposition, 1998–2017, Percent and Percentage Points ..................................................................................................................6 Figure 1.6: Labor Shares by Sector, 2006 and 2014, Percent.........................................................................................................................................................................8 Figure 1.7: Sectoral Labor Distribution, 2011–2015.......................................................................................................................................................................................9 Figure 1.8: Working Hours, Primary and Second Jobs, 2011–15, Percent..................................................................................................................................................10 Figure 1.9: Annual Hours Supplied by Workers to Different Sectors, 2015................................................................................................................................................11 Figure 1.10: Average Number of Hours Worked per Job Sector, 2015......................................................................................................................................................12 Figure 1.11: Productivity per Worker and per Hour, by Sector, 2015 (TZS).................................................................................................................................................14 Figure 1.12: Productivity Gaps by Sector, 2015............................................................................................................................................................................................15 Figure 1.13: Productivity Gaps by Economic Sector, 2015, Percent............................................................................................................................................................16 Figure 1.14: Productivity Gaps per Worker by Activity, 2015.......................................................................................................................................................................16 Figure 2.1: Firm Distribution by Sector and Size, Percent...........................................................................................................................................................................20 Figure 2.2: Firm Distribution by Age, Size, Sector, and Informality, Percent..............................................................................................................................................22 Figure 2.3: Employment by Gender and Firms Characteristics...................................................................................................................................................................23 Figure 2.4: Industrial Employment by Skills, Percent...................................................................................................................................................................................25 Figure 2.5: Value Added by Firm Characteristics.........................................................................................................................................................................................26 Figure 2.6: Major Challenges Faced by Industrial Firms, Percent...............................................................................................................................................................27 Figure 2.7: Firm and Job Distribution by Geographic Characteristics, Percent........................................................................................................................................28 Figure 2.8: Poverty and Firm Distribution by Geographic Zone.................................................................................................................................................................29 Figure 2.9: Share of Jobs Created by Firm Size, 2010–13, Percent.............................................................................................................................................................32 Figure 2.10: Jobs Created by Age Class, 2010–13, Percent of Firms..........................................................................................................................................................33 Figure 2.11: Net Job Creation by Firm Age and Size, 2010–13...................................................................................................................................................................34 Figure 2.12: Jobs Created by Sector, 2010–13, Percent...............................................................................................................................................................................35 Figure 3.1: Financial Inclusion in Tanzania, 2017, Percent............................................................................................................................................................................40 Figure 3.2: Trends in Financial Inclusion, 2014 and 2017, Percent..............................................................................................................................................................41 Figure 3.3: Financial Inclusion in Tanzania, 2014 and 2017, Percent...........................................................................................................................................................42 Figure 3.4: Financial Inclusion in Tanzania, 2017, Percent............................................................................................................................................................................44 Figure 3.5: Access to Finance of Formal Enterprises, 2006 and 2013, Percent..........................................................................................................................................46 Figure 3.6: Access to Finance of Industrial Enterprises, 2013, Percent.......................................................................................................................................................47 Figure 3.7: Access to Finance of Informal Household Enterprises, 2014, Percent.....................................................................................................................................48 Figure B.1: Distribution of Firms and Jobs by Firms’ Size and Age............................................................................................................................................................70 List of Tables Table 2.1: Productivity of Firms by Size and Age, 2010 and 2013................................................................................................................................................................31 Table 2.2: Production by Firm Size and Age, 2013, Percent........................................................................................................................................................................32 Table 2.3: Net Job Creation by Firm Size, 2010–13......................................................................................................................................................................................33 Table 2.4: Firm Size Transition Matrix, 2010–13, Percent..............................................................................................................................................................................35 Table A.1: Number of Primary and Secondary Jobs by Survey Round (employed population aged 15+)..............................................................................................64 Table A.2: Number of Primary and Secondary Jobs by Survey Round (employed population aged 15+)..............................................................................................64 Table A.3: Monthly Income by Sector, 2014, TZS..........................................................................................................................................................................................66 Table A.4: Productivity by Sector, 2014, TZS.................................................................................................................................................................................................66 Table A.5: Summary Statistics of Household and Individual Characteristics by Sector, 2014...................................................................................................................67 Table A.6: Alternative-Specific Mixed Logistic Regression for Sector Choice of Individuals ...................................................................................................................67 Table B.1: Firm Descriptive Statistics, 2013...................................................................................................................................................................................................69 Table B.2: Firm Descriptive Statistics, 2013...................................................................................................................................................................................................69 Table B.3: Employment by Firm Size and Age, 2013....................................................................................................................................................................................70 Table C.1: Correlates of Job Creation 2010–2013 – Size and Age..............................................................................................................................................................73 Table C.2: Correlates of Job Creation 2010–2013 – Productivity and Performance on All Firms.............................................................................................................73 Table C.3: Correlates of Job Creation 2010–2013 – Sector of Activity on All Firms...................................................................................................................................74 Table D.1: Determinants of the main Financial Inclusion indicators...........................................................................................................................................................77 Table D.2: Determinants of Saving (Formal and Informal)...........................................................................................................................................................................77 Table D.3: Determinants of Saving Motivation.............................................................................................................................................................................................77 Table D.4: Determinants of Loan-taking Motivation (Formal and Informal)...............................................................................................................................................77 Table D.5: Determinants of Sources of Borrowing.......................................................................................................................................................................................78 Table E.1: Access to Finance Constraint and Participation in Financial Market (mean)............................................................................................................................80 Table E.2: Effect of Access to Finance and Business Environment Constraints on Firms Growth, Subjective Measures, 2013.............................................................81 Table E.3: Effect of Objective Measure of Access to Finance on Firms Growth, 2013..............................................................................................................................83 Key Findings Structural transformation Tanzania recorded strong economic growth over the across sectors have been limited. Overall, over 60 percent past decade, particularly in non-agricultural sectors, of the employed labor force still works in agriculture, and slowly transforming the economy’s structure. Growth high-growth sectors have not witnessed complementary in gross domestic product (GDP) averaged 6.3 percent employment shifts. The transition away from agriculture per year since 2007. Between 2013 and 2017, the com- seems slightly faster when based on hours worked by sector. pounded yearly growth rate was 6.7 percent, dropping For instance, accounting for total hours worked in each to 3.4 percent when adjusted by population size. The sector, agriculture accounts for less than half of total hours fast-paced growth of industry and services has trans- worked. More importantly, even when agriculture is the formed the structure of the economy, even though all main sector of employment, workers supply a considerable sectors continue to be supply-side drivers of growth. On amount of time to nonfarm sectors. the demand side, economic growth was primarily driven by private consumption and investment; public consump- Accelerating the transition of labor away from agri- tion is modest. Finally, growth accounting disaggregation culture can potentially generate sizeable productivity shows that capital accumulation is the main driver of gains. Labor productivity is markedly higher in industry growth, followed by an expansion of the labor force. and services as compared to agriculture, suggesting that a faster sectoral transition can increase the overall However, transformation of the labor market’s structure productivity of the economy. In a more granular disag- was slower. Despite the sectoral shift from agriculture gregation, mining, transport, and trade are the most to manufacturing and services, employment transitions productive sectors in Tanzania. Firms’ profile The majority of Tanzanian firms are small and young, but be a severe barrier to the growth of larger firms. Overall, employment is primarily concentrated in large and older smaller and younger firms employ a higher proportion of firms. Firms’ landscape is primarily occupied by micro and unskilled labor: the share of skilled employees among the small-scale businesses, as well as young companies, operat- workforce of large and older firms is nearly twice that in ing in the manufacturing or trade sectors. Informality is also small and young firms. In terms of sectors, food and bever- widespread, particularly among younger and smaller firms. ages and wood processing account for the largest share of However, employment is disproportionately concentrated employment of unskilled workers. in older firms, and strongly skewed towards nonmarket ser- vices and manufacturing, followed by trade. Overall, net job There is considerable regional variation in firm loca- creation over the period 2010-13 was positively correlated tion and employment patterns, and this tends to be with firm age and firm size. correlated with the distribution of poverty across the country. More than one-fourth of businesses are concen- Large firms employ qualified workers and contribute trated in the Eastern zone, where Dar es Salaam is located. more to total value-added. Large industrial firms contrib- The city, which constitutes the main commercial and admin- ute twice as much to value-added as medium industrial istrative hub of Tanzania, hosts 25 percent and 39 percent of firms, and about 20 times more than micro industrial firms. medium and large businesses respectively, including most Nevertheless, shortage of qualified labor constitutes to manufacturing firms. Consequently, jobs are concentrated in the Eastern zone, which accounts for 30 percent of all robust correlation between the incidence of poverty and the jobs. There is strong evidence of a negative and statistically number of businesses at both regional and district levels. Access to finance Financial inclusion in Tanzania remains limited. A large ­ opulation – low-educated individuals or belong- of the p part of the population continues to lack any formal finan- ­ oorest households. Remittances have been the ing to p cial tool – namely the ownership of a formal bank account, underlying driving force as mobile banking constitute the the possession of formal savings, and the access to credit primary c­ hannel to send or receive transfers – 41 percent provided by a formal financial institution. Financial inclusion of the population used mobile banking either to receive is strongly determined by socio-economic characteristics. or send remittances, while only 1.5 percent used regular On average, older, better-off, and educated males are more bank accounts, 5.5 percent used other forms of transfer such likely to be financially included than young, uneducated, and as cash. poor women. In particular, education and wealth appear as very significant factors when explaining financial inclusion. Firms’ owners perceive financial access as a major con- Meanwhile, structured forms of savings, whether formal straint, but evidence does not show a significant impact or informal, remain scarce and subject to socio-economic of financial constraint on firms’ growth. Tanzania’s formal factors. Finally, access to credit is achieved primarily through firms perceive access to finance as a major and increasingly informal arrangements – namely relatives and saving clubs. important obstacle, in particular for smaller firms. More spe- cifically, a third of industrial firms feel financially constrained, Nevertheless, financial inclusion has progressed rapidly while only one tenth of informal household enterprises have over the last years, primarily driven by the development access to credit, which primarily comes from micro-credit of mobile banking and the increase of remittances. entities or informal credit channels. However, data and Between 2014 and 2017, the share of the population with econometric models show that the growth of Tanzanian mobile bank accounts increased by 6 percentage points firms does not appear to be significantly affected by finan- (pp), with fast increases recorded in vulnerable groups cial constraints. Economic choices by socially embedded individuals Data on Household Enterprises (HE) raises several funda- of HEs’ owners believe that ROSCAs have or could have a mental questions on their structure and their operations. positive impact on their businesses. It leads to the following Survey data shows that the major part – 80 percent—of questions: Is it possible to build on the current functioning the money borrowed to fund HEs comes from informal of ROSCAs in order to enhance their ability to channel sources, bearing the question of the implications for the informal credit, both at the intensive and at the extensive efficiency of these enterprises. Meanwhile, only a minority margin? And what explains the fact that individuals in ­ wners – around 10 percent—participate in Rotating of HEs’ o Tanzania resort more to borrowing and lending through rel- Savings and Credit Associations (ROSCA), but 80 percent atives, friends, and neighbors, rather than through ROSCAs? CHAPTER 1 Structural Transformation I. Economic Growth and Structural Transformation Tanzania’s economic growth has been robust since 1999, rate was at 6.7 percent (Figure 1.1). When adjusted for the with the economy growing at a steady compounded rate growth in population, Tanzania’s GDP per capita grew by of 6.3 percent annually. The figures are based on the new 3.2 percent annually from 1999 to 2017, and by 3.4 percent National Accounts series with a base year of 2015, released in for 2013–17. February 2019. Between 2013–17, the compounded growth The fast-paced growth of industry and services sector has transformed the structure of the economy, even though all sectors have been and are supply-side drivers of growth. The economy has gradually entered a sectoral trans- lesser extent services. While in 1998 agriculture accounted formation characterized by a decrease in the weight of for 40 percent of GDP, in 2017 it was down to 28 percent, agriculture and an increase in that of industry and to a with the decline being gradual but steady over the last two FIGURE 1.1: GDP Growth, 1999–2017, Percent 10 7.7 7.0 6.8 6.9 6.8 6.7 6.7 6.3 6.3 6.2 5.9 5.7 5.6 5.3 5 4.3 4.5 3.8 3.6 3.4 3.5 3.5 3.5 3.2 3.0 2.9 2.9 2.6 2.4 2.0 1.3 0 1998– 1998– 2003– 2008– 2013– 2008 2009 2010 2011 2012 2013 2014 2015 2016 2017 2017 2002 2007 2012 2017 5-year period Detail GDP growth GDP per capita growth Source: National Accounts, base year 2015. Tanzania National Bureau of Statistics (NBS) February 2019. 2 TA N Z A N I A M A I N L A N D P O V E R T Y A S S E S S M E N T decades (Figure 1.2A). Meanwhile, the weight of industry in growth in other sectors. The steady decline in the rela- GDP rose to 29 percent and of services to 43 percent. Though tive contribution of agriculture to GDP did not result from services account for the largest share of GDP, since 1999 their sectoral loss of value—throughout the period agricultural contribution to the economy went up by just 3 pp. Industry, output kept growing. For instance, between 2013 and 2017, in contrast, has grown much faster, increasing its contribution its annual compounded growth rate was 4.9 percent. It also to GDP by more than 50 percent. The curves for industry supported the growth of the whole economy, contributing and agriculture crossed in 2016, when both accounted for 1.3 pp in 2016 and 1.6 pp in 2017 (Figure 1.3B). However, 28 percent of Tanzania’s economy. with other sectors of the economy growing faster, the rel- ative weight of agriculture declined. For 2013–17, industry The primary beneficiaries of this structural shift were grew at a 9.4 percent annual compounded rate, and services the construction, manufacturing, financial, insurance, at 6.2 percent. and real estate subsectors. From 1998 to 2017, the weight of construction in GDP doubled from 7 to 14 percent Growth in industry followed by services sectors allowed (Figure 1.2B); over the last decade, its weight gained 4 pp. Tanzania to maintain high GDP growth in 2016 and Meanwhile, manufacturing, financial, insurance, and real 2017. Until 2016, services were the primary driver of estate each gained 2 pp in GDP and the contribution of ICT growth, contributing about 2.8 pp annually in 2008–16, went up by 1 pp. but in 2016 industry took the lead—primarily due to the construction subsector, which contributed about 1.7 pp The transformation in Tanzania’s economy was due not to overall economic growth in 2016 and 1.9 pp in 2017 so much to agricultural decline as to relatively higher (Figures 1.3A and B).1 FIGURE 1.2: Contribution of Economic Sectors to GDP, 1998–2017, Percent A. Sectors Aggregated                      B. Sectors Disaggregated, 5-year Periods 60 100 2 1 1 1 10 10 9 10 9 9 11 11 80 1 1 1 2 40 20 20 20 20 60 5 6 6 6 8 9 10 20 7 13 40 8 9 9 20 39 35 0 32 29 1998 2001 2004 2007 2010 2013 2016 Agriculture Industry Services 0 1998–2002 2003–2007 2008–2012 2013–2017 Agriculture and fishing Manufacturing Construction Mining and other industries Trade, transportation and ICT accommodation Other services Public administration Financial, insurance and real estate Source: National Accounts, 2019. 1 In other words, construction drove national growth by 25 percent in 2016 and 28 percent in 2017. C h a p t e r 1 S TR U C T U RA L TRA N S F O R M AT I O N 3 FIGURE 1.3: Contribution of Supply-side Factors to GDP Growth, 1998–2017, Percent and Percentage Points A. Contributions of sectors to economic growth, Percent and PP 8 7.7 7.0 6.8 6.9 6.8 6.7 6.7 6.3 6.3 6.2 5.9 5.7 5.6 6 5.3 3.7 3.2 2.6 2.2 2.7 2.7 2.8 4.5 1.9 3.2 3.6 2.5 2.8 4 2.6 0.8 1.4 2.6 2.9 2.5 2.6 2.9 3.0 2.1 1.5 3.0 1.3 2.3 2 1.7 1.0 2.1 2.4 1.0 1.4 1.8 1.4 1.3 1.4 1.7 1.8 1.4 1.6 1.0 0.9 0.9 1.2 1.3 0 1998– 1998– 2003– 2008– 2013– 2008 2009 2010 2011 2012 2013 2014 2015 2016 2017 2017 2002 2007 2012 2017 5-year period Detail Agriculture Industry Services GDP Growth B. Contributions of subsectors to economic growth, PP 1998–2017 1998–2002 2003–07 2008–12 2013–17 2013 2014 2015 2016 2017 Agriculture & fishing 1.4 1.8 1.4 1.3 1.4 1.2 1.8 1.4 1.3 1.6 Manufacturing 0.6 0.4 0.7 0.6 0.7 0.4 0.8 0.5 0.9 0.7 Construction 1.0 0.6 1.1 0.8 1.6 2.2 0.2 1.4 1.7 1.9 Mining & other industries 0.5 0.5 0.7 0.3 0.4 0.4 0.4 0.4 0.4 0.3 Trade transp. & accom. 1.2 1.1 1.4 1.2 1.1 1.1 1.6 0.7 1.0 1.1 ICT 0.2 0.0 0.2 0.3 0.1 0.2 0.2 0.1 0.0 0.1 Financial ins. & real estate 0.8 0.5 0.8 0.9 0.7 0.6 1.0 0.9 0.7 0.4 Public administration 0.6 0.7 0.8 0.4 0.6 0.5 0.7 0.7 0.6 0.4 Other services 0.1 0.1 0.0 0.1 0.1 0.1 0.1 0.1 0.1 0.1 GDP GROWTH 6.3 5.6 7.0 5.9 6.7 6.8 6.7 6.2 6.9 6.8 Source: National Accounts, 2019. On the demand side, economic growth is primarily driven by private consumption and investment; public consumption is modest. Over the last two decades, private consumption because of high goods imports, but the recent shift has and investment each contributed 3.1 pp to growth; been characterized by a lower value of imports and a now, services exports are also supporting growth. higher value of exports. Between 2013 and 2017, exports Between 2013 and 2017, private consumption contributed of services grew at an annual compounded rate of 2.4 pp to economic growth, accounting for 35 percent 6.8 percent while imports of services fell by 5.3 percent. of total GDP growth, and investment contributed 3.6 pp The resulting progressive reduction of the trade deficit to GDP growth, accounting for 52 percent (Figure 1.4). drove up economic growth. For instance, the reduction of Recently, the export-import profile has shifted. Tanzania the trade imbalance contributed 2.1 pp to GDP growth in continues to run a considerable trade deficit, mainly 2016 and 0.8 pp in 2017. 4 TA N Z A N I A M A I N L A N D P O V E R T Y A S S E S S M E N T FIGURE 1.4: Demand-side Contributions to GDP Growth, 1998–2017, Percentage Points 15 10 3.8 0.7 3.2 0.4 7.2 0.6 1.6 0.8 0.8 2.1 5 3.1 1.1 2.9 3.1 3.6 1.3 8.8 7.8 3.3 1.3 4.5 4.5 5.5 2.2 4.6 3.1 3.1 2.5 2.9 2.9 2.4 0.5 2.8 2.2 1.6 0.3 1.3 1.2 1.4 1.4 0 0.7 0.0 0.5 0.4 0.6 0.9 0.3 0.5 0.4 –1.1 –1.0 0.1 –0.3 –2.3 –0.3 –2.1 –0.6 –2.1 –3.5 –4.4 –5 1998– 1998– 2003– 2008– 2013– 2008 2009 2010 2011 2012 2013 2014 2015 2016 2017 2017 2002 2007 2012 2017 5-year period Detail Public consumption Private consumption Investment Net exports GDP growth Source: National Accounts, 2019. Capital accumulation is the main driver of growth, followed by expansion of the labor force. Capital accumulation accounts for most of it contributed on average 4.4 pp a year to economic Tanzania’s real economic growth, which relates to growth, 67 percent of the total; and for 2008–12 its the high investment rate previously noted. The rate ­ contribution averaged 4.5 pp, which constitutes 75 percent of capital accumulation has been steady. In 2013–17 of total growth (Figure 1.5).2 2 The results are based on the decomposition of economic growth using the Solow growth accounting model. C h a p t e r 1 S TR U C T U RA L TRA N S F O R M AT I O N 5 Additional workers also helped fuel growth, although by average years of schooling weighted by the return on total factor productivity (TFP) is still low. In 2016, labor education—has also been contributing to growth, indicating contributed 2 pp to economic growth and in 2017, this was constant progressive, though very slow, improvement in the 2.1 pp. Over 2013-17, added units of labor—mostly new quality of the workforce over the last 20 years. Conversely, young workers entering the labor market—contributed TFP is still low, contributing negatively in 2017 and averaging 1.6 pp, 25 percent of total growth. Human capital—proxied only 0.1 pp for 2013–17. FIGURE 1.5: Growth Accounting Decomposition, 1998–2017, Percent and Percentage Points 8 0.5 0.0 0.1 0.4 0.3 0.6 0.4 0.4 0.4 1.9 1.2 0.5 6 2.1 1.2 0.4 1.1 1.0 2.1 1.5 1.8 0.6 0.5 0.5 2.0 0.5 1.6 0.3 0.5 1.9 0.6 0.5 0.4 0.5 0.7 0.4 0.6 0.5 4 5.0 5.2 5.0 4.6 4.6 4.5 4.4 4.7 4.4 4.5 4.8 2 4.3 4.3 4.2 4.1 0 –0.4 –0.4 –0.8 –0.4 –0.4 –0.2 –1.1 –1.3 –2 1998– 1998– 2003– 2008– 2013– 2008 2009 2010 2011 2012 2013 2014 2015 2016 2017 2017 2002 2007 2012 2017 5-year period Detail TFP Capital Labor Human Capital per Labor Real GDP Growth Source: National Accounts, 2019, and World Bank staff calculations. 6 TA N Z A N I A M A I N L A N D P O V E R T Y A S S E S S M E N T II. Labor Market Transformation Microeconomic structural transformation was significantly slower than macroeconomic transformation. The employment pattern is slowly transitioning from industry by 5.5 pp. Thus, more households members seem to agriculture to services and industry. Integrated Labor Force have moved out of agriculture than households heads, prob- Surveys (ILFS) show that between 2006 and 2014, employment ably due to the increase of education among the younger in agriculture fell from 76 to 67 percent. Labor appears to be generation. National Panel Surveys (NPS) of 2011 to 2015 slowly shifting to services, where employment went up by corroborate the shift of labor away from agriculture, but the 9 pp, and to a lesser extent industry, up by 1.2 pp (Figure 1.6). decline was more modest, from about 72.7 to 70.8 percent. Within services, employment rose most in the restaurants and Industry gained about 1.4 pp and services about 0.5 pp, from accommodation subsector, followed by transport and com- 22.5 to 23 percent (Figure 1.7). The proportion of household munication, and wholesale and retail trade. In industry, mining heads working in agriculture fell by about 4 pp as the share in and quarrying followed by construction are the subsectors services rose by 2.4 pp and in industry by 1.6 pp. In contrast where jobs increased the most. However, they still account with HBS, NPS data suggest that household heads are leading for a very small share of total jobs. Moreover, the decline in the way out of agriculture. agricultural jobs is significantly lower than the decline in agri- culture’s contribution to GDP. Yet employment in the sectors driving economic growth is still very low. The 2018 HBS shows that the industry share of The shift away from agriculture also appears to be slow labor is 6.9 percent and services 34.8 percent. For household when using household surveys. Information about the heads, an estimated 54.3 percent are employed in agriculture, primary jobs of individuals gleaned from the Household 9.1 percent in industry, and 36.5 percent in services. A similar Budget Surveys (HBS) of 2012 and 2018 indicate a decline of pattern emerges from the 2015 NPS, though the labor share employment in agriculture from 75 to 58 percent and increase in agriculture is higher. Employment in industry is estimated of employment in services and industry by, respectively, 12 pp at 6.2 percent, and in services at about 23 percent. Household and 4 pp. The share of household heads employed in agricul- head shares are 61.2 percent in agriculture, 8.5 percent in ture fell by 9.6 pp as the share in services rose by 4.1 pp and in industry, and 30.3 percent in services.3 The transition away from agriculture seems faster when based on hours worked by sector. Accounting for total hours worked in each sector, agri- no matter what the job and that they only have one job. culture consumes less than half of Tanzanian work hours. Based on number of hours per job, the labor share of Measures based on hours worked in primary and in second- agriculture is a much lower at 46.3 percent, followed ary jobs provide a more accurate picture by relaxing the by services at 44.4 percent and industry at 9.3 percent assumptions that workers put in the same number of hours (Figure 1.7A). 3 Data from the 2018 HBS cannot be used in analyzing structural transformation because there is no way to look at productivity by sector and changing pattern of employment over time; for these analyses NPS is more appropriate. C h a p t e r 1 S TR U C T U RA L TRA N S F O R M AT I O N 7 FIGURE 1.6: Labor Shares by Sector, 2006 and 2014, 9 percent in industry, and 28 percent in services (Figure 1.8). Percent The new survey design (see Appendix A) changed the pattern drastically for 2015, during which 24.3 percent of total hours 100 worked in secondary jobs were in agriculture, 9.4 percent in 76.4 66.9 industry, and 66.4 percent in services (Figure I.8). 50 Hours-based measures of structural change indicate faster 26.8 19.3 movement out of agriculture even net of shifts in the share 4.3 6.3 of secondary jobs, which were largely driven by changes 0 Agriculture Industry Services in questionnaire design. Restricting the analysis to primary 2006 2014 jobs to insulate the results from the questionnaire changes, between 2011 and 2015 jobs in agriculture declined from 58.1 Source: ILFS 2006 and 2014. to 49.1 percent, with most going to services (Figure I.8). This confirms the idea that labor is not used intensively in agricul- 2006 2014 VARIATION ture, and workers there are likely to be underemployed. # OF JOBS % # OF JOBS % (%) Agriculture/ hunting/ forestry and fishing 12 719 757 76.4 13 400 163 66.9 5.4 Mining & quarrying 83 136 0.5 220 332 1.1 165.0 Manufacturing 432 305 2.6 620 934 3.1 43.6 Construction 182 898 1.1 420 633 2.1 130.0 Electricity, gas & water 16 627 0.1 Wholesale & retail trade 1 263 662 7.6 2 543 828 12.7 101.3 Transport/storage & communication 249 407 1.5 520 784 2.6 108.8 Hotels & restaurants 332 543 2.0 781 175 3.9 134.9 Real estate/renting & business activities 99 763 0.6 Public admin & defense 182 898 1.1 120 181 0.6 –34.3 Education 232 780 1.4 420 633 2.1 80.7 Health & social service 99 763 0.6 160 241 0.8 60.6 Other community/social & personal service activities 748 221 4.5 821 236 4.1 9.8 Total 16 643 760 100 20 030 139 100 20.3 Estimates based on hours worked in two jobs suggest a faster transition out of agriculture. Between 2011 and 2015, hours worked in agriculture declined from 58.3 to 46.3 percent (–12 pp) and rose in industry (+1.7 pp) and services (+10.3 pp); and services accounted for 44.4 percent of total labor input (Figures 1.7B and C).4 The transition was fastest between 2011 and 2013 and largely ascribable to secondary jobs, which changed far more than primary activities; however, it may also be attributable to a statistical artefact associated with changes in survey instruments.5 Services have been the main beneficiary of the exodus from agriculture: by 2015 service jobs had expanded by 28 pp and accounted for 36.4 percent of second jobs. Such changes are mirrored in the sectoral distribution of working hours. In 2011, about 63 percent of hours worked in secondary jobs were in agriculture, less than 4 Measures based on hours worked in all jobs, with missing hours imputed, confirm movements of labor from agriculture to services. Estimated sectoral shares based on imputed working hours are smaller for agriculture and larger for services because more services values are missing. For example, in 2015, the agricultural labor share drops from 46 to 36 percent when imputed rather than reported hours are used, and labor supplied to services rises from 44.4 to 54.9 percent (see Appendix A). 5 First, the number of workers with a second job doubled from 8.7 to 16.3 percent between 2011 and 2013, when the questionnaire was changed. Between 2013 and 2015, when the two questionnaires were similar, that percentage went up by less than 2 pp, from 16.3 to 18.1 percent. Second, the participation-based share of workers with a second job in agriculture declined from 51.1 percent in 2011 to 39.7 percent in 2015, bouncing back from a low of 36 percent in 2013. See Appendix A for more details. 8 TA N Z A N I A M A I N L A N D P O V E R T Y A S S E S S M E N T FIGURE 1.7: Sectoral Labor Distribution, 2011–2015 A. Sectoral Employment Distribution, Percent 100 23 21 23 28 28 30 34 80 43 6 45 44 45 5 6 55 7 9 9 60 8 9 9 11 11 40 9 73 73 71 65 63 61 58 44 46 48 44 20 36 0 2011 2013 2015 2011 2013 2015 2011 2013 2015 2011 2013 2015 Participation Participation - Houshold Working hours Working hours imputed head Agriculture Industry Services B. Cumulative Changes in Distribution of Labor, PP 15 11.8 10.3 10 Percentage points 5 1.6 2.4 1.7 1.4 0.5 0 –0.1 –1.9 –5 –4.0 –10 –12.0 –11.7 –15 2011-2015 2011-2015 2011-2015 2011-2015 Participation Participation Working hours Working hours Household head imputed Agriculture Industry Services Continue on next page C h a p t e r 1 S TR U C T U RA L TRA N S F O R M AT I O N 9 FIGURE 1.7C. Detailed Sectoral Employment Distribution, Percent 100 10 11 12 14 16 17 19 21 21 23 23 21 12 10 11 80 2 14 12 3 1 3 3 13 1 1 2 13 1 2 3 4 4 5 23 20 23 2 22 2 2 2 36 2 4 60 1 3 2 3 4 5 6 4 6 2 2 2 2 3 40 4 73 73 71 1 65 63 61 58 44 46 48 44 20 36 0 2011 2013 2015 2011 2013 2015 2011 2013 2015 2011 2013 2015 Participation Participation - Houshold Head Working hours Working hours imputed Agriculture Mining Manufacturing Other industry Trade Other services Source: National Panel Surveys (NPS) 2010–11, 2012–13, and 2014–15. FIGURE 1.8: Working Hours, Primary and Second Jobs, 2011–15, Percent 100 28 34 34 80 44 42 42 45 9 64 66 8 8 60 11 9 11 9 40 63 12 58 58 9 46 47 49 44 20 24 24 0 2011 2013 2015 2011 2013 2015 2011 2013 2015 Working hours Working hours Working hours All jobs Primary jobs Secondary jobs Agriculture Industry Services Source: NPS 2010-11, 2012-13 and 2014-15. 10 TA N Z A N I A M A I N L A N D P O V E R T Y A S S E S S M E N T Even when agriculture is the main sector of employment, workers supply a considerable amount of time to nonfarm sectors. In 2015 workers categorized as agricultural spent about as agricultural workers. McCullough (2017) confirmed that 26 percent of their time in agriculture, 32 percent in average hours worked by those whose primary job is non- industry, and 42 percent in services (Figure 1.9). In con- agricultural is higher than the hours worked by agricultural trast, workers employed primarily in industry spent about workers.6 She found that differences ranged from 2.3–2.5 in 52 percent of their time in industry, 35 percent in services, Malawi to 2.4–2.6 in Ethiopia and 2.1–2.2 in Tanzania. Our and a meagre 13 percent in agriculture. Similarly, workers analysis reveals that, in 2015, nonfarm-to-agriculture annual employed primarily in services gave it about 72 percent of working-hour ratios are estimated at 2.2 for industry-to-­ ­ their time. agriculture and at 2.6 for services-to-agriculture. Restricting the sample to primary jobs does not substantially affect these On average, workers allocate far more time to nonagri- estimates: 2.3 for the former and 2.8 for the latter. Thus, cultural sectors than to agriculture. In 2015, agricultural although agriculture is the main employer, agricultural work- workers worked about 650 hours annually, industrial work- ers work far fewer hours than workers in other sectors and, in ers 1,600, and services workers 2,400 (Figure I.10). Gollin, parallel, supply a considerable amount of time to nonagricul- Lagakos and Waugh (2014) estimated that nonagricultural tural sectors. workers gave their sectors about 1.3 times as many hours FIGURE 1.9: Annual Hours Supplied by Workers to Different Sectors, 2015        A. Average (number of hour)                  B. Annual Distribution, Percent 3,500 100 3,000 34.8 80 41.6 2,500 1066 71.8 1093 60 2,000 1902 1,500 32.5 40 51.8 1586 855 1,000 20 13.9 368 500 682 25.9 411 380 14.3 13.4 0 Agriculture Industry Services 0 Agriculture Industry Services Primary sector of employment Primary sector of employment Agriculture Industry Services Agriculture Industry Services Source: NPS 2010–11, 2012–13 and 2014–15. 6 McCullough used micro-based productivity measures from LSMS-ISA datasets from four Sub-Saharan African countries to investigate the productivity gaps that households face, and to assess key structural change parameters—sector participation, time use, and labor productivity—from a micro perspective. Results mentioned above are for 2010-11. C h a p t e r 1 S TR U C T U RA L TRA N S F O R M AT I O N 11 FIGURE 1.10: Average Number of Hours Worked per Job Sector, 2015 2428 2,500 2,000 1618 1,500 1,000 656 500 0 Agriculture Industry Services Source: NPS 2010–11, 2012–13 and 2014–15. 12 TA N Z A N I A M A I N L A N D P O V E R T Y A S S E S S M E N T III. Productivity Gaps Sizable productivity gaps between agriculture and services suggest potential for enormous productivity gains from accelerating the transition of labor away from agriculture. The enormous potential for productivity growth through productivity gaps are very wide (Figure 1.12A). Productivity structural change has not yet been realized due to the is estimated to be about 5.6 times higher in industry than in slow transition of labor. Large productivity gains could be agriculture, and about 9.5 times higher in services. achieved by shifting labor across sectors because the pro- ductivity gaps between sectors are significant, especially Adjusting productivity gaps for hours worked confirms between agriculture and manufacturing or services (McMillan that there are considerable differences in productivity and Headey 2014). Given the persistent concentration of between sectors, although narrower than those based Tanzania’s labor in agriculture despite the widespread poverty on productivity per worker. Adjusting productivity gaps and underemployment among farmers, the question is why for hours worked in all sectors is expected to produce more agricultural workers are not moving away faster. One possible accurate measures (Box 1.2). In 2015, productivity per hour explanation is that farmers, and rural populations generally, worked in agriculture was estimated at TZS 463 ($0.6 in PPP) are confronted by barriers to entering more productive and compared with TZS 1,131 ($1.40 PPP) in industry and TZS remunerative sectors and activities—mainly too little human 1,494 ($1.90 PPP) in services (Figure 1.11B).8 It is not surprising capital, experience, or physical or financial assets. It may also that these results are smaller than per-worker results due to be that productivity differentials between agriculture and violation of the one-sector and equal-hours-across-sector nonfarm sectors are not as high as the literature suggests. assumptions. Therefore, one hour worked in industry is And even given the productivity gap, higher risks and uncer- estimated to be on average 2.4 times as productive as one tainties may offset differences in expected returns from more hour worked in agriculture, and one hour worked in services is productive sectors. This section examines the second pos- about 3.2 times as productive (Figure 1.12B).9 sibility, productivity differences between sectors using labor In a more granular disaggregation, commerce, mining, productivity measures from the NPS 2011 to 2015 (Box 1.1). and transport come out to be the most productive sec- Two measures of labor productivity, per-worker and per hour, tors. Using productivity gaps based on participation (total are estimated for different types of activities in agriculture, output per worker), mining is the most productive industrial industry, and services and the main subsectors (for details, see subsector no matter what it is compared to (Figure 1.13). On Appendix A). average, each mining worker produces 8 times as much as Labor productivity is markedly higher in industry and ser- each worker in agriculture; each worker in manufacturing pro- vices than in agriculture. In 2015, annual output per worker duces 4.8 times as much, and each worker in construction or was estimated at TZS 1.9 million (about $2,439 in Purchasing utilities produces 5.4 times as much as an agricultural worker. Power Parity -PPP) in industry and TZS 3.2 million ($4,123 PPP) Services are even more productive relative to agriculture: in services (Figure I.11A).7 The estimate for agriculture is a average productivity in commerce is about 8.6 times higher much lower TZS 337,500 (about $433 PPP). Thus, per-worker and in other services, such as transport and financial services, 7 Output per-worker and per-hour-worked is measured in Tanzanian shillings (TZS) and expressed in 2015 prices. 8 Estimates of productivity per hour worked using imputed hours indicate roughly similar productivity in agriculture whereas after imputation, productivity in industry and services declined due to a larger number of hours. Precisely, hourly productivity is estimated at TZS 890 in industry and TZS 929 in services (Appendix A). 9 When imputed instead of working hours are used, productivity gaps narrow from 2.4 to 2 for industry and from 3.2 to 2.1 for services (See Appendix A). C h a p t e r 1 S TR U C T U RA L TRA N S F O R M AT I O N 13 it is 10.5 times higher. When productivity is measured in terms Industry and services are more productive than farming of output per hour worked, the gaps between subsectors regardless of whether the labor is supplied by nonagri- are smaller and less heterogenous. For example, one hour of cultural household enterprises or wage jobs. However, the mining is 3.1 times more productive than one hour of farming. magnitude of the gap with agriculture differs (Figure 1.14). After farming, manufacturing is the least productive—only The gap between wage employment in industry and farming 1.7 times more productive than agriculture. In services, one is estimated at 4 times and between services and farming hour of work in commerce is 3 times and in other services 10 times. For both sectors, when labor is in nonfarm house- 3.5 times more productive than one hour in agriculture. hold enterprises, the output-per-worker gap is about 7 times. BOX 1.1 Measuring Labor Productivity Structural transformation is a dynamic process, in terms labor is misallocated, why workers are gradually moving out of both reallocation of labor from less to more productive of agriculture, and what is likely to be driving the modest sectors, and the economic growth resulting from the move. movements of labor between sectors. Among factors driving the process are productivity levels within sectors and the productivity gaps between them. In this analysis labor productivity will be considered The wider the productivity gap between sectors, especially per worker and per hour and estimated for the three between agriculture and manufacturing or services, the main sectors of the economy—agriculture, industry, and greater the opportunity to achieve larger productivity gains services—and by subsector and economic activity. Labor as labor shifts between sectors (see Appendix A). The pres- information will cover both primary and secondary jobs. This ent analysis focuses on shifts of workers between sectors is very important in countries like Tanzania where a consid- as measured by changes in the sectoral share of labor. The erable share of workers have more than one job. Sectoral analysis examines the existence and extent of such labor labor share based on household survey data will be con- productivity gaps from a microeconomic perspective. trasted with a common value-added measure of structural Estimates of productivity gaps help illuminate whether transformation based on national accounts. FIGURE 1.11: Productivity per Worker and per Hour, by Sector, 2015 (TZS) A. Value of Output Per Worker                   B. Value of Output Per Hour Worked 3,500,000 1,500 3,000,000 1,250 2,500,000 1,000 2,000,000 750 1,500,000 500 1,000,000 500,000 250 0 0 Source: NPS 2010-11, 2012-13 and 2014-15. Note: Per-worker productivity gaps are calculated as ratios between productivity in industry/services and productivity in agriculture. Ratio for agriculture is therefore 1. 14 TA N Z A N I A M A I N L A N D P O V E R T Y A S S E S S M E N T FIGURE 1.12: Productivity Gaps by Sector, 2015 A. Per Worker Productivity                     B. Per Hour Worked Productivity 10 9.5 4 9.2 3.2 8 3 2.4 6 5.6 2 4 1.0 1 2 1.0 0 Agriculture Industry 0 Services Non Agriculture sector (VA) Agriculture Industry Services Source: NPS 2010-11, 2012-13 and 2014-15. Note: Per-worker productivity gaps are calculated as ratios between productivity in industry/services and productivity in agriculture. Ratio for agriculture is therefore 1. BOX 1.2 Biases of Productivity Measures in Tanzania Labor productivity measures based on the primary sector agriculture and hours spent in industry and services differ of employment assume that workers do not supply labor to substantially. Furthermore, some workers have jobs in other sectors, and all systematically work the same number more than one sector, especially agricultural workers, who of hours whatever their primary sector may be. It is there- provide a considerable amount of labor to other sectors. For fore important to characterize patterns of labor supply in Tanzania, then, productivity measures based on the primary terms of annual working hours. job are biased and likely to overestimate sectoral differ- ences. Figure 1.7 makes clear the magnitude of the bias by For instance, if more labor is supplied to nonagricultural sec- drawing in parallel the sectoral distribution based on annual tors by agricultural workers than is supplied to agriculture hours supplied by workers to primary and secondary jobs by nonagricultural workers, using per-worker productivity and the sectoral distribution based on participation in pri- measures based only on a worker’s primary activity may mary jobs. In all NPS rounds, the results of the two measures be biased. They are likely to overestimate the amount labor differ, with agriculture’s labor share being lower in both supplied to agriculture, and underestimate productivity rounds. Thus, agricultural productivity might be underesti- in agriculture relative to other sectors. This is in fact true mated relative to other sectors when participation-based for Tanzania. Using data from the NPS, our analysis clearly measures are used. In contrast, using hours-based measures rejects the cross-sector equal-hours assumption. We find are likely to provide higher estimates of agricultural produc- that on average, workers in nonagricultural sectors put in tivity and in turn smaller productivity gaps. This is extremely considerably more hours than agricultural workers. important for understanding structural change at the macro level on one hand and push and pull factors that affect This evidence is a warning against adopting either the household and individual decisions about where to allocate one-sector or the cross-sector equal-hours assump- their labor on the other. tion because the microdata rejects both. Hours spent in C h a p t e r 1 S TR U C T U RA L TRA N S F O R M AT I O N 15 FIGURE 1.13: Productivity Gaps by Economic Sector, 2015, Percent 12 10.5 10 8.6 7.9 8 6 5.4 4.8 3.1 3.5 4 2.9 3.0 1.7 2 1.0 1.0 0 Based on output per worker Based on output per hour worked Agriculture Mining Manufacturing Other secondary (Utilities and Construction) Commerce Other services (Transport, Finance and Other) Source: NPS 2010-11, 2012-13 and 2014-15. Note: The sector of reference is agriculture (ratio = 1). FIGURE 1.14: Productivity Gaps per Worker by Activity, 2015 12 10.0 10 8 7.3 7.3 6 4.1 4 2 1.0 0.9 0 Household Agriculture Industry Services Industry Services Farm Wage Household non-farm business business Source: NPS 2010-11, 2012-13 and 2014-15. Note: The activity of reference is household farming (ratio = 1). In the case of both household and nonfarm businesses, hired labor is added to the labor provided by household members. 16 TA N Z A N I A M A I N L A N D P O V E R T Y A S S E S S M E N T IV.  Occupation Decision Based on the 2014 ILFS, the industrial sector has the Over 90 percent of employees in transportation and s­ torage highest median wages and median productivity – defined are men while 82 percent of workers in accommodation as income per hour worked - of all three sectors in the and food are women. Thus, gender norms which dictate the economy. Median incomes in the industrial sector are around proper social role for women may cause barriers for women 173,000 TZS per month, which is slightly higher than what is entering into higher-productivity jobs. found in services (see Appendix A for the methodology and results). Incomes in the industrial sector also appear to vary As expected, individuals in rural areas are less likely to access less than in other sectors. The lowest incomes are found in jobs in the industrial and services sectors compared to urban the agricultural sector where the mean monthly income is ones. Only twenty five percent of employees in industry and just under 67,000 TZS (Table A.3 in Appendix A). This is a 18 percent of those in services lived in rural areas. Furthermore, quarter of the mean income in the industrial sector and a fifth the industrial and services jobs available in the rural areas tend to of the mean income in the service sector. Productivity is also be much lower paying than in urban areas. Difference in means the lowest in the agricultural sector. On average, individuals tests of total monthly income by locality across the two sectors working in agriculture make only 510 TZS per hour worked show that monthly mean incomes in the industrial and in the while individuals working in services make nearly three times services sectors are, respectively, around 90,000 TZS and 113,000 as much on overage (Table A.4 in Appendix A). TZS lower in rural areas than in urban areas. Individuals choosing to work in the industrial sector are Individuals choosing to work in agriculture appear to have more likely to have higher education levels than individuals less access to credit than individuals working in other in the agricultural and services sectors. This may indicate sectors. Only 1 percent of individuals working in agriculture that access to higher and more stable wages in the sector are obtained credit over the 12 months preceding the survey dependent on technical skills and knowledge gained through period compared to 3 percent in industry and 6 percent in continued education. In fact, a difference in means test of edu- services, which in itself is quite low. The dominant sources of cational attainment on sector indicates that individuals working credit are savings cooperatives, banks, and financial institu- in industry are significantly more likely to have completed tions. However, use of these services appears to be extremely secondary and tertiary education than all other individuals. This low among agricultural workers. This may be due to problems points to the lack of education being an important barrier to of proximity (i.e., banks are far away from where the farmers entering the industrial sector (Tables A.5 and A.6 in Appendix A). live), but it may also point to the existence of other institu- tional barriers which prevent farmers from accessing formal Age has a statistically significant, but minor effect on an lines of credit. It may also be telling that several individuals individual’s probability of working in any sector. While the operating in the private non-agricultural sector were able to average age of individuals working in agriculture is slightly access credit through relatives and private money lenders. lower than the average age of individuals working in the Thus, there may be a self-reinforcing relationship between industrial and service sectors, it does not appear that older access to credit and household or family wealth. individuals are more likely to choose employment in those sectors. In fact, the negative coefficient produced in the esti- There is evidence that lower household dependency mation model indicates that older individuals are less likely to ratios are associated with higher likelihoods of choosing opt into employment in industry and services. employment in the industrial and services sectors. High Being female is associated with a slightly higher like- dependency ratios are typical of rural and poor households lihood of choosing to work in the services sector than who are predominantly employed in agriculture. Thus, this agricultural. Around 11 percent of women are employed in result is not surprising. Large number of dependents places wholesale and retail jobs and another 6 percent are engaged a larger strain on those individuals who are in the labor force in employment related to accommodation and food. and may prevent risk taking and investing in transitioning to However, these services sector jobs are highly gendered. higher-productivity sectors. C h a p t e r 1 S TR U C T U RA L TRA N S F O R M AT I O N 17 CHAPTER 2 Firm Profiles and Structural Transformation I. Profile of Tanzanian Firms Tanzania’s industrial landscape is primarily occupied by micro and small-scale businesses. Two-thirds of businesses in Tanzania are in manufacturing (30 percent, mostly in apparel); and furniture (14 percent).1 or trade. The rest are essentially in services, mainly non- Very few firms are in high-value-added and knowledge- market (Figure 2.1A). Less than 1 percent of firms are intensive industries. Only 1 percent manufacture computers, in agriculture, as owners usually operate farms without machinery, or electrical and transport equipment. establishing an enterprise. The number of businesses in mining and quarrying is very small, about 550. Manufacturing Micro- or small-scale enterprises account for almost all firms are primarily in food and beverage processing businesses. Of the 154,618 businesses, 96 percent have fewer (39 percent); textiles, wearing apparel, and leather than 10 workers, and about 60 percent of these have only FIGURE 2.1: Firm Distribution by Sector and Size, Percent A. Distribution by Sector of Activity 20.1 10.4 1.3 0.1 9.5 3.7 34.9 0.4 Manufacturing 4.8 13.7 0.6 34.2 0.7 0.7 Agriculture and mining Food, beverage and tobacco Utilities and construction Textiles, wearing apparel and leather Trade (wholesale and retail Wood and wood products Market services Petroleum, chemical and pharmaceuticals Non-market services Plastic, mineral and metals Equipment Furniture Other manufacturing 1 Fig. 2.1C covers manufacturing firms, 2.1B all businesses. 20 TA N Z A N I A M A I N L A N D P O V E R T Y A S S E S S M E N T FIGURE 2.1B. Distribution by Sector and Size Manufacturing       FIGURE 2.1C. Firm Distribution by Subsector and Size 100 5 1 3 0 3 1 4 0 100 0 0 1 2 0 2 10 5 0 6 6 0 1 13 6 0 6 0 0 13 1 2 16 19 17 19 23 24 32 80 80 26 21 19 60 38 48 60 43 30 89 94 93 94 40 77 40 82 79 75 64 66 70 20 43 37 28 20 38 0 0 . t. es es g e a ct ns i ad in p. an . . vic vic fa v. re m m f. r d in co ui he be Tr u u nz tu oo ar ar er er m an an Eq at i nd ph ph Ta rn ts ts d W d rm M le an an Fu sa ke ke d nd d e an od ar ar . tie an th ric .a M m l. Fo O ili Ag t. tro in Ut on x m Te Pe N ., Pl Micro Small Medium Large Micro Small Medium Large Source: Tanzania Statistical Business Register (SBR) 2014/15. Notes: - The SBR covers 154,618 establishments operating in all sectors (NBS 2016). - Market services include transportation, accommodation & food, information & communication, etc. - Non-market services include scientific & technical activities, health, and education & social work activities, and public administration. - Equipment includes manufacturing of computer, electrical, machinery, and transport equipment. - Sectors and subsectors are grouped based on ISIC Rev.4 code. - Micro firms employ 1–4 workers, small firms: 5–19, medium firms: 20–99, large firms: 100+. one or two workers; only 1 percent of firms have more than highest in manufacturing, where about 61 percent of firms are 50 workers (Figure 2.1B). Large businesses account for not registered; informality is predominant in textile, apparel, 0.5 percent of all firms but are 13 percent of chemical (mainly and leather (82 percent); furniture (76 percent); plastics, soap and detergents) and pharmaceutical manufacturing. minerals, and metal (62 percent); computer and machinery Large firms tend to operate in food processing, even though (55 percent); and wood processing (50 percent). In trade and micro-scale firms dominate the sector.2 market services, about half the firms are not registered, but in nonmarket services only 14 percent are not. Informality is Most Tanzanian firms are privately owned: Although negatively correlated with firm size. While 56 percent of micro 13 percent are state-owned, 78 percent have sole pro- firms are not registered, about 95 percent of medium and prietors and the rest are owned by other companies, large firms are (Figure 2.2D). religious institutions, cooperatives, and NGOs. Less than 1 percent are publicly traded. Over 80 percent of the busi- Tanzanian firms are young, with an average age of nesses are owned by Tanzanians and a clear majority have 8.5 years and a median age of 4; half of the firms have no branches. Of privately-owned businesses, 90 percent are been in existence less than 5 years. The proportion of micro sized and operate in manufacturing and trade. State- young firms decreases as firm size increases (Figure II.2A). owned companies average 15 employees, and many have Thus, most micro enterprises are young and most large more; they are concentrated in social services (education enterprises older. About 22 percent of firms have been in and health) and administration. operation for less than a year, over 90 percent are micro enterprises, and most are not registered, which suggests Informality is prevalent. Nearly half of Tanzanian firms that very small-scale informal activities are common. Within (48.5 percent) are not registered by any government. The sectors, trade and then manufacturing have the largest prevalence of informality varies by sector and firm size. It is shares of very young firms; nonmarket services have the 2 About half the large enterprises process foods and beverages, but there are only 90 of these; meanwhile 20,000 micro-firms operate in the sector. C h a p t e r 2 F i r m P r ofil e s a n d S t r uc t u r a l T r a nsfo r m at ion 21 FIGURE 2.2: Firm Distribution by Age, Size, Sector, and Informality, Percent.        A. By size and age             B. By sector and age 100 6.8 100 5.7 3.2 7.7 13.1 20.8 14.4 12.7 14.0 33.1 16.6 80 40.5 15.9 80 20.2 48.1 18.5 45.2 20.6 25.6 27.6 25.0 60 21.2 20.6 60 22.8 26.1 24.7 17.4 40 30.9 40 23.7 20.9 65.6 58.7 50.4 52.8 15.3 20 20.9 20 49.2 12.3 28.0 35.3 24.8 22.1 12.5 8.7 0 0 Micro Small Medium Large Tanzania Agric. and Manufact. Utilities Trade Market Non Mining and Services Market Less than 5 years 5–9 years Const. Services 10–19 years 20 and more Less than 5 years 5–9 years 10–19 years 20 and more       C. By age and informality          D. By size and informality 100 100 4.5 5.7 15.4 19.6 80 45.0 37.7 80 48.5 61.9 57.8 60 60 95.6 94.3 40 84.6 40 80.4 55.0 62.3 51.5 20 38.1 20 42.2 0 0 Less than 5–9 years 10–19 years 20 and Total Miro Small Medium Large 5 years more Formal Informal Formal Informal Source: SBR 2014/15. largest share of older firms, indicating that state-owned it is for firms that start out self-capitalized in micro-scale enterprises and firms operating in public sectors have activities to survive and grow, and to contribute to societal higher chances of survival and growth than private ones welfare and economic prosperity. (Figure II.2B). Over half of new entrants (57 percent) are in wholesale and retail trade, 22 percent in manufacturing, and 10 percent in market services. Informality is also per- vasive among young firms (Figure 2.2C). Most businesses have little startup capital; fewer than 10 percent had initial capital investment of TZS 50 million or more. The clear majority of capital investment, particularly among micro and young firms, is from personal income. Only 20 percent of businesses secured funding from banks or microfi- nance institutions, with micro firms relying more on Credit Co-Operative Society (SACCOS) (3.3 percent) and other microfinance institutions (2.6 percent); larger firms rely more on bank financing (about 25 percent). The predominance of micro, young, and informal firms demonstrates how hard 22 TA N Z A N I A M A I N L A N D P O V E R T Y A S S E S S M E N T Employment is disproportionately concentrated in older and larger firms and is acutely biased against women. Despite the predominance of micro firms in the economy, slightly higher in small and young firms. In these firms, the they account for just 24 percent of employment, and average number of men is two times higher than that of women, large firms, which represent less than 1 percent of total while it is 1.6 times higher in larger and older firms. firms, account for 28 percent. About half of Tanzanian jobs (49 percent) are in medium and large firms, which account Employment is skewed to nonmarket services and man- for about 4 percent of all businesses. Thus, the very few large ufacturing, followed by trade. Farm employment is not firms employ on average quite a large number of workers. reflected in these data because so few businesses operate in Even though fewer than 0.5 percent of all firms employ more agriculture. Nonmarket services, where most public enter- than 100 workers, they account for about 28 percent of all prises, which are larger and older, operate, account for about employment—the average number of their workers is over 370, 46 percent of total jobs (Figure 2.3D). Manufacturing and compared to no more than 2 in micro firms and 9 in small ones trade follow, because more firms operate in these sectors. (Figure 2.3A). New firms in business for less than 5 years account Mining and construction contribute less than 2 percent to for less than 25 percent of jobs, compared to 36 percent in total employment, but they tend to average a higher number firms in business for more than 20 years (Figure 2.3B). Formal of workers per firm than in the other sectors. These sectors, firms, despite their low share in the economy, account for over and to a lesser extent manufacturing, principally employ men, 80 percent of jobs (Figure 2.3C). Overall, firms tend to employ who constitute more than 75 percent of their workers. Only twice as many men as women, with the gender discrepancy market services tend to have slightly more women than men. Industrial firms employ more unskilled workers. Over 40 percent of industrial workers are unskilled or terms of size, employment, age, and structure of ownership unpaid. The 2013 Census of Industrial Production (CIP) to the distribution for all businesses in the 2014/15 Tanzania shows the distribution of industrial firms to be similar in Statistical Business Register (SBR). The CIP provides more FIGURE 2.3: Employment by Gender and Firms Characteristics          A. By firm size             B. By firm age 40 400 40 20 Contribution to total employement (%) Contribution to total employement (%) Average number of workers Average number of workers 30 300 30 15 20 200 20 10 10 100 10 5 0 0 0 0 Micro Small Medium Large Less than 5 5–9 years 10–19 years 20 and % of employment Av. # of workers years more Av. # of women Av. # of men % of employment Av. # of workers Av. # of women Av. # of men C h a p t e r 2 F i r m P r ofil e s a n d S t r uc t u r a l T r a nsfo r m at ion 23 FIGURE 2.3C. By informality status FIGURE 2.3D. By firm sector 100 10 60 30 Contribution to total employement (%) Contribution to total employement (%) Average number of workers Average number of workers 80 8 40 20 60 6 40 4 20 10 20 2 0 0 0 0 Formal Informal . t. e g es es ct ad ns in fa vic vic in co Tr u % of employment Av. # of workers er er m an s& ts ts nd M Av. # of women Av. # of men ke ke tie .a ar ar ili ric M m Ut Ag on N % of employment Av. # of workers Av. # of women Av. # of men Source: SBR 2014/15. details on workers and industrial establishments. It shows Manufacturers of foods and beverages, followed by those that about 28 percent of the workers are working propri- in wood processing, account for the largest shares of etors, managers, or professionals and 29 percent are skilled unskilled labor. The sectors dominating industrial production (Figure 2.4A). The rest are essentially unskilled or unpaid are food and beverage processing, 40 percent of firms, and workers. About 77 percent are men, who constitute a slightly manufacture of textiles and wearing apparel, 29 percent. The larger proportion of managerial and skilled professionals vast majority of these firms are young and micro-sized.3 Nearly than women. Only 1.5 percent of workers are non-Tanzanian, 50 percent of those making textiles and wearing apparel and and most of them are in professional and skilled positions, 28 percent of those in food and beverage processing are which suggests their employment is a way to compensate one-person firms, and about 50 percent of workers in the for the skills deficit. latter are unskilled or unpaid. This proportion holds even for medium and large firms in the sector. The lower share of Large and older firms tend to rely more on skilled labor unskilled workers in textiles and wearing apparel is due to than small and younger ones. The share of managerial and the predominance of self-employment. However, in the few professional positions may seem to be significantly higher firms—less than 1 percent—that have managed to survive in micro and very young firms (Figures 2.4A and 2.4B), but and grow in this sector, skilled workers now account for about that is because self-employment predominates in these 65 percent of the workers. The situation is similar among firms. About a third of these are one-person firms where furniture manufacturers. About 14 percent of firms operate the worker is the proprietor (i.e., self-employed). Although in this industry, of which about one-fourth are self-employed medium and larger firms, which are also older, have signifi- sole proprietors. Skilled workers account for over 50 percent cantly more skilled workers than smaller and younger ones, of those employed by the very few firms (0.4 percent) that more than 40 percent of their workers are unskilled. These have managed to become medium and large in this industry, firms tend also to hire more skilled males, often foreigners. compared to less than 20 percent employed by smaller firms. About 8 percent of workers are in metal manufacturing and 3 About 42 percent of firms in food processing and 30 percent of those in textiles and wearing apparel are one year old or less. 24 TA N Z A N I A M A I N L A N D P O V E R T Y A S S E S S M E N T FIGURE 2.4: Industrial Employment by Skills, Percent           A. By gender            B. By firm size 100 8.1 6.9 3.4 1.7 100 3.7 1.5 12.1 13.1 17.0 80 35.1 36.6 80 39.0 35.3 35.7 27.0 45.4 60 60 36.6 12.6 40 28.9 29.7 26.1 40 19.2 29.1 46.0 58.2 20 20 47.3 27.7 28.2 26.1 27.2 21.9 13.5 0 0 Total Men Women Foreign Micro Small Medium Large Managerial Skilled operatives Managerial Skilled operatives Unskilled operatives Unpaid and other Foreign Unskilled operatives Unpaid and other Foreign          C. By firm age           D. By firm activity 100 9.9 7.5 4.2 100 2 7 7 3 2 11.4 14 12 14 22 80 37.2 80 38 19 44 32.2 37.2 35.9 43 41 36 24 60 60 19.1 32 21 55 40 22.6 33.1 38.8 40 23 23 29 21 51 20 37.3 20 35 41 30.4 23.5 26 29 24 31 19.8 20 0 0 8 Less than 5–9 years 10–19 years 20 and more v. re s g g r al d tie he be in in itu oo et 5 years in ar ili ot M rn d W M Ut we an nd Fu Managerial Skilled operatives d od .a an ic Fo em Unskilled operatives Unpaid and other Foreign e til Ch x Te Managerial Skilled operatives Unskilled operatives Unpaid and other Foreign Source: Census of Industrial Production (CIP) 2013. Note: Chemical & other includes manufacture of pharmaceutical, chemical, paper, electronic, plastic, repair of machinery etc. they are grouped together due to the very limited number of firms operating in these sectors. The predominance of unskilled employment in this sector is due to the prevalence of unskilled workers in the plastic industry and manufacture of detergents. 4 percent in wood manufacturing. Wood manufacturing with more skills, which are lacking in Tanzania. The lack of firms are larger and older than those previously discussed. skills particularly affects mining, utilities, and more techni- However, the fact that more than 50 percent of their workers cally advanced industries like machinery, electric equipment, are unskilled, suggests that furniture processing is still very medical and pharmaceuticals, chemicals, and plastic. It basic. Mining and utilities, followed by manufacture of also severely affects large furniture manufacturers, nearly appliances and machinery, which account for significantly 40 percent of whom report that the shortage of qualified greater shares of large and older firms, rely more on skilled labor is a major challenge. employees. Mining has more foreign workers, 4.5 percent, than the national average of 1.5 percent. However, no more than 1 percent of industrial firms operate in mining or utilities. The shortage of qualified labor severely challenges the growth of large industries. About 13 percent of all industrial firms—and 33 percent of medium and large firms—consider it a major problem. This suggests that as firms grow, they engage in more sophisticated activities that require workers C h a p t e r 2 F i r m P r ofil e s a n d S t r uc t u r a l T r a nsfo r m at ion 25 The largest share of value added in industry is generated by large firms and in sectors where there are more of these firms. Large firms contribute the highest proportion of val- Manufacture of food and beverages, followed by min- ue-added in industry. Total value- added in industry in 2013 ing and carrying, contribute the most to Tanzanian was TZS 8,220,560 million, 84 percent of it contributed by value-added. Food and beverage processing is the largest large firms (Figure 2.5). Value-added per worker in large firms manufacturing sector and value-adding subsector of the is 2 times higher than in medium firms and about 20 times economy, at about 41 percent, due to both its large number higher than in micro-firms. It appears that firms that manage of firms and the high value-added generated by such indus- to grow over time become markedly more productive. Yet, tries as those making sugar, malt liquor, and soft drinks, where they account for only 0.5 percent of industrial firms, which may most firms are large or medium. However, the per-worker explain why industry in general is persistently unproductive. value-added is relatively low because there are also numerous FIGURE 2.5: Value Added by Firm Characteristics        A. By firm size (billion TZS)        B. By firm activity (billion TZS) 223 3% 160 2% 392 5% 320 4% 105 1% 731 9% 2920 35% 970 12% Mining Food and beverage 87 1% Textile and wearing Micro 247 3% Wood Small Chemical and other Medium Metal Large Furniture Utilities 3339 41% 6948 84%       C. Per worker by size (million TZS)        D. Per worker by activity (million TZS) 60 20 50.4 16.8 15 40 11.2 10 8.3 25.4 4.7 5 3.6 3.1 2.3 2.6 20 0 5.4 v. re s g 3.5 ng r al d 2.6 ie he be in itu oo et i t ar in ili ot M d rn W M we Ut 0 an d Fu an d od Micro Small Medium Large Tanzania an al Fo ic e em til x Te Ch Source: CIP 2013. 26 TA N Z A N I A M A I N L A N D P O V E R T Y A S S E S S M E N T FIGURE 2.6: Major Challenges Faced by Industrial Firms, Percent High cost of production Uncertain economic environment Inadequate technology Infant Private Sector with weak support Insuf cient demand Insuf cient production capacity Shortage of raw materials Environmental challenges Inadequate infrastructure Inadequate nancial services Unfair competition Taxes Shortage of quali ed labour Administrative procedures Foreign currency uctuations 0 20 40 60 80 100 All industrial rms Micro and small Medium and large Source: CIP 2013. micro-firms. In mining, where 40 percent of the firms are performance by both small and large firms. However, small medium or large, total and per-worker value-added are very and large firms differ in their perception of the importance high. Value-added is also quite high in manufacture of elec- of other problems. More than 30 percent of micro- and small tronic components, nonmetallic minerals, and plastic products, firms cite the uncertain economic environment, inadequate but there are few firms in these areas. Although manufacturers technology, lack of support to the private sector, and insuffi- of wood and metals contribute very little to total value-added, cient demand and production capacity as their major concern their value-added per worker is relatively important. (Figure 2.6). However, more than 30 percent of medium and large enterprises report inadequate physical infrastructure, Small firms are more hemmed in by the surrounding eco- currency fluctuations, unfair competition, and lack of raw nomic environment and lack of support for entrepreneurs, materials and qualified labor as the main blocks. Inadequate large firms by factors related to competitiveness. The high financial services also rank high as a challenge to industry, but cost of production is perceived as the leading block to better the problem affects small firms more than larger ones. The distribution of firms and employment across Tanzania is uneven, resonating with the distribution of poverty. Disparity in the geographical distribution of firms is pro- distribution of businesses is even more pronounced; at nearly nounced. Businesses are concentrated in the Eastern zone, 20 percent Dar es Salaam has by far the highest number of with 27.3 percent, and the Lake zone, with 16.7 percent. In firms (Figure 2.7B). Arusha, Mbeya, Morogoro, and Ruvuma contrast, 5 percent or fewer businesses are in the Southern house 5–7 percent each, while less than 2 percent are located and Western zones (Figure 2.7A). Disparity in the regional in Geita, Katavi, Kigoma, and Simiyu, and there are relatively C h a p t e r 2 F i r m P r ofil e s a n d S t r uc t u r a l T r a nsfo r m at ion 27 FIGURE 2.7: Firm and Job Distribution by Geographic Characteristics, Percent         A. Firms by zone         B. Firms by administrative region Dar es Salaam Eastern Mbeya Morogoro Ruvuma Lake Arusha Mara Singida Southern Highlands Dodoma Iringa Kilimanjaro South West Highlands Kagera Manyara Rukwa Shinyanga Northern Njombe Tabora Tanga Central Mwanza Lindi Mtwara Southern Pwani Simiyu Kigoma Western Geita Katavi 0 25,000 50,000 0 10,000 20,000 30,000 Micro Small Medium Large Micro Small Medium Large C. Jobs by zone, Percent 29.6 30 15 Share of total employment 20 10 Workers per rm 16.4 8.3 14.1 11.8 7.4 6.3 6.6 5.9 5.3 8.3 9.0 10 4.7 5 4.3 6.4 4.4 0 0 Southern Western Southern South West Central Northern Lake Eastern Highlands Highlands Share of total employment Average # of workers per rm Source: SBR 2014/15. Note: Tanzania Mainland is divided into eight zones based on DHS 2015/16 classification: (1) Western: Tabora, Kigoma; (2) Northern: Kilimanjaro, Tanga, Arusha; (3) Central: Dodoma, Singida, Manyara; (4) Southern Highlands: Iringa, Njombe, Ruvuma; (5) Southern: Lindi, Mtwara; (6) South West Highlands: Mbeya, Rukwa, Katavi; (7) Lake: Kagera, Mwanza, Geita, Mara, Simiyu, Shinyanga; and (8) Eastern: Dar es Salaam, Pwani, Morogoro. 28 TA N Z A N I A M A I N L A N D P O V E R T Y A S S E S S M E N T few in Lindi, Mtwara, or Pwani despite their favorable coastal zone with 14 percent. The Western and Southern zones are location. It is important to note that there are large disparities both home to no more than 6 percent of jobs. Yet firms in the in the distribution of businesses across the administrative Western zone have the highest average number of workers regions within the Lake zone: most businesses are in Mara, (8.3), ahead of firms in the Northern (7.4), Eastern (6.6), and while Geita and Simiyu have the fewest number of firms.4 The Lake (5.9) zones. The highlands zones have the fewest workers distribution of manufacturing firms by both geographic zones per firm, because the businesses operating there are very and regions is very similar to the pattern for all businesses. small. Regionally, Dar es Salaam accounts for 21.2 percent of all jobs, trailed far behind by Mbeya, Kilimanjaro, and Micro-and small-scale businesses predominate in all Arusha. The regions of Tabora, Dodoma, Mwanza, Manyara, regions, but some regions have larger proportions of and Ruvuma each have just over 4 percent of jobs. The lowest medium and large businesses. The Eastern zone, because shares are in Katavi, Geita, Njombe, and Rukwa, which each Dar es Salaam is there, has 25 percent of medium and 39 per- account for less than 2 percent. Firms in Mwanza have the cent of large businesses, including most manufacturing firms.5 highest average number of jobs at 10.2, followed by Kilimand- In the Central zone, particularly in Manyara, and the Northern jaro, Kigoma, Pwani, and Ruvuma, where the average number zone, in Arusha and Kilimanjaro, over 5 percent of all busi- of jobs per firm is about 9. Katavi, Njombe, Rukwa, Iringa, and nesses are medium and large and are manufacturing firms.6 Lindi have the lowest average number of jobs per firm at less Southern Highlands and South West Highlands are third in than 4. terms of where Tanzanian businesses are concentrated, but they have the largest shares of micro-sized (90 percent) and The geographic distribution of firms resonates with the informal businesses (60 percent). As for informality, about 55 spatial distribution of poverty (Figure 2.8). This indicates percent of firms in Dar es Salaam are not registered. that the more businesses are present, whether manufacturers or other types, the more poverty is reduced in the district or As a result of the uneven distribution of businesses, jobs region. Even though micro and small firms are known for not are concentrated in the Eastern and Lake zones, where paying well, their presence makes a significant contribution most firms are located. Because of Dar es Salaam, the to improving living standards and reducing poverty. Not Eastern zone accounts for 30 percent of all jobs (Figure 2.7C), surprisingly, however, poverty reacts more positively—drops followed by the Lake zone with 16 percent and the Northern more—to the presence of medium and large businesses. FIGURE 2.8: Poverty and Firm Distribution by Geographic Zone B. … and the decline is even stronger with the number A. Poverty declines significantly with the number of micro and medium and small businesses… of large businesses 100 100 80 80 Poverty headcount (%) Poverty headcount (%) 60 60 40 40 20 20 0 0 5 10 15 0.1 0.2 0.3 0.4 0.5 Number of micro and small businesses (thousand) Number of medium and large businesses (thousand) 4 The disparity of the distribution of businesses within the Lake zone is in line with the disparity of poverty. Mara, which contains the largest number of firms, has a poverty rate of 23 percent. Geita and Simiyu, where the number of businesses is the lowest, have poverty rates at, respectively, 37 and 39 percent. 5 About 33.5 percent of medium and large manufacturing firms operate in Dar es Salaam. They account for nearly 4 percent of all manufacturing firms in the city. 6 About 7 percent of medium and large firms operate in Arusha and represent around 5 percent of all firms there. About 7 percent are in Manyara and nearly 8 percent in Kilimanjaro. C h a p t e r 2 F i r m P r ofil e s a n d S t r uc t u r a l T r a nsfo r m at ion 29 FIGURE 2.8C: Districts with higher concentration of businesses…       FIGURE 2.8CD: … also have less poverty Sources: SBR 2014/15, 2012 Population Census data, HBS 2018, and Tanzania Jobs Diagnostic (2017). Notes: The number of poor has strong and significant correlations with the number of micro/small businesses, estimated at –0.04, and of medium/large businesses, estimated at –0.98, both significant at 1 percent, with the latter relationship being stronger (R2 = 0.08, p-value = 2*10–5 compared to R2 = 0.05, p-value = 4*10–3). These relationships are robust to controlling for district or region characteristics such as access to basic services, distance to main services, etc. The results are also robust to restriction of the sample to industrial firms. These results are supported by the 2015–16 Demographic wealth quintiles. Conversely, more than 7 in 10 people in the Health Survey (DHS), which found that in the Western and Eastern zone (75 percent) are in the two highest quintiles, and Southern zones, which have the fewest firms, in the former 2 in over 50 percent are in the richest quintile. 3 people and in the latter about 5 in 10 are in the two lowest 30 TA N Z A N I A M A I N L A N D P O V E R T Y A S S E S S M E N T II. Which Firms Create the Most Jobs The objective of this section is to establish the profile percent of all formal firms but account for 35 percent of total by activity, size, and age of firms that created the most employment in formal firms.7 Conversely, small firms, those jobs between 2010 and 2013—the reference years for with 5–19 workers, constitute 63 percent of formal compa- the 2013 Enterprise Survey. Most of Tanzania’s jobs are nies but account for only 32 percent of jobs, and micro-firms, concentrated in large firms, but large firms are relatively those with 4 or fewer employees, constitute 17 percent of rare. According to Tanzania’s 2013 Enterprise Survey, formal all firms but account for just 2 percent of total employment enterprises with more than 100 employees represent just 2 (see Appendix B for more details). Between 2010 and 2013 average productivity decreased slightly for of all firms, but firm performance is still quite good. The slight dip in firm productivity between 2010 and fell by 4.3 percent, but there is no correlation with firm size 2013 had no clear relationship to firm size or age. Average (Table 2.1). The largest firms seemed to be not the most firm productivity, calculated as real gross output per worker, productive, ranking third after one- and two-worker firms. TABLE 2.1: Productivity of Firms by Size and Age, 2010 and 2013 PRODUCTIVITY 2010 – LN(Y/L) PRODUCTIVITY 2013 – LN(Y/L) VARIATIONS (%) MEAN MEDIAN MEAN MEDIAN MEAN MEDIAN By size 1 16.3 16.5 17.8 18.1 9.3 9.8 2 16.4 16.8 16.2 16.1 –1.4 –4.1 [3–4] 16.1 16.5 14.8 15.2 –7.8 –8.1 [5–9] 16.2 16.2 15.3 15.3 –5.5 –5.9 [10–19] 15.8 15.9 15.1 15.2 –4.4 –4.4 [20–49] 15.6 15.5 15.2 15.2 –2.4 –1.9 [50–99] 15.8 16.2 15.6 16.1 –0.9 –0.3 [>=100] 16.5 17.2 16.1 16.3 –2.6 –5.5 By age (years) 1 16.5 16.5 2 14.7 16.9 16.3 16.2 10.7 –4.1 3 16.1 16.5 15.2 15.3 –5.6 –7.4 4 15.1 15.8 14.7 15.4 –2.7 –2.4 [5–9] 16.4 16.2 15.2 15.2 –7.1 –6.6 [10–14] 16.1 16.0 15.6 15.8 –2.9 –1.6 [15–19] 15.9 15.6 15.2 15.2 –4.2 –2.9 [>=20] 15.8 15.6 15.5 15.3 –1.7 –1.5 Total 16.0 16.0 15.3 15.3 –4.3 –4.5 Source: Enterprise Survey 2013. Note: Y is measured as the gross output reported by the firm during the survey; L is the total of permanent full-time workers at the end of the fiscal year preceding the survey. 7 This section builds on data from the 2013 Enterprise Survey and uses the approach developed by Rijkers et al. (2014). See Appendix B for more details on the methodology. C h a p t e r 2 F i r m P r ofil e s a n d S t r uc t u r a l T r a nsfo r m at ion 31 TABLE 2.2: Production by Firm Size and Age, 2013, youngest firms, 2 years old, performed better than the rest Percent with productivity rising by 10.7 percent. In general, all firms, AVERAGE NUMBER OF OPERATING regardless of size, perform at about the same level. CAPACITY UTILIZATION HOURS PER WEEK MEAN MEDIAN MEAN MEDIAN On average, Tanzanian firms use 80 percent of their By size production capacity--the extent to which a firm can use 1 77.4 82 30.7 12 all its potentiality for production. Capacity is related to 2 33.1 25 22.1 10 neither firm size nor firm age. Self-employers use more than [3–4] 86.2 100 83.4 84 [5–9] 81.8 100 32.5 11 three-quarters of their capacity, firms of two workers use only [10–19] 91.9 100 27.8 11 a third (Table 2.2). All firms with more than three workers [20–49] 81.3 100 33.4 11 use on average at least 80 percent of their capacity, peaking [50–99] 87.7 80 49.2 40 at 92 percent for 10–19-worker firms. Due to missing data, [>=100] 85.5 90 62.5 45 utilization capacity was not computed for firms in business By age (years) for only a year. Firms with 2 years of operation use their total 1 production capacity but firms in business for 3 years use less 2 100.0 100 63.0 63 than 50 percent; 4–15-year-old firms about 75 percent, and 3 46.1 33 60.7 24 4 76.8 80 59.6 12 the oldest firms more than 80 percent. [5–9] 76.2 90 35.7 12 Tanzanian firms operate on average 35 hours a week, and [10–14] 74.4 85 43.1 11 [15–19] 87.6 100 30.1 11 hours in operation tend to decrease as firms age. Firms of >=20 83.1 100 32.1 11 2 workers operate just 22 hours a week (Table 2.2). This could Total 80.2 100 35.3 11 explain why they use only a third of their capacity. Mean- Source: Enterprise Survey 2013. while, 3–4-worker firms, which use most of their capacity, also operate longer each week—much more than the largest firms. The largest increases in productivity occurred for firms with As the age of firms increases, their weekly operating hours a single worker (+9.3 percent). Conversely, productivity of tend to decrease. While the oldest firms average 30–32 hours 3-to-4-worker firms fell by 5.5 percent. Similarly, no relation- a week, young firms tend to work overtime, operating more ship can be drawn between firm age and productivity. The than 60 hours a week. FIGURE 2.9: Share of Jobs Created by Firm Size, 2010–13, Percent Net job creation has been primarily 40 36.6 driven by large and older firms. 33.5 30.5 Net job creation is positively and significantly correlated with increased firm size. Small firms created 33.5 percent 20 of new jobs, medium firms 36.6 percent, and large firms 30.5 percent (Figure 2.9 and Table 2.3).8 However, micro firms lost jobs; they were down by about 1 percent. Regression 0 results of net job creation on firm size dummies suggest that –0.7 job creation is the highest among firm employing 50 workers and more: the net job creation rate in these firms is about 50 percent higher than that of one-person firms (Table C.1 –20 in Appendix C). Micro Small Medium Large Source: Enterprise Survey 2013. Notes: Micro firms employ 1–4 workers, small firms 5–19, medium 20–99, and large: Excluding new entrants does not affect the pattern. 100 or more. When excluding firms created between 2010 and 2013, larger Net job creation is the difference between firm size in 2013 and 2010. Job creation without new entrants is the difference between firm size in 2013 and 2010 for firms firms still appear to contribute the most to net job creation. set up before 2011. Job creation with new entrants is the difference between firm size in 2013 and at the startup year for firms set up after 2010. 8 Net job creation is calculated as the difference between firm size in 2013 and in the start-up year for firms that opened in 2011, 2012, or 2013, and between firm size in 2013 and in 2010 for firms that existed before 2011. See Appendix C for more details. 32 TA N Z A N I A M A I N L A N D P O V E R T Y A S S E S S M E N T TABLE 2.3: Net Job Creation by Firm Size, 2010–13 FIGURE 2.10: Jobs Created by Age Class, 2010–13, NET JOB CREATION NET JOB CREATION NET JOB Percent of Firms ALL FIRMS WITHOUT NEW ENTRANTS CREATION NEW 40 ENTRANTS (#) (%) (#) (%) (#) 31.0 Firm size 30 1 –319 –1.1 –178 –0.6 –141 2 9 0.0 9 0.0 0 22.8 20.0 [3-4] 113 0.4 27 0.1 86 20 17.8 [5-9] 4,253 14.2 3,829 13.1 424 [10-19] 5,821 19.4 5,539 18.9 282 [20-49] 5,253 17.5 5,178 17.7 76 10 8.4 [50-99] 5,745 19.1 5,745 19.6 0 [>=100] 9,178 30.5 9,178 31.3 0 Total 30,053 100 29,326 100 726 0 [1–4] [5–9] [10–14] [15–19] [>=20] Source: Enterprise Survey 2013. Firms larger than 100 employees accounted for 31.3 percent Job creation is also mostly attributable to older firms, but of net job creation between 2010 and 2013, and all firms the effect of age on job creation is not statistically signifi- larger than 20 employees created nearly 70 percent of cant. Firms in business less than 5 years contributed 8.4 percent the new jobs during the period (Table 2.3). Regression of jobs created between 2010 and 2013 (Figure 2.10). Con- results support this finding and show that creation of new versely, job creation appears to be concentrated in firms more jobs is still primarily and significantly driven by medium than 15 years old, which accounted for more than 50 percent and large firms of 50 workers and more (Table C.1). Firms of the jobs created over the period. However, once statistical launched after 2010 also contributed to job creation, but artefacts resulting from new entrant firms are removed, no sig- to a much lower extent. Furthermore, within new entrants, nificant relationship emerges between net job creation and the self-employment firms—those with only 1 employee—seem length of time a firm has been in existence (Table C.1). to have instead destroyed jobs. C h a p t e r 2 F i r m P r ofil e s a n d S t r uc t u r a l T r a nsfo r m at ion 33 FIGURE 2.11: Net Job Creation by Firm Age and Size, 2010–13 5000 4000 Net job creation (#) 3000 2000 1000 0 –1000 [>=100] [50–99] 2 [20–49] 3 [10–19] 4 [5–9] [5–9] [3–4] [10–14] Firms' 2 [15–19] age ' size 1 [>=20] Firms –1000–0 0–1000 1000–2000 2000–3000 3000–4000 4000–5000 Source: Enterprise Survey 2013. Clearly, net jobs are primarily created by older and larger Mobility in terms of firm size is relatively high among firms. Across firm age classes, net jobs were mainly created micro and small firms but more limited among large by medium and largest firms. Between 2010 and 2013 most firms. Between 2010 and 2013 most micro firms, which man- new jobs in formal firms were created by larger firms with aged to survive, grew relatively quickly. Only 39.4 percent longer operating histories (Figure 2.11). of one-person firms in 2010 were still the same size in 2013, 15 percent of them had expanded to 2 workers, and 34 TA N Z A N I A M A I N L A N D P O V E R T Y A S S E S S M E N T TABLE 2.4: Firm Size Transition Matrix, 2010–13, Percent (Figure 2.12). Manufacturing followed with nearly one-third SIZE IN 2013 (J ) of the new jobs created. The trade sector—wholesale and 1 2 [3–4] [5–9] [10–19] [20–49] [50–99] [>=100] TOTAL retail—accounted for 21 percent of jobs creation. Within Size in manufacturing, most of the job creation was in furniture 2010 (i ) (11 percent), food processing (6 percent) and transformation 1 39.4 15.2 6.2 24.6 11.1 3.6 0.0 0.0 100 industries (i.e. plastic, metal, chemicals) (6 percent). How- 2 10.7 35.5 2.1 23.5 11.3 7.4 9.5 0.0 100 [3–4] 2.2 0.4 49.0 42.8 5.0 0.3 0.2 0.0 100 ever, regression results show that the influence of sectors of [5–9] 0.8 0.9 1.0 86.4 10.8 0.1 0.1 0.0 100 activity on job creation is not significant when controlling for [10–19] 0.0 0.1 0.1 9.5 81.0 8.8 0.6 0.0 100 other firm characteristics, such as size and length of exis- [20–49] 0.0 0.0 0.0 0.1 2.7 95.2 1.5 0.6 100 tence (Table C.3 in Appendix C). [50–99] 0.0 0.0 0.0 0.0 0.0 13.6 64.4 21.9 100 [>=100] 0.0 0.0 0.0 0.0 0.0 1.5 2.3 96.2 100 Source: Enterprise Survey 2013. Note: Cells indicate what percentage of firms in row category i in 2010 end up in column category j in 2013. 25 percent had expanded to 5–9 workers (Table 2.4). Simi- larly, more than half of firms that had 2 workers in 2010 had more employees in 2013. Conversely, job growth in medium and large firms seems to have slowed once firms reached at least 5 workers. In large firms, mobility between 2010 and 2013 was much lower: for instance, 95 percent of firms that had 20–49 employees in 2010 were the same size in 2013, and very few of those with more than 100 workers had changed size in 2013. The vast majority of jobs created between 2010 and 2013 were in services. Between 2010 and 2013, 41 percent of the jobs created were in market services, which comprises hotels, restaurants, and transport services FIGURE 2.12: Jobs Created by Sector, 2010–13, Percent Market services 40.6 Trade 20.7 Furniture 11.3 29.7 Utilities and construction 9.1 40.6 Food and beverage 5.8 Plastic, mining and metal 5.6 Textile and leather 2.2 Petroleum and pharmaceutical 2.1 9.1 Wood 1.7 20.7 Equipment 1.0 0 10 20 30 40 50 Manufacturing Utilities and construction Trade Market services Source: Enterprise Survey 2013. C h a p t e r 2 F i r m P r ofil e s a n d S t r uc t u r a l T r a nsfo r m at ion 35 36 TA N Z A N I A M A I N L A N D P O V E R T Y A S S E S S M E N T CHAPTER 3 Access to Finance The critical role of financial inclusion in alleviating pov- of different reasons. It is therefore important to measure erty and enhancing inclusive growth has been broadly financial inclusion and identify opportunities to remove the recognized in the literature. There is growing evidence that barriers that may be preventing the access to and use of participating in the financial system, increases people capacity financial services. to start and expand businesses, invest in human capital, and cope with shocks (Demirgüc¸Kunt et al., 2015). It also contrib- During the recent years, there has been a rapid surge utes to empower women, and to boost productive investment in mobile banking in Tanzania as well as a moderate and consumption. The benefits go beyond individuals, as expansion of bank accounts usage. However, the access to access to finance is inextricably linked to firms’ performance financial services remains dominated by the heavy reliance and growth, and therefore to jobs creations (Fowowe, 2017). on informal financing, and the needs of the poor and women Expansion of financial inclusion is gradually becoming a for financial inclusion are still unmet. Limited access to formal priority in economic development. finance services has contributed heavy concentration of busi- nesses in microenterprises and limited growth of medium and Access to finance can be limited by a number of factors large enterprises. Expanding access to finance and address- including limited number of financial institutions, heavy ing unmet needs for financial inclusion have been identified collateral requirements, prohibitive costs, high levels among the priority policy actions to accelerate poverty reduc- of business informality, regulations or other market tion and shared prosperity in Tanzania. The aim of this section failures. The use of financial service, or financial inclusion, is to contribute to a better understanding of the potential can be low even in the absence of lack of access as drivers and barriers to foster financial inclusion in Tanzania, people may choose not to use financial services for a set both from an individuals’ and a firms’ perspective. 38 TA N Z A N I A M A I N L A N D P O V E R T Y A S S E S S M E N T I. Extent and Determinants of Individuals’ Financial Inclusion While half of Tanzania’s population lacks financial inclusion, access to financial tools remains strongly dependent on socio-economic factors such as education and wealth. Financial inclusion in Tanzania remains moderate as a individuals with a tertiary educational level possessed a formal large part of the population continues to lack any formal bank account, compared to only 41 percent of the individuals financial tool. We measure financial inclusion through three that had completed primary education or less. key indicators that are essential when measuring one’s ability to access basic and critical financial tools – the ownership of Older individuals appear to be more likely financially a formal bank account, the possession of formal savings, and included. The bell-shaped curve of the share of individu- the access to credit provided by a formal financial institution als with formal bank account related to age in Figure 3.1D (See Appendix D). According to the 2017 World Bank’s Global suggests a potential existence of a generational shift in Findex data for Tanzania, about 47 percent of the population terms of access to formal bank account, and therefore in possessed a formal bank account (Figure 3.1A). The rate terms of financial inclusion. This effect was more pronounced of coverage of the Tanzanian population for the two other when using 2014 Findex data but seems to have continued financial indicators were much lower, with 6.2 and 5.3 percent throughout 2017. The curve based on 2017 data confirms of the population that respectively had formal savings, and increased financial inclusion over time and with age, as access to a formal credit. almost all age classes had higher rates of access to formal bank account in 2017 than in 2014 and as access continues to Financial inclusion is strongly determined by socio-eco- be slightly higher among older groups, except oldest ones, nomic characteristics. In 2017, 52 percent of men had than younger groups. In the near future, these rates for the a formal account, compared to 42 percent of women left tail of the curve will, most probably, continue to improve ­(Figure 3.1A). On average, older, better-off and more edu- as many individuals from the youngest age groups are still cated Tanzanians are more likely to be financially included in the process of acquiring formal financial tools – such as than their younger, less educated and poorer counterparts. opening a formal account in a financial institution. They are In particular, education and income appear as very signifi- also acquiring more education, which will likely increase their cant factors when explaining financial inclusion (Table D.2 in financial inclusion. Regression results support the positive and Appendix D). The richer an individual, the more likely he/she statistically significant effect of age on financial inclusion, with is to hold a formal account, to possess formal savings and to a higher likelihood for older individuals to have formal bank have access to formal credit. Only 32 percent of individuals account, to hold savings, and to benefit from access to formal from the lowest income quintile possessed a formal bank credit (Table D.2 in Appendix D). Yet, age appears to have a account, whereas 63 percent of individuals from the richest nonlinear relation with all indicators of financial inclusion, with quintile had one (Figure 3.1B). The same discrepancies can be a negative coefficient for age squared, suggesting that older observed when it comes to formal savings and formal credit. people are more likely to be financially included, but after For instance, the share of individuals with formal savings is a certain age, the probability of being financially included almost 7 pp higher in the richest quintile than in the poorest declines. This is supported by the decline of the blue curve in one. Likewise, as educational attainment increases, access Figure 3.1D for those aged 60 years old and above. to key financial tools increases (Figure 3.1C). In 2017, all C h a p t e r 3 Acc e ss t o F in a nc e 39 FIGURE 3.1: Financial Inclusion in Tanzania, 2017, Percent      A. Financial inclusion by gender     B. Financial inclusion by income quintiles 60 80 51.6 46.8 63.2 42.2 60 52.9 40 43.1 42.5 40 32.0 20 20 12.0 6.0 4.1 8.3 8.0 7.9 6.1 6.2 5.2 2.9 2.9 3.4 4.9 4.6 4.6 5.3 0 0 1 2 3 4 5 Men Women Tanzania Income Quintiles Formal account Formal saving Formal credit Formal account Formal saving Formal credit      C. Financial inclusion by education   D. Formal bank account by age, 2014 and 2017 100.0 60 100 80 65.0 40 60 41.0 44.1 40 20 17.9 20 12.4 4.5 7.6 0 3.9 0 9 4 9 4 9 4 9 4 9 4 + –1 –2 –2 –3 –3 –4 –4 –5 –5 –6 65 Primary Secondary Tertiary 15 20 25 30 35 40 45 50 55 60 Formal account Formal saving Formal credit 2014 2017 Sources: Global Findex 2014 and 2017. Financial inclusion has progressed rapidly over the last years, primarily driven by the development of mobile banking. Access to formal banking accounts continues to progress, The development of mobile banking particularly primarily because of the rapid development of mobile benefitted the most vulnerable individuals – low- banking. Between 2014 and 2017, the share of the population educated and poorest individuals. An analysis of the having access to a formal bank account increased by 7 pp trends in mobile banking across a set of socio-economic (Figure 3.2A). Most of the increase was driven by the devel- indicators underlines the fact that the fastest progresses opment mobile money banking. Between 2014 and 2017, the were recorded in vulnerable groups of the population. share of the population with mobile bank accounts increased Between 2014 and 2017, the share of primary-educated-or- by 6 pp, to reach 38.5 percent. less individuals with access to mobile banking increased 40 TA N Z A N I A M A I N L A N D P O V E R T Y A S S E S S M E N T FIGURE 3.2: Trends in Financial Inclusion, 2014 and 2017, Percent        A. Formal bank account   B. Trends in mobile money banking by socio- economic characteristics 60 100 94 93 46.8 80 39.8 40 38.5 60 54 55 55 55 44 44 32.4 38 40 40 34 33 33 33 34 27 26 26 21.0 18 16 19.0 20 20 0 1 2 3 4 5 en en y y y ar ar ar M om nd rti im Te co W Pr Se 0 Formal account Financial inst. Mobile banking Gender Education Income quintiles 2014 2017 2014 2017 Sources: Global Findex 2014 and 2017. 7 pp, compared to 1 pp for tertiary-educated individuals the population sent remittances. 87.5 percent of them (Figure 3.2B). Likewise, the share of individuals from the used mobile banking to issue the transfer, which represents poorest and second-­ poorest quintiles with mobile banking 26.5 percent of the total population. Likewise, 36 percent of access increased by respectively 8 and 18 pp. Conversely, the population received remittances, 88.7 percent of which the increase was much more modest or inexistent for received the transfer through mobile banking, accounting individuals from the other income quintiles. for 32.2 percent of the total population. Overall, 41 percent of the population used mobile banking either to receive or The central driving force behind the development of send remittances, while only 1.5 percent used regular bank mobile banking is the sending and receiving of remit- accounts, 5.5 percent used other forms of transfer such as tances through a mobile phone. In 2017, 30 percent of cash. Barriers to financial inclusion are slowly decreasing, accompanying the improved financial inclusion observed. A number of subjective barriers continue to constrain center (35 percent) (Figure 3.3A). However, compared access to formal financial accounts, but the impact of to 2014, it seems that the proportion of the population some barriers is diminishing. Within the individuals that affected by some of these barriers is slightly receding. For do not have a banking account registered at a formal instance, in 2014, nearly 46 percent of individuals without financial institution, the most significant barriers cited are a bank account cited the cost as a significant barrier, and the lack of money (81 percent), the cost of opening a bank 50 percent considered the distance to be an impediment account (39 percent), and the distance from the financial in 2014. C h a p t e r 3 Acc e ss t o F in a nc e 41 FIGURE 3.3: Financial Inclusion in Tanzania, 2014 and 2017, Percent A. Barriers to financial inclusion B. Type of saving and saving motivation 59.2 60 74.1 Lack of money 80.7 48.4 45.8 Too expensive 40 39.0 32.9 50.2 Too far away 23.0 35.0 18.3 29.0 20 13.3 Lack of documentation 9.2 30.5 7.3 5.8 6.2 No need for nancial 23.1 services 12.2 0 e s s 12.5 g g th es ag Lack of trust in in on sin av 11.3 av ld m s s bu ro al al Family member has an 12 4.1 rm Fo rm or st account 6.2 fo Fo m la In ar e rf th 3.1 Religious reasons Fo in 3.2 d ve Sa 0 20 40 60 80 100 Saving motivation Type of saving 2017 2014 2014 2017      C. Saving by income quintiles         D. Saving by education 80 100 93.3 66.2 80 60 54.1 50.3 63.5 43.4 60 40 44.1 43.6 27.7 26.8 40 21.0 20 16.9 16.0 12.0 22.0 19.0 10.7 17.3 5.2 6.0 4.9 20 12.4 2.9 3.9 0 1 2 3 4 5 0 Primary Secondary Tertiary Income quintiles Saved in the last 12 months Formal saving Informal saving Saved in the last 12 months Formal saving Informal saving Source: Global Findex 2014 and 2017. Despite half of the population saving, structured forms of saving, whether formal or informal, remain scarce, and the saving rate has been decreasing over the last couple of years. Nearly half of the population save some money at some having saved some money at some point over the last point, but the saving rate has decreased between 2014 12 months, even if it was outside of the framework of a and 2017. In 2017, 48.4 percent of Tanzanians declared structured formal or informal scheme. Between 2014 and 42 TA N Z A N I A M A I N L A N D P O V E R T Y A S S E S S M E N T 2017, the saving rate decreased by more than 10 pp, down all, the likelihood of informal saving is very strongly cor- from 59.2 percent in 2014 (Figure 3.3B). The likelihood of related with gender, undermining the role played by women saving is strongly linked to socio-economic factors: saving in developing and using informal saving schemes such as appears primarily and significantly correlated to older age, rotating savings and credit groups upatu or other saving higher income, and higher education (See Table D.5 in clubs. While around 13 percent of men resort to an informal Appendix D), highlighting the role played by financial literacy saving scheme, the proportion rises to 24 percent in the case and education in the choices made by individuals with of women. regards to the saving of their income. While the saving rates of the individuals from the poorest quintiles stands at about Saving is mainly driven by business purposes, while 28 percent, it reaches 66 percent for the richest one (Figure saving’s strategies for retirement underscore the role 3.3C). Likewise, 44 percent of primary-educated individuals played by education. The objective of saving is primarily saved some money over the last 12 months, compared to 93 to pursue a business opportunity. About 23 percent of percent of tertiary-educated individuals (Figure 3.3D). Tanzanians save in order to invest in future economic assets – in a farm or in a business. Retirement purposes only Meanwhile, structured forms of saving, whether formal concerns 6 percent of the surveyed individuals (Figure 3.3B). or informal, remain scarce and subject to socio-economic The latter motive is particularly important for tertiary- factors. The use of structured forms of saving remains very educated individuals as 26 percent of them save for old age limited (Figure 3.3B). Only 6.2 percent of individuals used a compared to only 5 percent of primary-educated individuals, formal saving scheme, down from 9.2 percent in 2014, and underscoring the effect of education on the ability to plan 18.3 percent of them saved or set aside money by using an the future and on the knowledge of the benefits of saving. informal saving club or a person outside the family (+ 5 pp The regression results show that the likelihood of saving between 2041 and 2017), suggesting most of the saving for retirement is significantly higher for better-educated occurs within the family circle and primarily consist of setting individuals, and the likelihood of saving for business money aside, therefore not benefiting from any remuner- purposes is significantly higher for individuals from the ation. Formal and informal savings are driven by distinct richest income quintiles (Table D.6 in Appendix D). Indeed, socio-economic dynamics. Formal saving is primarily and the proportion of individuals from the richest quintile that significantly correlated to higher education; informal saving save for business or retirement purposes is much higher schemes are more likely for individuals with lower educational than in the lowest quintile (respectively +18 and +6 pp attainment (Figure 3.3D and Table D.5 in Appendix D). Above differences) Access to credit remains widely informal, primarily relying on family, friends, and saving clubs. Access to credit is achieved primarily through informal access to credit across income quintiles exist but are not as arrangements, across all educational and income groups. important as between education groups. In particular, access While around 42 percent of the population has access to to informal credit is similar across all quintiles, concerning some form of credit, 32 percent of Tanzanians rely on informal around one-third of individuals in each income quintile credit sources, compared to only 5 percent who use credit (Figure 3.4B). However, richer individuals appear more likely arrangements from formal financial institutions (Figure 3.4A). to secure a formal loan with the rate of access to formal credit Educated individuals are more likely to access credit (39 increasing as income increases (poorest quintile: 3 percent; percent for primary-educated individuals compared to 83 richest quintile: 8 percent). percent for tertiary-educated ones), with a similar pattern and significant likelihood observed for both formal and informal The role played by saving clubs as well as families and credit (Table D.8 in Appendix D). Only 5 percent of primary- friends to provide such informal credit is very important. educated people are able to secure formal loans, compared The informal credit in Tanzania primarily relies on family to 18 percent of tertiary-educated individuals. Looking sources and saving clubs. In 2017, about 28 percent of through the wealth perspective shows that differences in individuals had borrowed in the last 12 months from family C h a p t e r 3 Acc e ss t o F in a nc e 43 or friends, and more than half of the population (53 percent) from the poorest quintiles, underlining the role played by resorted to saving clubs. Interestingly, the share of individuals informal saving schemes to provide credit to a large part of resorting to saving clubs is higher for those the population with not enough collateral nor resources to seek a formal loan (Figure 3.4D). FIGURE 3.4: Financial Inclusion in Tanzania, 2017, Percent A. Credit sources by education B. Credit sources by income quintiles 100 60 82.7 80 46.0 41.6 40.0 41.6 61.2 40 36.8 35.6 34.2 60 30.7 48.8 29.0 29.4 38.5 41.2 40 37.2 29.9 31.8 20 17.9 8.3 8.0 20 7.6 2.9 3.4 4.1 4.5 5.3 0 0 Primary Secondary Tertiary Tanzania 1 2 3 4 5 Education Income quintiles Credit Informal credit Formal credit Credit Informal credit Formal credit C. Informal credit sources by education D. Informal credit sources by income quintiles 100.0 100 100 80 80 67.6 61.5 61.2 60 50.1 60 55.2 52.7 44.7 45.4 42.7 40 35.8 30.0 26.8 27.0 27.8 40 26.1 24.8 27.5 20 20 0 0 1 2 3 4 5 Primary Secondary Tertiary Tanzania Income quintiles Education Family and friends Saving club Family and friends Saving club Source: Global Findex 2017. 44 TA N Z A N I A M A I N L A N D P O V E R T Y A S S E S S M E N T II. The Effects of Access to Finance on the Growth of Tanzanian Firms Access to finance is perceived as a major obstacle to formal firms’ operations. Tanzania’s formal firms perceive access to finance as credit constraints. Even though the large majority of Tanza- a major and increasingly important obstacle. The 2013 nia’s firms perceive to be financially constrained, the empirical Enterprise Survey collects data for formal firms operating analysis of the effects of access to finance on their perfor- across all sectors. When asked to name the most severe mance shows that credit constraints do not have a significant obstacle to their current operations, 38 percent of the adverse effect on their growth –proxied by the growth of the 2013 Enterprise Survey’s respondents identified Access number of employees (See Appendix E for more details). to Finance, ranking the constraint first out of all the Performing econometric estimations — using regressions possibilities (Figure 3.5A). Furthermore, the perception on data from the 2013 Enterprise Survey —we find that the of access to finance from 2006 and 2013 seems to have subjective measure of access to finance, obtained from the worsened. In the 2006 Enterprise Survey, around 46 percent ranking of access to finance as no obstacle or severe obstacle of the respondents identified access to finance as not being to business operations, has a statistically significant negative an obstacle or only a minor one. In 2013, the proportion effect on firms growth, suggesting that inadequate financing had decreased to 29 percent. Conversely, the share of is a serious constraint that Tanzanian firms face (Table E.2 respondents considering access to finance as a moderate, in Appendix E). However, objective measures of access to major, or very sever obstacle had increased from 54 percent finance based on whether firms are constrained in obtaining to 71 percent (Figure 3.5B). credit or not indicate that access to finance constraint does not exert a statistically significant negative effect on firms In particular, access to finance tends to be perceived as a growth.1 The only conclusive and significant results obtained major obstacle by smaller firms. Not surprisingly, the per- are for the credit constrained status, which indicate that firms ception of access to finance as a constraint and an obstacle that are moderately credit constrained tend to grow more tends to decrease with size of the firms. Only 28 percent of than firms that are fully credit constrained (Table E.3 in Appen- Tanzanian companies that have 10 employees or less do not dix E). Nevertheless, the results do not reveal any significant consider access to finance as an obstacle or only a minor one, effect of loans, credit lines or overdraft on firms growth. This while the rate increases to 69 percent in the case of compa- is probably due to both the small size of the sample and the nies with more than 200 employees (Figure 3.5C). Conversely, overall limited access to credit, as only few large firms benefit 23 percent of Large companies see access to finance as a from this access and their potential to grow is lower than small major or severe obstacle, compared to 39 and 52 percent of and medium firms. Micro and Small firms respectively. However, the empirical analysis shows that the growth of Tanzanian firms is not significantly affected by 1 The descriptive statistics from the 2013 Enterprise Survey show that large firms tend to have more access to finance, in particular, to evolved financial tools such as overdraft facilities and credit lines in formal financial establishments (Table E.1). C h a p t e r 3 Acc e ss t o F in a nc e 45 FIGURE 3.5: Access to Finance of Formal Enterprises, 2006 and 2013, Percent A. Most severe obstacle to operations B. Perceived vs objective financial constraint, 2006 and 2013 Access to nance 37.9 100 100 8.6 Electricity 24.9 17.3 16.1 19.0 Tax rates 8.3 80 80 Informal sector competitors 6.1 23.4 27.8 Access to land 5.1 60 58.0 Customs and trade 3.2 60 13.4 52.4 Business licensing and permits 3.1 Corruption 2.5 27.0 40 40 Crime and theft 1.9 Transportation 1.9 14.2 7.1 45.9 20 20 Low educ. labour 1.7 29.1 Tax administration 19.1 21.5 1.3 Labour regulations 1.2 0 0 Political instability 1.1 2006 2013 2006 2013 Telecommunications 0.0 No obstacle/minor Moderate NCC MCC 0 10 20 30 40 Major Severe PCC FCC C. Perceived financial constraint by firms’ size D. Objective financial constraint by firms’ size 100 100 3.0 7.6 11.8 15.8 16.8 17.8 23.7 24.0 23.9 15.4 80 80 23.4 7.6 33.0 35.6 23.8 60 60 55.7 44.3 49.5 33.3 17.7 40 28.0 40 15.9 69.4 8.1 20 6.1 20 20.4 23.6 27.3 23.5 31.7 34.8 27.6 0 0 Micro Small Medium Large Micro Small Medium Large Non credit constraint Maybe credit constraint No obstacle/minor Moderate Major Severe Partial credit constraint Full credit constraint Source: Enterprise Surveys 2006 and 2013. Notes: - Firms’ size is as follow: Micro (10 employees or less), Small (11 to 50 employees), Medium (51 to 200 employees), and Large (more than 200 employees). - Objective measure of financial constrain is as follow: NCC (Non-Credit Constraint), MCC (Maybe Credit Constraint), PCC (Partial Credit Constraint), and FCC (Full Credit Constraint) – See Appendix E for more details. 46 TA N Z A N I A M A I N L A N D P O V E R T Y A S S E S S M E N T Around one third of industrial firms, whether formal or informal, report facing inadequate financial services. Around one third of industrial firms feel financially con- For instance, while 29 percent of Micro industrial companies straint, with higher constraints perceived by smaller – between 1 and 10 employees—report inadequate financial firms. Overall, 29 percent of Tanzania’s industrial firms from services as an obstacle, only 16 percent of Large industrial the 2013 Census of Industrial Production (CIP) declare that enterprises – more than 200 employees— feel financial con- inadequate financial services constitute a major challenge. strained. The prevalence of inadequate financial services also Disaggregation by firms’ size shows that the financial constrain appears slightly higher for informal industrial companies than tends to decrease as enterprises’ size increases (Figure 3.6A). for formal ones. Very few informal households’ enterprises are able to secure a loan for their operations, and mostly rely on micro-credit organizations. Only one tenth of informal household enterprises have resorted to micro-credit entities or informal credit channels access to credit, which primarily comes from micro-credit such as SACCOS (Savings and Credit Co-Operative Society), entities or informal credit channels. At the national level, private money lenders, or upatu schemes: 40 percent of the only 9 percent of informal household enterprises were able to household enterprises that contracted a loan did so through secure a loan to develop or support their business operations SACCO and 20 percent through private money lenders. In (Figure 3.7A). The rate increases to 12 percent in the case of comparison, on top of the low lending rate, only 18 percent of household enterprises operating in urban environment, and household enterprises that took a loan resorted to a tradi- to 12 percent for household enterprises in the services’ sector. tional and formal bank (Figure 3.7B). Most informal household enterprises that obtained a loan FIGURE 3.6: Access to Finance of Industrial Enterprises, 2013, Percent A. Perceived financial constrain by firms’ size B. Perceived financial constrain by manufacturing sub-sector 40 Utilities 32.6 Textile 30.5 29.4 30 29 29.2 29 29 30 26 Petroleum, chemical and pharm. 30.0 25 23.6 23 Furniture 29.9 19.2 19 20 16.4 18 Food and beverage 29.0 Plastic, mineral and metal 28.8 10 Equipment 27.4 Mining and quarying 26.8 0 Water and waste 24.3 Micro Small Medium Large Tanzania 22.7 Wood Firms' size Other manufacturing 19.3 Industrial enterprises Formal Informal 0 10 20 30 40 Source: CIP 2013. C h a p t e r 3 Acc e ss t o F in a nc e 47 FIGURE 3.7: Access to Finance of Informal Household Enterprises, 2014, Percent A. HH Enterprises with a loan B. Sources of loan (% of HH Enterprises with a loan) 50 Tanzania 9.0 40.0 40 Urban 11.6 Area 30 Rural 5.8 20.3 20 18.0 Agriculture 5.8 9.4 10 5.2 5.5 5.6 Manufacturing 1.6 Sector 0 r U O ds er S nk e O AT Construction and G 5.5 th nd en Ba CC N O UP le fri utilities SA ey d an on es m Services 12.2 tiv te la iva Re Pr 0 5 10 15 Source: 2014 Integrated Labor Force Survey. 48 TA N Z A N I A M A I N L A N D P O V E R T Y A S S E S S M E N T CHAPTER 4 Economic Choices by Socially Embedded Individuals: recent advances in economics research and their potential implications for informal finance in Tanzania Occupational choices and structural change are central issues within poverty and development discourses. The relationship between poverty and structural change to information and to markets, one must keep in mind that has been central to the debate about economic develop- access to such technology is by no means universal. As such ment in developing countries. The economic development it may, in theory, exacerbate rather than alleviate inequalities, of most European countries over the past few centuries was although a recent study suggests that the overall impact of accompanied by two major structural changes: the population fast internet on job creation and on average income in Africa became more urbanized, and the rural population became has so far been positive (Hjort and Poulsen, 2019). more involved in non-agricultural activities (Allen, 2000). Within a few generations, the diversity of possible occupations Occupational choice is at the heart of structural change. and income sources was expanded, and large swaths of the Ultimately, however, as was the case in historical Europe, the population chose to transition from farming to other types future of SSA will be determined by its inhabitants’ occupational of economic activities. While a qualitatively similar evolution choices. These will determine not only the supply of goods and can be observed over the past few decades in Sub-Saharan services, but also the overall distribution of risk in the economy, African (SSA) countries, this evolution is still far from having the land use patterns with its ensuing ecological consequences, brought these countries to par with developed ones (McMillan, the migration patterns with its implications for the evolution Margaret, and Kenneth Harttgen, 2015; Davis, Di Giuseppe, of urbanization rates and other social dynamics, as well as the and Zezza, 2017). Furthermore, it is important to keep in mind long-term returns to human capital. A thorough and com- that structural change does not necessarily reduce economic prehensive understanding of the determinants as well as the inequality, as has been shown in work on historical data from effects of occupational choices is thus key to achieve effective Europe and the Americas (Lindert and Williamson, 2017). and desirable policy design. This chapter makes an attempt at summarizing some important issues which are still not well It is important to consider the global environment in which understood. The next section exposes the general arguments. structural change in Sub-Saharan Africa takes place today. It will be followed by an examination of the situation in Tanzania. Now, it may be erroneous to believe that the evolution in SSA must necessarily follow the exact same path as Europe to reach similar levels of economic development (McMillan, Rodrik, and Sepúlveda, 2015). The contexts are simply too dif- ferent, for many reasons, of which we here name three. First, while in historical Europe trade in agricultural produce was largely inexistent, for the past 20 years SSA’s share in global exports was higher for agricultural products than for goods in services in general (Badiane, Odjo, and Collins, 2018). Second, today’s overall economic landscape cannot be compared to the one in historical Europe, in which large multi-national play- ers in the manufacturing and financial sectors were essentially inexistent. Third, while modern technology facilitates access 50 TA N Z A N I A M A I N L A N D P O V E R T Y A S S E S S M E N T I.  What do We Know About Occupational Choices? What are the drivers of an individual’s occupational by economists to approach this question, and then examines choice? This section summarizes the traditional model used complementary approaches. A. The Classical Model of Occupational Choice in Economics Is the lack of access to credit the real problem? The early economics literature on occupational choice a combination of (1) the presence of predatory moneylenders, focused on the lack of access to finance as a source and (2) a lack of incentives for individuals to pay back their of inefficient occupational choice. The most prominent loans unless lenders could access information about borrow- classical models used to analyze the determinants of occu- ers’ ability to reimburse. Micro-finance was believed to solve pational choice typically share two features.1 First, focus is these two issues: interest rates would be much lower than on the choice between a small number of options, which those charged by the traditional moneylenders, and credit differ qualitatively in certain economic dimensions. Second, extended to groups rather than to individuals was expected individuals are taken to care solely about the consumption to provide incentives for individuals to reimburse (since the benefits (utility) that their net income confers on them. In group members could monitor each other’s activities). the influential model proposed by Banerjee and Newman (1993), there are three occupational statuses (that require the Traditional micro-credit institutions did not prove to be individual to supply labor): (i) self-employed; (ii) entrepreneur as efficient as expected, but recent research has unveiled with employees; (iii) employee. All categories may appear in potential for improvement. In its early phases, micro- any sector of the economy, including the agricultural one. The credit failed to deliver the improvements it was expected occupational categories differ in the inputs that they require: to generate; while it increased the consumption of durable capital is required for (i) and (ii) but not for (iii), and more cap- goods and had a positive impact on business creation, ital is required for (ii) than for (i). The set of choices accessible these effects were often modest, and they did not lead to an to an individual thus depends on his or her access to funds overall increase in consumption levels (Banerjee, 2013). Some (either through savings and/or through credit). The focus of recent experimental research, however, suggests that certain the analysis is on how constrained access to credit affects modifications in the design of micro-credit institutions may occupational choice and the ensuing wealth dynamics. turn out to significantly improve their economic impact. For example, in a field experiment in India, Maitra et al. (2017) find The early economics literature on micro-finance was highly that when local traders (typically the owner of a local shop) optimistic about the potential of micro-credit institu- were given the opportunity to have a say in which individuals tions. The belief that restricted access to finance is a major could benefit from a micro-finance loan (what they call cause behind the lack of structural change, induced massive trader-agent-intermediated lending, or TRAIL), the perfor- attempts to enhance access to credit by way of setting up mance of the micro-credit was enhanced. In the villages they micro-credit institutions. In particular, economists believed study, they find that, compared to traditional group-based that the lack of access to credit was mainly explained by lending, TRAIL enhanced the incomes of farmers on average 1 The standard economics model of individual choice is straightforward: among the options available in the individual’s feasible set, the individual selects the one ranking highest according to his or her personal preferences. C h a p t e r 4 E C O N O M I C C H O I C E S B Y S O C I A L LY E M B EDDED I N D I V I D U A L S 51 by 22 percent. There is nonetheless still widespread belief are correlated with the goal of income maximization. Neglect that the determinants of the performance of credit markets in of these other dimensions then leads to biased results, even if developing societies are far from being well understood, and the scientist or the policymaker is solely interested in under- this issue will be examined in greater detail below. standing how individuals go about to reach income goals. This neglect may turn out to be particularly troublesome By relying on a simplistic view of human motivation, the when it comes to understanding the early phases of economic classical model of economic choice fails to adequately development, which are complex and associated with import- inform policy design. The traditional analysis may still miss ant structural and societal changes, as mentioned above.2 key driving forces behind the choices an individual makes. The standard approach in economics, which consists in view- The theory of human behavior is increasingly influencing ing individuals as simply seeking to maximize own income, the modeling of individual decision-making. In the past is at odds with the way any other scientific discipline views the decades, the economics discipline has increasingly relied on a drivers of human behavior. If the goal is to distill the effects theory of human behavior which includes human motivational of income maximization from those of other motivations, factors beyond income maximization. The following discus- one can defend the modeling approach of asking: how does sion summarizes some of the contributions to this literature, income maximization affect occupational choice (and more with a focus on motivational factors which are thought to be generally behavior). However, approach can be problematic of particular interest for developing economies. This summary if behavior induced by other dimensions of human motivation is by no means exhaustive, however. B. Occupational Choice Under a More Complex Set of Motivations Human beings belong to social networks. The classical eco- minority of the population is employed by some formal institu- nomics model assumes that individual operates independently tion, such as a private firm or a public agency. Ties with social from others. Specifically, both the income (or more generally networks can be particularly strong among population groups the set of options an individual has access to) and the personal exposed to risks and vulnerability, and where formal insurance preferences of any given individual are presumed to depend mechanisms and strong institutions are lacking. Membership in only on his or her own actions. But, in essence, all human such networks may have a non-trivial impact on: (i) the options beings belong to one or several social networks, including to which an individual has access; and (ii) his or her goals. Taken the family, networks based on friendship or physical proximity, together, these two effects imply that membership in such perhaps a religious community and so forth (Jackson, 2019). networks may have a significant impact on individual behavior. Arguably, this is particularly true in societies where only a The two effects are examined in turn. Membership in social networks: Do informal transfers matter for occupational choices? Social networks are important sources of informal risk many ways. For example, the set of job opportunities can sharing and informal credit: this may in turn impact depend on the individual’s network contacts (Granovetter, productive efforts. Membership in a network can potentially 1973, Lalanne and Seabright, 2011); more specifically, rela- affect the set of options that an individual has access to in tives or friends of individuals in remote villages, who live in 2 The traditional model of occupational choice described above can be criticized on other grounds as well. In particular two comments are in order. First, in most variants of such models it is assumed that individuals have access to complete and accurate information about the options that exist, and about the investment costs and income distributions associated with them. This is clearly a very strong assumption, even when applied to individuals in developed countries. Second, the classical model disregards the effects of malfunctioning institutions and inadequate infrastructure (Laffont, 1998) as well as attitudes towards institutions (Alesina and Giuliano, 2015). A useful version of the baseline model must accommodate these factors. If good enough data is available, such a model may serve as a basis for empirical analysis of individual occupational choice. 52 TA N Z A N I A M A I N L A N D P O V E R T Y A S S E S S M E N T cities, may open doors to urban occupations. However, and et al. (2016) show that individuals who are on the receiving arguably more relevant in the context of developing coun- end of transfers (from emigrated relatives in the former study, tries, networks are a source of risk sharing (Rosenzweig, 1988, from older siblings in the latter study) work less compared Udry, 1990, Townsend, 1994), and more generally of infor- to those who are not. Di Falco and Bulte (2013) find a nega- mal transfers (Cox, Galasso and Jimenez, 2006, Fafchamps, tive correlation between the presence of relatives and levels 2011). The anticipation of being on either the receiving or of self-protection. A number of empirical and experimental the giving end of such transfers may impact an individual’s studies3 suggest that individuals undertake strategies to avoid incentives to invest or undertake productive efforts, or to having to share with kin (i.e., family, relatives).4 Finally, some make a particular occupational choice. It may seem intuitive empirical studies5 indicate that kin taxes may be responsible for for economists to hypothesize that informal transfers that a substantial number of “missing entrepreneurs”, i.e., individ- reduce the risk to which an individual is exposed, necessarily uals who due to the kin taxes choose to refrain from engaging induce lower efforts, as in the classical moral hazard model in entrepreneurial activities. In sum, this research underlines (Arrow, 1963, Pauly, 1968, Arnott and Stiglitz, 1988); indeed, the need for a better understanding of the effects of informal intuition suggests that someone who expects to be fully or transfers in networks of family, friends and neighbors. almost fully compensated in case of a loss, has a low incentive to self-protect against the loss in the first place. However, if There are costs and benefits of leaving an informal net- informal transfers are driven by an intrinsic desire to share, work; these are not yet well understood. Social networks such as altruism or warm glow, individuals may be expected are not fixed, and they depend on the individuals’ decisions to internalize the external effects of their efforts on others; for to participate in such networks (Barr and Genicot, 2008, example, an individual who is happy to help out his relatives, Attanasio et al., 2012, Banerjee et al., 2013). Although the or who derives social status from doing so, may work more factors that influence these decisions are not yet well under- than he/she would absent such helping behavior. Recent stood, there is now a growing body of research on this. In this theoretical research shows that the moral hazard effect may respect, the contributions by Kranton (1996) and Banerjee be outweighed by this positive incentive effect, resulting in et al. (2018) offer interesting insights. In Kranton’s (1996) the- a net positive effect of informal transfers on productive efforts oretical model, individuals face a choice between remaining (Alger and Weibull, 2008, 2010). in an informal network with reciprocal and personalized exchange of goods and services and leaving the network Omitting the potential effects of informal transfers in deci- to trade goods and services anonymously in a market. It is sion making can be misleading. Whether the effect turns out shown that the informal network can be sustained even if to be positive or negative in practice, the key point is that any everybody would be better off if they switched to market relationship between informal transfers and efforts implies exchange. The intuition is straightforward: when markets are that the classical model of individualistic decision-making thin, then an individual who leaves the reciprocal network has is misleading. Thus, if transfers have a detrimental effect on to spend a lot of time searching for the goods (s)he wants. In effort, they may hamper economic development, even though other words, the cost of leaving the reciprocal network is too they enable risk sharing; if they have a positive effect on high for any given individual. However, if everybody moved effort, they help promote economic development above and to market exchange, then it would be easy to find goods, beyond the benefits of risk sharing. and specialization would imply that more goods would be available than in the reciprocal network. Perhaps this theory Recent empirical and experimental work conducted in some may help explain occupational choice in some developing African countries suggests that “kin taxes” impact produc- economies, and the ensuing dynamics of structural change. tive efforts and/or savings patterns. Studies based on data collected in certain parts of Africa suggest that the expectation Data on the endogenous formation of networks, and that one should share income with kin (i.e., the expectation that on how this process depends on economic factors, one should pay a “kin tax”) on average tends to have either is necessary for a better understanding of their effects no effect or a detrimental effect on productive efforts, invest- on the decision-making process.6 The data collected and ments, and/or savings. Azam and Gubert (2005) and Baland analyzed by Banerjee et al. (2018) provides one of the first 3 See Baland, Guirkinger, and Mali (2011), Di Falco and Bulte (2011), Dupas and Robinson (2013), Jakiela and Ozier (2016), and Boltz, Marazyan, and Villar (2016). 4 By contrast, a study in rural Mexico based on a novel experimental design, which allows the authors to compare real effort choices in the presence and the absence of trans- fers, finds that on average transfers have a positive or no effect on efforts chosen by the participants in the study (Alger et al., 2018). 5 See Grimm, Hartwig, and Lay (2016), Squires (2016), and Alby, Auriol, and Nguimkeu (2019). 6 See Jackson and Storms (2019) for a recent methodological contribution on how network structure can be inferred from behavioral data. C h a p t e r 4 E C O N O M I C C H O I C E S B Y S O C I A L LY E M B EDDED I N D I V I D U A L S 53 steps in this direction. 7 The authors exploit variation in the 2006, Alesina and Giuliano, 2010). Relatedly, some recent rolling out of microfinance in villages in India to compare the theoretical work suggests that the strength of family ties network dynamics in exposed villages to that in non-exposed may have evolved differently in different parts of the world villages. They find that exposure to micro-finance induces a in pre-industrial times (Alger and Weibull, 2010), and there significant reduction in the number of network links over time. is evidence that family structures as well as inheritance patterns still vary systematically between different parts There is mounting evidence that family structure and of the world (Todd, 2011). Taken together, these strands interactions vary across the world; the consequences of of literature further weaken the case for economic mod- this variation are poorly understood. In recent decades, els in which individuals operate independently of their economists have also increasingly recognized the role that social environment. the family plays in shaping individuals’ economic deci- sions, and ultimately economic growth (Becker, 1991, Greif, How do ties with social networks impact preferences? It is increasingly recognized that individuals care not only implications of such aspects of human motivation for about their absolute income but also about how their occupational choice in emerging economies? While this income compares to that of others. Seventy years ago, is ultimately an empirical question, theory indicates that Duesenberry (1949) proposed in his PhD thesis that indi- the implications can be significant. To wit, consider the viduals care not only about their own absolute income, but model proposed by Genicot and Ray (2017) on the long- also about their income relative to that of others. While this run dynamics of the wealth distribution in a society.9 In the relative income hypothesis did not make it into the classical model, individuals allocate income between consumption economics models (Frank, 2005), in recent decades similar and savings for their children. Each individual is driven by ideas have been taken into account in several strands of eco- a desire to achieve a balance between on the one hand, nomic research, including behavioral economics, research on utility derived from own consumption, and on the other incentives in organizations, and in more general accounts of hand, utility derived from seeing one’s child surpass some human behavior.8 Furthermore, the idea that identity matters aspirational goal. The model encompasses many kinds has also made its way into economics (Akerlof and Kranton, of aspirational goals, including the objective to climb the 2000), as has the notion that homophily, i.e., the tendency income ladder. The key point of the model is that while for individuals to interact preferentially with those who share realistic aspirations foster savings and therefore also an some characteristics with them, may influence labor market increase in wealth across generations, highly ambitious outcomes. For example, a tendency for individuals to interact aspirations render the cost of attaining them so high that preferentially with people who have the same occupational savings are discouraged. It is shown that, under quite gen- status as them, may exacerbate unemployment by rendering eral conditions, this can lead to income stagnation at the it more difficult for unemployed individuals to create social lower end of the income distribution and an ever-increasing ties with employed people (Bramoullé and Saint-Paul, 2010). widening of the income gap across generations. Concretely, Similarly, people from poor backgrounds may be at a disad- aspirations may have a negative impact on the productive vantage when it comes to building social ties with individuals efforts of individuals at the lower end of the income scale, from higher socio-economic status groups, and this may in by (1) either inducing a low level of aspiration in them (this turn put them at a disadvantage when it comes to acquiring would be the case if they simply seek to keep up with their information about employment opportunities. neighbors’ incomes, and housing is segregated), (2) or by inducing a feeling of discouragement (this would be An individual’s social environment may influence his the case if they try to aim too high). Ultimately, however, or her aspirations, and therefore also his or her deci- sions on savings and productive efforts. What are the 7 The data can be downloaded from https://dataverse.harvard.edu/dataset.xhtml?persistentId=hdl:1902.1/21538. 8 See for example Fehr and Schmidt, 1999); Bandiera, Barankay, and Rasul (2010), and Frank (1985). 9 See also Bénabou and Tirole (2006). 54 TA N Z A N I A M A I N L A N D P O V E R T Y A S S E S S M E N T whether aspirations have a positive or a negative impact is women changed, women’s autonomy increased, and fertility an empirical question. decreased). The presumed mechanism is that exposure to social norms and standards of living in other parts of the world More research on how participation in social networks through TV shows influences attitudes (La Ferrara, Chong, (as well as exposure to media and social media) affects and Duryea, 2012). Moreover, it is increasingly recognized individuals’ preferences and goals is necessary to achieve and documented that pro-social preferences vary between a better understanding of occupational choice in devel- different parts of the world (Falk et al., 2018). Together with oping countries. Empirical evidence shows that exposure the theoretical model described above, this evidence begs through television and social media to social milieus and the question of how preferences are shaped in the first place, behaviors that they are rarely (if ever) exposed to in real and how the evolution of preferences in a society contributes life, has an impact on behavior. For example, Jensen and to the long-run dynamics of behaviors such as savings, Oster (2009) show that the introduction of cable and sat- productive efforts, and occupational choice. ellite television had a positive impact on women’s status in rural India (attitudes towards domestic violence toward C h a p t e r 4 E C O N O M I C C H O I C E S B Y S O C I A L LY E M B EDDED I N D I V I D U A L S 55 II.  Structural Change in Tanzania: The Role of Household Enterprises Household enterprises offer a potential for structural novel ones. In particular, the lack of finance is perceived as change, but their role is limited by the lack of finance. a major issue by HE owners.11 Hence, the discussion in this As seen in the previous chapters, Tanzania’s population section centers on the lack of access to credit. Two aspects is increasingly relying on non-agricultural activities and revealed by the data pertaining to access to funding are Household Enterprises (HEs) as a source of income. HEs examined below, in light of the recent developments in can drive the process of structural transformation, and economic research summarized above. The overarching hence poverty reduction and development.10 Yet, these goal is to provide food for thought, that may ultimately enterprises appear to face a number of challenges, which help improve access to finance for current and prospective hamper the functioning of existing HEs and the creation of HE owners in Tanzania. HEs are often funded through informal credit from relatives, neighbors, and friends: what are the implications of this state of affairs for the efficiency of HEs? Survey data shows that the major part of the money productive efforts or save in the first place, compared to credit borrowed to fund HEs comes from informal sources. In from formal sources. In particular, little is known about the the absence of access to a formal banking sector, individuals implications of the fact that credit to HEs stem mostly from may resort to a number of alternative informal arrangements neighbors, friends, and relatives, although several questions to enhance their access to funds (see Chapter 6 in Part I and appear of high relevance. Chapter 3 in Part II of this report). In a survey conducted by Sánchez Puerta et al. (2018), the respondents report that more First, does the ease of access to informal credit from than 80 percent of their borrowed money comes from infor- neighbors, friends and relatives imply that debtors make mal sources, of which half comes from friends and neighbors, more or less effort to ensure prompt and full reimburse- and a quarter from relatives.12 These figures clearly mean ment, compared to if they had access to credit only from that friends, neighbors, and relatives are the most import- formal sources? An individual who borrows from a relative ant source of funding for HE owners. In light of the recent may be more or less prone to reimburse the loan than if literature on the effects of informal transfers (summarized (s) he had instead borrowed from a formal financial institution. in the previous section), this fact begs the question of how On the one hand, (s)he may care more for the relation (s)he such informal funding affects the incentives to undertake has with the relative, and will therefore make a greater effort to 10 Kweka and Fox (2011) report that 24 percent of households have a HE as a primary source of employment, and that another 42 percent of households have a HE as a sec- ondary source of employment; Morisset and Haji (2014) identify small non-farm businesses as one of three key sources of job creation in Tanzania. HEs are arguably a major channel of structural change, in Tanzania and other SSA countries (Fox and Sohnesen, 2012). 11 Recent data collected through surveys, interviews, and focus group discussions reveal that, on average, the lack of finance is perceived as representing the main challenge (Sánchez Puerta et al., 2018). Other challenges include weak markets, inadequate infrastructure, and, to a lesser extent, lack of skilled labor. 12 The survey gathered data from 7,400 households in Tanzania and included a specific section on non-farm household enterprises useful to create a profile of HEs and identify their constraints. 56 TA N Z A N I A M A I N L A N D P O V E R T Y A S S E S S M E N T reimburse, perhaps by making a greater effort at generating Fourth, does the high prevalence of informal credit from income. On the other hand, however, relatives may also be networks of relatives, friends, and neighbors imply that more forgiving than banks, and this may reduce the borrower’s HE owners feel the pressure to conduct business and/or effort to reimburse. On balance, it is ultimately an empirical given discounts within these networks? If so, what does this question whether individuals are better at reimbursing loans imply for the economic viability of such HEs? to relatives than to banks. Knowledge about this would help evaluate the full set of benefits and costs of promoting formal There is a clear need for more data on how formal credit markets. It would thus be relevant for policy design. and informal credit impact productive behaviors (e.g., productive efforts, investment decisions, hir- Second, do individuals who expect that some neighbor, ing decisions, and more generally the entrepreneurial friend, or relative, may be in need of credit in the future, process). The preceding discussion highlights an immense save more in order to be able to meet this need, or less in gap in the economics literature. So far, the literature on order to avoid having to lend? Answers to this question would informal transfers (summarized in the previous section) has shed light on whether this type of informal credit enhances focused primarily on the role that these transfers play in or reduces aggregate savings, compared to a situation where sharing risk, and the idea that these transfers are sometimes individuals would save only to meet own future needs. perceived as a “kin tax”. Yet informal transfers seem also to be a key source of informal credit. More research on how Third, does the participation in networks of friends, productive behaviors are affected by borrowing and lending neighbors, and relatives, in which borrowing and lending between relatives, friends, and neighbors, as opposed to takes place, affect an individual’s occupational choice? borrowing from formal financial institutions, is thus needed. In particular, does it affect the allocation of time between While such research needs to clarify some theoretical issues, farming and non-farming activities? Since these activities may ultimately it is necessary to examine whether large effects imply different levels and kinds of risk, intuition suggests that appear in the data. In particular, it would seem important there may be a link. Of particular relevance is the question of to understand whether individuals tend to distort their whether resorting to informal borrowing and lending induces productive behaviors, thereby underexploiting economic a different time allocation than would resorting to formal, and opportunities, when resorting to credit from relatives, thus impersonal, credit and savings channels. friends, and neighbors. HEs in Tanzania: Missing ROSCAs? Rotating Savings and Credit Associations (ROSCAs) are its participants to get access to the total amount of money (s) believed to provide an efficient way for individuals to save he contributes over the cycle of the ROSCA, before (s)he has in order to invest in small businesses. ROSCAs are informal contributed the full amount. groups in which individuals contribute a certain amount on a regular basis and the sum thus collected is handed to one of ROSCAs are believed to entail many other benefits, too. the participants. In the academic literature they represent one The community involved in any given ROSCA may also be of the most well-known informal savings arrangements to fund a valuable source of information (for example about prices, business creation (see the summary of the theoretical litera- business practices, or administrative rules), of social support, ture in Appendix F). and of informal insurance for its members.13 Some of the respondents’ answers in the study of Sánchez Puerta et al. ROSCAs are believed to make it possible for individuals (2018) confirm that ROSCAs indeed are perceived as deliv- who do not have access to the formal banking sector ering these benefits.14 Furthermore, ROSCAs may serve as to save amounts of money large enough for business a commitment device for individuals who suffer from a limited investments. As is well known, a ROSCA allows all but one of ability to commit to saving; indeed, the fact that the ROSCA 13 For example, see Udry (1990), Bouman (1995), and Fafchamps (2011). 14 Businesses, including HEs, benefit from information pertaining to market conditions, prices, suppliers, and regulation. While this suggests that the information benefit of ROSCAs to HE owners may be maximal if all the participants are active in the same branch, it is not clear whether competitors would be willing to share information with each other. C h a p t e r 4 E C O N O M I C C H O I C E S B Y S O C I A L LY E M B EDDED I N D I V I D U A L S 57 meets on a regular basis helps mitigate procrastination in overall contribution to the ROSCA help alleviate trust issues? savings behaviors.15 And, by extension, can this increase ROSCA participation (by spreading adequate and trustworthy information)? Indeed, In spite of these potential benefits, only a small minority as indicated above, there appears to be a gap between of HE owners in Tanzania participate in ROSCAs. Accord- individuals’ perception of the impact of ROSCAs on HEs, ing to the data collected by Sánchez Puerta et al. (2018), and the rate of participation in such ROSCAs. Among the only a minority of HE owners participate in ROSCAs: around respondents in the study by Sánchez Puerta et al. (2018), the 17 percent of respondents declared that they have partici- main reasons cited for not participating in ROSCAs include pated in credit and savings groups and 10 percent declared the fear of not being able to contribute on a regular basis, that they have participated in savings groups only. Instead, as the belief that the poor are not accepted into ROSCAs, lack noted above, more than 80 percent of their borrowed money of information, and lack of trust. Hence, there is some reason comes from informal sources, of which half comes from friends to believe that some ROSCAs are “missing” due to a lack of and neighbors, and a quarter from relatives. adequate conditions. Websites and/or mobile phone apps that allow individuals to sign up for ROSCAs online (like The rates of participation in ROSCAs are at odds with matontine in Francophone Africa) may be a valuable tool to how HE owners in Tanzania perceive ROSCAs. According overcome some of these issues. to Sánchez Puerta et al. (2018), around 80 percent of all the respondents, including those who do not currently participate Further empirical research may help identify ways in which in ROSCAs, believe that savings groups have or could have a the current ROSCA formation processes in Tanzania may positive impact on their businesses. These observations lead be improved. To sum up the preceding paragraphs, there to the following questions: is it possible to build on the cur- appears to be a serious mismatch between, on the one hand, rent functioning of ROSCAs in order to enhance their ability the actual rates of participation by HE owners in Tanzania in to channel informal credit, both at the intensive and at the ROSCAs, and on the other hand, the large number of benefits extensive margin? And what explains the fact that individuals thought to be entailed by ROSCAs as well as the perception in Tanzania resort more to borrowing and lending through rel- by HE owners in Tanzania that ROSCA participation would be atives, friends, and neighbors, rather than through ROSCAs?. beneficial. This mismatch strongly suggests that more data should be collected on why HE owners choose not to partici- On top of its ability to improve access to funds for pate in ROSCAs. In particular, is this choice related to the fact investment, ROSCAs are also believed to mitigate certain that HE owners typically obtain credit from relatives, friends, behavioral problems. ROSCAs may have untapped potential and neighbors? In other words, are HE owners reluctant to when it comes to responding to certain “behavioral” issues modify their behavior with respect to their social networks by (Datta and Mullainathan, 2014). One such issue is trust, a fac- instead turning to ROSCAs? If so, can the size of the eco- tor which is key for economic development (Algan and Cahuc, nomic consequences of this reluctance be estimated? Some 2014). Another potentially important issue is aspiration (Ray, research conducted by academics in other countries may 2006, Genicot and Ray, 2017). be relevant for such empirical analysis of the current ROSCA Can ROSCAs be instrumental when it comes to enhancing formation processing in Tanzania; in particular, Angelucci et trust and aspiration? In particular, is it possible to identify al. (2009) and Attanasio et al. (2012) study the formation of ways in which the widespread access to mobile phones in risk-sharing groups in Mexico and Colombia, respectively, Tanzania may be used to improve ROSCA design? For exam- while Banerjee et al. (2013) examine the diffusion of informa- ple, can a system whereby individuals may rate each other’s tion about microfinance in social networks in India. Concluding remarks The preceding sections have put forward some evidence Tanzania and the funding thereof, as well as a summary of regarding the prevalence of household enterprises in recent advances in economics that appears to be highly 15 See Ambec and Treich (2006). 58 TA N Z A N I A M A I N L A N D P O V E R T Y A S S E S S M E N T relevant for developing economies. Taken together, this Finally, while ROSCAs have caught the attention of many material leads to the conclusion that individuals’ occupational economists, they are certainly not the only semi-formal choices, sources of funding, and savings decisions, are likely financial institutions (Bouman, 1995). In particular, the the result of a set of complex economic and social factors. closely related but nonetheless qualitatively distinct Accu- The importance of these factors is expected to be particularly mulated Savings and Credit Associations (ASCRAs) may true in developing countries, where formal employment and merit more attention, especially in light of the evidence that access to formal credit is still the rule rather than the excep- it may be difficult for households to ensure adequate forms tion, and where informal networks instead dominate life. of savings (Dupas and Robinson, 2013). The ASCRA operates Hence, it is argued that policy-makers might in general bene- in a similar way to ROSCAs (i.e., members contribute to a fit from research that seeks to improve our understanding of common pool). However, instead of distributing the pot to the interactions between economic and social factors. one member at each meeting, the savings are accumulated, and members may request loans from the pot. Contributions In particular, it would be useful to empirically estimate the plus interest plus fines make up the pool of funds available. extent to which there are “missing ROSCAs” in Tanzania. An advantage of the ASCRA over the ROSCA is that it is Besides quantitative analyses, normative ones are also called more flexible: it typically does not require that individuals for. Such analyses could compare the implications of switching contribute on a regular basis, it can accommodate a more HE funding from its current main source, i.e., relatives, friends, heterogeneous set or members, and it is typically larger. and neighbors, to ROSCAs. Would such a switch entail an Furthermore, according to Bouman (1995), formalized increase in productive efforts, in savings, and in the way in ASRCAs can help people “extricate themselves from the which individuals allocate their time between agricultural oppressive influence of traditional village elders” (p 377). activities and HEs? Would it enable individuals to develop A disadvantage is that ASCRAs have higher overhead costs a greater sense of trust in individuals outside of their tradi- than ROSCAs, due to the paperwork that it entails. In a tional networks? Answers to these questions are necessary to similar vein, and as already mentioned, modifications to develop a constructive discussion about the improvements micro-credit institutions, such as Savings and Credit Coop- that modifications to the design of ROSCAs or the technology erative Organizations (SACCOs), may turn out to be fruitful used to implement ROSCAs, might bring about. (Maitra et al., 2017). C h a p t e r 4 E C O N O M I C C H O I C E S B Y S O C I A L LY E M B EDDED I N D I V I D U A L S 59 APPENDIX A Measuring Structural Transformation Definitions The term “structural transformation” is used extensively in and the variables Y, L, and K denote output, labor input, and the literature but definitions and measurements adopted capital input. Free labor mobility implies that in equilibrium often differ. The most commonly used definition refers to the wage should be equalized across sectors. The assumption reallocation of economic activity across three broad sectors of competitive labor markets implies that workers are paid (agriculture, manufacturing, and services) that accompanies the value of their marginal product and firms hire up to the the process of economic development. Three indicators are point where the marginal product of labor equal the wage. largely used to measure changes in economic activity at the Therefore, marginal value products are also equalized. It is sectoral level: shares of (a) employment, (b) value added, then possible to derive the ratio of output per worker in non- and (c) final consumption expenditure. This study focuses agricultural sectors to that in agriculture as: primarily on employment shares as the indicator of structural transformation for two reasons. First, structural transformation Yn is the main transition channel between growth and poverty Ln =1 Ya reduction in contexts characterized by high employment- La to-population ratios. Second, the employment measure of structural transformation is increasingly seen as preferable to those based on shares in value added and consumption In case this condition is not met, labor is misallocated expenditure because it does not suffer from the difficulty relative to the competitive model. For example, if the value of separating out changes in quantities from changes in of the ratio is above 1, then productivity in non-agricultural prices. This approach finds its roots in models of economic sectors is higher than in agriculture and this would be an development that understand changes in the allocation of incentive for workers to move from agriculture to non- labor across sectors as key to economic development (for agricultural sectors. The process is expected to continue example, Harris and Todaro (1970) and Lewis (1954)). The until the products are about the same in the two sectors. It is assumption is that a surplus of agricultural labor characterizes important to note that this condition does not depend on any most developing countries and movements of labor out of assumption about other factors markets. If the condition does agriculture into higher-productivity sectors are the main driver not hold, the explanation could be in measurement issues of growth due to increases in overall labor productivity. related to labor inputs (or value added) or some frictions that prevent the labor market from clearing. Assuming for now After defining structural change as changes in sectoral that the labor market is perfectly competitive, the first part of employment shares, the next important question concerns the study will focus on potential measurement problems in the definition of sector. The concept of structural transfor- labor inputs or value added. Total labor inputs can vary along mation can be applied to different levels of disaggregation. the extensive (participation in each sector) and intensive This analysis will primary look at structural transformation margin (conditional on participation, number of working across broad economic sectors, namely agriculture, industry, hours in each sector). Therefore, following McCullough (2017), and services, although additional breakdowns by detailed the productivity gap in output per worker between non- sectoral categories and by activity will be discussed. agricultural sectors and agriculture can be decomposed into a gap in productivity per hour worked (PGAP) and a gap in Next, what does theory imply for cross-sector productiv- employment levels (EGAP): ity gaps? Following Gollin, Lagakos and Waugh (2014), the neoclassical two-sector model envisages a Cobb-Douglas Yn Yn Hn * production function in agriculture and in non-agricultural Ln Hn Ln GAP = = = PGAP * EGAP sectors, assumes free labor mobility across sectors, and Ya Ya Ha * competitive labor markets. The production functions can be La Ha La characterized as follow: where Ln denotes the number of workers in non-agricultural Ya = AaLθ 1−θ θ 1−θ aK a   and Yn = AnLnK n sectors and Hn refers to the total annual hours worked by a worker in non-agricultural sectors. Therefore, the presence where θ indicates the labor share in production in each sector, of large gaps in productivity per worker across sectors a and n indicates agriculture and nonagricultural sectors, might not necessarily be a sign of labor misallocation, 62 TA N Z A N I A M A I N L A N D P O V E R T Y A S S E S S M E N T but rather of sizable differences in total labor inputs, as Typically, productivity gaps are measured using value- measured by working hours, provided to each sector. This added from national accounts. This study will focus primarily has important implications in terms of possible drivers of on measures based on micro-data, whereby the output is labor movements across sectors. In presence of differences measured in terms of total labor income, thus excluding in working hours across sectors with comparable hourly returns to capital which are instead captured in indicators that returns to labor, the driving force of labor reallocation might use value-added measures from national accounts. lie in the search for full employment rather than for higher returns per hour. Data sources To examine productivity gaps, labor productivity measures The first comprises gross total wages earned by wage and other key variables are constructed using micro data workers, including in-kind payments and gratuities. The from three rounds of the Tanzania National Panel Survey second group includes returns to operating a non-agricultural (2010/11 or NPS2, 2012/13 or NPS3, and 2014/15 or business that are measured using reported enterprise profits NPS4).1 NPSs are Living Standard Measurement Surveys – since they are considered more reliable relative to measures Integrated Surveys in Agriculture (LSMS-ISA), collected by the constructed as reported gross revenues minus costs.2 In National Bureau of Statistics with technical support from the addition, there are profits from family farms that are based World Bank. The surveys collect information on a wide range on net revenues and are constructed by the Rural Income of topics including agricultural production, non-farm income Generating Activities (RIGA) project and include the value of generating activities, consumption expenditures, and a wealth auto-consumed farm output.3 All nominal values are adjusted of other socioeconomic characteristics. for cost-of-living differences by using temporal and spatial price adjustments within rounds. All values are then deflated NPS data are used to derive two variables that are key to to 2015 prices using Consumer Price Index (CPI) data from the measurement of productivity gaps across sectors. First, World Development Indicators. one needs to construct total output by sector as captured by total labor income. Second, labor supply is measured along The second key data input is a measure of labor supply. both the extensive and intensive margin, i.e. in terms of NPSs capture individuals’ employment by using both a 7-day number of workers participating to each sector and number and 12-month recall period. However, the structure of the of working hours supplied to each sector in all jobs. Tanzania questionnaire changes over time, thus making comparability NPSs’ questionnaires allow to identify workers, working hours, across rounds challenging to some extent. The most and labor income for both wage workers and non-wage important change between NPS2 on the one hand and NPS3 workers, including own-account workers and contributing and NPS4 on the other hand lies in the employment roster, family workers that help out in the family business and/or on which captures information about individuals’ labor market the household-operated farm. participation. NPS3 and NPS4 questionnaires include an extended battery of questions aimed at capturing all types of Starting with output, labor income is composed of two main work carried out by respondents over the last 12 months as items: (i) wages and (ii) and profits of household-operated well as over the last 7 days and to identify main and secondary farms and/or non-agricultural household enterprises. activity in each case. 1 NPS1 is not used in this study due to important changes in the survey instrument relative to following rounds, which might compromise the interpretation of patterns observed over time. NPS1, for example, only elicits information about the main wage job, whereas the three following rounds allow workers to report their secondary wage activity. Based on the last 3 rounds, between 5 and 8 percent of workers aged 15+ reported to have a second wage job. Although this is apparently not a major restriction, it could affect the pace of structural change estimated by comparing NPS1 with other rounds. Precisely, the main fieldwork of the second round of the NPS started in October 2010 and finished in September 2011, with specialized tracking teams remaining in the field until November 2011. Similarly, the duration and timing of the field work for the third round of NPS was from October 2012 to November 2013. Field work for the fourth round started in October 2014 and lasted until January 2016. 2 Whenever, profits are not reported, they are substituted by gross revenues minus costs instead of losing information about the enterprise and its labor. 3 Income from fishing is not accounted for in the calculation of total income since the way the survey captures fishing production change significantly over time, with separate dedicated modules in NPS3 and NPS4 only. Also, income from livestock activities is excluded from household total income as the corresponding labor inputs information is not captured as precisely as it is done for other activities. Livestock income corresponds to about 7/10 percent of total household income. A p p e n d i x A M EA S U R I N G S TR U C T U RA L TRA N S F O R M AT I O N 63 Despite efforts to improve comparability across survey with missing hours is about 19.6 percent of the sampled rounds, such changes have sizeable implications in terms individuals aged 15+, categorized as employed, in 2010/11, of overall employment and secondary jobs number. First, 3.6 percent in 2012/13, and 7.4 percent in 2014/15. Those the employed population aged 15 and above increases from with missing earnings decline from about 17.8 percent in about 20,2 million in 2010/11 to 21.8 million in 2012/13 and 2010/11 to 11 percent in 2014/15. Patterns are quite differ- 22.3 million in 2014/15. This implies a jump in the employ- ent by whether the activity is a primary or secondary job and ment-to-population ratio between 2010/11 and 2012/13 whether it is a wage or a non-wage job. The share of missing as opposed to what is estimated for the following period. hours is quite high in 2010/11 for both primary and second- The employment-to-population ratio goes from 79 percent ary jobs in non-agricultural household enterprises. Missing in 2010/11 up to 83.9 percent in 2012/13 (+4.9 percentage values for both hours and earnings are particularly high for points) and 85.4 percent in 2014/15 (+1.5 percentage points secondary jobs. In the case of hours, the share of secondary with respect to the 2012/13 estimate) (Table A.1). Second, jobs observations with missing values is about 36 percent in the pattern of secondary jobs is quite remarkable and likely 2010/11, 13 percent in 2012/13 and 21 percent in 2014/15. In to be a statistical artifact. Since information on labor supply the case of earnings, the share is higher and declining from will not be restricted to primary sector of employment, cor- 45 to 25 percent between 2010/11 and 2014/15 (Table A.2). rectly estimating the number of secondary jobs is extremely important. The number of secondary jobs booms between Because of the issued described so far and in order the second and the third round from about 1.9 million (or to avoid interpreting changes estimated over time as 7 percent of all jobs) in 2010/11 to 4.2 million (or 14 percent of structural changes of the economy rather than a mere all jobs) in 2012/13 and 4.9 million (or 16 percent of all jobs) in statistical artifact, the analysis of labor input, measured 2014/15 (Table A.1).4 Such humongous increase in the share of in terms of number of workers and number of working secondary jobs has a large impact of the sectoral distribution hours across all jobs, as well as of productivity gaps will of employment in secondary jobs: the share of agricultural be restricted to the last survey round. Yet, the first part of secondary jobs declines from over 50 percent in 2010/11 to the analysis will illustrate patterns of structural transformation 36 percent in 2012/13 when commerce explodes and reaches exploiting information from all the last three NPS rounds. 32 percent compared with 8 percent 2 years earlier. Labor supply variables and labor income will then be used In addition, patterns of missing values in annual working to construct measures of average labor productivity at hours and earnings variables show considerable varia- sectoral level and on a per-worker and per-hour basis. tion over time. In particular, the number of observations TABLE A.1: Number of Primary and Secondary Jobs by TABLE A.2: Number of Primary and Secondary Jobs by Survey Round (employed population aged 15+) Survey Round (employed population aged 15+) PRIMARY JOBS SECONDARY JOBS ALL JOBS SHARE SECONDARY JOBS MISSING HOURS MISSING EARNINGS 2011 20,246,261 1,914,745 27,535,654 9.5% PRIMARY JOB SECONDARY JOB PRIMARY JOB SECONDARY JOB 2013 21,812,977 4,219,831 30,205,567 19.3% 2011 18.0 36.0 15.2 44.9 2015 22,358,815 4,949,037 31,124,911 22.1% 2013 1.9 13.2 12.3 30.6 Source: Tanzania National Bureau of Statistics. 2015 4.7 21.2 8.2 24.7 Source: Tanzania National Bureau of Statistics. Measuring labor supply In order to measure labor supply in terms of number to each worker the number of hours worked in their of workers and total number of hours worked by each primary and secondary job. Starting with employment, individual in primary and secondary jobs, one needs to NPSs capture individuals’ employment by using both a 7-day first identify the employed population and then assign and 12-month recall period. However, the structure of the 4 In 2015, when the reference population is restricted to 25+ about 20 percent of workers have a secondary job. 64 TA N Z A N I A M A I N L A N D P O V E R T Y A S S E S S M E N T questionnaire changed over time, thus making comparability Individual annualized labor supply measures are across rounds challenging to some extent. The most constructed for wage, own-account, and contributing important change between NPS2 on the one hand and NPS3 family workers. Wage labor supply is generated with and NPS4 on the other hand lies in the employment roster, reference to a 12-month period using reported number which captures information about individuals’ labor market of months worked over the years preceding the interview, participation. NPS3 and NPS4 questionnaires include an typical number of weeks worked per month, and usual extended battery of questions aimed at capturing all types number of hours worked per week. Information about of work carried out by respondents over the last 12 months number of months and weeks worked is not available for as well as over the last 7 days and to identify main and secondary wage jobs in NPS4, so median values from secondary activity in each case. In NPS2, the main question NPS3 are used instead. In the case of non-agricultural reads as follows: “Did you do any work of any type for pay, own account and contributing family work, the number profit, barter or home use during the last 7 days?”. In NPS3 of annual working hours is calculated for all individuals, and NPS4 instead, employed is captured by a set of different whose employment roster and NFE module information is questions, one for each specific activity and including matched, by multiplying the number of months the business wage job, non-farm family business and help therein, family has been in operation over the past year times the total farm and help therein, and unpaid apprenticeships. Such number of working weeks in a month (4.33) and the number questions are asked first with reference to a 12-month of hours worked last 12 months (or last 7 days if the first recall period and then to a 7-day recall period. In addition, is missing) in the NFE from the employment module. In in NPS3 and NPS4, respondents classified as employed the detailed agricultural module, information about labor are allowed to list upfront the type of employment (wage participation and typical daily hours worked on the family work, own account work, contributing family work, unpaid farm is collected for each plot and separately for the long apprenticeship) they were engaged in during the last 7 days and short rainy season (the latter only in NPS3 and NPS4). and 12 months in both their primary and secondary activity. Agricultural labor inputs are then aggregated at individual This possibility was precluded to respondents in the first level assuming agricultural workers worked every day of the two rounds. long and short season an average of 10 working hours per day in order to generate the total number of annual hours Moreover, NPS2 allows household members to list a worked on the farm. NPS2 does not collect information maximum of 2 non-farm enterprises per household, regarding typical daily working hours on household farm NPS3 and NPS4 do not limit the number of enterprises in the agricultural model. For the sake of comparability, each household can report. For the sake of comparability number of weekly working hours on the household farm is over time, in NPS3 and NPS4 the number of non-farm estimated in NPS3 and NPS4 and used to construct annual enterprises considered in the analysis is restricted to working hours in all rounds. In NPS3 and NPS4, the median the first two NFE listed by each household. This is not a number of working hours spent in a typical working day on major departure from what is observed in the data as less the all crop-related activities is 10. As some workers report than 5 percent of sampled households that run non-farm to be farm workers in the employment roster and are not enterprises report more than two enterprises and the found in the agricultural module, their annual number of share of household income derived from these additional working hours on the farm has been imputed using a set enterprises is negligible. In NPS3 and NPS4, for each non- of non-missing average values estimated in the following farm enterprise, the person responsible of the business order: (i) mean value of annual working hours estimated is requested to list up to 6 household members working by gender and age group and primary job within the therein, whereas in NPS2 this information is not collected. household; (ii) if still missing, mean value of annual working For the sake of comparability, all individuals who report hours estimated by gender, age group, primary job, and to be own-account workers in the employment roster are region interacted with urban dummy; (iii) if still missing, matched with the identifier of the person in charge of the mean value of annual working hours estimated by gender, non-agricultural business in the NFE module. In addition, age group, and region interacted with urban dummy. all individuals reporting to be non-agricultural contributing family workers in the employment roster are assigned to the household main non-agricultural enterprise. A p p e n d i x A M EA S U R I N G S TR U C T U RA L TRA N S F O R M AT I O N 65 Occupational Choices Predicted incomes in each sector are fed into the second Methodology stage of the model, which is the alternative-specific mixed The alternative-specific logistic model is used to model logistic regression model described above, in order to the probability that an individual chooses one of several determine the probability of choosing to work in each unordered alternatives for sector of work. This model can sector for each individual. The dependent variable of the incorporate case-specific variables such as age and educational second stage is a binary choice outcome for agriculture, attainment as well as attributes that vary by the alternative – in industry, and services. The alternative-specific variable is the this case, projected income – in order to model a discrete choice. predicted income from the first stage of the regression and the case-specific variables are a vector of household and For the alternative-specific mixed logit model, the utility that individual characteristics. individual i receives from alternative a, denoted by Uia, is The base alternative in the model is the agricultural sector, Uia = xiabi + wiaa + zida + ¨ia a = 1,…,A to which all other results are compared. From this process, the individual probabilities for households can be computed bi are the set of random coefficients that vary over in an unbiased in efficient way. Probabilistic outcomes are not individuals in the population, and xia is the vector of the guaranteed to map to observed choices of sector of employ- alternative-specific variable for predicted income. α are ment, but rather provide a likelihood that an individual falls fixed coefficients on wia, a vector of the alternative-specific into one of the three categories based on their observed variable. δ are the alternative-specific coefficients on zi, which characteristics. is a vector of the case-specific variables. ¨ia is the random term. The probabilities of an individual selecting each discrete The average maximum probability of selecting any choice of sector are standard logistic probabilities integrated given sector is 0.665 with a standard deviation of 0.123. over the density f( b ). To determine if the model is producing reasonable esti- mates, it is expected that the highest predicted sectoral The first stage of the model involves predicting incomes outcome should match the observed result in approx- for each observation using case-specific individual and imately the same proportion as the average maximum household characteristics based on the actual data of indi- probability computed above. This analysis shows that viduals working in each sector. These case-specific variables the model is assigning the correct occupation, that is the are drawn from the methodology used in McCullough (2015). observed matches the predicted, in 76.5 percent of cases. Three models are generated with this strategy depending For well distributed probabilities across the sectors, the on the discrete choices of sector of employment for each model is effectively guessing at which occupation the individual – agriculture, industry, or services. household has. TABLE A.3: Monthly Income by Sector, 2014, TZS TABLE A.4: Productivity by Sector, 2014, TZS SECTOR MEAN MONTHLY MEDIAN MONTHLY SD N SECTOR MEAN PRODUCTIVITY (MONTHLY MEDIAN SD N INCOME (NET) INCOME (NET) INCOME PER HOURS WORKED) PRODUCTIVITY Agriculture 66848.7 0 548921.6 32436 Agriculture 510.5 0 2912 32436 Industry 272672.3 172500 451550.9 6813 Industry 1472.8 808.6 3137 6813 Services 313643.3 168000 591747.2 28758 Services 1516.5 735.2 3858.5 28758 Total 131981.8 28000 563816.9 68007 Total 785.4 135 3189.4 68007 Source: Integrated Labor Survey 2014. Source: Integrated Labor Survey 2014. 66 TA N Z A N I A M A I N L A N D P O V E R T Y A S S E S S M E N T TABLE A.5: Summary Statistics of Household and Individual Characteristics by Sector, 2014 SECTOR STATISTIC TOTAL MONTHLY COMPLETED AGE FEMALE RURAL ACCESS TO WEALTH HH RECEIVED EDU HH FEMALE HH HH INCOME (NET) PRIMARY CREDIT INDEX TRANSFER HEAD HH HEAD DEPENDENCY SIZE RATIO Agriculture mean 66848.7 0.8 31.7 0.5 0.8 0.0 0.1 0.1 2.6 0.2 0.5 6.1 n=32436 p50 0 – 28 – – – 0 – 3 – 0.5 6 Industry mean 272672.3 1.0 34.9 0.3 0.2 0.0 0.2 0.1 3.1 0.2 0.3 5.1 n=6813 p50 172500 – 33 – – – 0.19 – 3 – 0.36 4 Services mean 313643.3 1.0 35.3 0.5 0.2 0.1 0.2 0.2 3.3 0.3 0.4 4.6 n=28758 p50 168000 – 34 – – – 0.13 – 3 – 0.33 4 Total mean 131,981.8 0.8 32.7 0.5 0.6 0.0 0.1 0.1 2.8 0.2 0.4 5.7 n=68007 p50 28,000 – 30 – – – – – 3 – 0.5 5 Source: Integrated Labor Survey 2014. TABLE A.6: Alternative-Specific Mixed Logistic Regression for Sector Choice of Individuals OCCUPATIONAL CHOICE OF HOUSEHOLD COEFFICIENT STANDARD ERROR Z P>|Z| 95% CONFIDENCE INTERVAL Agriculture Base Alternative Industry Completed Primary Education 0.530 0.150 3.530 0.000 0.236 0.825 Age –0.007 0.002 –2.910 0.004 –0.011 –0.002 Female –0.869 0.072 –12.140 0.000 –1.009 –0.729 Rural –1.794 0.097 –18.520 0.000 –1.984 –1.604 Access to Credit in Past 12 Months 1.283 0.211 6.090 0.000 0.870 1.696 Wealth Index 2.765 0.215 12.880 0.000 2.344 3.186 Household Received Transfers –0.268 0.076 –3.530 0.000 –0.417 –0.119 Household Head Education 0.451 0.055 8.230 0.000 0.344 0.559 Household Female Head 0.315 0.079 3.980 0.000 0.160 0.470 Household Dependency Ratio –1.047 0.140 –7.460 0.000 –1.322 –0.772 Household Size –0.060 0.013 –4.640 0.000 –0.085 –0.034 [District Fixed Effects] Services Completed Primary Education 0.256 0.089 2.860 0.004 0.081 0.431 Age –0.006 0.002 –3.030 0.002 –0.009 –0.002 Female 0.367 0.050 7.380 0.000 0.269 0.464 Rural –2.090 0.060 –34.620 0.000 –2.208 –1.971 Access to Credit in Past 12 Months 2.143 0.169 12.680 0.000 1.812 2.474 Wealth Index 3.377 0.174 19.360 0.000 3.035 3.718 Household Received Transfers –0.084 0.058 –1.440 0.150 –0.198 0.030 Household Head Education 0.691 0.050 13.830 0.000 0.593 0.789 Household Female Head 0.271 0.055 4.930 0.000 0.163 0.379 Household Dependency Ratio –1.142 0.110 –10.420 0.000 –1.357 –0.928 Household Size –0.048 0.012 –4.030 0.000 –0.072 –0.025 [District Fixed Effects] Source: Integrated Labor Survey 2014. APPE N D I X A M EA S U R I N G S TR U C T U RA L TRA N S F O R M AT I O N 67 APPENDIX B Data and Descriptive Statistics of the 2013 Enterprise Survey 68 TA N Z A N I A M A I N L A N D P O V E R T Y A S S E S S M E N T The survey was administered to a representative sample of 50 workers or more represent less than five percent of all 813 firms in the non-agricultural formal private economy. firms (4.5 percent) but account for 43 percent of employment The universe of the survey is consistently defined and includes (Table B.2). Medium firms with 20 to 99 employees represent the entire manufacturing sector, the services sector, and 18 percent of firms and 31 percent of employment. Therefore, the transportation and construction sector. Public utilities, firms with 20 workers or more represent 20 percent of all firms government services, health care, and financial service sectors but account for around two thirds (66 percent) of employment. are not included in the universe. Besides firm’s characteristics This leaves less than one third of employment to micro and such as location, year firms started operations, legal status, small firms (firm with less than 20 employees) though they and sector of activities, the data encompasses detailed represent 80 percent of the firms in Tanzania. information on total labor cost, employment, intermediate inputs, obstacles to expansion, and production costs. The Young firms of less than 5 years account for 14 of survey also reported the size of firms in 2010 and 2013 all firms and concentrate 9 percent of employment. allowing the possibility to compute the net job creation as the Firms in the age group of [5-9] years represent more than difference of the two. one quarter of all firms and concentrate 24 percent of employment. The age group of [10-19] concentrate most On average, firms in Tanzania were aged 13 years in firms (41 percent) and most employment (36 percent). Firms 2013 and concentrated around 18 workers (Table B.1). of 20 years or more represent 17 percent of all firms, and However, 50 percent of firms are less than 12 years old and account only for 31 percent of jobs. have less than 8 workers, an indication of the high prevalence of micro and small firms. Labor cost including wages, bonuses The entry rate of firms between 2010 and 2013 is 3.5 and in-kind advantages to workers reaches on average percent representing the percentage of firms created 102.7 million TZS and the sales during a fiscal year 4,773.4 within that period. In 2013, single-worker firms, meaning million TZS. Around 10 percent of the firms export their the self-employed, account for approximately 5 percent of products and only three percent use technology license all firms and only 0.3 percent of employment. Those self- acquired from a foreign-owned company. Finally, two-third of employed firms have, on average, 11.6 years of existence and firms are located in Dar es Salam. 15.3 percent of them are new firms that started operations after 2010. Two-worker firms exhibit almost the same pattern The survey did not report the size and the age of some accounting for just 4.6 percent of all firms, concentrating firms. Firms for which the size is unknown are excluded in the 0.5 percent of employment, and averaging 11.6 years of analysis of the number of employment. Firms are categorized existence but no new entry. in four groups: micro firms which employ 1-4 workers, small firms with 5-19 workers, medium firms with 20-99 workers, and TABLE B.2: Firm Descriptive Statistics, 2013 large firms with 100 workers or more. FIRMS EMPLOYMENT AGE ENTRY RATES Overall, medium and large firms concentrate most jobs # OF FIRMS % OF FIRMS # JOBS % OF JOBS YEARS 2011–2013 but represent few companies. Most firms are concentrated Size category 1 752 5.5 752 0.3 11.6 15.3 in micro and small groups. Large firms with 100 employees 2 627 4.6 1,254 0.5 11.6 0.0 or more represent just 2 percent of the firms in Tanzania but [3–4] 940 6.9 3,231 1.3 14.3 4.8 account for 35 percent of employment. Moreover, firms with [5–9] 5,521 40.6 35,937 15.0 12.2 3.7 [10–19] 3,058 22.5 40,547 16.9 13.3 2.3 [20–49] 2,104 15.5 56,395 23.5 13.8 2.2 TABLE B.1: Firm Descriptive Statistics, 2013 [50–99] 325 2.4 19,055 7.9 17.4 0.0 MEAN MEDIAN OBSERVATIONS (#) [>=100] 283 2.1 83,010 34.6 20.4 0.0 Total 13,610 100 240,182 100 13.1 3.5 Age (year) 13.1 12 776 Source: Enterprise Survey 2013. Size (# workers) 17.7 8 792 Note: The statistics presented in this table are based on the 2013 enterprises Labor cost (million TZ shilling) 102.7 8 334 survey. For each firm, jobs are measured as the sum of all employment. The Sales (million TZ shilling) 4,773.4 35 512 number in the fourth column presents the aggregate total by firm size category. Age is measured as the average age of firm in the size category. The age of a firm Exporter (dummy) 0.11 792 is the difference between 2013 and the year of startup. Entry rates are measured Use of High-technology (dummy) 0.03 792 as the share of new firms created between 2011 and 2013 of the total number of Dar es Salam (dummy) 0.65 792 firms that are operational in 2013. Source: Enterprise Survey 2013. Note: Labor cost includes wages, bonuses and in-kind advantages to workers. A p p e n d i x B DATA A N D DE S C R I PT I V E S TAT I S T I C S O F THE 2 0 1 3 E N TERPR I S E S U R V E Y 69 Large firms (100+ employees) accounting for most jobs are concentrated in relatively old large firms. Young micro and tend to be the oldest with 20.4 years of existence on aver- small firms concentrate fewer jobs. This is the general pattern age and no new entry. also followed by medium firms where most of the jobs are concentrated in medium old firms. Apart from firms of one Table B.3 below presents the distribution of employment and two years, most jobs are concentrated in medium and by firm size and age in 2013. It documents that most jobs large firms regardless of the age groups. FIGURE B.1: Distribution of Firms and Jobs by Firms’ Size and Age           A. By firm size                         B. By firm age 80 50 41.4 63.0 40 36.4 60 30.8 30 28.1 23.6 40 34.6 31.8 31.4 20 17.0 13.6 17.0 17.9 20 9.1 10 2.2 2.1 0 0 Micro Small Medium Large [<5] [5–9] [10–19] [>=20] % Firms % Jobs % Firms % Jobs Source: Enterprise Survey 2013. Note: Micro firms employ 1–4 workers, small firms: 5-19, medium firms: 20-99, large firms: 100+. TABLE B.3: Employment by Firm Size and Age, 2013 FIRM SIZE CATEGORY 1 2 [3–4] [5–9] [10–19] [20–49] [50–99] [>=100] TOTAL SHARE Age (years) 1 0 0 141 861 529 202 0 0 1,732 0.8 2 110 0 22 347 564 863 0 0 1,907 0.8 3 41 159 106 931 1,027 0 246 1,072 3,581 1.6 4 150 99 177 1,935 1,269 6,400 2,355 1,263 13,648 6.0 [5–9] 80 408 454 10,328 13,976 13,844 1,157 13,714 53,961 23.6 [10–14] 158 302 1,263 8,434 8,672 10,992 3,052 8,308 41,181 18.0 [15–19] 89 86 526 6,540 7,742 9,134 7,649 10,234 42,000 18.4 >=20 89 200 384 4,880 6,197 14,830 4,300 39,331 70,211 30.8 Total 717 1,254 3,074 34,257 39,975 56,265 18,758 73,922 228,221 100 Share 0.3 0.5 1.3 15.0 17.5 24.7 8.2 32.4 100 Source: Enterprise Survey 2013. 70 TA N Z A N I A M A I N L A N D P O V E R T Y A S S E S S M E N T APPENDIX C Determinants of Job Creation Methodology The methodology to estimate job creation that uses Rijkers et al. (2014) assess the extent of observed rela- employment level at the end year at the dependent tionship between firm size and firm growth that is due to variable induces the attendant biases. Davis et al. (1996) firm size per se or to firm characteristics by progressively suggest using the average of the firm size between the start considering elaborate sets of explanatory variables. and the end year. Rijkers et al. (2014) state that this last meth- Inspired from Haltiwanger et al. (2013), they first include odology reduces the bias despite some limitations since firms size and age dummies separately and subsequently jointly. that traverse size classes are counted as having originated In this paper, we follow the same approach where both size in a size class that is an average of the starting and ending dummies based on year t and year t-1 are used to examine size class. The contribution of firms in size classes on either the impact of measurement error and the determinants of extreme of the size distribution is then likely to be underes- firm growth. Then, for 2010 and 2013, we investigate the timated. From Davis et al. (1996), aggregate measures of job impact of productivity (Y/L), and firm performance measured creation and destruction can be generated at any level of by two variables that are the utilization capacity of the firm aggregation by using appropriately employment weighted and hours worked per week. summations of firm-level employment growth, gist, that’s the change in employment from year t-1 (2010 for this paper) The general specification of the model takes the following to year t (2013 for this paper), divided by average size. For form: this paper, employment-weighted firm-level regressions of bS Size + bA Age + bp Productivity + bUC (Utilization gist =  net employment growth is estimated using the approach by Capacity) + bhw (Hours worked) + bττ + b1I + eit Rijkers et al. (2014). Eist − Eist −1 Where Size is the vector of size dummies, Age is the vector of gist = 2 (Eist + Eist −1) age dummies, t is the vector of time dummies, I is the vector of industry dummies, and Productivity, Utilization Capacity This is the dependent variable used in the regressions. and Hours are variables representing those concepts. Time It is the Davis-Haltiwanger-Schuh growth rate in which Eist and industry are control variables. Data are subject to mea- denotes employment in firm i of types s at year t. The vari- surement errors especially when estimating the hours worked able gist is symmetric bounded by -2 and 2. If a firm entered per week that may be affected by sporadic interruptions. the market after 2010, Eist-1=0 and gist=2. If a firm that existed Also, measurement of firm size can be affected by extreme in 2010 has disappeared in 2013, its size equals zero in 2013, values that induce another bias. To reduce those biases, we Eist=0, and gist=-2. Hence, gist accommodates both entry use the log of output per worker for productivity and rank for and exit (Rijkers et al., 2014). However, for entrant firms that hours worked. Those strategies help minimizing the impact of started operations after 2010, we choose the start-up year potential measurement errors (Rijkers et al., 2014). as the t-1 year. Results Tables C.1 and C.2 present the results of regressions The size of firms has a positive impact on job creation. of net job creation on firm-size and age dummies. Regression results of net job creation on firm size dummies Regressions have been executed for the year 2013, first for reveals that job creation by firms of 3-4 workers is 27.6 percent all firms and then for firms excluding new entrants – i.e., firms higher than job creation by one-person firms. The coefficient created between 2010 and 2013 (Table C.1). Omitted cate- estimates suggest that job creation is the highest among firm gories for firm size and firm age are respectively size=1 and employing between 50 and 99 workers: the net job creation age=0 or 1. The regression coefficients are then expressed rate in these firms is around 52 percent higher than that for relatively to these categories. one-person firms. 72 TA N Z A N I A M A I N L A N D P O V E R T Y A S S E S S M E N T TABLE C.1: Correlates of Job Creation TABLE C.2: Correlates of Job Creation 2010–2013 – 2010–2013 – Size and Age Productivity and Performance on All Firms ALL FIRMS WITHOUT NEW ENTRANTS (1) (2) (3) (4) (5) (1) (2) (3) (4) (5) (6) 0.014 0.020 0.023 Productivity Firms’ Size (omitted: 1 worker) (0.01) (0.01) (0.02) 0.131 0.131 0.172 0.172 0.002 0.001 0.001 2 Utilization capacity (0.13) (0.13) (0.13) (0.13) (0.00) (0.00) (0.00) –0.001 –0.001 –0.001 0.276** 0.276** 0.318** 0.318** Rank of hours worked [3–4] (0.00) (0.00) (0.00) (0.11) (0.11) (0.11) (0.11) Firms’ Size (omitted: 1 worker) 0.299** 0.299** 0.343*** 0.343*** [5–9] 0.353 0.279 0.000 (0.09) (0.09) (0.10) (0.10) 2 (0.18) (0.35) (.) 0.402*** 0.402*** 0.444*** 0.444*** [10–19] 0.497** 0.406 0.216 (0.10) (0.10) (0.10) (0.10) [3–4] (0.17) (0.33) (0.20) 0.299** 0.299** 0.341*** 0.341*** [20–49] 0.438** 0.240 0.036 (0.10) (0.10) (0.10) (0.10) [5–9] (0.15) (0.32) (0.16) 0.518*** 0.518*** 0.558*** 0.558*** 0.557*** 0.305 0.094 [50–99] [10–19] (0.11) (0.11) (0.11) (0.11) (0.15) (0.33) (0.19) 0.418*** 0.418*** 0.457*** 0.457*** 0.484** 0.282 0.143 [>=100] [20–49] (0.11) (0.11) (0.11) (0.11) (0.15) (0.32) (0.18) Firms’ Age (omitted: 1 year old / 3 years old without new entrants) 0.687*** 0.546 0.391* [50–99] –0.021 0.093 (0.16) (0.33) (0.18) 2 (0.23) (0.23) 0.541*** 0.419 0.214 [>=100] –1.900*** –1.862*** (0.16) (0.33) (0.19) 3 (0.21) (0.21) Firms’ Age (omitted: 1 year old) –1.792*** –1.775*** 0.107 0.084 –0.058 0.000 0.000 4 2 (0.20) (0.19) (0.14) (0.13) (0.49) (.) (.) –1.827*** –1.819*** 0.073 0.040 –1.867*** –1.946*** –2.106*** [5–9] 3 (0.20) (0.19) (0.13) (0.13) (0.44) (0.49) (0.51) –1.860*** –1.676*** –1.662*** –1.748*** –1.775*** 0.150 0.080 4 [10–14] (0.44) (0.43) (0.44) (0.20) (0.19) (0.13) (0.13) –1.751*** –1.985*** –1.820*** –1.934*** –1.964*** –0.034 –0.108 [5–9] [15–19] (0.44) (0.41) (0.41) (0.20) (0.19) (0.13) (0.13) –1.917*** –1.966*** –1.889*** –1.933*** –2.022*** –0.034 –0.165 [10–14] >=20 (0.44) (0.39) (0.38) (0.48) (0.48) (0.47) (0.46) –2.008*** –1.917*** –1.821*** Sector [15–19] (0.43) (0.39) (0.38) dummies Yes Yes Yes Yes –2.167*** –1.994*** –2.024*** Year >=20 (0.61) (0.54) (0.52) dummies Yes Yes Yes Yes Sector dummies Yes Yes Yes Yes Yes N 776 776 776 759 759 759 Year dummies Yes Yes Yes Yes Yes R2 0.365 0.333 0.365 0.137 0.092 0.137 N 507 250 507 250 204 Note: * Significant at the 10 percent level; ** significant at the 5 percent level; R2 0.242 0.394 0.281 0.431 0.501 *** significant at the 1 percent level. Note: * Significant at the 10 percent level; ** significant at the 5 percent level; *** significant at the 1 percent level. Age appears to significantly affect job creation capaci- (6) removes the statistical artefact as well as the significance ties of firms in a negative way. Results suggest that older of age. firms create less jobs. However, it reflects a statistical artefact stemming from the creation of the dependent variable (Gist) The results are not sensitive to the cut-offs of firms’ size. for which firms created after 2010 were attributed a maximal When considering large firms as those with 50 employees or value of 2. Removing new entrants as in models (4), (5), and more, and medium firms as those with 20 to 49 employees, A p p e n d i x C DETER M I N A N T S O F J O B C REAT I O N 73 results did not change. Change in the cut-offs increases the TABLE C.3: Correlates of Job Creation 2010–2013 – Sector weight of large firms. of Activity on All Firms (1) (2) The second regression model (Table C.2) attempts to Sector (omitted: Manufacturing) determine whether variables such as firms’ productivity 0.249 0.247 and performance have an impact on firms’ growth and Utilities & construction (0.15) (0.15) job creation. The performance of firms is measured with two −0.004 0.015 Trade variables namely (i) the utilization capacity and (ii) the number (0.04) (0.04) of hours worked. 0.003 0.009 Market services (0.04) (0.04) To minimize the impact of measurement error and mis- Size (omitted: 1 worker) reporting, productivity is measured as the log of output 2 0.160 per worker, and hours worked is replaced by its rank (0.12) 0.291** following Rijkers et al. 2014. These latter draw attention [3–4] (0.11) on potential endogeneity of the productivity and perfor- 0.319*** mance variables and suggest interpreting the regression [5–9] (0.09) coefficients as conditional correlations rather than causal 0.418*** relationships. [10–19] (0.10) 0.320*** Results from (1) and (2) show that firm productivity as well [20–49] (0.10) as their performance in 2013 do not influence job growth. 0.538*** [50–99] Controlling the regression on firms’ productivity for firms’ size (0.11) and age, size appear to have some significant positive impact 0.433*** [>=100] on firm growth, while firms’ age seems to be significantly (0.11) negatively correlated with firm’s growth – albeit reflecting a Age (omitted: 1 year old) 0.104 statistical artefact (See discussion above). 2 (0.23) The third regression model (Table C.3) attempts to −1.871*** 3 (0.21) determine whether firms in some sectors of activities are −1.767*** more dynamics than those in other sectors. Firms have 4 (0.19) been grouped in four sectors: Manufacturing which includes −1.824*** all manufacturing firms; Utilities & construction; Trade that [5–9] (0.19) includes firms working wholesale, retail, and services of −1.778*** [10–14] motor vehicles; and Market services that includes transport (0.19) and hotels, restaurants and other accommodation, and IT. −1.961*** [15–19] Manufacturing is taken as the omitted category. (0.19) −2.093*** >=20 Regressions, even after controlling for firm size, firm age and (0.48) year dummies show that no particular sector is particularly Sector dummies No No dynamic. Year dummies Yes Yes N 776 776 R2 0.326 0.358 Note: * Significant at the 10 percent level; ** significant at the 5 percent level; *** significant at the 1 percent level. 74 TA N Z A N I A M A I N L A N D P O V E R T Y A S S E S S M E N T APPENDIX D Determinants of Financial Inclusion The analysis of the determinants of financial inclu- is restricted to Tanzania and contains 1,008 individuals, of sion in Tanzania relies on the paper of Zins and Weill which 507 are women and 501 are men. (2016) and uses the 2017 World Bank’s Global Findex database. The database is obtained from surveys real- Following Zins and Weill, the analysis uses probit models to ized in 143 countries and covering almost 150,000 persons examine the factors influencing financial inclusion, includ- worldwide. It provides a large number of indicators on ing formal finance, mobile money banking and informal financial inclusion enabling to assess the amount of account finance. More specifically, it explores the impact of individuals’ penetration, the use of financial services, the purposes and socio-­economic and demographic characteristics – gender, motivations, the alternatives to formal finance, etc. It also age, income, education, employment, geographic location – on provides micro-level information – gender, age, income ­ ­nclusion. Financial inclusion is measured using three financial i and education – that is used in our estimations. The sample main ­indicators: formal account, formal saving and formal credit. Financial inclusion Indicators: Dependent variables − Formal account refers to the fact that the individual − Formal credit refers to the fact that the individual has an account either at a financial institution or borrowed from a financial institution in the past through a mobile money provider. 12 months. − Formal saving refers to the fact that the individual All these variables are dummies equal to one if the person saved using an account at a financial institution in the responded “yes” and zero elsewise. past 12 months. Determinants of financial inclusion: Explanatory variables Explanatory variables include the individual characteristics: the omitted dummy variable. Two dummy variables are used for Gender is a dummy variable equal to one if the individual education: Secondary education, equal to one if the individual is woman (Female) and zero else. Age is represented with has completed secondary education, and Tertiary education, the number of years (Age) and its squared (Age2) in order equal to one if the individual has completed tertiary education to control for a possible nonlinear relation between age and or more. The omitted dummy variable is primary school or less. financial inclusion. Four dummy variables are used to take income into household income quintile (poorest 20%, second Table D.1 shows the results and the marginal effects of 20%, third 20%, fourth 20% and richest 20%). Each quintile is a the probit estimations for the main indicators of financial dummy variable equal to one if the individual/household is in ­inclusion. Tables D.2 to D.5 further examine the effect of this income quintile, zero elsewise. The fifth richest quintile is individual characteristics on saving and credit behavior. ­ 76 TA N Z A N I A M A I N L A N D P O V E R T Y A S S E S S M E N T TABLE D.1: Determinants of the main Financial TABLE D.2: Determinants of Saving (Formal and Informal) Inclusion indicators FORMAL SAVING INFORMAL SAVING SAVED ANY MONEY IN FORMAL ACCOUNT FORMAL SAVING FORMAL CREDIT THE PAST 12 MONTHS Female –0.181 0.105 0.074 Female 0.105 0.167 –0.106 (0.099) (0.128) (0.148) (0.128) (0.118) (0.101) Age 0.044** 0.060* 0.061** Age 0.060* 0.066*** 0.055*** (0.017) (0.024) (0.022) (0.024) (0.016) (0.014) Age squared –0.000* –0.001* –0.001* Age squared –0.001* –0.001*** –0.001*** (0.000) (0.000) (0.000) (0.000) (0.000) (0.000) Income - poorest 20% –0.957*** –0.795** –0.327 Income - poorest 20% –0.795** 0.139 –0.292 (0.168) (0.243) (0.244) (0.243) (0.190) (0.165) Income - second 20% –0.835*** –0.732*** –0.800*** Income - second 20% –0.732*** –0.181 –0.222 (0.156) (0.222) (0.223) (0.222) (0.180) (0.163) Income - third 20% –0.591*** –0.893*** –0.182 Income - third 20% –0.893*** 0.095 –0.014 (0.148) (0.192) (0.197) (0.192) (0.172) (0.156) Income - fourth 20% –0.293 –0.423* –0.688** Income - fourth 20% –0.423* –0.223 –0.202 (0.150) (0.180) (0.220) (0.180) (0.177) (0.155) Secondary education 0.771*** 0.862*** 0.215 Secondary education 0.862*** –0.075 0.330** (0.120) (0.150) (0.162) (0.150) (0.140) (0.122) Tertiary education 1.765*** 1.111*** 0.645 Tertiary education 1.111*** –0.217 1.018** (0.496) (0.313) (0.356) (0.313) (0.412) (0.379) Observations 1006 992 997 Observations 992 997 1006 Pseudo R 2 0.136 0.182 0.076 Pseudo R2 0.182 0.045 0.042 Log likelihood –584.585 –248.103 –223.273 Log likelihood –248.103 –371.537 –651.708 Standard errors in parentheses, * p < 0.05, ** p < 0.01, *** p < 0.001 Standard errors in parentheses, * p < 0.05, ** p < 0.01, *** p < 0.001 Source: Findex 2017. Source: Findex 2017. TABLE D.3: Determinants of Saving Motivation TABLE D.4: Determinants of Loan-taking Motivation FOR FARM OR BUSINESS FOR OLD AGE FOR EDUCATION (Formal and Informal) Female –0.221* –0.056 0.011 FOR EDUCATION FOR MEDICAL FOR FARM OR TO PURCHASE A (0.101) (0.136) (0.106) PURPOSES BUSINESS HOME OR LAND Age Female 0.017 –0.083 –0.236* 0.211 0.049** 0.043* 0.041** (0.115) (0.101) (0.105) (0.148) (0.015) (0.019) (0.015) Age 0.031 0.039** 0.066*** 0.032 Age squared –0.001** –0.000 –0.000* (0.017) (0.015) (0.017) (0.032) (0.000) (0.000) (0.000) Age squared –0.000 –0.000* –0.001*** –0.000 Income - poorest 20% –0.671*** –0.525* –0.141 (0.000) (0.000) (0.000) (0.000) (0.167) (0.221) (0.168) Income - poorest 20% 0.243 0.709*** –0.531** –0.661** Income - second 20% –0.319* –0.212 –0.180 (0.186) (0.160) (0.170) (0.251) (0.156) (0.212) (0.166) Income - second 20% –0.217 0.042 –0.477** –0.711** Income - third 20% –0.314* –0.331 0.049 (0.183) (0.153) (0.162) (0.253) (0.145) (0.205) (0.148) Income - third 20% 0.014 0.254 –0.249 –0.601** Income - fourth 20% –0.150 –0.309 0.058 (0.177) (0.146) (0.152) (0.210) (0.144) (0.194) (0.153) Income - fourth 20% –0.031 0.047 –0.351* –0.449 Secondary education (0.181) (0.154) (0.159) (0.309) 0.197 0.662*** 0.310** (0.114) (0.159) (0.119) Secondary education 0.444*** 0.222 0.009 0.584** (0.134) (0.119) (0.123) (0.194) Tertiary education 0.876 * 1.706 *** 0.339 Tertiary education 0.170 0.400 –0.214 1.287*** (0.351) (0.325) (0.326) (0.397) (0.364) (0.325) (0.352) Observations 1004 997 1000 Observations 1000 1000 1002 999 Pseudo R 2 0.058 0.123 0.025 Pseudo R2 0.028 0.041 0.052 0.130 Log likelihood –599.405 –228.194 –526.741 Log likelihood –448.187 –561.838 –505.868 –161.045 Standard errors in parentheses, * p < 0.05, ** p < 0.01, *** p < 0.001 Standard errors in parentheses, * p < 0.05, ** p < 0.01, *** p < 0.001 Source: Findex 2017. Source: Findex 2017. A p p e n d i x D D e t e r min a n t s of F in a nci a l I nclusion 77 TABLE D.5: Determinants of Sources of Borrowing A STORE FAMILY AND FRIENDS ANOTHER PRIVATE LENDER INFORMAL CREDIT FORMAL CREDIT ALL SOURCES Female –0.463 –0.133 0.016 –0.115 0.074 –0.029 (0.382) (0.097) (0.148) (0.097) (0.148) (0.098) Age 0.038 0.015 0.018 0.022 0.061** 0.027 (0.042) (0.014) (0.023) (0.014) (0.022) (0.014) Age squared –0.000 –0.000 –0.000 –0.000 –0.001* –0.000* (0.000) (0.000) (0.000) (0.000) (0.000) (0.000) Income - poorest 20% 1.055* 0.248 0.074 0.285 –0.327 0.234 (0.504) (0.156) (0.225) (0.155) (0.244) (0.161) Income - second 20% 0.593 0.198 –0.214 0.216 –0.800*** 0.078 (0.519) (0.150) (0.214) (0.150) (0.223) (0.154) Income - third 20% 0.736 0.083 –0.440* 0.040 –0.182 –0.016 (0.469) (0.143) (0.213) (0.142) (0.197) (0.147) Income - fourth 20% 0.716 –0.079 –0.038 –0.111 –0.688** –0.220 (0.462) (0.143) (0.221) (0.142) (0.220) (0.146) Secondary education 0.775** 0.212 –0.002 0.249* 0.215 0.325** (0.242) (0.113) (0.169) (0.113) (0.162) (0.116) Tertiary education 0.000 0.285 0.000 0.235 0.645 0.545 (.) (0.332) (.) (0.332) (0.356) (0.316) Observations 980 1003 979 1006 997 1006 Pseudo R2 0.108 0.011 0.018 0.015 0.076 0.020 Log likelihood –32.299 –667.047 –237.833 –673.824 –223.273 –677.711 Standard errors in parentheses, * p < 0.05, ** p < 0.01, *** p < 0.001 Source: Findex 2017. 78 TA N Z A N I A M A I N L A N D P O V E R T Y A S S E S S M E N T APPENDIX E Access to Finance and Tanzania’s Firms Growth We conduct an empirical investigation of the effects is a dummy variable with a value of 1 if the firm has a loan or of access to finance on the growth of firms in Tanzania creditline, and 0 otherwise; the second is Overdraft, which is using data from the 2006 and 2013 Enterprise Surveys a dummy variable with a value of 1 if the firm has an overdraft and following the econometric approach of Fowowe facility and 0 otherwise. (2017). Following Fowowe, firm growth is calculated as the log difference between the current number of permanent Table E.1 presents mean values of the access to finance employees and the number of permanent employees three constraint and objective access to finance variables based years before the survey year, divided by the difference on the 2013 Enterprise Survey. We observe interesting between the survey years: variability across firms based on their age and size. The access to finance constraint is lower among large firms. FGit = [(lnSit − lnSi,t−3)/3], where FGit is firm growth and Sit It is interesting to note that both subjective and objective is firm size, represented by permanent employment. measures suggest that larger firms (and to a lesser extent older ones) face less severe constraints in accessing finance Enterprise surveys also contain data on sales of firms, but than micro and small ones (and younger firms). they were not considered as a measure of growth due to their high volatility and measurement biases. We employ TABLE E.1: Access to Finance Constraint and Participation both subjective and objective measures of access to finance in Financial Market (mean) as explanatory variables. The subjective measure of access to ACCESS TO FINANCE CREDIT CONSTRAINED OVERDRAFT CREDITLINE finance is obtained from the ranking of access to finance as no CONSTRAINT STATUS (CCS) obstacle or severe obstacle to business operations. Although Tanzania 3.20 2.31 0.06 0.15 Age subjective measures offer useful insight into the severity   <5 years 3.23 2.20 0.11 0.15 of access to finance constraints and business environment   6 to 15 years 3.19 2.39 0.05 0.16 obstacles, they represent firm perceptions of the business   > 16 years 3.32 2.39 0.09 0.17 environment, which could reflect idiosyncratic differences in Size the degree of optimism or pessimism of the respondents.  Micro 3.20 2.29 0.03 0.13 Answers could also be influenced by the experience and  Small 3.24 2.31 0.10 0.18 performance of the firm. To address these limitations and  Medium 3.33 2.71 0.20 0.35 ensure the robustness of the results, objective measures are  Large 2.58 2.94 0.28 0.63 used in addition to subjective ones to examine how access to Source: Enterprise Survey 2013. Note: finance affects firm performance. The objective measure of - Access to Finance Constraint is a subjective evaluation of the severity access to finance is a variable which measures whether firms of obstacles that firms face in accessing finance on a scale of 1–5 (1 = no obstacle; 5 = very severe obstacle). are constrained in obtaining credit or not. - Credit Constrained Status is an ordinal variable that ranges from 1 to 4 (1 = FCC, 2 = PCC , 3 = MCC, and 4 = NCC). Three objective measures were created: The first mea- - Overdraft and Creditline are dummy variables (0 and 1). sure uses the definition of credit constrained status to The regression model controls for general business construct four groups based on the extent to which conditions, firm characteristics, and regional controls. firms were credit constrained during the fiscal year of General business conditions are proxied by business survey. Following Fowowe, we create an ordinal variable regulatory conditions and corruption. Firms characteristics Credit Constrained Status(CCS) for which: 1 = Full Credit include information on size, age, and sector. Constrained (FCC), for firms that applied for a loan and were rejected and do not have any type of external finance; We estimate models where employment growth is the 2 = Partially Credit Constrained (PCC), which includes firms dependent variable and variables measuring access to that applied for a loan and were rejected but that managed finance constraints and participation in financial markets to find some other forms of external finance; 3 = Maybe are the primary explanatory variables. Additional variables Credit Constrained (MCC), for firms that have had access to measuring business conditions, corruption, firm characteristics, external finance and there is evidence of them having bank and regional dummies are included as control variables. The finance; 4 = Non Credit Constrained (NCC), for firms that did models were estimated using the panel Enterprise survey not apply for a loan during the previous year because they for 2006 and 2013 and using the 2013 dataset only. However, have enough capital for the firm’s needs. Thus, higher values the limited quality of the panel survey data undermined the of CCS denote higher values of access to finance. Objective robustness of the results. Estimation results using the 2013 measures include two additional variables: Creditline, which dataset are summarized in Tables E.2 and E.3. 80 TA N Z A N I A M A I N L A N D P O V E R T Y A S S E S S M E N T TABLE E.2: Effect of Access to Finance and Business TABLE E.2: Effect of Access to Finance and Business Environment Constraints on Firms Growth, Subjective Environment Constraints on Firms Growth, Subjective Measures, 2013 Measures, 2013 (Continued) NO CONTROLS FIRM CONTROLS NO CONTROLS FIRM CONTROLS Access to Finance (omitted: no obstacle) Major obstacle 0.15 0.033 Minor obstacle 2.14** –2.09* (0.25) (0.26) (1.12) (1.13) Very severe obstacle 0.16 0.17 Moderate obstacle –2.8* –2.63* (0.57) (0.56) (1.5) (1.5) Access to land (omitted: no obstacle) Major obstacle –3.52*** –3.63*** Minor obstacle 0.040 0.15 (1.40) (1.6) (0.26) (0.27) Very severe obstacle –1.6 –1.5 Moderate obstacle –0.35 –0.29 (1.9) (1.9) (0.62) (0.65) Electricity (omitted: no obstacle) Major obstacle –0.12 –0.11 Minor obstacle 1.00 1.13 (0.34) (0.34) (1.61) (1.66) Very severe obstacle –0.80 –0.69 Moderate obstacle 1.99 2.20 (0.71) (0.71) (1.50) (1.60) Crime and theft (omitted: no obstacle) Major obstacle 0.34 0.44 Minor obstacle –0.070 –0.084 (1.05) (1.00) (0.38) (0.37) Very severe obstacle 0.72 0.81 Moderate obstacle –0.72** –0.72** (1.20) (1.05) (0.36) (0.37) Telecommunications (omitted: no obstacle) Major obstacle –0.30 –0.24 Minor obstacle –0.024 –0.072 (0.43) (0.43) (0.29) (0.29) Very severe obstacle 1.58* 1.64* Moderate obstacle –0.13 –0.12 (0.88) (0.90) (0.62) (0.56) Tax Rate (omitted: no obstacle) Major obstacle 0.017 –0.19 Minor obstacle –0.42 –0.43 (0.48) (0.49) (0.69) (0.71) Very severe obstacle –2.59 –2.62 Moderate obstacle 0.67 0.74 (1.74) (1.61) (0.72) (0.72) Transportation (omitted: no obstacle) Major obstacle 0.34 0.47 Minor obstacle 0.71 0.78 (0.58) (0.61) (0.49) (0.53) Very severe obstacle 0.080 0.15 Moderate obstacle 0.23 0.37 (0.54) (0.54) (0.38) (0.46) Tax administration (omitted: no obstacle) Major obstacle 0.45 0.69 Minor obstacle –0.43 –0.51 (0.55) (0.61) (0.35) (0.39) Very severe obstacle 0.27 0.30 Moderate obstacle –1.20 –1.30 (0.86) (0.86) (0.80) (0.83) Customs and trade (omitted: no obstacle) Major obstacle –0.66 –0.76 Minor obstacle 0.75 0.72 (0.54) (0.59) (0.47) (0.45) Very severe obstacle –0.76 –0.94 Moderate obstacle 1.31** 1.24** (0.60) (0.67) (0.63) (0.60) Business licensing and permits (omitted: no obstacle) Major obstacle 0.98 0.87 Minor obstacle –0.24 –0.29 (0.61) (0.60) (0.31) (0.33) Very severe obstacle 1.23 1.27 Moderate obstacle 0.021 0.048 (1.45) (1.42) (0.32) (0.31) Informal sector competitors (omitted: no obstacle) Major obstacle 0.40 0.54 Minor obstacle –0.14 –0.26 (0.39) (0.42) (0.41) (0.43) Very severe obstacle –0.12 –0.12 Moderate obstacle 0.11 0.049 (0.43) (0.45) (0.29) (0.32) (Table Continued on next page) (Table Continued on next page) Appendix E Acc e ss t o F in a nc e a n d Ta n z a ni a’ s F i r ms G r o w t h 81 TABLE E.2: Effect of Access to Finance and Business TABLE E.2: Effect of Access to Finance and Business Environment Constraints on Firms Growth, Subjective Environment Constraints on Firms Growth, Subjective Measures, 2013 (Continued) Measures, 2013 (Continued) NO CONTROLS FIRM CONTROLS NO CONTROLS FIRM CONTROLS Political instability (omitted: no obstacle) Region (omitted: Arusha) Minor obstacle 0.23 0.26 Dar Es Salaam –0.035 0.0016 (0.58) (0.72) (0.29) (0.43) Moderate obstacle –1.74* –1.64* Mbeya –0.50 –0.32 (1.04) (0.84) (0.56) (0.72) Major obstacle –0.080 0.076 Mwanza –1.30 –1.31 (0.44) (0.57) (0.96) (0.96) Very severe obstacle 0.59 0.61 Zanzibar 2.22 2.18 (0.75) (0.78) (1.79) (1.84) Corruption (omitted: no obstacle) Sector Minor obstacle –0.30 –0.26 Textiles and garments –0.25 (0.42) (0.41) (0.31) Moderate obstacle –0.14 –0.24 Other manufacturing –0.75** (0.50) (0.53) (0.30) Major obstacle 0.59 0.59 Size of the enterprise (omitted: micro) (0.51) (0.58) Small 0.57** Very severe obstacle –0.13 0.011 (0.27) (0.57) (0.55) Mature –0.69** Courts (omitted: no obstacle) (0.35) Minor obstacle 0.33 0.35 Older –0.062 (0.33) (0.35) (0.38) Moderate obstacle –0.0010 0.070 Constant –0.38 –0.18 (0.34) (0.37) (1.21) (1.20) Major obstacle –0.39 –0.39 Observations 902 898 (0.50) (0.51) R2 0.175 0.190 Very severe obstacle –0.91 –0.89 Adjusted R 2 0.108 0.117 (0.83) (0.86) Source: Enterprise Survey 2013. Notes: Standard errors in parentheses, * p < 0.10, ** p < 0.05, *** p < 0.01 Labor regulations (omitted: no obstacle) Minor obstacle 0.30 0.50 (0.34) (0.37) Moderate obstacle –1.06 –1.00 (1.02) (1.01) Major obstacle –0.37 –0.25 (0.62) (0.59) Very severe obstacle –2.22* –2.54* (1.33) (1.38) Low educational level of labor force (omitted: no obstacle) Minor obstacle –0.29 –0.32 (0.57) (0.57) Moderate obstacle 0.17 0.19 (0.34) (0.34) Major obstacle 0.058 –0.070 (0.27) (0.27) Very severe obstacle 0.48 0.39 (0.56) (0.55) (Table Continued on next page) 82 TA N Z A N I A M A I N L A N D P O V E R T Y A S S E S S M E N T TABLE E.3: Effect of Objective Measure of Access to Finance on Firms Growth, 2013 1 2 3 4 Credit Constraint Status (CCS) (omitted: Full Credit Const.) Partial Credit Constrained (PCC) 0.31 0.31 0.36 0.33 (0.24) (0.24) (0.24) (0.24) Maybe Credit Constrained (MCC) 0.61** 0.61** 0.84** 0.72** (0.30) (0.30) (0.39) (0.32) Non Credit Constrained (NCC) 0.24 0.23 0.27 0.25 (0.41) (0.41) (0.39) (0.40) Sector Textiles and garments –0.18 –0.18 –0.17 –0.19 (0.20) (0.20) (0.21) (0.20) Other manufacturing –0.64** –0.64** –0.66** –0.63** (0.29) (0.30) (0.30) (0.30) Region (omitted: Arusha) Dar Es Salaam –0.51* –0.51* –0.52* –0.50* (0.30) (0.30) (0.30) (0.29) Mbeya –0.34 –0.34 –0.38 –0.35 (0.25) (0.25) (0.25) (0.25) Mwanza 0.14 0.14 0.14 0.14 (0.29) (0.29) (0.29) (0.29) Zanzibar –0.47 –0.47 –0.55 –0.51 (0.51) (0.51) (0.52) (0.51) Size of the enterprise (omitted: micro) Small 0.39** 0.39** 0.40** 0.41** (0.17) (0.17) (0.18) (0.18) Mature –0.36** –0.36** –0.38** –0.36** (0.14) (0.14) (0.16) (0.14) Older –0.14 –0.14 –0.16 –0.13 (0.19) (0.20) (0.19) (0.20) Regulation (# of visits) 0.00058 0.00061 0.00011 0.00027 (0.0011) (0.0011) (0.0010) (0.00099) Overdraft 0.028 (0.15) Creditline –0.51 (0.64) Creditline and overdraft –0.16 (0.24) Constant 0.19 0.19 0.16 0.19 (0.25) (0.25) (0.25) (0.25) Observations 764 764 764 764 R2 0.030 0.030 0.033 0.031 Adjusted R 2 0.011 0.010 0.013 0.011 Source: Enterprise Survey 2013 Notes: Column one contains the results of estimations using credit constrained status (CCS) and firm controls, column 2 adds overdraft only, column 3 adds loans and credit lines only, and column 4 adds both creditline and overdraft. Standard errors in parentheses, * p < 0.10, ** p < 0.05, *** p < 0.01 Appendix E Acc e ss t o F in a nc e a n d Ta n z a ni a’ s F i r ms G r o w t h 83 APPENDIX F A Summary of Theoretical Contributions on ROSCAs 84 TA N Z A N I A M A I N L A N D P O V E R T Y A S S E S S M E N T A ROSCA may be described in formal terms as follows. save for the durable good a fait accompli. Anderson Some number n of individuals agree to meet on a regular and Baland (2002) show that this may trigger mem- basis. At each meeting, each individual brings an amount of bership in a ROSCA, even if there is a fixed cost cash B/n. The collected amount, or pot, B, is then handed associated with such membership. over to one participant, either drawn at random among those who have not yet received the pot (this is the random Being informal, ROSCAs must be self-enforcing. This ROSCA), or according to a pre-determined order (fixed issue is not trivial. For instance, in a random ROSCA, the ROSCA). The ROSCA ends when all the participants have participant who receives the pot first has no incentive to received the pot exactly once. If participants prefer to receive continue paying his remaining dues. In a fixed ROSCA, the the pot earlier, perhaps to be able to enjoy the benefits participant who is designated to be the last one to receive associated with it (consumption, investment, etc.), they would the pot may prefer to not participate, if his main motive for be willing to contribute more than B/n to obtain the pot early. participating is the early pot motive. Anderson, Baland, and Bidding ROSCAs (many variants of which can be imagined) Moene (2009) examine this issue in detail for random and allow participants to do so. fixed ROSCAs, when these are repeated and individuals are infinitely lived. They ask whether enforcement may be Three motivations for participating in a ROSCA have been achieved either through exclusion from future ROSCAs, identified in the theoretical literature: through the order in which the pot is distributed, or through a membership fee that is lost if a participant defaults. Their • The early pot motive. ROSCAs enable all but one of findings inspire pessimism. As an illustration, they show that the participants to get access to the amount B earlier even the strongest possible punishment through exclusion, than under autarky. In their seminal theoretical analysis in the form of permanent exclusion from any future ROSCA of random and bidding ROSCAs, Besley, Coate, and following the first default, does not ensure enforcement in Loury (1993) formalize this argument in a model where random ROSCAs, when the desire to participate is driven individuals use the amount B to buy a durable good. by the early pot motive. The logic is clear. Consider a partic- In this setting, they also find that random ROSCAs ipant who receives the pot at the first meeting. If he defaults dominate bidding ROSCAs from a welfare perspective he avoids paying the remaining (n-1)B/n and he can save an if individuals are homogenous, while the reverse may amount B every n periods on his own. If he does not default, be true if there is heterogeneity in the marginal con- he pays the remaining (n-1)B/n and the expected wait time sumption utility derived from the durable good. until he gets the pot again is (n-1)+n/2>n. Nonetheless, as Anderson, Baland, and Moene (2009), and also Besley, • The savings commitment motive. Even absent con- Coate, and Loury (1993) point out, enforcement may be sumption of durable goods, ROSCAs can provide a ensured if large enough social sanctions can be inflicted useful commitment device for individuals with limited on defaulters. self-control. Ambec and Treich (2007) formalize this idea in a model where an individual would spend cash on his hands on a good that he would rather refrain from buying from an ex ante perspective. Such an individual is better off in an arrangement that leaves him with cash on his hands once instead of n times, since this thus reduces his consumption of the super- fluous good. Gugerty (2007) makes a similar point in a different model. • The household conflict motive. A ROSCA may be useful for an individual who (i) belongs to a house- hold in which decisions to purchase durable goods are joint, (ii) has a stronger preference for the durable good than the other household member(s), and (iii) has limited power in the intra-household bargaining process. Membership in a ROSCA empow- ers such an individual by rendering the decision to Appendix F 85 References Alby, Philippe, Emmanuelle Auriol, and Pierre Nguimkeu. Arnott, Richard and Joseph Stiglitz. 1990. “The Welfare 2019. “Does Social Pressure Hinder Entrepreneurship in Economics of Moral Hazard,” NBER Working Paper 3316. Africa? The Forced Mutual Help Hypothesis”, Economica, forthcoming. Arrow, Kenneth J. (1963) “Uncertainty and the Welfare Economics of Medical Care,” American Economic Review, Alesina, Alberto, and Paola Giuliano. 2010. “The Power of the 53, 941–73. Family.” Journal of Economic Growth, 15, 93–125. Attanasio, Orazio, Abigail Barr, Juan Camilo Cardenas, Alesina, Alberto, and Paola Giuliano. 2015. “Culture and Garance Genicot, and Costas Meghir. 2012. “Risk Pooling, Institutions,” Journal of Economic Literature, 53, 898–944. Risk Preferences, and Social Networks,” American Economic Journal: Applied Economics, 4, 134–167. Algan, Yann, and Pierre Cahuc. 2014. “Trust, Well-being and Growth: New Evidence and Policy Implications,” in Azam, Jean-Paul, and Flore Gubert. 2005. “Those in Kayes. Aghion, Philippe, and Steven Durlauf (eds.) Handbook of The Impact of Remittances on their Recipients in Africa,” Economic Growth. New York: Elsevier Science. Revue Économique, 56, 1331–1358. Alger, Ingela, Laura Juarez, Miriam Juarez-Torres, and Josepa Badiane, Ousmane, Odjo, Sunday, and Collins, Julia (Eds). Miquel-Florensa. 2019. “Do Informal Transfers Induce Lower 2018. Africa Agriculture Trade Monitor Report 2018. Efforts? Evidence from Lab-in-the-field Experiments in Rural Washington, DC: International Food Policy Research Mexico,” Economic Development and Cultural Change. Institute (IFPRI). Alger, Ingela, and Jörgen W. Weibull. 2008. “Family Ties, Baland, Jean-Marie, Isabelle Bonjean, Catherine Guirkinger, Incentives and Development: A Model of Coerced and Roberta Ziparo. 2016. “The Economic Consequences Altruism.” In Basu, Kaushik and Ravi Kanbur, eds., of Mutual Help in Extended Families,” Journal of Arguments for a Better World: Essays in Honor of Amartya Development Economics, 123, 38–56. Sen, Volume II: Society, Institutions and Development. Oxford: Oxford University Press. Baland, Jean-Marie, Catherine Guirkinger, and Charlotte Mali. 2011. “Pretending to Be Poor: Borrowing to Escape Forced Alger, Ingela, and Jörgen W. Weibull. 2010. “Kinship, Solidarity in Cameroon,” Economic Development & Incentives and Evolution.” American Economic Review, Cultural Change, 60, 1–16. 100, 1725–1758. Banerjee, Abhijit, Arun Chandrasekhar, Esther Duflo, and Allen, Robert C. (2000). “Economic structure and agricultural Matthew Jackson. 2013. “The Diffusion of Microfinance,” productivity in Europe, 1300–1 800,” European Review of Science, 341(6144). Economic History, 3, 1–25. Banerjee, Abhijit, Arun Chandrasekhar, Esther Duflo, and Ambec, Stefan, and Nicolas Treich. 2006. “ROSCAs Matthew Jackson. 2018. “Changes in Social Network as Financial Agreements to Cope with Self-control Structure in Response to Exposure to Formal Credit Problems,” Journal of Development Economics, Markets.” Working paper MIT and Stanford. 82, 120–137. Becker, Gary S. 1974. “A Theory of Social Interactions,” Angelucci, Manuela, Giacomo De Giorgi, Marcos A. Rangel, Journal of Political Economy, 82, 1063–1093. and Imran Rasul. 2009. “Village Economies and the Structure of Extended Family Networks,” The Berkeley Gary S. Becker (1981). A Treatise on the Family. Cambridge, Electronic Journal of Economic Analysis & Policy, 9 (1). MA: Harvard University Press. Andreoni, James. 1990. “Impure Altruism and Donations to Bénabou, Roland, and Jean Tirole. 2006. “Belief in a Just Public Goods: A Theory of Warm-Glow Giving,” Economic World and Redistributive Politics,” Quarterly Journal of Journal, 100, 464–477. Economics, 121, 699–746. Arnott, Richard and Joseph Stiglitz. 1988. “The Basic Analytics Blau, P. M., Gustad, J. W., Jessor, R., Parnes, H. S., & Wilcock, of Moral Hazard,” Scandinavian Journal of Economics, R. C. 1956. “Occupational Choice: A Conceptual 90, 383–413. Framework,” ILR Review, 9, 531–543. 88 TA N Z A N I A M A I N L A N D P O V E R T Y A S S E S S M E N T Boltz, Marie, Karine Marazyan, and Paola Villar. 2015. Grimm, Michael, Flore Gubert, Ousman Koriko, Jann Lay, “Income Hiding and Informal Redistribution: A Lab in and Christopher J. Nordman. 2013. “Kinship-ties and the Field Experiment in Senegal,” PSE Working Papers n Entrepreneurship in Western Africa,” Journal of Small 2015–15. Business and Entrepreneurship, 26, 125–150. Bramoullé, Yann, and Gilles Saint-Paul. 2010. “Social Networks Grimm, Michael, Renate Hartwig, and Jann Lay. 2016. and Labor Market Transitions,” Labour Economics, “Does Forced Solidarity Hamper Investment 17, 188–195. in Small and Micro Enterprises?” Journal of Comparative Economics, online July 2016. Christiaensen, Luc. 2017. “Agriculture in Africa – Telling myths from facts: A synthesis,” Food Policy, 67, 1–11. Haltiwanger, J.C., Jarmin, R.S., Miranda, J., (2013). Who ­ creates jobs? Rev. Econ. Stat. 95(2), 347–361. Datta, Saugato, and Sendhil Mullainathan. 2014. “Behavioral Design: A New Approach to Development Policy,” Helpman, Elhanan and Jean-Jacques Laffont. 1975. Review of Income and Health, 60, 7–35. “On Moral Hazard in General Equilibrium Theory,” Journal of Economic Theory, 10, 8–23. Davis, Benjamin, Stefania Di Giuseppe, and Alberto Zezza. 2017. “Are African households (not) leaving agricul- Jakiela, Pamela, and Owen Ozier. 2016. “Does Africa Need a ture? Patterns of households’ income sources in rural Rotten Kin Theorem? Experimental Evidence from Village Sub-Saharan Africa,” Food Policy, 67, 153–174, Economies,” Review of Economic Studies, 83, 231–268. Di Falco, Salvatore, and Erwin Bulte. 2011. “A Dark Side of Jensen, Robert, and Emily Oster. 2009. “The Power of Social Capital? Kinship, Consumption, and Savings,” TV: Cable Television and Women’s Status in India,” Journal of Development Studies, 47, 1128–1151. Quarterly Journal of Economics, 124, 1057–1094. Di Falco, Salvatore, and Erwin Bulte. 2013. “The Impact of Jütting, J., Parlevliet, J. and Xenogiani, T. (2008), Informal Kinship Networks on the Adoption of Risk-Mitigating Employment Re-loaded. IDS Bulletin, 39: 28–36. Strategies in Ethiopia,” World Development, 43, 100–110. Kweka, Josaphat, and Louise Fox. 2011. The Household Duesenberry, James S. 1949. Income, Saving and the Enterprise Sector in Tanzania: Why It Matters and Who Theory of Consumer Behavior. Cambridge, MA: Cares? Poverty Reduction and Economic Management. Harvard University Press. The World Bank. Washington DC. Fafchamps, Marcel. 2011. “Risk Sharing Between Kranton, Rachel E, 1996. “Reciprocal Exchange: Households,” in Benhabib, Jess , Alberto Bisin, and A Self-Sustaining System,” American Economic Matthew O. Jackson (eds.). Handbook of Social Review, 86, 830–851. Economics, Amsterdam: North-Holland. Kranton, Rachel, E., and Deborah F. Minehart. 2001. “A Theory Falk, A., Becker, A., Dohmen, T., Enke, B., Huffman, D., of Buyer-Seller Networks,” American Economic Review, & Sunde, U. 2018. “Global evidence on economic 91, 485–508. ­preferences,” Quarterly Journal of Economics, 133, 1645–1692. La Ferrara, Eliana, Alberto Chong, and Suzanne Duryea. Fowowe, Babajide (2017). Access to finance and firm 2012. “Soap Operas and Fertility: Evidence from Brazil,” ­ performance: Evidence from African countries. Review American Economic Journal: Applied Economics, 4, 1–31. of Development Finance 7, 6–17. Levine, Ross & Beck, Thorsten & Demirguc-Kunt, Asli. (2005). Fox, Louise, and Thomas Pave Sohnesen. 2012. “Household SMEs, Growth, and Poverty: Cross-Country Evidence. Enterprises in Sub-Saharan Africa: Why They Matter for Journal of Economic Growth. 10. 199-229. 10.1007/ Growth, Jobs, and Livelihoods.” Policy Research Working s10887-005-3533-5. Paper No. 6184. World Bank, Washington, DC. Maksimov, Vladislav, Wang, Stephanie Lu and Luo, Yadong, Frank, Robert H. Choosing the Right Pond: Human Behavior (2017), Reducing poverty in the least developed countries: and the Quest for Status. New York: Oxford University The role of small and medium enterprises, Journal of Press, 1985. World Business, 52, issue 2, p. 244–257. REFERENCES 89 Mead, Donald C. and Liedholm, Carl, (1998), The dynamics of Ray, Debraj. 2006. “Aspirations, Poverty and Economic micro and small enterprises in developing countries, World Change,” in Banerjee, Abhijit, Roland Bénabou, and Dilip Development, 26, issue 1, p. 61–74. Mookherjee (eds.) What Have We Learned About Poverty? Oxford: Oxford University Press. McCullough, E. B. 2015. Understanding Agricultural Labor Exits in Tanzania. In 2015 AAEA & WAEA Joint Annual Rijkers, B., Freund, C., Nufifora, A., (2014). Which firms create Meeting, July 26-28, San Francisco, California (No. 206080). the most jobs in developing countries? Evidence from Agricultural and Applied Economics Association & Tunisia. Labour Economics 31(2014), 84–102. Western Agricultural Economics Association. Sánchez Puerta, Maria Laura, M. Julia Granata, Odette Maciel McCullough, Ellen B. 2017. “Labor productivity and Becerril, Gwendolyn Heaner, and Mohamed Ihsan Ajwad. ­ employment gaps in Sub-Saharan Africa”, Food Policy, 2007. “Untapped potential: giving household enterprises 67, 133–152. the chance to succeed in Tanzania.” World Bank Report. Washington, D.C. McMillan, Margaret, and Kenneth Harttgen. 2015. “Africa’s Quiet Revolution,” in Monga, Celestin and Lin, Justin Sleuwaegen, Leo and Goedhuys, Micheline, (2002), Growth of Yifu, (2015), The Oxford Handbook of Africa and firms in developing countries, evidence from Cote d’Ivoire, Economics: Volume 2: Policies and Practices, Oxford Journal of Development Economics, 68, issue 1, p. 117–135. University Press. Squires, Munir. 2016. “Kinship Taxation as a Constraint to McMillan, Margaret S., Dani Rodrik, and Claudia Sepúlveda Microentreprise Growth: Experimental Evidence from (2017) Structural change, fundamentals, and growth: Kenya,” mimeo, London School of Economics. A framework and case studies. Washington, D.C.: International Food Policy Research Institue (IFPRI). Steven J. Davis and John Haltiwanger (1992), Gross Job Creation, Gross Job Destruction, and Employment Morisset, Jacques, and Mahjabeen Haji. 2014. “Tanzania: Reallocation. The Quarterly Journal of Economics, Vol. 107, Productive Jobs Wanted,” The World Bank Group. No. 3, pp. 819–863. Washington DC. Todd, Emmanuel. 2011. L’origine des systèmes familiaux. Neumark, D., Wall, B., Zhang, J., (2011). Do Small Businesses Tome 1 : L’Eurasie. Paris : Gallimard. Create More Jobs? New Evidence for the United States from the National Establishment Time Series. Rev. Econ. Udry, Christopher. 1990. “Rural Credit in Northern Nigeria: Stat. 93(1), 16–29. Credit as Insurance in a Rural Economy,” World Bank Economic Review, 4, 251–269. Nichter, Simeon and Goldmark, Lara, (2009), Small Firm Growth in Developing Countries, World Development, Van Biesebroeck, Johannes. 2005. Firm Size Matters: Growth 37, issue 9, p. 1453–1464. and Productivity Growth in African Manufacturing, Economic Development and Cultural Change, 53, issue 3, p. 545–83. Pauly, Mark V. 1968. “The Economics of Moral Hazard: Comment,” American Economic Review, 58, 531–537. Zins, A. and Weill, L. 2016. The determinants of financial inclusion in Africa. Review of Development Finance 6, 46–57. ­ 90 TA N Z A N I A M A I N L A N D P O V E R T Y A S S E S S M E N T