Policy Research Working Paper 8980 Industrialization on a Knife’s Edge Productivity, Labor Costs and the Rise of Manufacturing in Ethiopia Stefano Caria Africa Region Office of the Chief Economist August 2019 Policy Research Working Paper 8980 Abstract The latest push for industrialization in Ethiopia has However, Ethiopia appears competitive when compared attracted much academic and public interest. This paper to Bangladesh. Capital, firm size, or sectoral composition assesses Ethiopia’s competitiveness and attractiveness as an do not explain the low productivity of the Ethiopian man- investment destination by comparing domestic productiv- ufacturing sector. Ethiopian firms, however, have worse ity and input costs to a sample of manufacturing exporting management, particularly in the area of labor management. countries. The paper documents that, in a comparison The paper concludes by discussing the potential for labor with Kenya, India or Vietnam, the labor cost advantage interventions to increase productivity and create the con- of Ethiopian firms is more than offset by low productivity. dition for further industrialization. This paper is a product of the Office of the Chief Economist, Africa Region, commissioned as part of the ‘Industrialization for Jobs in Africa’ regional study. It is part of a larger effort by the World Bank to provide open access to its research and make a contribution to development policy discussions around the world. Policy Research Working Papers are also posted on the Web at http://www.worldbank.org/prwp. The author may be contacted at stefano.caria@bristol.ac.uk. The Policy Research Working Paper Series disseminates the findings of work in progress to encourage the exchange of ideas about development issues. An objective of the series is to get the findings out quickly, even if the presentations are less than fully polished. The papers carry the names of the authors and should be cited accordingly. The findings, interpretations, and conclusions expressed in this paper are entirely those of the authors. They do not necessarily represent the views of the International Bank for Reconstruction and Development/World Bank and its affiliated organizations, or those of the Executive Directors of the World Bank or the governments they represent. Produced by the Research Support Team Industrialization on a Knife’s Edge Productivity, labor costs and the rise of manufacturing in Ethiopia Stefano Caria∗ JEL codes: O14, O15, O55. Key words: Industrialization, productivity, labor costs, Ethiopia. 1 Introduction The latest push for industrialisation in Ethiopia has attracted much academic and public interest. While the domestic manufacturing sector is still relatively small, Ethiopia has made large investments to promote industrial development and has witnessed some early success (Oqubay, 2015). The flagship project is the creation of several indus- trial parks across the country. The most famous of these parks, in the southern city of Hawassa, has attracted significant foreign direct investment from international gar- ment manufacturers. This park alone currently employs more than 15,000 workers and exports 100 million USD worth of goods, doubling the country’s foreign exchange earn- ings from garment manufacturing (Butler, 2018). These early results are surprising in the context of the recent literature on ‘prema- ture deindustrialization’ (Rodrik, 2015). This literature argues that in the last decades developing countries have moved out of manufacturing much faster than today’s rich countries did in the past, when they had comparable levels of income. This early exit from a labour-intensive sector is feared to be an obstacle to generating employment and broad-based gains from growth. Also, the international experience with special economic zones and industrial parks has been mixed (Khandelwal and Teachout, 2016). In this paper, I assess Ethiopia’s competitiveness and attractiveness as an invest- ment destination by comparing the productivity and input costs faced by firms based in Ethiopia to a sample of manufacturing exporting countries. Given the current strate- gic importance of the garment sector, I select comparison countries that are garment exporters in Africa (Kenya), and Asia (Bangladesh, India, Vietnam). I use harmonized data from the World Bank Enterprise Survey, which enables me to construct aggregates figures for sales for worker (a measure of productivity), total labor costs, capital stock, firm size, and a somewhat noisy measure of value added. I compute these figures for the whole manufacturing sector and for the garment sector. Further, I use data from the World Management Survey to study the quality of management. My main finding is that Ethiopia’s labor cost advantage is more than offset by low productivity with respect to all but one of the comparison countries (Bangladesh). Ethiopia’s industrialization is thus on a knife’s edge. Its fundamentals make it competitive with respect to the countries at the bottom of the productivity distribution. In recent years, the increase in labour costs has also been modest compared to these countries. How- ever, given the likely pressure on wages that will result from further expansion of the manufacturing sector, additional gains in productivity will be crucial in order to secure Ethiopia’s position as an attractive investment destination. The low productivity of Ethiopian firms is not explained by differences in size, cap- 2 ital, or sector. However, Ethiopian firms score particularly low on a measure of the quality of management. The most critical area is that related to the management of workers. In particular, the quality of selection, incentives and retention practices is lower than in all comparison countries. This suggests a need for reformed labor man- agement practices. Further, given the scarcity of rigorous evidence on the merits of specific interventions, this highlights the need for more experimental work on labor management in Ethiopia. What is particularly interesting is that the historical experience of other nations also points to the importance of developing the right labor institutions and practices to make industrialization possible and to distribute its gains. Scholars of both the industrial revolution in the UK (Thompson, 1967; Clark, 1994) and of mid 20th century indus- trialization in Asia and Latin America (Kohli, 2004) agree on this point. Research on early industrialization in Africa also points to labor management problems, in particu- lar, retention (Elkan, 1971). Finally, newly available historical data has highlighted the important role of unions in securing a wider distribution of the gains from industrial- ization (Farber et al., 2018). This paper is also related to and complements two recent studies by Gelb et al. (2016) and Gelb et al. (2017), which argue that high labor costs hold back industrialization in Africa. These studies identify Ethiopia as a country with relatively low labor costs. Our analysis further points out that this labor cost advantage is partly offset by low productivity. The rest of this paper is organized as follows. Section 2 describes the data. I then present the comparison of productivity and input costs and discuss the determinants of Ethiopia’s low productivity in Section 3 and 4. Finally, in Section 5, I discuss evidence from recent studies of labor interventions. Section 6 concludes. 2 Data Most of the analysis is based on data from the World Bank Enterprise Survey – an har- monised representative survey of firms covering about 139 countries in the world. For the analysis, I select five countries of interest: Ethiopia, Kenya, Bangladesh, India and Vietnam.1 Kenya is a geographical neighbour of Ethiopia and a natural competitor within east Africa, with a key advantage coming from direct access to the sea. I then select three Asian countries, covering the range of the productivity distribution among 1 I use the latest survey available for each country. For Ethiopia and Vietnam, this is the 2015 survey; for India this is the 2014 survey, and for Bangladesh and Kenya this is the 2012 survey. 3 major manufacturing economies: from relatively high-productivity Vietnam to rela- tively low-productivity Bangladesh. I focus the analysis on manufacturing firms (two-digit ISIC codes 15-37). I also re- port data specific to the garment sector (ISIC codes 17-19). The key variables I extract from the data are number of workers (defined as the sum of employees on permanent and temporary contracts), total sales, total labor costs (including wages, payroll taxes, benefit payments, etc..), total cost of materials (the cost of physical inputs in produc- tion), and total value of physical capital. I measure productivity as sales per worker. Further, I calculate a measure of values added that takes into account rental payments for capital and land. All figures are converted to 2010 USD by using exchanges rate from the relevant year and US CPI data.2 I use sampling weights to compute averages representative of the underlying population.3 Missing data is not common, with a few exceptions – in particular for material costs. All variables are truncated at the 95th percentile of the country-specific distribution. I report all key descriptives in Tables A.1-A.5. I also use data from two other sources to complement the analysis. First, I use origi- nal survey data collected for manufacturing firms in Addis Ababa by Abebe et al. (2017) to check the robustness of the sales figures reported for Ethiopia. Second, I use data from the World Management Survey to compare the quality of management across the countries in my sample (with the exception of Bangladesh, which is not included in the World Management Survey). This survey studies management practices in manufac- turing firms in 34 countries.4 Firm managers are asked to describe several management practices in the firms over a phone interview and are then scored by a trained enumer- ator. The instrument produces an overall management score, as well as specific scores for the quality of operations management, monitoring of performance, target setting, and HR practices (including recruitment and retention strategies). 2 I used exchange rate data from the World Bank, which is available here https://data.worldbank. org/indicator/PA.NUS.FCRF?locations=ET. For each country, I use the exchange rate for the year prior to the date when the survey was released to the public (e.g. for Ethiopia, I use the 2014 exchange rate). The CPI data is available here https://www.measuringworth.com/datasets/uscpi/. 3 The World Bank Enterprise Survey offers three types of weights. I use the recommended median weights. 4 I use the 2004-14 combined manufacturing survey data available here: https:// worldmanagementsurvey.org 4 3 A comparison of productivity and labor costs Fact 1. Ethiopia and Bangladesh have the lowest productivity in the sample. The average Indian firm is twice as productive as the average Ethiopian firm. Manufacturing firms in Ethiopia have relatively low levels of productivity. The data confirms a large productivity gap with Vietnam and India, and, surprisingly, also with Kenya. The average Ethiopian firm sells 14,180 USD worth of output per worker per year. This is about half of the sales per worker of the average firm in India (30,875 USD) and Vietnam (48,755 USD). The difference with Kenya, where the average firm sells 46,420 USD of output per year, is also striking. If we restrict attention to the gar- ment sector, we find similar differences (e.g. sales per worker are close to 7350 USD in Ethiopia and to 26,400 USD in India). However, the data also shows that Ethiopian firms have very similar productivity to those of a key competitor — Bangladesh. Aver- age productivity in the whole manufacturing sector is 40 percent higher (14,180 USD against 10,055 USD of sales per worker per year). In the garment sector, productivity in Ethiopia is about 16 percent lower (7,700 USD against 9,200 USD). This is a surpris- ing finding, as Bangladesh has a larger, more established manufacturing industry. We report these findings in Figures 1 and A.1. Figure 1: Average sales per worker and costs per worker All manufacturing 5 The productivity gap highlighted above occurs mostly at the top of the distribution. In Figures A.4 and A.5 we plot the differences in sales per worker between Ethiopia and each competitor country at five different percentiles. Differences tend to be small up to the median firm. However, at the 75th and 90th percentiles of the distribution, we observe a large productivity gap. Thus, what is distinctive about Ethiopia is the lack of high productivity firms. Fact 2. Ethiopia has the lowest labor costs in the sample, but faces intermediate material input costs. Manufacturing firms in Ethiopia also face low labor costs. This confirms a wide perception that abundant labor and competitive wages can help Ethiopia expand its manufacturing sector. Monthly labor costs per worker are about 69 USD in Ethiopia, 79 USD in Bangladesh, 175 USD in Kenya, 180 USD in India, and 201 USD in Vietnam. In the garment sector, the differences are even more striking: average labor costs in Bangladesh and India are 20 percent and 128 percent higher than in Ethiopia, respec- tively. On the other hand, Ethiopian firms are in the middle of the distribution in terms of the cost they pay to source inputs of production. This is not surprising, given that Ethiopia is the only land-locked country in the sample, with limited options to access the sea. Further, material costs may also reflect the mix of sectors of the economy. If I restrict attention to the garment sector, for example, I find that material costs are higher than neighbouring Kenya, but lower than Bangladesh (India and Vietnam have much higher material costs, probably due to specialisation in higher quality garment). Fact 3. The value added of Ethiopian firms is above that of Bangladesh, but below India, Kenya and Vietnam. Overall, low labor costs cannot fully compensate for low productivity. As a result, Ethiopian firms are among the least profitable (in terms of value added per worker) in our sample (Figure 2 and A.6). However, value added per worker in Ethiopia is actually relatively close to those of the next most profitable country – Bangladesh. If we look at the whole manufacturing sector, value added per worker is higher than in Bangladesh. If we focus on the garment sector, then we find that the profitability of Ethiopian firms is just below that of firms in Bangladesh. There are three important caveats related the value added data I have just presented. First, the large number of missing values in the material costs variable means that I am working with a selected sample. Second, the surveys for Ethiopia and Vietnam do not record data on rental payments for land or capital. I impute these payments by using 6 predicted values from a model that regresses rents on capital, capital square, and firm size. Third, it is important to state that all figures are before tax, and that taxation regimes may differ across countries. The data on value added per worker should thus be interpreted with caution. Figure 2: Median value added per worker All manufacturing Fact 4. Ethiopian garment firms have substantially improved their competitiveness (in terms of relative sales per worker and cost per worker) with respect to Bangladeshi garment firms over a period of 10 years. The final result in this section is related to the trends in productivity and cost of labour. For this, I assembled WBES surveys from 2005 to 2007 for all countries but Vietnam.5 I show average sales and cost per worker in Figure A.2 and Table A.6. In Figure A.3 I plot the average annual change in these variables. This figure shows that both Ethiopia and Bangladesh have witnessed a fast growth in productivity (about 14 percent per year). However, the annual growth in labour costs in Ethiopia has been 5 I use the 2005 survey for India, the 2006 survey for Ethiopia, the 2007 surveys for Kenya and Bangladesh. I focus on the garment sector and report sample sizes and statistics on missing data points in table A.3. I convert all values to 2010 USD, as described in the section above. The Ethiopian dataset does not include sampling weight, so the results presented for all countries are not reweighted. 7 about half of that of Bangladesh (7 percent versus 15 percent). While this highlights the growing attractiveness of Ethiopia as an investment destination, it may also highlight that Ethiopian employers in this sector have higher monopsony power compared to their counterparts in Bangladesh. 3.1 Robustness to use of alternative data sources The productivity figure used for Ethiopia are consistent with those obtained by an in- dependent survey run in Addis Ababa at a similar point in time for the study of Abebe et al. (2017). I show this by comparing the distribution of sales for worker in the two surveys in Figure A.7 and by showing means and medians in Table A.7. Average sale for workers in Addis Ababa is remarkably similar across the two surveys (22,708 USD and 22,462 USD). Median sales are somewhat higher in the survey of Abebe et al. (2017), but the difference is not excessively large. Finally, the figures reported in the previous section align well with the perception of foreign firm managers. I have carried out qualitative field in the Hawassa Industrial Park at several points during 2018. During these visits I had several personal conversa- tions with firm managers about productivity and labor management. On average, firm managers believed that productivity in Ethiopia’s garment factories is between a third and a half of that of comparable factories producing similar products in India; and that wage costs are about one third of those in India. These figures confirm the aggregate statistics reported in Figure A.1. 4 Where does Ethiopia’s low productivity come from? Fact 5. Capital stock, firm size and sectoral composition do not explain the low productivity of Ethiopian firms. To study the factors that determine the low productivity of manufacturing firms in Ethiopia we run a regression model where we regress sales per worker on firm size, sector, and value of the capital stock. Specifically, we create dummies for quartiles of firm size and capital stock. We then create dummies for each sector and introduce them in the model on their own, and interacted with the dummies for firm size and capital stock. We compute the residual productivity for this regression model, and standardise this value using the average and standard deviation of the residual productivity for firms in Vietnam. We present the result of this exercise in Figures 3 and A.8. The main conclusion from this analysis, is that, after accounting for size, sector, and capital, Ethiopian firms are still less productive than firms in competitor countries. In 8 particular, average residual productivity is about .22 standard deviation below Viet- nam. By comparison, Bangladesh has residual productivity that is about .18 standard deviations below Vietnam and India is about .08 standard deviations below Vietnam. A similar picture emerges if we look at the garment sector. This is consistent with the fact Ethiopian firms do not appear to be exceptionally small or to operate much less physical capital compared to the other countries in the sample (Figures A.9 and A.10). For example, on average Ethiopian firms appear to use more physical capital than firms in India or Bangladesh. Ethiopian garment firms are smaller than those in Bangladesh (which are exceptionally large), but have a similar size to those in India and Kenya. Capital, firm size and sector also do not explain the lack of high productivity firms. To show this, we regress a dummy for being a high productivity firm (ie a firm with sales per worker above 20,000 USD) on capital, firm size, sector dummies and the inter- actions described above. We calculate the residual probability of having a high produc- tivity firm and normalise it by subtracting the average residual probability of Vietnam. We find that after accounting for the influence of these variables, Ethiopia still has an unusually low share of high productivity firms: 30 percentage points fewer than Viet- nam and 10 percentage points fewer than Bangladesh (Figures A.11 and A.12). Figure 3: Sales per worker controlling for sector, size, and capital All manufacturing 9 Fact 6. Management scores in Ethiopia are lower than in competitor countries, particularly in the area of labor management. What could explain the residual difference in productivity? The recent literature has emphasised the role of good management in determining productivity (Bloom et al., 2013, 2018). However, the World Bank Enterprise Survey does not have enough infor- mation to measure management practices. We thus turn to the World Management Sur- vey. Using this dataset, we are able to document two facts. First, Ethiopian firms have worse management compared to India, Kenya and Vietnam. Second, the main driver of the low management scores is poor HR practices. Ethiopian firms scores below Viet- namese firms on all dimension of management (operations, monitoring, targets, and HR). However, the gap in HR practices is about twice as large as the gap in the other three dimensions of good management (almost one standard deviation versus .5 of a standard deviation). We document these results in Figures 4 and 5. Figure 4: Management scores Labor management consists of a host of practices including recruitment policies, performance incentives and retention strategies (Oyer and Schaefer, 2010; Bandiera et al., 2011). In Figure A.13 we show that Ethiopian firms perform poorly in terms of their strategies to identify and retain high performers within the organisation, and to deal with poor performers. This is consistent with the fact that, in a survey of HR man- agers, Abebe et al. (2017) show that the two most pressing HR problems reported are 10 Figure 5: Dimensions of management performance (compared to Vietnam) worker selection and retention. Further, Blattman and Dercon (2017) document high rates of turnover in a sample of Ethiopian factories. 5 Possible labor management interventions The recent experimental literature in development economics confirms that labor in- terventions can increase firm productivity. For example, (Adhvaryu et al., 2018) eval- uate the effects of soft skills training on the productivity of garment workers in India; (Kaur et al., 2015) study the effect of policy where Indian data entry workers can choose their own targets; (Dean, 2018) measures the effect of changes in background noise. We present the estimates from these studies in Figure 6. Treatment effects on output per worker range from less than 1 percent (for the target intervention, which, however had a 9 percent treatment effect on a subgroup of workers with self-control problems) to about 13 percent (for soft skills training). The soft skills intervention costed about 90 USD per worker, but generated much larger benefits over time. The estimated net rate of return for the firm was 256%. The literature in this area, however, is still limited. In particular, we know little about how workers can be retained by private firms. In the context of government and non- profit institutions, Deserranno (2014) and Ashraf et al. (2014) show that information about the nature of the job enables organisations to attract workers that are retained for 11 Figure 6: Treatment effects on output from recent studies longer. However, we have no evidence on whether these findings generalise to private sector firms. 6 Conclusion In this paper, I assess the competitiveness of the Ethiopian manufacturing sector rela- tive to other manufacturing exporters. I document that, in a comparison with Kenya, India or Vietnam, the labor cost advantage of Ethiopian firms is more than offset by low productivity. However, Ethiopia appears competitive when compared to Bangladesh. The prospects for further industrialisation in Ethiopia seem to critically depend on achieving higher productivity. Poor management, particularly in the area of labor, is a key factor that may be holding Ethiopian firms back. The evidence on how labor management can be improved, however, is scarce. To develop this evidence, we need systematic experimentation with different interventions targeting how workers are se- lected, trained, incentivised, and retained. 12 References Abebe, G., S. Caria, M. Fafchamps, P. Falco, S. Franklin, S. Quinn, and F. Shilpi (2017). Job Fairs: Matching Firms and Workers in a Field Experiment in Ethiopia. Working paper. Abebe, G., S. Caria, and E. Ortiz-Ospina (2017). The Selection of Talent: Experimental and Structural Evidence from Ethiopia. Working paper. Adhvaryu, A., N. Kala, and A. Nyshadham (2018). The Skills to Pay the Bills: Returns to On-The-Job Soft Skills Training. Working paper. Ashraf, N., O. Bandiera, S. S. Lee, et al. (2014). Do-Gooders and Go-Getters: Career Incentives, Selection, and Performance in Public Service Delivery. Working Paper. Bandiera, O., I. Barankay, and I. Rasul (2011). Field Experiments with Firms. The Journal of Economic Perspectives 25(3), 63–82. Blattman, C. and S. Dercon (2017). The Impacts of Industrial and Entrepreneurial Work on Income and Health: Experimental Evidence from Ethiopia. Working paper. Bloom, N., B. Eifert, A. Mahajan, D. McKenzie, and J. Roberts (2013). Does Management Matter? Evidence from India. The Quarterly Journal of Economics 128(1), 1–51. Bloom, N., A. Mahajan, D. McKenzie, and J. Roberts (2018). Do Management Interven- tions Last? Evidence from India. Working paper. Butler, M. (2018). Labour Challenges in Hawassa. Clark, G. (1994). Factory Discipline. The Journal of Economic History 54(1), 128–163. Dean, J. (2018). Noise, Cognitive Function, and Worker Productivity. Working paper. Deserranno, E. (2014). Financial Incentives as Signals: Experimental Evidence from the Recruitment of Health Workers. Working Paper. Elkan, W. (1971). An African Labour Force. In Developing the Underdeveloped Countries, pp. 240–246. Springer. Farber, H. S., D. Herbst, I. Kuziemko, and S. Naidu (2018). Unions and Inequality Over the Twentieth Century: New Evidence from Survey Data. NBER Working Paper. 13 Gelb, A., C. Meyer, and V. Ramachandran (2016). Does Poor Mean Cheap? A Compara- tive Look at Africa?s Industrial Labor Costs. Revue d?économie du développement 24(2), 51–92. Gelb, A., C. Meyer, V. Ramachandran, and D. Wadhwa (2017). Can Africa be a Manu- facturing Destination? Labor Costs in Comparative Perspective. Working paper. Kaur, S., M. Kremer, and S. Mullainathan (2015). Self-control at Work. Journal of Political Economy 123(6), 1227–1277. Khandelwal, A. and M. Teachout (2016). Special Economic Zones for Myanmar. Kohli, A. (2004). State-Directed Development: Political Power and Industrialization in the Global Periphery. Cambridge University Press. Oqubay, A. (2015). Made in Africa: Industrial Policy in Ethiopia. Oxford University Press, USA. Oyer, P. and S. Schaefer (2010). Personnel Economics: Hiring and Incentives. NBER Working Paper No. 15977. Rodrik, D. (2015). Premature Deindustrialization. NBER Working Paper No. 20935. Thompson, E. P. (1967). Time, Work-Discipline, and Industrial Capitalism. Past & present (38), 56–97. A.1 Appendix A.2 A.1 Figures Figure A.1: Average sales per worker and cost per worker Garment sector A.3 Figure A.2: Average sales per worker and cost per worker Garment sector: mid 2000s A.4 Figure A.3: Change in average sales per worker and cost per worker between two waves A.5 Figure A.4: Differences in sales per worker across the distribution All manufacturing A.6 Figure A.5: Differences in sales per worker across the distribution Garment sector A.7 Figure A.6: Median value added per worker Garment sector A.8 Figure A.7: Comparison of data from two surveys A.9 Figure A.8: Sales per worker controlling for sector, size, and capital Garment sector A.10 Figure A.9: Firm size and capital stock All manufacturing A.11 Figure A.10: Firm size and capital stock Garment sector A.12 Figure A.11: High productivity firms controlling for sector, size, and capital All manufacturing A.13 Figure A.12: High productivity firms controlling for sector, size, and capital Garment sector A.14 Figure A.13: Quality of HR practices (compared to Vietnam) A.15 A.2 Tables Table A.1: Sample size and missing data All manufacturing Sample size (no.) Missing data (%) Worker no. Sales Labor costs Material costs (1) (2) (3) (4) (5) Ethiopia 383 0.01 0.09 0.08 0.17 Kenya 414 0.02 0.12 0.19 0.28 Bangladesh 1179 0.00 0.03 0.02 0.04 India 7163 0.00 0.02 0.04 0.06 Vietnam 694 0.00 0.01 0.09 0.20 Table A.2: Sample size and missing data Garment sector Sample size (no.) Missing data (%) Worker no. Sales Labor costs Material costs (1) (2) (3) (4) (5) Ethiopia 70 0.00 0.04 0.09 0.17 Kenya 58 0.03 0.09 0.21 0.21 Bangladesh 467 0.00 0.01 0.01 0.03 India 912 0.00 0.04 0.06 0.08 Vietnam 178 0.01 0.01 0.09 0.22 A.16 Table A.3: Sample size and missing data Garment sector: mid 2000s Sample size (no.) Missing data (%) Worker no. Sales Labor costs Material costs (1) (2) (3) (4) (5) Ethiopia 105 0.00 0.00 0.00 0.09 Kenya 29 0.00 0.00 0.00 0.00 Bangladesh 296 0.00 0.00 0.00 0.00 India 275 0.07 0.07 0.08 0.08 A.17 Table A.4: Descriptives All manufacturing Mean St.Dev. Percentiles 10th 25th 50th 75th 90th Ethiopia Number of workers 159.94 1060.41 7.00 17.00 52.00 140.00 341.00 Sales per worker 13725.23 17110.34 1175.71 2588.93 6206.52 14093.96 31165.12 Labor costs per worker 805.51 696.94 130.23 268.73 699.03 1128.68 1626.67 Material costs per worker 6950.81 10821.35 376.23 940.56 2565.18 7683.99 17566.12 Physical capital per worker 2851.29 7127.28 2.99 331.90 1508.20 3846.36 7901.64 Kenya Number of workers 145.01 593.51 10.00 18.00 45.00 140.00 356.00 Sales per worker 42960.24 1.2e+05 2028.72 5373.74 14045.01 31326.89 67416.06 Labor costs per worker 1927.80 2790.24 140.45 325.75 1123.60 2331.72 5185.85 Material costs per worker 7485.05 36595.53 97.08 421.35 1797.76 5618.01 15730.42 Physical capital per worker 4809.24 9662.65 51.07 449.44 2247.20 6741.61 14045.01 Bangladesh Number of workers 298.78 734.57 9.00 19.00 40.00 224.00 650.00 Sales per worker 9858.11 37717.17 1353.57 2320.41 5220.93 11447.37 25137.81 Labor costs per worker 925.90 1115.29 174.03 464.08 756.66 1044.19 1546.94 Material costs per worker 5751.81 35400.55 232.04 725.13 2320.41 5801.03 13535.75 Physical capital per worker 1804.03 4612.95 77.35 232.04 773.47 2417.10 5801.03 India Number of workers 76.78 310.97 10.00 18.00 40.00 100.00 250.00 Sales per worker 30508.62 51232.20 3833.74 7487.78 15973.93 35142.65 69930.34 Labor costs per worker 2139.47 3586.33 492.85 958.44 1597.39 2475.96 4312.96 Material costs per worker 19032.04 34671.05 1448.30 3194.79 8519.43 20446.63 44727.01 Physical capital per worker 4100.20 6874.08 239.61 745.45 2007.83 4792.18 9983.71 Vietnam Number of workers 224.12 1049.23 13.00 22.00 56.00 150.00 400.00 Sales per worker 47482.83 57533.16 2395.72 4718.85 12099.61 33602.33 94895.48 Labor costs per worker 2384.72 2011.57 301.56 871.17 1742.34 2756.33 4188.32 Material costs per worker 20003.75 32481.13 194.40 871.17 3555.80 13938.74 43558.58 Physical capital per worker 6038.23 9550.89 164.37 622.27 2177.93 5782.11 13814.29 A.18 Table A.5: Descriptives Garment sector Mean St.Dev. Percentiles 10th 25th 50th 75th 90th Ethiopia Number of workers 316.74 963.66 6.00 14.00 69.00 272.00 783.00 Sales per worker 8026.21 10939.86 1175.71 2653.31 5023.56 9936.79 21583.00 Labor costs per worker 874.96 626.31 157.50 453.25 750.42 1080.31 1626.74 Material costs per worker 4065.42 5302.97 352.71 817.98 2137.65 5206.70 13781.62 Physical capital per worker 2251.48 4141.68 0.00 124.14 1149.25 3630.86 5681.37 Kenya Number of workers 259.74 815.35 12.00 25.00 68.00 168.00 500.00 Sales per worker 28544.99 83634.90 1940.08 4369.56 8863.96 19975.13 44471.86 Labor costs per worker 1724.27 2042.17 123.60 224.72 1123.60 2496.89 3606.62 Material costs per worker 22135.72 87844.56 157.30 468.17 1123.60 3862.38 6938.19 Physical capital per worker 3636.56 5857.69 86.43 290.89 1726.47 5477.56 9987.57 Bangladesh Number of workers 456.90 1005.04 14.00 24.00 115.00 450.00 1050.00 Sales per worker 8978.94 15905.52 1325.95 2677.40 5801.03 11976.33 24063.55 Labor costs per worker 1030.42 1281.02 135.36 556.90 835.35 1160.21 1740.31 Material costs per worker 4459.05 10591.25 283.15 966.84 2747.09 5994.40 12709.54 Physical capital per worker 1420.82 3909.78 79.11 248.62 709.67 2578.24 5273.67 India Number of workers 119.52 579.60 14.50 26.00 60.00 167.00 427.50 Sales per worker 26481.62 35929.71 4472.70 7720.73 14912.53 31947.87 58571.09 Labor costs per worker 1955.41 2879.70 588.51 1037.86 1536.67 2236.35 3422.99 Material costs per worker 15559.22 25555.54 1464.28 3288.75 7768.55 17838.43 37012.77 Physical capital per worker 3275.27 5432.36 159.74 501.66 1597.39 3993.48 8407.33 Vietnam Number of workers 312.48 1120.31 15.00 30.00 90.00 250.00 800.00 Sales per worker 29667.93 42984.62 2177.93 3629.88 7599.08 19165.78 77044.23 Labor costs per worker 2437.67 2647.62 241.59 614.85 1718.29 2480.55 3798.07 Material costs per worker 13206.43 23576.74 114.63 435.59 3097.50 9333.98 40207.92 Physical capital per worker 3791.38 6836.35 65.50 248.91 1028.94 2903.91 6556.47 A.19 Table A.6: Descriptives Garment sector: mid 2000s Mean St.Dev. Percentiles 10th 25th 50th 75th 90th Ethiopia Number of workers 110.70 364.62 5.00 7.00 11.00 50.00 200.00 Sales per worker 2452.07 3069.46 389.42 856.68 1522.12 2760.31 4831.53 Labor costs per worker 467.84 448.25 89.32 193.26 397.57 618.44 821.36 Material costs per worker 1060.75 1575.87 100.21 225.47 445.79 1380.44 2576.81 Kenya Number of workers 152.17 313.28 10.00 13.00 80.00 150.00 350.00 Sales per worker 25943.92 35994.88 5000.61 8673.37 12858.71 30003.66 44123.02 Labor costs per worker 3857.40 4597.00 750.09 1393.70 2474.30 4552.53 8182.82 Material costs per worker 10016.56 15697.87 1750.21 2500.30 3475.25 9478.43 23574.30 Bangladesh Number of workers 606.57 714.68 110.00 250.00 400.00 650.00 1350.00 Sales per worker 4042.69 3403.03 871.74 1824.57 3303.43 5579.12 7398.13 Labor costs per worker 445.87 195.13 203.99 313.83 441.60 564.89 696.07 Material costs per worker 2511.09 2398.83 222.76 792.72 1824.44 3536.07 5265.37 India Number of workers 68.91 116.49 5.00 9.50 28.00 92.50 145.00 Sales per worker 21534.18 89165.36 1401.05 3056.85 8253.49 16035.27 22044.57 Labor costs per worker 1320.87 3522.43 185.83 509.47 1017.13 1264.13 1728.14 Material costs per worker 9823.08 21307.47 407.58 1557.15 4275.95 9486.60 14710.10 Table A.7: Comparison of data from two surveys All manufacturing, Addis Ababa Sample size Mean Median (1) (2) (3) Abebe et al. survey 90 22708.90 14874.57 World Bank Manucturing Survey 393 22462.02 11029.03 A.20