Policy Research Working Paper 8980 Sources of Manufacturing Productivity Growth in Africa Patricia Jones Emmanuel K.K. Lartey Taye Mengistae Albert Zeufack Africa Region Office of the Chief Economist August 2019 Policy Research Working Paper 8980 Abstract This paper investigates the sources of growth in manufactur- Reallocation of market share plays an important role in ing productivity in Cote D’Ivoire, Ethiopia and Tanzania raising aggregate productivity in Côte d’Ivoire as well. But in comparison with the case of Bangladesh. Based on the the pattern here is opposite to that in Ethiopia in that in analysis of establishment census data since the mid-1990s, Côte d’Ivoire entering (or newly opening) plants have larger it finds that reallocation of market share between firms impact on aggregate productivity growth than closing (or contributed substantially to productivity growth in each of exiting) plants. Unlike the case with Cote D’Ivoire and of the four countries, although to a varying extent. In Ethiopia, Ethiopia, the reallocation of market share among surviving the impact of market share reallocations among survivors plants is a smaller source of manufacturing productivity tended to be larger than those associated with increases in growth in Tanzania than the new plant openings and plant within-plant productivity. In addition, plant closure (or closure. The data suggest that the reallocation of market exit) boosted productivity more than new plant openings share among surviving plants and exiting plants has larger (or entry) did in the sense that the relative productivity of impact on productivity growth in Bangladesh than the pro- survivors (or continuing plants) was higher relative to that ductivity gap between new plants and survivors, as in the of closing plants (or exit cases) than it was relative to the case of Ethiopia. productivity of newly opening plants (or new entrants). 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 authors may be contacted at jonesp@newpaltz.edu, elartey@worldbank.org, tmengistae@worldbank.org, and azeufack@worldbank.org. 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             Sources of Manufacturing Productivity Growth in Africa Patricia Jones SUNY New Paltz Emmanuel K.K. Lartey World Bank Taye Mengistae World Bank Albert Zeufack World Bank August 2019       JEL codes: L6, 014, 012 Keywords: Industry Studies-Manufacturing; Industrialization, Productivity, Growth, Microeconomic Analysis of Development     I. Introduction In this paper, we focus on three African countries at different rates of industrialization of their economies and assess the different sources of productivity growth in each country. Specifically, we estimate the size and sources of productivity growth for Ethiopia, Côte d’Ivoire, and Tanzania—and then compare these productivity gains to that in an external comparator— Bangladesh. To do this, we compiled establishment-level data from a wide range of sources including manufacturing censuses, business registrars, and economic censuses. Using these data, we investigate whether differences in the performance of manufacturing establishments can be accounted for by differences in the efficiency with which resources are allocated to their most productive use. We carry out both the static Olley-Pakes (1996) decomposition and the dynamic Olley-Pakes decomposition (DOPD) proposed by Melitz and Polanec (2015). These decomposition methods break down aggregate productivity changes into two components: one component that captures shifts in the productivity distribution (via changes in the unweighted mean of establishment-level productivity); and another component that captures market share reallocations (via changes in the covariance between market shares and establishment-level productivity). This allows us to identify patterns of productivity growth across different types of firms (e.g., by firm size and industry) as well as the efficiency of the factor reallocation in allocating resources to a sector’s most productive firms. A growing literature has begun to examine the extent of factor misallocation in developing countries (Alfaro, Charlton, and Kanczuk, 2008; Alfaro and Chari, 2014; Barltesman, Haltiwanger, 2    and Scarpetta, 2009, 2013; Hsieh and Klenow, 2009; Restuccia and Rogerson, 2008, 2017). This paper contributes to this literature in two primary ways. First, it is the first paper to our knowledge to apply the Olley-Pakes decomposition methods to African data. To carry out this decomposition, we have generated new longitudinal data for Ethiopia (2012-2014) and Tanzania (2008-2012) and gained access to new panel data from Côte d’Ivoire.1 In the case of Ethiopia, our dataset adds to the existing panel which covers the period 1996 to 2009.2 Plus, we have generated a short panel for Bangladesh that covers every other year between 1995 and 2001. Using these data, we examine the sources of manufacturing productivity growth including the relative importance of factor reallocation for African countries. Several interesting findings emerge from this analysis. First, there are clear differences in the size and sources of productivity growth across both countries and plants within countries. In terms of country performance, Ethiopia outperformed both Côte d’Ivoire and Tanzania on nearly every metric that we estimated. For example, aggregate manufacturing productivity increased by 75 percent in Ethiopia between 2006 to 2016 while it increased by 6 percent in Côte d’Ivoire over a similar period (2004 to 2016). Aggregate manufacturing productivity fell in Tanzania over the period 2008 to 2012. Second, we find several similarities between Ethiopia’s pattern of productivity growth in recent years and that of Bangladesh during its early years of industrialization. For example, Bangladesh’s aggregate manufacturing productivity increased by 52 percent between 2001 and 2012 which is comparable to that achieved by Ethiopia between                                                              1  We’d like to thank Nouhoum Traore for sharing the Ivorian panel data with us.  2  In addition, we have establishment-level data for 2010, 2011, 2015, and 2016. We hope to match the establishments in those years to our existing dataset so that we have a panel covering the period 1996 to 2016. Similarly, we have one additional year of data for Tanzania (2013) and hope to extend the Tanzanian panel from 2008 to 2013. 3    2006 and 2016. Also, Ethiopia’s garment and textile sector achieved comparable productivity growth rates to that achieved by Bangladesh’s garment and textile industry. Perhaps most interesting are the results that emerge when we estimate the sources of manufacturing productivity growth by firm size and industry. We find strong evidence, for example, that entrants are the main driver of manufacturing productivity growth among firms with less than 200 workers but not among larger firms. This finding raises the possibility that selection mechanisms—particularly the role of entry—may differ by firm size. In addition, we reveal that market share reallocations play an important role in driving productivity growth among garment and textile producers in both Ethiopia and Bangladesh. Since garment and textile producers often face significant competition from abroad, we estimate the impact of trade exposure on plant-level total factor productivity (TFP). This is accomplished by employing semi-parametric estimation techniques to estimate plant-level total TFP and then use these TFP measures in a second stage regression (following Pavcnik, 2002) to estimate the effects of increased trade exposure on plant-level productivity. Importantly, our identification method separates the variation in productivity that arises due to changes in trade exposure from those emanating from other sources. Two important results stand out from this analysis. First, we find strong evidence that increased trade exposure significantly raises plant- level TFP in all three African countries. However, we find no evidence that increased trade exposure results in stronger selection mechanisms that weed out less efficient firms. The rest of the paper is organized as follows. Section II highlights the important role that industrialization plays in the development process. In addition, this section discusses how productivity gaps between plants might persist in long run equilibrium. Section III describes the decomposition methods that we use—both the static Olley-Pakes productivity decomposition and 4    the dynamic Olley-Pakes productivity decomposition. Section IV describes our data and their sources. Section V highlights our main results. Finally, Section VI concludes the paper. II. Conceptual Background An economy’s aggregate productivity is the weighted sum of plant-level productivity where the weights are the market shares of individual plants. Economic growth occurs when individual plants become more productive but also when plants with above average productivity expand their market shares. This second component was well understood by the first generation of development economists who built multi-sector growth models based on this stylized fact (Lewis, 1956; Kuznets, 1961; Fei and Ranis, 1964). But, despite the passage of more than 50 years, empirical support of the underlying mechanisms that accelerate factor reallocation remains limited. Instead, much of the empirical growth literature during the past two decades has been based on one-sector models that seek to identify the “best” set of variables that are correlated with cross-country variations in income and growth. The weaknesses of such growth regressions are well known (see, for example, the critique by Durlauf, Johnson, and Temple, 2005). More recently, increased attention has been given to identifying the mechanisms that facilitate and constrain the efficient allocation of resources across heterogenous production units (see Hopenhayn, 2014 and Restuccia and Rogerson, 2017 for good reviews of this literature). Recent estimates suggest that such misallocation can explain up to 60 percent of the aggregate TFP differences between rich and poor countries (Kalemli-Ozcan and Sorensen, 2012). The magnitude of this dispersion and its persistence over time raises the question of how low-productivity 5    producers survive in the same industry as high-productivity producers, particularly in the long run. One view is that “the observed dispersion reflects the frictions and perhaps distortions that prevent resources from being immediately reallocated to the most productive firms” (Haltiwanger, 2015, p. 343). Examples include regulations that levy higher taxes on producers that are larger in terms of employment, sales or capital (Hopenhayn and Rogerson, 1993; Guner, Ventura, and Xu, 2008); financial frictions that distort the allocation of capital across producers making the same good (Buera et al, 2011); and trade policies that generate a wedge that prevents the equalization of marginal products across heterogenous producers (Eaton and Kortum, 2002; Melitz, 2003; Eaton, Kortum, and Kramarz, 2011). Another view is that “there may be sources of curvature in the profit function, so the most productive firms do not take over the market” (Haltiwanger, 2015, p. 343). For example, heterogeneous firms may face downward sloping demand curves due to product differentiation or high transport costs that reduce the scale of market selection via firm entry and exit (Melitz, 2003; Syverson, 2004a, 2004b; Melitiz and Ottaviano, Jones et al, 2018). Such models have become increasingly popular over the last decade as new studies find evidence documenting the existence of substantial price dispersion across producers, even those operaing within the same narrowly defined industry. In addition, productivity gaps might arise due to selection and learning dynamics among young firms. According to Jovanovic (1982), new firms do not fully know their own productivity but instead enter the market with some prior belief about their relative position in their sector’s productivity distribution. Each period, they update their beliefs as new information becomes available. Efficient firms grow and survive while inefficient firms exit the market. In this model, “firms differ in size not because of the fixity of capital, but because some learn they are more 6    efficient than others” (Jovanovic, 1982, p. 649). As a result, learning dynamics and selection mechanisms that weed out inefficient firms play a key role in explaining industry dynamics. The Jovanovic model is consistent with several empirical findings including the observation that young firms have higher and more variable growth rates than older, more established firms. Extensions of the Jovanovic model posit that learning dynamics are important not only to young firms but to older firms as well. Hopenhayn (1992), for example, argues that firms are subject to idiosyncratic shocks over their lifespan and must adapt to changing economic circumstances to survive. While a firm’s past successes are correlated to its future successes, the ability to adapt to changing market conditions is what separates successful firms from unsuccessful firms. This combination of idiosyncratic firm shocks, learning dynamics, and firm selection results in resources being constantly reallocated across firms. Over time, new firms enter the market, surviving firms expand or contract, and inefficient firms exit the market. Such reallocation results in high turnover rates across both firms and jobs. For instance, in the United States during the 1990s, about one-third of the stock of jobs and over forty percent of manufacturing firms exited the market and then were replaced by new entrants during each five-year period (Hopenhayn, 1992). Such “churning” in the market results from productivity changes at the establishment level and lies at the heart of all firm dynamics. III. Productivity Decomposition In this paper, we use both the static and dynamic Olley-Pakes decomposition (Melitz and Polanec 2015) to examine the size and sources of productivity changes in four developing countries: Bangladesh, Côte d’Ivoire, Ethiopia, and Tanzania. These decompositions breakdown aggregate 7    productivity Φ into two components: 1) the contribution due to a shift in the firm-level distribution of productivity, ; and 2) the contribution due to market share reallocations, . That is, Φ ̅ ,  (1)    where the bar over a variable denotes the mean over all establishments in a given year. In this framework, the covariance represents the contribution to aggregate productivity that results from a reallocation of resources across establishments with different productivity levels. If the covariance is positive, it indicates that resources are moving from relatively less-productive establishments to relatively more-productive. Similarly, if the covariance is negative, it indicates that resources are moving from relatively more-productive establishments to relatively less- productive establishments. Following Melitz and Polanec (2015), we can decompose such aggregate productivity changes, ∆Φ, over time in terms of the contribution of different groups of firms. For example, say we have three groups of firms: survivors (S), entrants (E), and exiting firms (X). The resulting decomposition is: ΔΦ Φ Φ Φ Φ Φ Φ   (2)    = ∆ ∆ Φ Φ Φ Φ (3)  The first line decomposes aggregate productivity changes into the contributions made by each group. The first component accounts for changes in aggregate productivity ΔΦ that arise due to changes in plant-level productivity among surviving firms as well as changes in the relative productivity between survivors and other firms (e.g., entrants and exits). For example, aggregate 8    productivity rises when surviving firms become more productive but also when new firms with higher productivity enter the market and when inefficient firms with lower productivity exit the market. The second line further breaks down each group’s contribution. Specifically, the contribution of survivors can be decomposed into a component that captures the change in average productivity among survivors, Δ , and another component that captures the change in the covariance between the productivity of survivors and their market shares, . Similarly, the contribution of entrants can be decomposed into a component that captures the relative difference in the productivity of entrants and survivors, φ φ , and another component that captures the relative difference in the covariance between market shares and productivity across the two groups ( . The contribution of exiting firms is analogous to that of entrants: one component that captures the relative difference in the productivity of exiting firms and survivors, , and another component that captures the relative difference in the covariance between market shares and productivity across the two groups, . The intuition is simple: the contribution of entrants (or exiting firms) to aggregate productivity is the change in aggregate productivity that would have been generated if we were to add entrants (or remove exiting firms) to some initial distribution of firms. Since we cannot observe entrants in period 1 nor exiting firms in period 2, we cannot apply the same counterfactual in both periods. Instead, the DOPD method uses “the set of surviving firms as a benchmark and [then] asks how adding the group of entrants (or exiters) affects the aggregate productivity change” (Melitz and Polanec, 2014). Using a different reference period for each group is critical because of the timing of entry and exit. In the DOPD method, entrants only contribute to productivity growth if their productivity is higher than survivors when entry occurs (period 2). Likewise, exiting 9    firms only contribute to productivity growth if their productivity is higher than survivors when exit occurs (period 1). The fact that the DOPD specification uses different reference periods when estimating the impact of entry and exit is a clear advantage over other decomposition methods that use the same reference period (e.g., Griliches and Regev, 1995; Foster, Haltiwanger, and Krizan, 2001). These alternative methods are likely to result in measurement bias when measuring the contribution of one group or the other. While the dynamic decomposition has clear advantages over the static decomposition, we use both methods to take advantage of the cross-sectional data that we have for the period since 2016. These data are included in our analysis to highlight current trends as well as the different sources of manufacturing growth over the past decade. IV. Data The analysis of business dynamics requires data that both tracks individual establishments over time and is representative of different firm types (e.g., survivors, entrants, and exiting plants). Typically, these criteria are met only by manufacturing censuses and business registers as survey- based data are unlikely to be representative of entrants and exiters as well as firms at different stages of their life-cycle (e.g, young and mature plants). Table 1 lists the data sources that we use to conduct our analysis. In total, we have data on more than 67,000 establishments from 45 country-year samples. Its breakdown is as follows: data covering 21 years for Ethiopia (1996- 2016), 12 years for Côte d’Ivoire (2003-2014), six years for Tanzania (2008-2013), and six years for Bangladesh (1995, 1997, 1999, 2001, 2005, and 2012). 10    Our first step in the productivity decomposition is to classify each establishment as either an entrant, exiter, or a survivor. Following Bartelsman, Haltiwanger, and Scarpetta (2013), we adopt the following definitions: Entrant (E): Entrants and their employees are defined by the first year they are observed in the registry. Entrants are those observed as (out, in) in the registry at time (t-1, t). Exiter (X): Exiters and their employees are defined by the last year they are observed in the registry. Exiters are those observed as (in, out) in the registry at time (t-1, t). Survivor (S): Survivors and their employees are those establishment that were in the registry for two consecutive years. Survivors are those observed as (in, in) at time (t-1, t). Given these definitions, there is a clear link between the change in the stock of survivors and the number of entrants and exits in any two-year period. This relationship can be expressed as: . (4)  Given that survivors and entrants both exist in time , the total number of establishments in any year is defined as the sum of survivors and entrants. That is, . (5)  This implies that the change in the total number of establishments between year and 1 can be defined as the difference between the number of entrants and exits. We write this relationship as: . (6)  11    In other words, the total change in the number of establishments, ∆, between period and 1 is simply the difference between the cumulative sum of entrants and the cumulative sum of exits over the period. That is, ∆ Using these definitions, we can classify each establishment into one of three categories: survivors (S), entrants (E), and exiting plants (X). Several patterns are worth noting. First, all four countries exhibit considerable firm turnover. About 20% of operating plants “enter” the market each year while another 18% of plants “exit.” Given the fact that entry rates exceed exit rates, the number of plants in each country rises over time. While these increases are relatively modest in the African countries, Bangladesh ends the period with more than 10,000 new manufacturing plants. Table 2 provides some descriptive statistics on the manufacturing sector in each country. The first pattern to notice is that African establishments are smaller than those in Bangladesh. In 2012, the median establishment in 2012 employed 33 workers in Côte d’Ivoire, 20 workers in Ethiopia, 20 workers in Tanzania, and 119 workers in Bangladesh.3 Perhaps more important, the average scale of plants in Africa does not appear to be rising whereas it is clearly rising in Bangladesh. A quick look at the size distribution of African manufacturing establishments (Figures 1a-1d) reveals that the largest share of establishments employs less than 50 workers and increases in size over time. For example, the share of establishments employing less than 50 workers increases from 69 percent to 76 percent in Ethiopia (between 1996 and 2016); from 47 percent to 60 percent in Côte d’Ivoire (between 2003 and 2014); and from 62 percent to 68 percent in                                                              3  In each country, these numbers are based on establishments with 10+ workers. 12    Tanzania (between 2008 and 2013). In addition, all African countries experience a monotonic decline in the share of larger establishments over these time periods. The opposite trend occurs in Bangladesh where the share of establishments employing less than 50 workers falls from 79 percent to 62 percent and the share of larger establishments (e.g., those employing 50 to 199 workers) rises by 12 percent 25 percent over the period 2001 to 2012. The failure of the African manufacturing sector to generate scale should concern policy makers, particularly given recent evidence that larger establishments are highly correlated with both increased industrialization and higher income per capita (Buera and Kaboski, 2011; Buera et al, 2012; Bento and Restuccia, 2017). These cross-country differences in scale may be partially driven by differences in the sectoral composition of the manufacturing sector in each country. As revealed in Figures 3a-3d, there are clear differences between manufacturing activities pursued by African and Asian establishments. In Bangladesh, 43 percent of all manufacturing establishments produce either textiles or garments (in 2012) whereas this share is just 13 percent in Ethiopia (in 2016), 5 percent in Tanzania (in 2013) and 2 percent in Côte d’Ivoire (2014). Since garments and textiles are often produced for global markets, this suggests that African establishments may not have a comparative advantage in these sectors. By contrast, the share of African establishments engaged in manufacturing activities that cater to domestic markets (e.g., food and beverages, furniture) remains high. In all three African countries, establishments that produce food and beverages comprise the largest share of manufacturing firms and the largest share of manufacturing employment (see Figures 5a-5d). Forty-three percent of manufacturing workers are employed by the food and beverage industry in Tanzania (in 2013) while 23 percent are employed in Ethiopia (in 2016) and 22 percent are employed in Côte d’Ivoire (in 2014). In Bangladesh, this percentage is only 6 percent. 13    Are these differences in sectoral composition and scale being driven by differences in plant-level productivity across establishments? We next turn to this issue by examining the sources of productivity growth among different categories of plants. Using both the static Olley-Pakes (1996) decomposition and the dynamic Olley-Pakes decomposition (DOPD) developed by Melitz and Polanec (2015). Our preferred method is the Melitz and Polanec (2015) decomposition but utilize the static decomposition method when panel data are not available. As discussed by Olley- Pakes (1996), this method is more directly linked to theoretical models of firm dynamics developed to analyze the sources of market share reallocations across heterogenous firms (see, for example, Hsieh and Klenow, 2009; Collard-Wexler, Asker and de Loecker, Bartlesman, Haltiwanger, and Scarpetta, 2011; and Restuccia and Rogerson, 2017). V. Results Tables 3 and 4 report the results of the Olley-Pakes decomposition methods. Several patterns are worth noting. First, aggregate productivity growth in Ethiopia (Table 3) outstripped that of both Côte d’Ivoire and Tanzania (Table 4). In Ethiopia, aggregate productivity increased by 76 percent between 2006 and 2016 whereas it increased by only 6 percent in Côte d’Ivoire over a similar period (2004 to 2014).4 The weak performance of the Ivorian manufacturing sector is not surprising given the country’s two civil wars and extreme political instability during this period. However, growth rates rebounded following the peace agreement that ended the Second Ivorian Civil War (2010-2011). As reported in Table 4, aggregate productivity increased by 30 percent in                                                              4  Table 5 reports that aggregate productivity growth increased by 47 percent between 1996 and 2009. Note our estimates contradict earlier findings by Newman et al (2016) drawn from the Groningen Africa Database who estimate manufacturing productivity changes based on aggregate (sectoral) data. 14    2012 and by 27 percent in 2013. Somewhat surprisingly, we find that aggregate manufacturing productivity in Tanzania fell each year between 2009 and 2012. Second, we find evidence that market share reallocations contributed substantially to productivity growth in each country, although with vary magnitudes. In Ethiopia, the impact of market share reallocations among survivors (represented by Δcov) tended to be larger than those associated with increases in within-plant productivity (represented by the Δ in unweighted productivity). In addition, we find evidence that exiting plants boost productivity more than entering plants. The relative productivity of survivors is higher than that of exiting plants, suggesting that selection mechanisms are weeding out the least productive plants. In a related paper, Jones et al (2018) find that exiting plants in Ethiopia have lower physical total factor productivity (TFPQ) than surviving plants, but only when after controlling for producers’ transport costs. In Côte d’Ivoire, market share reallocations also play an important role in raising aggregate productivity. But, the pattern is opposite to what we find for Ethiopia. In Côte d’Ivoire, entering plants—not exiting plants—tend to have a larger impact on aggregate productivity growth. The relative productivity of entering plants is higher than that of surviving plants in 8 out of the 10 years between 2004 and 2014. By contrast, exiting plants reduce productivity in 7 out of 10 years during this period. Once again, a different pattern emerges in Tanzania. In Tanzania, we find that market share reallocations among surviving plants appear to have a smaller impact on aggregate productivity growth than both entry or exit. So how do these patterns compare to those in Bangladesh? Table 5 presents the decomposition of aggregate productivity growth for Bangladesh. While our data are limited, the evidence suggests that market share reallocations among surviving 15    firms and exiting firms have the largest impact on productivity growth in Bangladesh. This pattern is similar to what we find for Ethiopia. Next, we examine how the sources of productivity growth differ by firm size and industry. Given the large number of years covered, we present the results in roughly five-year intervals for each country. The results are reported in Tables 6 and 7. Several interesting findings emerge. First, the smallest plants (those with less than 50 workers) experienced the largest productivity growth in three out of the four countries (Ethiopia, Tanzania, and Bangladesh). This is important as plants in this size category comprise the majority of formal sector plants with at least 10 workers in each country. Côte d’Ivoire is the exception with slightly larger plants (those with 50 to 199 workers) experiencing the fastest productivity growth. Another key finding is that entrants are the main source productivity growth among plants with less than 200 workers and exiting plants have depress productivity in plants w200 to 499 workers. This raises the possibility that selection mechanisms may have different effects on plants that vary in size. Finally, we conduct the productivity decomposition for plants in different industries. Initially, we focus on three industries: 1) garments and textiles; 2) food and beverages; and 3) furniture. While garments and textiles tend to be export-intensive industries, firms in the food and furniture industries tend to sell their products locally. Among garment and textile producers, Ethiopian plants achieved the highest productivity growth of the four countries. Aggregate productivity in Ethiopia rose by 24 percent between 1996 and 2001 and by 30 percent between 1996 and 2006. In Bangladesh, aggregate productivity increased by 33 percent between 1995 and 2001. Interestingly, this productivity growth was largely driven mainly by more efficient survivors gaining market shares and by less efficient plants exiting the market. A similar pattern emerges among garment and textile plants in Ethiopia. During the first period, market share reallocations 16    were the primary source of productivity growth whereas, during the second period, the entry of more efficient plants was the driving force. Neither Côte d’Ivoire nor Tanzania experienced positive productivity growth in their garments and textile industries over the periods for which we have data. Mixed results are found for firms in the food and beverage industry as well as for firms in the furniture industry. Overall aggregate productivity in the food and beverage industry grew by 11 percent in Ethiopia between 1996 and 2006 and by 33 percent in Bangladesh between 1995 and 2001. This productivity growth was driven by market share reallocations among survivors in Ethiopia and by selection mechanisms in Bangladesh. By contrast, aggregate productivity fell for plants producing food and beverages in both Côte d’Ivoire and Tanzania. For Côte d’Ivoire, this fall in productivity was due to surviving firms becoming less productive, a likely result of the political unrest in the country. In Tanzania, however, productivity fell as an increasing share of output was shifted to less productive plants in the industry. Lastly, we examine the sources of productivity in the furniture industry in each country. Both Ethiopia and Côte d’Ivoire experienced a rise in aggregate productivity among furniture producers. Aggregate productivity rose by 36 percent between 1996 and 2006 in Ethiopia and by 19 percent in Côte d’Ivoire between 2004 and 2014. The main source of productivity growth in Ethiopia was due to surviving plants becoming more productive and gaining market share. By contrast, the main source of productivity growth in Côte d’Ivoire was due to more productive survivors gaining market share and less efficient plants exiting the market. It is often argued that increased trade exposure in a market strengthens the selection mechanisms that weed out less efficient firms. Is there any evidence that increased trade exposure has this effect on African manufacturing firms? We test this hypothesis using a two-stage 17    estimation technique similar to that developed by Pavnik (2002). In the first stage, we estimate plant-level TFP using the Levinsohn-Petrin (2003) technique for estimating production functions. This estimation technique controls for the possible correlation between a plant’s chosen input levels and its unobserved plant-level productivity shocks by using intermediate inputs as a proxy for the unobserved productivity shock. Once we have estimated plant-specific TFP, we use these TFP measures in a second stage regression (following Pavcnik) to estimate the effects of increased trade exposure on plant-level productivity. Importantly, our identification method separates the variation in productivity that arises due to changes in trade exposure from those emanating from other sources. We do this by exploiting both the variation in productivity over time and across plants with different exposures to trade. We estimate the following equation:   ∗ (7)  ∗ + where is the unweighted productivity estimate for plant in year , is a vector of year indicators, is a dummy variable that indicates whether plant has positive foreign sales, and is a vector of plant characteristics including industry fixed effects and whether the plant exits the market in any given year. In this specification, the direct effects of trade are represented by the coefficient and the indirect effects of trade (via exit) are represented by the coefficient . Tables 8, 9 and 10 report the results of the second-stage regression. Two important results stand out from this analysis. First, we find strong evidence that increased trade exposure significantly raises plant-level TFP. For each country, the coefficient on exporter is both positive and significant at the 1% level. However, we find no evidence that increased trade exposure results in stronger selection mechanisms that weed out less efficient firms. 18    VI. Conclusion This paper examines the size and sources of productivity growth for three African countries— Ethiopia, Côte d’Ivoire, and Tanzania—and then compares these productivity gains to those of an external comparator: Bangladesh. We utilize static and dynamic Olley-Pakes decomposition and to obtain several interesting findings emerge from this analysis. First, there are clear differences in the size and sources of productivity growth across both countries and plants within countries. In terms of country performance, Ethiopia outperformed both Côte d’Ivoire and Tanzania on nearly every metric that we estimated. Second, we find strong evidence that entrants are the main driver of productivity growth among small and medium-sized firms (those with less than 200 workers) but not among the countries’ largest firms (those with 200+ employees). This finding raises the possibility that selection mechanisms—particularly the role of entry—may differ by firm size. In addition, the results reveal that market share reallocations play an important role in driving productivity growth among garment and textile producers in both Ethiopia and Bangladesh. Furthermore, the findings provide strong evidence that increased trade exposure significantly raises plant-level TFP but do not show that increased trade exposure in a market results in stronger selection mechanisms that weed out less efficient firms. 19    References Alfaro, Laura, Andrew Charlton, and Fabio Kanczuk. 2008. “Plant Size Distribution and Cross- Country Income Differences.” NBER Working Paper 14060. Alfaro, Laura and Anusha Chari. 2014. “Deregulation, Misallocation and Size: Evidence from India,” Journal of Law & Economics 57(4): 897-936. Asker, John, Collard-Wexler, Alan, and Jan de Loecker. 2014. “Dynamic Inputs and Resource (Mis)Allocation,” Journal of Political Economy 122(5): 1013-1063. Banerjee and Duflo. 2004. “Growth Theory through the Lens of Development Economics,” in Aghion Philippe and Steven Durlauf (Eds.) Handbook of Economic Growth, Vol. 1. Amsterdam: Elseveier. Bartelsman, Eric, John Haltiwanger, and Stefano Scarpetta. 2009. “Measuring and Analyzing Cross-country Differences in Firm Dynamics,” in Dunne, Timothy, J. Jensen, and Mark J. Roberts (Eds.) Producer Dynamics: New Evidence from Micro Data. Chicago: University of Chicago. Bartelsman, Eric, John Haltiwanger, and Stefano Scarpetta. 2013. “Cross-Country Differences in Productivity: The Role of Allocation and Selection,” American Economic Review 103(1): 305-334. Bento, Pedro, and Diego Restuccia. 2017. “Misallocation, Establishment Size, and Productivity,” American Economic Journal: Macroeconomics 9(3): 267-303. Buera, Francisco J., Joseph P. Kaboski, and Yongseok Shin. 2011. “Finance and Development: A Tale of Two Sectors.” American Economic Review 101(5): 1964-2002. Buera, Francisco J. and Joseph P. Kaboski. 2012. “Scale and the Origins of Structural Change,” Journal of Economic Theory 147:684-712. Durlauf, Steven D, Paul A. Johnson, and Jonathan R.W. Temple. 2005. “Growth Econometrics,” in Aghion Philippe and Steven N. Durlauf (Eds). Handbook of Economic Growth, Vol. 1. Amsterdam: Elsevier. Eaton, Jonathan and Samuel Kortum. 2002. “Technology, Geography, and Trade,” Econometrica 70(5): 1741-1779. Eaton, Jonathan, Samuel Kortum, and Francis Kramarz. 2011. “An Anatomy of International Trade: Evidence from French Firms,” Econometrica 79(5): 1453-1498. Foster, Lucia, John C. Haltiwanger, and C.J. Krizan. 2001. “Aggregate Productivity Growth: Lessons from Microeconomic Evidence,” in Hulten Charles R., Edwin R. Dean, and 20    Michael J. Harper (Eds), New Developments in Productivity Analysis. Chicago: University of Chicago Press. Gollin, Douglas, Stephen L. Parente, and Richard Rogerson. 2004. “Farm Work, Home Work, and International Productivity Differences,” Review of Economic Dynamics (7): 827-850. Griliches, Zvi and Haim Regev. 1995. “Firm Productivity in Israeli Industry: 1979-1988.” Journal of Econometrics 65: 175-203. Gunner, Nezih, Gustavo, Ventura, and Yi Xu. 2008. “Macroeconomic Implications of Size- Dependent Policies,” Review of Economic Dynamics 11(4): 721-744. Haltiwanger, John. 2015. “Job Creation, Job Destruction, and Productivity Growth: The Role of Young Businesses,” Annual Review of Economics 7: 341-358. Hopenhayn, Hugo. 1992. “Entry, Exit, and Firm Dynamics in Long run Equilibrium,” Econometrica 60(5): 1127-1150. Hopenhayn, Hugo. 2014. “Firms, Misallocation, and Aggregate Productivity: A Review,” Annual Review of Economics 6: 735-770. Hopenhayn, Hugo and Richard Rogerson. 1993. “Job Turnover and Policy Evaluation: A General Equilibrium Analysis,” Journal of Political Economy 101(5): 915-938. Hsieh, Chang-Tai and Peter J. Klenow. 2014. “The Life Cycle of Plants in India and Mexico,” Quarterly Journal of Economics 129(3): 1035-1084. Jovanovic, Boyan. 1982. “Selection and the Evolution of Industry,” Econometrica 50(3):649-670. Jones, Patricia, Taye Mengistae, and Albert Zeufack. 2018. “Selection, Firm Turnover, and Productivity Growth,” World Bank Policy Research Working Paper 8291. Kalemli-Ozcan, Sebnem and Bent E. Sorenson. 2012. “Misallocation, Property Rights, and Access to Finance: Evidence from Within and Across Africa.” NBER Working Paper 18030. Levinsohn, James and Amil Petrin. 2003. “Estimating Production Functions using Inputs to Control for Unobservables,” Review of Economic Studies 70(2): 317-341. Melitz, Marc J. “The Impact of Trade on Intra-industry Reallocations and Aggregate Industry Productivity,” Econometrica 70(6): 1695-1725. Melitz, Marc J. and G.I.P. Ottaviano. 2008. “Market Size, Trade and Productivity”, Review of Economic Studies 75: 295-316. Melitz, Marc J. and G.I.P. Ottaviano. 2008. “Market Size, Trade and Productivity”, Review of Economic Studies 75: 295-316. 21    Newman, Carol, John Page, John Rand, Abebe Shimeles, Måns Söderbom, and Finn Tarp. 2016. Made in Africa: Learning to Compete in Industry. Washington, DC: Brookings Institution Press. Page, John. 2012. “Can Africa Industrialize?” Journal of African Economies, Vol. 21. AERC Supplement 2, pp. ii86-ii125. Page, John. 2016. “Are Certain Countries Doomed to Remain Emerging? Globalization, Structural Change, and the Uneven Geography of Convergence.” Report #6 in 11 Global Debates. Brookings Institute. Page, John. 2018. “Rethinking Africa’s Structural Transformation: The Rise of New Industries.” In Brahima S. Couliby (Ed.) Foresight Africa: Top Priorities for the Continent in 2018 (pp. 64-80). Washington, DC: Brookings Institute. Olley, G. Steven and Ariel Pakes. 1996. “The Dynamics of Productivity in the Telecommunications Equipment Industry.” Econometrica 64: 1263–1297. Pavcnik, Nina. 2002. “Trade Liberalization, Exit, and Productivity Improvements: Evidence from Chilean Plants,” Review of Economic Studies 69(1): 245-276. Melitz, Marc and Sašo Polanec. 2015. “Dynamic Olley-Pakes Productivity Decomposition with Entry and Exit.” Rand Journal of Economics 46(2): 362-375. Restuccia, Diego and Richard Rogerson. 2008. “Policy Distortions and Aggregate Productivity with Heterogenous Establishments,” Review of Economic Dynamics 11(4): 707-720. Restuccia, Diego and Richard Rogerson. 2017. “The Causes and Costs of Misallocation,” Journal of Economic Perspectives 31(3): 151-174. Rodrik, Dani. 2015. “Premature Deindustrization.” NBER Working Paper 20935. Rodrik, Dani. 2017. “Growth without Industrialization?” Project Syndicate (October 10, 2017). Syverson, C. 2004a. “Market Structure and Productivity: A Concrete Example,” Journal of Political Economy 112(6) : 1181-1222. Syverson, C. 2004b. “Product Substitutability and Productivity Dispersion,” Review of Economics and Statistics 86(2): 534-550. Syverson, C. 2011. “What Determines Productivity?” Journal of Economic Literature 49(2): 326- 365. United Nationas Industrial Development Organization (UNIDO). 2009. Industrial Development Report. Breaking In and Moving Up: New Industrial Challenges for the Bottom Billion and the Middle Income Countries. Vienna, Austria. 22    23    TABLE 1: Data Sources Used for Productivity Decomposition Country Name of Survey/Census Source Period Sectors Panel Threshold Côte d’Ivoire Registrar of Companies for the Business 2003-2014 All yes None Modern Enterprise Sectors Registrar sectors Ethiopia Survey of Large & Medium Scale Manufacturing 1996-2009 Manufacturing yes Employment ≥ 10 Enterprises Census 2012-2014 only Survey of Large & Medium Scale Manufacturing 2010-2016 Manufacturing no Employment ≥ 10 Enterprises Census only Tanzania Annual Survey of Industrial Manufacturing 2008-2012 Industrial sectors yes Employment ≥ 10 Production (ASIP) Census only Census of Industrial Production Industrial 2013 All no None (CIP) Census sectors Bangladesh Census of Manufacturing Industries Manufacturing 1995, 1997, Manufacturing yes Employment ≥ 10 (CMI) Census 1999, 2001 only Census of Manufacturing Industries Manufacturing 2005, 2012 Manufacturing no Employment ≥ 10 (CMI) Census only 24    TABLE 2: Descriptive Statistics Ethiopia: Côte d’Ivoire: Tanzania: Bangladesh: 1996-2016 2003-2014 2008-2013 1995-2012 # of # of Median # of # of Median # of # of Median # of # of Median Year firms workers firm size firms workers firm size firms workers firm size firms workers firm size 1995 28,769 2,016,638 70 1996 623 82,079 21 1997 697 83,138 20 28,297 2,175,025 82 1998 725 81,764 20 1999 725 89,508 21 26,969 2,230,087 103 2000 739 79,839 21 2001 722 73,568 22 27, 276 2,801,441 86 2002 883 80,126 19 2003 939 83,655 20 332 69,306 56 2004 997 86,916 22 342 65,729 51 2005* 763 86,161 35 339 60,596 54 27,461 2,347,769 90 2006 1,153 102,655 24 328 62,872 54 2007 1,339 111,763 20 335 62,110 54 2008 1,734 112,997 17 366 74,878 50 649 107,388 29 2009 1,948 126,837 16 398 63,130 42 659 93,573 30 2010 1,958 157,817 21 413 75,043 41 1,079 105,752 18 2011 1,934 123,997 16 468 74,345 34 1,087 109,957 18 2012 2,129 178,708 20 526 81,230 33 1,091 117,562 20 41,716 4,982,446 119 2013 2,388 202,347 21 569 74,820 33 998 107,732 22 2014 2,405 196,637 20 591 91,373 32 2015 2,718 200,390 15 2016 2,719 227,971 17 25    Table 3: Ethiopia Decomposition of Aggregate Productivity Growth Ethiopia Surviving Firms Dynamic OP Aggregate Δ Unweighted Δ Entering Exiting T=1 T=2 Productivity Productivity Covariance Firms Firms 1996 1997 0.06 0.02 0.04 -0.02 0.02 1996 1998 0.01 -0.12 0.13 -0.03 0.03 1996 1999 0.09 -0.15 0.22 -0.03 0.04 1996 2000 0.19 -0.05 0.22 0.02 0.00 1996 2001 0.20 -0.20 0.21 0.11 0.09 1996 2002 0.02 -0.08 0.07 -0.01 0.04 1996 2003 0.20 -0.04 0.21 -0.01 0.05 1996 2004 0.28 0.06 0.19 -0.03 0.06 1996 2005 0.26 0.07 0.07 0.08 0.05 1996 2006 0.37 0.03 0.13 0.16 0.04 1996 2007 0.55 0.15 0.24 0.09 0.07 1996 2008 0.58 0.25 .023 -0.04 0.15 1996 2009 0.47 0.07 0.47 -0.17 0.09 … 2012 2013 0.08 0.02 0.01 0.02 0.02 2013 2014 0.22 0.17 0.02 -0.08 0.11 Ethiopia Static OP Aggregate Δ Unweighted Δ T=1 Productivity Productivity Covariance T=2 2006 2007 0.18 -0.05 0.23 2006 2008 0.22 -0.07 0.29 2006 2009 0.10 -0.19 0.29 2006 2010 0.08 0.33 -0.25 2006 2011 0.80 0.84 -0.04 2006 2012 0.17 0.32 -0.15 2006 2013 0.29 0.46 -0.17 2006 2014 0.39 0.56 -0.16 2006 2015 0.61 0.34 0.27 2006 2016 0.76 -0.07 0.76 26    Table 4: Côte d’Ivoire & Tanzania Decomposition of Aggregate Productivity Growth Côte Surviving Firms d’Ivoire Aggregate Δ Unweighted Δ Entering Exiting Dynamic OP Productivity Productivity covariance Firms Firms T=1 T=2 2004 2005 -0.02 -0.10 0.03 0.07 -0.01 2004 2006 -0.07 -0.10 0.00 -0.01 0.05 2004 2007 0.00 -0.01 -0.03 0.11 -0.07 2004 2008 -0.01 -0.05 0.01 0.08 -0.05 2004 2009 0.03 -0.11 -0.03 0.19 -0.02 2004 2010 0.20 -0.09 0.04 0.26 -0.01 2004 2011 -0.03 -0.27 0.12 0.10 0.02 2004 2012 0.27 -0.26 0.30 0.20 0.03 2004 2013 0.47 -0.20 0.28 0.39 0.00 2004 2014 0.06 -0.27 0.40 -0.01 -0.05 Tanzania Surviving Firms Dynamic OP Aggregate Δ Unweighted Δ Entering Exiting T=1 T=2 Productivity Productivity covariance Firms Firms 2008 2009 0.22 0.22 -0.05 0.02 0.06 2008 2012 -0.14 -0.14 0.14 0.10 0.03 2008 2013 -0.08 -0.08 -0.02 0.25 0.01 2008 2014 -0.16 -0.16 0.05 0.17 -0.03 27    Table 5: Bangladesh Decomposition of Aggregate Productivity Growth Bangladesh Surviving Firms Dynamic OP Aggregate Δ Unweighted Δ Entering Exiting T=1 T=2 Productivity Productivity covariance Firms Firms 1995 1997 0.13 -0.04 0.30 -0.35 0.21 1995 1999 0.01 -0.12 0.28 -0.21 0.06 1995 2001 0.33 -0.44 0.81 -0.19 0.16 Bangladesh Static OP Aggregate Unweighted T=1 T=2 Productivity Productivity Covariance 2001 2005 0.46 0.22 0.24 2001 2012 0.52 0.57 -0.05 28    Table 6: Decomposition of Aggregate Productivity Growth by Firm Size Plants with 10-49 Workers Plants with 50-199 Workers Aggregate Δ Unweighted Entering Exiting Aggregate Δ Unweighted Entering Exiting Country Year Productivity Productivity Δ Cov Firms Firms Productivity Productivity Δ Cov Firms Firms Ethiopia 1996-2001 1.16 -0.20 0.29 1.18 -0.12 0.11 -0.34 0.02 0.34 0.09 1996-2006 0.48 0.35 -0.03 0.21 -0.05 0.44 -0.04 0.07 0.29 0.12 Côte d’Ivoire 2004-2009 -0.16 -0.13 0.08 0.41 -0.52 0.22 0.14 -0.27 0.48 -0.13 2004-2014 -0.65 -0.20 0.10 0.15 -0.70 0.21 -0.17 0.18 0.43 -0.23 Tanzania 2008-2012 0.47 0.23 0.27 0.27 -0.31 -1.43 -0.59 -1.37 0.15 0.37 Bangladesh 1995-2001 0.13 -0.29 -0.21 0.32 0.31 0.02 -0.37 -0.09 0.27 0.20 Plants with 200-499 Workers Plants with 500+ Workers Aggregate Δ Unweighted Entering Exiting Aggregate Δ Unweighted Entering Exiting Country Year Productivity Productivity Δ Cov Firms Firms Productivity Productivity Δ Cov Firms Firms Ethiopia 1996-2001 -0.22 -0.14 -0.08 -0.13 0.013 0.30 -0.07 0.25 0.07 0.06 1996-2006 0.02 0.13 -0.14 0.23 -0.20 0.55 -0.02 0.43 0.13 -0.01 Côte d’Ivoire 2004-2009 -0.06 -0.06 -0.35 0.28 0.07 -0.18 -0.10 -0.17 -0.03 0.12 2004-2014 -0.48 .024 -0.08 -0.20 0.03 -1.47 -0.29 -0.14 -0.13 -0.91 Tanzania 2008-2012 -0.54 0.05 -0.73 -0.13 0.28 -0.34 0.19 -0.25 0.05 -0.33 Bangladesh 1995-2001 0.12 -0.04 -0.15 0.29 0.01 -0.06 0.06 0.27 -0.36 -0.03 29    Table 7: Decomposition of Aggregate Productivity Growth by Industry Garments & Textiles Food & Beverages Aggregate Δ Unweighted Entering Exiting Aggregate Δ Unweighted ΔCov Entering Exiting Country Year Productivity Productivity Δ Cov Firms Firms Productivity Productivity Firms Firms Ethiopia 1996-2001 0.25 -0.27 0.35 0.19 -0.03 0.01 -0.32 0.04 0.16 0.12 1996-2006 0.54 0.05 -0.07 0.56 -0.02 0.11 -0.17 0.26 -0.04 0.07 Côte d’Ivoire 2004-2009 -0.40 -0.52 0.04 0.09 -0.01 -0.24 -0.13 -0.32 0.05 0.16 2004-2014 -1.03 -1.26 0.00 -0.82 1.05 -1.25 -0.36 0.10 -0.42 0.56 Tanzania 2008-2012 -0.50 -0.14 -0.36 0.02 -0.03 -0.70 0.16 -0.93 0.05 0.03 Bangladesh 1995-2001 0.33 -0.44 0.81 -0.19 0.16 0.47 0.01 -0.28 0.64 0.10 Furniture Aggregate Δ Unweighted Entering Exiting Country Year Productivity Productivity ΔCov Firms Firms Ethiopia 1996-2001 -0.02 -0.47 0.72 -0.25 -0.02 1996-2006 0.36 0.59 0.55 -0.59 -0.20 Côte d’Ivoire 2004-2009 -0.24 -0.05 0.07 -0.31 0.05 2004-2014 0.19 -0.27 0.39 -0.25 0.32 Tanzania 2008-2012 -0.47 0.34 -0.36 0.00 -0.45 Bangladesh 1995-2001 -2.34 0.06 -2.25 0.58 -0.72 30    Table 8: Impact of Trade Exposure on Productivity: Ethiopia Plant-Level TFP Plant-Level TFP Exporter 0.631*** 0.709*** (0.134) (0.140) Exporter*2013 -0.231 -0.167 (0.182) (0.185) Exporter*2014 -0.0328 -0.111 (0.178) (0.183) Exiting firm -0.215*** -0.197*** (0.0438) (0.0448) Exporter*Exiting Firm -0.376 (0.203) Year Fixed Effects Yes Yes Industry Fixed Effects Yes Yes Observations 6,004 6,004 R2 0.113 0.114 Standard errors in parentheses * p < 0.05, ** p < 0.01, *** p < 0.001   31 Table 9: Impact of Trade Exposure on Productivity: Cote d'Ivoire Plant-Level TFP Plant-Level TFP Exporter 0.514*** 0.489*** (0.0549) (0.0603) Exporter*2011 -0.0696 -0.0646 (0.218) (0.218) Exporter*2012 0.107 0.0962 (0.215) (0.215) Exporter*2013 0.109 0.0994 (0.220) (0.220) Exporter*2014 0.0635 0.0864 (0.228) (0.229) Exiting firm -0.0246 -0.0409 (0.0397) (0.0431) Exporter*Exiting Firm 0.105 (0.109) Year Fixed Effects Yes Yes Industry Fixed Effects Yes Yes Observations 6,515 6,515 R2 0.161 0.161 Standard errors in parentheses * p < 0.05, ** p < 0.01, *** p < 0.001 32 Table 10: Impact of Trade Exposure on Productivity: Tanzania Plant-Level TFP Plant-Level TFP Exporter 0.539* 0.526* (0.253) (0.262) Exporter*2009 -0.489 -0.497 (0.394) (0.397) Exporter*2010 -0.197 -0.192 (0.398) (0.399) Exporter*2011 0.252 0.251 (0.381) (0.381) Exporter*2012 -0.299 -0.286 (0.384) (0.390) Exiting firm -0.0891 -0.101 (0.136) (0.150) Exporter*Exiting Firm 0.0699 (0.362) Year Fixed Effects Yes Yes Industry Fixed Effects Yes Yes Observations 789 789 R2 0.095 0.095 Standard errors in parentheses * p < 0.05, ** p < 0.01, *** p < 0.001 33 34 Figure 1a: Ethiopia: Firm Count & Employment by Firm Size Number of Firms Share of Firms 1996 432 1996 0.69 10-49 Workers 2006 795 10-49 Workers 2006 0.69 2016 2,131 2016 0.76 1996 109 1996 0.17 50-199 Workers 2006 250 50-199 Workers 2006 0.22 2016 434 2016 0.16 1996 42 1996 0.07 200-499 Workers 2006 65 200-499 Workers 2006 0.06 2016 150 2016 0.05 1996 40 1996 0.06 > 500 Workers 2006 43 > 500 Workers 2006 0.04 2016 76 2016 0.03 Source: Ethiopia Survey of Large & Medium Scale Manufacturing Industries, Source: Ethiopia Survey of Large & Medium Scale Manufacturing Industries, 1996, 2006, & 2016. 1996, 2006, & 2016. Shares may not add to one due to rounding errors. Total Employment Employment Share 1996 7,461 1996 0.09 10-49 Workers 2006 14,809 10-49 Workers 2006 0.14 2016 36,820 2016 0.16 1996 11,973 1996 0.15 50-199 Workers 2006 23,431 50-199 Workers 2006 0.23 2016 43,789 2016 0.19 1996 13,808 1996 0.17 200-499 Workers 2006 20,650 200-499 Workers 2006 0.20 2016 44,957 2016 0.20 1996 48,838 1996 0.60 > 500 Workers 2006 43,765 > 500 Workers 2006 0.43 2016 102,405 2016 0.45 Source: Ethiopia Survey of Large & Medium Scale Manufacturing Industries, Source: Ethiopia Survey of Large & Medium Scale Manufacturing Industries, 1996, 2006, & 2016. 1996, 2006, & 2016. Shares may not add to one due to rounding errors 35 Figure 1b: Côte d'Ivoire Firm Count & Employment by Firm Size Number of Firms Share of Firms 2003 160 2003 0.47 10-49 Workers 2008 190 10-49 Workers 2008 0.50 2014 374 2014 0.60 2003 107 2003 0.31 50-199 Workers 2008 108 50-199 Workers 2008 0.28 2014 151 2014 0.24 2003 44 2003 0.13 200-499 Workers 2008 49 200-499 Workers 2008 0.13 2014 61 2014 0.10 2003 0.09 2003 32 > 500 Workers 2008 33 > 500 Workers 2008 0.09 2014 0.05 2014 33 Source: Côte d'Ivoire Registrar of Companies for the Modern Enterprise Source: Côte d'Ivoire Registrar of Companies for the Modern Enterprise Sectors, 2003, 2008, & 2014. Shares may not sum to one due to rounding Sectors, 2003, 2008, & 2014. errors Total Employment Employment Share 2003 3,662 2003 0.05 10-49 Workers 2008 4,503 10-49 Workers 2008 0.06 2014 7,943 2014 0.09 2003 11,114 2003 0.16 50-199 Workers 2008 10,099 50-199 Workers 2008 0.13 2014 14,276 2014 0.16 2003 13,466 2003 0.19 200-499 Workers 2008 14,078 200-499 Workers 2008 0.19 2014 18,472 2014 0.20 2003 0.59 2003 41,064 > 500 Workers 2008 46,198 > 500 Workers 2008 0.62 2014 0.55 2014 50,682 Source: Côte d'Ivoire Registrar of Companies for the Modern Enterprise Source: Côte d'Ivoire Registrar of Companies for the Modern Enterprise Sectors, 2003, 2008, & 2014. Shares may not sum to one due to rounding Sectors, 2003, 2008, & 2014. errors 36 Figure 1c: Tanzania Firm Count & Employment by Firm Size Number of Firms Share of Firms 2008 422 2008 0.62 10-49 Workers 10-49 Workers 2013 2013 682 0.68 2008 163 2008 0.24 50-199 Workers 50-199 Workers 2013 0.21 2013 212 2008 54 2008 0.08 200-499 Workers 200-499 Workers 2013 0.07 2013 67 2008 0.06 2008 41 > 500 Workers > 500 Workers 2013 0.04 2013 38 Source: Tanzania Annual Survey of Industrial Production, 2008 & Census Source: Tanzania Annual Survey of Industrial Production, 2008 of Industrial Establishments, 2013. Shares may not sum to one due to & Census of Industrial Establishments, 2013. rounding errors. Total Employment Employment Share 2008 8,477 2008 0.08 10-49 Workers 10-49 Workers 2013 13,863 2013 0.13 2008 15,946 2008 0.15 50-199 Workers 50-199 Workers 2013 21,110 2013 0.20 2008 17,322 2008 0.16 200-499 Workers 200-499 Workers 2013 0.20 2013 21,386 2008 0.61 2008 65,643 > 500 Workers > 500 Workers 2013 0.48 2013 51,374 Source: Tanzania Annual Survey of Industrial Production, 2008 & Census Source: Tanzania Annual Survey of Industrial Production, 2008 of Industrial Establishments, 2013. Shares may not sum to one due to & Census of Industrial Establishments, 2013. rounding errors. 37 Figure 1e: Bangladesh Firm Count & Employment by Firm Size Number of Firms Share of Firms 1995 22,830 1995 0.79 10-49 Workers 2005 20,444 10-49 Workers 2005 0.74 2012 25,974 2012 0.62 1995 3,532 1995 0.12 50-199 Workers 2005 3,270 50-199 Workers 2005 0.12 2012 10,251 2012 0.25 1995 1,640 1995 0.06 200-499 Workers 2005 1,416 200-499 Workers 2005 0.05 2012 3,791 2012 0.09 1995 767 1995 0.03 > 500 Workers 2005 2,333 > 500 Workers 2005 0.08 2012 1,700 2012 0.04 Source: Census of Manufacturing Industries, Source: Census of Manufacturing Industries, 1995, 2005, & 2012. Shares may not 1995, 2005, & 2012. sum to one due to rounding errors. Total Employment Employment Share 1995 384,460 1995 0.19 10-49 Workers 2005 419,730 10-49 Workers 2005 0.18 2012 573,713 2012 0.12 1995 361,069 1995 0.18 50-199 Workers 2005 297,136 50-199 Workers 2005 0.13 2012 1,043,740 2012 0.21 1995 520,058 1995 0.26 200-499 Workers 2005 462,905 200-499 Workers 2005 0.20 2012 1,100,902 2012 0.22 1995 751,051 1995 0.37 > 500 Workers 2005 1,168,044 > 500 Workers 2005 0.50 2012 2,264,091 2012 0.45 Source: Census of Manufacturing Industries, Source: Census of Manufacturing Industries, 1995, 2005, & 2012. Shares may not 1995, 2005, & 2012. sum to one due to rounding errors. 38 Table 2a: Ethiopia: Firm Count by Industry Number of Firms Number of Firms 1996 2016 Tobacco 1 Petroleum 1 Elec. Machinery 1 Recycling 1 Tobacco 2 Basic Metal 3 Precision Instr. 3 Paper 5 Media Equip. 10 Vehicles 11 Vehicles 10 Machinery 14 Elec. Machinery 12 Machinery 20 Rubber & Plastics 15 Paper 24 Garments 23 Basic Metal 63 Wood, not furn 26 Wood, not furn 64 Textiles 32 Leather 66 Chemicals 89 Chemicals 35 Garments 105 Printing 38 Printing 107 Fab. metal 39 Fab. metal 174 Leather 63 Rubber & Plastics 186 Furniture 75 Textiles 258 Furniture 395 Non-metallic metal 82 Non-metallic metal 533 Food & Bev 160 Food & Bev 668 0 100 200 300 400 500 600 700 0 100 200 300 400 500 600 700 Graphs by Year Graphs by Year Table 3a: Ethiopia: Firm Shares by Industry Share of Firms Share of Firms 1996 2016 Tobacco 0.2 Petroleum 0.0 Elec. Machinery 0.2 Recycling 0.0 Tobacco 0.1 Basic Metal 0.5 Precision Instr. 0.1 Paper 0.8 Media Equip. 0.4 Vehicles 1.8 Vehicles 0.4 Machinery 2.2 Elec. Machinery 0.4 Machinery 0.7 Rubber & Plastics 2.4 Paper 0.9 Garments 3.7 Basic Metal 2.3 Wood, not furn 4.2 Wood, not furn 2.3 Textiles 5.1 Leather 2.4 Chemicals 3.2 Chemicals 5.6 Garments 3.8 Printing 6.1 Printing 3.8 Fab. metal 6.3 Fab. metal 6.2 Leather 10.1 Rubber & Plastics 6.7 Furniture 12.0 Textiles 9.2 Furniture 14.2 Non-metallic metal 13.2 Non-metallic metal 19.1 Food & Bev 25.7 Food & Bev 23.9 0 5 10 15 20 25 30 0 5 10 15 20 25 30 Graphs by Year Graphs by Year 39 Table 2b: Côte d'Ivoire Firm Count by Industry Number of Firms Number of Firms 2003 2014 Tobacco 2 Tobacco 2 Paper 5 Paper 4 Other Transport 7 Other Transport 5 Textiles 9 Furniture 7 Non-metallic metal 10 Non-metallic metal 10 Leather 12 Textiles 12 Petroleum 12 Petroleum 26 Electrical Machinery 12 Leather 35 Furniture 13 Electrical Machinery 45 Fabricated metal 31 Chemicals 54 Chemicals 35 Rubber & Plastics 55 Printing 37 Fabricated metal 57 Rubber & Plastics 41 Wood, not Furniture 69 Wood, not Furniture 54 Printing 70 Food & Bev 63 Food & Bev 168 0 50 100 150 200 0 50 100 150 200 Graphs by year Graphs by year Table 3b: Côte d'Ivoire Firm Shares by Industry Share of Firms Share of Firms 2003 2014 Tobacco 0.6 Tobacco 0.3 Paper 1.5 Paper 0.6 Other Transport 2.0 Other Transport 0.8 Textiles 2.6 Furniture 1.1 Non-metallic metal 2.9 Non-metallic metal 1.6 Leather 3.5 Textiles 1.9 Petroleum 3.5 Petroleum 4.2 Electrical Machinery 3.5 Leather 5.7 Furniture 3.8 Electrical Machinery 7.3 Fabricated metal 9.0 Chemicals 8.7 Chemicals 10.2 Rubber & Plastics 8.9 Printing 10.8 Fabricated metal 9.2 Rubber & Plastics 12.0 Wood, not Furniture 11.1 Wood, not Furniture 15.7 Printing 11.3 Food & Bev 18.4 Food & Bev 27.1 0 5 10 15 20 0 10 20 30 Graphs by year Graphs by year 40 Table 2c: Tanzania Firm Count by Industry Number of Firms Number of Firms 2008 2013 Garments 1 Precision Instr. 1 Elec. Machinery 1 Leather 2 Vehicles 1 Oth. Transport 2 Leather 3 Tobacco 4 Oth. Transport 3 Garments 4 Tobacco 4 Elec. Machinery 7 Media Equip. 5 Basic Metal 14 Precision Instr. 5 Paper 14 Paper 7 Petroleum 15 Petroleum 7 Media Equip. 21 Basic Metal 7 Chemicals 32 Machinery 13 Rubber & Plastics 15 Machinery 33 Wood, not furn 17 Rubber & Plastics 43 Non-metallic metal 26 Textiles 45 Chemicals 30 Printing 46 Fab. metal 31 Fab. metal 54 Textiles 40 Wood, not furn 68 Printing 51 Furniture 76 Furniture 177 Non-metallic metal 88 Food & Bev 236 Food & Bev 430 0 100 200 300 400 500 0 100 200 300 400 500 Graphs by year Graphs by year Table 3c: Tanzania Firm Shares by Industry Share of Firms Share of Firms 2008 2013 Garments 0.1 Precision Instr. 0.1 Elec. Machinery 0.1 Leather 0.2 Vehicles 0.1 Oth. Transport 0.2 Leather 0.4 Tobacco 0.4 Oth. Transport 0.4 Garments 0.4 Tobacco 0.6 Elec. Machinery 0.7 Media Equip. 0.7 Basic Metal 1.4 Precision Instr. 0.7 Paper 1.4 Paper 1.0 Petroleum 1.5 Petroleum 1.0 Media Equip. 2.1 Basic Metal 1.0 Chemicals 3.2 Machinery 1.9 Rubber & Plastics 2.2 Machinery 3.3 Wood, not furn 2.5 Rubber & Plastics 4.3 Non-metallic metal 3.8 Textiles 4.5 Chemicals 4.4 Printing 4.7 Fab. metal 4.6 Fab. metal 5.4 Textiles 5.9 Wood, not furn 6.8 Printing 7.5 Furniture 7.6 Furniture 26.0 Non-metallic metal 8.8 Food & Bev 34.7 Food & Bev 43.1 0 5 10 15 20 25 30 35 40 45 50 0 5 10 15 20 25 30 35 40 45 50 Graphs by year Graphs by year 41 Table 2e: Bangladesh Firm Count by Industry Number of Firms Number of Firms 2001 2012 Petroleum 10 Office Equipment 13 Precision Instruments 32 Petroleum 19 Media Equipments 40 Media Equipments 136 Vehicles 63 Vehicles 137 Other Transport 110 Furniture 235 Electrical Machinery 115 Other Transport 276 Paper 127 Wood, not Furniture 298 Machinery 162 Machinery 461 Basic Metal 189 Tobacco 487 Tobacco 206 Electrical Machinery 738 Wood, not Furniture 250 Paper 902 Rubber & Plastics 355 Printing 904 Leather 368 Leather 930 Fabricated metal 531 Rubber & Plastics 1,036 Chemicals 596 Chemicals 1,057 Printing 848 Basic Metal 1,205 Furniture 1,255 Fabricated metal 1,449 Non-metallic metal 2,271 Non-metallic metal 4,652 Garments 3,654 Garments 4,670 Food & Bev 5,416 Food & Bev 8,805 Textiles 6,841 Textiles 13,297 0 5,000 10,000 15,000 0 5,000 10,000 15,000 Graphs by year Graphs by year Table 3e: Bangladesh Firm Shares by Industry Share of Firms Share of Firms 2001 2012 Petroleum 0.0 Office Equipment 0.0 Precision Instruments 0.1 Petroleum 0.0 Media Equipments 0.2 Media Equipments 0.3 Vehicles 0.3 Vehicles 0.3 Other Transport 0.5 Furniture 0.6 Electrical Machinery 0.5 Other Transport 0.7 Paper 0.5 Wood, not Furniture 0.7 Machinery 0.7 Machinery 1.1 Basic Metal 0.8 Tobacco 1.2 Tobacco 0.9 Electrical Machinery 1.8 Wood, not Furniture 1.1 Paper 2.2 Rubber & Plastics 1.5 Printing 2.2 Leather 1.6 Leather 2.2 Fabricated metal 2.3 Rubber & Plastics 2.5 Chemicals 2.5 Chemicals 2.5 Printing 3.6 Basic Metal 2.9 Furniture 5.4 Fabricated metal 3.5 Non-metallic metal 9.7 Non-metallic metal 11.2 Garments 15.6 Garments 11.2 Food & Bev 23.1 Food & Bev 21.1 Textiles 29.2 Textiles 31.9 0 10 20 30 0 10 20 30 Graphs by year Graphs by year 42 Table 4a: Ethiopia: Employment by Industry Industry Employment Industry Employment 1996 2016 Elec. Machinery 88 Petroleum 0 Machinery 371 Recycling 0 Precision Instr. 60 Vehicles 546 Machinery 552 Tobacco 965 Tobacco 570 Basic Metal 1,082 Media Equip. 805 Paper 1,204 Paper 1,969 Elec. Machinery 2,193 Rubber & Plastics 1,853 Wood, not furn 2,479 Fab. metal 1,902 Basic Metal 5,390 Wood, not furn 2,217 Printing 6,777 Furniture 2,240 Vehicles 6,791 Fab. metal 9,046 Chemicals 2,403 Garments 12,134 Printing 3,875 Furniture 12,388 Garments 3,959 Chemicals 12,434 Non-metallic metal 4,345 Leather 13,361 Leather 7,285 Rubber & Plastics 20,481 Textiles 26,085 Food & Bev 20,563 Non-metallic metal 42,942 Textiles 27,181 Food & Bev 51,514 0 20,000 40,000 60,000 0 20,000 40,000 60,000 Graphs by Year Graphs by Year Table 5a: Ethiopia: Employment Shares by Industry Employment Share Employment Share 1996 2016 Elec. Machinery 0.1 Petroleum 0.0 Machinery 0.5 Recycling 0.0 Precision Instr. 0.0 Vehicles 0.7 Machinery 0.2 Tobacco 1.2 Tobacco 0.3 Basic Metal 1.3 Media Equip. 0.4 Paper 1.5 Paper 0.9 Elec. Machinery 1.0 Rubber & Plastics 2.3 Wood, not furn 1.1 Fab. metal 2.3 Basic Metal 2.4 Wood, not furn 2.7 Printing 3.0 Furniture 2.7 Vehicles 3.0 Fab. metal 4.0 Chemicals 2.9 Garments 5.3 Printing 4.7 Furniture 5.4 Garments 4.8 Chemicals 5.5 Non-metallic metal 5.3 Leather 5.9 Leather 8.9 Rubber & Plastics 9.0 Textiles 11.4 Food & Bev 25.1 Non-metallic metal 18.8 Textiles 33.1 Food & Bev 22.6 0 10 20 30 40 0 5 10 15 20 25 Graphs by Year Graphs by Year 43 Table 4b: Côte d'Ivoire Employment by Industry Industry Employment Industry Employment 2003 2014 Tobacco 313 Tobacco 295 Machinery 576 Furniture 432 Non-metallic metal 705 Paper 488 Other Transport 719 Other Transport 584 Paper 1,162 Non-metallic metal 1,522 Furniture 1,355 Machinery 2,557 Printing 1,936 Leather 3,249 Leather 2,347 Printing 3,403 Fabricated metal 2,805 Textiles 3,933 Rubber & Plastics 4,264 Fabricated metal 5,691 Textiles 4,918 Wood, not Furniture 10,118 Chemicals 5,105 Chemicals 12,204 Wood, not Furniture 13,140 Rubber & Plastics 13,257 Petroleum 14,329 Petroleum 13,324 Food & Bev 15,632 Food & Bev 20,316 0 5,000 10,000 15,000 20,000 25,000 30,000 0 5,000 10,00015,000 20,000 25,000 30,000 Graphs by AN_EXO Graphs by AN_EXO Table 5b: Côte d'Ivoire Employment Shares by Industry Employment Share Employment Share 2003 2014 Tobacco 0.5 Tobacco 0.3 Machinery 0.8 Furniture 0.5 Non-metallic metal 1.0 Paper 0.5 Other Transport 1.0 Other Transport 0.6 Paper 1.7 Non-metallic metal 1.7 Furniture 2.0 Machinery 2.8 Printing 2.8 Leather 3.6 Leather 3.4 Printing 3.7 Fabricated metal 4.0 Textiles 4.3 Rubber & Plastics 6.2 Fabricated metal 6.2 Textiles 7.1 Wood, not Furniture 11.1 Chemicals 7.4 Chemicals 13.4 Wood, not Furniture 19.0 Rubber & Plastics 14.5 Petroleum 20.7 Petroleum 14.6 Food & Bev 22.6 Food & Bev 22.2 0 5 10 15 20 25 0 5 10 15 20 25 Graphs by AN_EXO Graphs by AN_EXO 44 Table 4c: Tanzania Employment by Industry Industry Employment Industry Employment 2008 2013 Elec. Machinery 39 Precision Instr. 13 Oth. Transport 44 Leather 33 Garments 46 Garments 193 Vehicles 65 Oth. Transport 199 Precision Instr. 89 Elec. Machinery 900 Media Equip. 367 Media Equip. 1,064 Leather 593 Petroleum 1,376 Wood, not furn 779 Machinery 1,720 Basic Metal 818 Basic Metal 2,174 Petroleum 1,140 Fab. metal 2,189 Machinery 1,233 Paper 2,303 Paper 2,026 Fab. metal 2,395 Printing 2,773 Non-metallic metal 2,560 Wood, not furn 3,769 Printing 3,531 Rubber & Plastics 4,243 Chemicals 3,547 Furniture 4,447 Rubber & Plastics 4,125 Non-metallic metal 4,548 Tobacco 6,747 Tobacco 5,061 Textiles 11,983 Chemicals 6,026 Furniture 15,623 Textiles 17,810 Food & Bev 49,638 Food & Bev 46,889 0 20,000 40,000 60,000 0 20,000 40,000 60,000 Graphs by year Graphs by year Table 5c: Tanzania Employment Shares by Industry Employment Share Employment Share 2008 2013 Elec. Machinery 0.0 Precision Instr. 0.0 Oth. Transport 0.0 Leather 0.0 Garments 0.0 Garments 0.2 Vehicles 0.1 Oth. Transport 0.2 Precision Instr. 0.1 Elec. Machinery 0.8 Media Equip. 0.3 Media Equip. 1.0 Leather 0.6 Petroleum 1.3 Wood, not furn 0.7 Machinery 1.6 Basic Metal 0.8 Basic Metal 2.0 Petroleum 1.1 Fab. metal 2.0 Machinery 1.1 Paper 2.1 Paper 1.9 Fab. metal 2.2 Printing 2.6 Non-metallic metal 2.4 Wood, not furn 3.5 Printing 3.3 Rubber & Plastics 3.9 Chemicals 3.3 Furniture 4.1 Rubber & Plastics 3.8 Non-metallic metal 4.2 Tobacco 6.3 Tobacco 4.7 Textiles 11.2 Chemicals 5.6 Furniture 14.5 Textiles 16.5 Food & Bev 46.2 Food & Bev 43.5 0 10 20 30 40 50 0 10 20 30 40 50 Graphs by year Graphs by year 45 Table 4e: Bangladesh Employment by Industry Industry Employment Industry Employment 2001 2012 Precision Instruments 469 Office Equipment 1,027 Petroleum 557 Petroleum 1,417 Media Equipments 1,826 Vehicles 4,906 Machinery 5,535 Wood, not Furniture 8,495 Wood, not Furniture 5,957 Furniture 9,471 Vehicles 6,473 Media Equipments 15,363 Other Transport 10,010 Other Transport 17,921 Paper 11,288 Machinery 18,620 Electrical Machinery 12,281 Printing 26,667 Rubber & Plastics 12,428 Electrical Machinery 37,495 Basic Metal 13,119 Rubber & Plastics 41,139 Fabricated metal 15,562 Paper 42,376 Leather 29,474 Fabricated metal 44,462 Furniture 45,810 Tobacco 52,204 Tobacco 54,596 Leather 75,524 Printing 88,377 Basic Metal 120,965 Chemicals 152,493 Chemicals 123,978 Food & Bev 211,746 Food & Bev 300,675 Non-metallic metal 230,807 Non-metallic metal 471,816 Textiles 511,591 Textiles 1,715,008 Garments 1,352,008 Garments 1,852,834 0 750,000 1,500,000 2,250,000 0 750,000 1,500,000 2,250,000 Graphs by year Graphs by year Table 5e: Bangladesh Employment Shares by Industry Employment Share Employment Share 2001 2012 Precision Instruments 0.0 Office Equipment 0.0 Petroleum 0.0 Petroleum 0.0 Media Equipments 0.1 Vehicles 0.1 Machinery 0.2 Wood, not Furniture 0.2 Wood, not Furniture 0.2 Furniture 0.2 Vehicles 0.2 Media Equipments 0.3 Other Transport 0.4 Other Transport 0.4 Paper 0.4 Machinery 0.4 Electrical Machinery 0.4 Printing 0.5 Rubber & Plastics 0.4 Electrical Machinery 0.8 Basic Metal 0.5 Rubber & Plastics 0.8 Fabricated metal 0.6 Paper 0.9 Leather 1.1 Fabricated metal 0.9 Furniture 1.7 Tobacco 1.0 Tobacco 2.0 Leather 1.5 Printing 3.2 Basic Metal 2.4 Chemicals 5.5 Chemicals 2.5 Food & Bev 7.6 Food & Bev 6.0 Non-metallic metal 8.3 Non-metallic metal 9.5 Textiles 18.5 Textiles 34.4 Garments 48.8 Garments 37.2 0 10 20 30 40 50 0 10 20 30 40 Graphs by year Graphs by year 46