WPS7780 Policy Research Working Paper 7780 Resource Misallocation in Turkey Ha Nguyen Temel Taskin Ayberk Yilmaz Development Research Group Macroeconomics and Growth Team & Macroeconomics and Fiscal Management Global Practice Group August 2016 Policy Research Working Paper 7780 Abstract This paper examines resource misallocation within narrow is also examined. Improvement in allocative efficiency was industries in Turkey. It finds that resource misallocation in sizable between 2003 and 2013, but significantly slower Turkey is substantial. The hypothetical gain from moving after 2007. However, the earlier trend reversed in 2014 and to “U.S. efficiency” is 24.5 percent of manufacturing total resource misallocation worsened in Turkey’s manufactur- factor productivity in 2014. The evolution of resource ing. The cross-sector analysis reveals that misallocation is misallocation over time and across disaggregated sectors most pronounced in textiles, transport, food, and leather. This paper is a product of the Macroeconomics and Growth Team, Development Research Group and the Macroeconomics and Fiscal Management Global Practice Group. 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://econ.worldbank.org. The authors may be contacted at hanguyen@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 RESOURCE MISALLOCATION IN TURKEY Ha Nguyen, Temel Taskin, Ayberk Yilmaz 1 Keywords: Total factor productivity, across-firm misallocation, within-sector misallocation, firm-level data, firm heterogeneity, Turkey. J.E.L Codes: D24, L25, O12.                                                              1 Ha Nguyen is with Development Research Group; Temel Taskin is with Development Prospects Group; Ayberk Yilmaz is with Macroeconomics and Fiscal Management Global Practice, The World Bank. We thank Ulrich Bartsch, Roberto Fattal Jaef, Hans Timmer, Luis Serven for comments and feedback. Contact: hanguyen@worldbank.org , or ayilmaz@worldbank.org. 1. Introduction Turkey’s economy expanded rapidly in the 2000s, and per capita income is now close to the threshold beyond which the World Bank classifies countries as high income. While part of this economic growth was attributable to human and physical capital accumulation, an important fraction was due to total factor productivity (TFP) growth (World Bank, 2014). Growth in TFP can be divided into three different sources. The first one is via growth in technology, which is the most traditional thinking about TFP. The second source is the movement of factors from sectors with lower productivity (such as agriculture) to sectors with higher productivity (such as manufacturing). About two-thirds of the overall productivity gains in Turkey came from the shift of labor out of agriculture and into higher-productivity manufacturing and service industries (World Bank, 2014). The third source of improving a country’s TFP is via better allocation of resources within industries. This takes place when resources from firms with lower productivity move to firms with higher productivity. Since the seminal work of Restuccia and Rogerson (2008) and Hsieh and Klenow (2009), we know that across-firm resource misallocation within industries can lead to lower aggregate TFP. Across-firm resource misallocation is a consequence of highly productive firms not obtaining sufficient resources (in terms of capital and labor) to expand production, while firms with low productivity continue employing resources instead of shrinking and eventually exiting. This could be a result, for example, of politically connected firms having easier access to finance and therefore expanding production, despite their productivity being lower than that of less connected firms. This phenomenon could substantially reduce a country's total output and productivity because highly productive firms would be smaller and less productive ones would be larger than they should be at optimal allocation. Hsieh and Klenow (2009) estimate that if the problem of resource allocation in China and India is eliminated, that is, if capital and labor are hypothetically reallocated to equalize marginal products to the extent observed in the United States, manufacturing TFP can increase 30% to 50% in China and 40% to 60% in India. In this paper, using Hsieh and Klenow’s (2009) framework (HK, henceforth), we measure how much aggregate manufacturing TFP in Turkey could increase if capital and labor were reallocated to equalize marginal products across firms within each four-digit sector to the extent observed in the United States in 1997. For Turkey, moving to ‘1997 U.S. efficiency’ could have boosted manufacturing TFP by 24.5% in 2014. Thanks to the availability of Turkish manufacturing data covering 11 years, from 2003 to 2014, we can track the improvement of resource allocation in Turkey over time. We find that resource allocation improved quite significantly. In 2003, the manufacturing TFP gap from the US caused by resource misallocation is 56.1%, while that in 2013 is only 18.4%. Finally, we break down resource misallocation 2   by industries and by regions. The results reveal that resource misallocation is larger among textile, transport, food, and leather industries. The rest of the paper is organized as follows: section 2 provides some background on Turkish economy; section 3 reviews the related literature; section 4 explains the analytical framework; section 5 describes the data set; section 6 presents the results; and finally, section 7 concludes. 2. Background on Turkey’s economy2 Turkey is an upper middle income country and a member of G20. Turkey has the world's 17th largest nominal GDP (in PPP units) and stands at the threshold to high income today, with GNI per capita (Atlas method, current US$) at $10,830 in 2014.3 Turkey’s economic development after 2001 resulted in impressive economic achievements. After a banking crisis in 2001, the country embarked on a concerted path of structural reforms supported by strong fiscal consolidation, strengthened banking supervision, and a shift to a flexible exchange rate regime with an independent central bank responsible for inflation targeting. The pro-market reform process was further enhanced and anchored by the EU Accession process. In the following period, Turkey`s economy grew on average by 6.9% annually until the Global Economic and Financial crisis. Turkey had already started to exhaust the benefits of the reform momentum of the early 2000s. However, the global crisis and subsequent rapid recovery in Turkey had diverted attention from remaining structural weaknesses and considerably complicated macroeconomic management (World Bank, 2014). Following a swift rebound from the recession in the crisis in 2008-09, concerns over Turkey’s vulnerability to tightening global liquidity and deteriorating political and regulatory environment caused a lack of private investment spending (World Bank, 2014). The interventions in independent regulatory institutions suggest that the principles of arms’ length regulation had not yet put down deep roots. Furthermore, there have been growing concerns over the transparency in public tenders and the allocation of land development rights in Turkey (World Bank, 2014). The anchor provided by the EU Accession process had been weaker ever since the mid-2000s and Turkey had been losing market share in FDI to emerging markets as a whole since 2007. Against this backdrop, private investment, one of the main drivers of growth in the pre-2008 period, dropped sharply in 2012 and has stagnated since then. As a result, economic growth slowed since 2012.                                                              2 This section draws mostly from the World Bank Turkey`s Transitions Report (World Bank, 2014). 3 The World Bank defines high income economies as those with a GNI per capita of $12,736 or more. Please see World Bank WDI database for Turkey`s historical GNI per capita performance and GDP ranking. 3   3. Literature review This study is related to an overarching question on countries’ TFP (total factor productivity): why are some countries’ productivity levels higher than others? A traditional answer is the country’s technology level. For example, in some countries, firms adopt advanced technologies and are innovative in producing new ones. In other countries, because of several reasons such as lack of access to finance or low levels of human capital, firms are not as good in terms of learning and producing new technologies. Conventional policy implications hence focus on factors affecting a country’s aggregate technological absorption capacity. Examples of a country’s technological absorption capacity include access to finance, education, FDI, and openness. More recently, a new strand of literature has explored the role of resource misallocation for countries’ aggregate TFP. The idea is that in addition to primitive technology, resource misallocation can hurt a country’s aggregate productivity. As discussed in the introduction, resource misallocation refers to productive firms not being able to expand, and unproductive firms being larger than they should be. Because resources flow to the wrong firms, the country’s total output is lower given the same input. This means the country has lower productivity. Policy implications for this approach are drastically unrelated to technology. The focuses are now competition policies and political economy. Restuccia and Rogerson (2008), in a standard neoclassical growth model with heterogeneous firms, provide the first framework to examine how resource misallocation can affect aggregate productivity. They consider distortions that generate differences in the prices faced by individual firms. For example, politically connected firms may have lower interest rates on loans than unconnected firms. Restuccia and Rogerson (2008) term these policies as “idiosyncratic distortions”. They emphasize that the productivity losses due to misallocation would be larger if the “distortions” are positively correlated with the level of productivity of firms. For example, if highly productive firms happen to be politically unconnected and have to pay higher interest rates, they will not have sufficient resources to expand. As a result, the country’s aggregate productivity is reduced. On the other hand, if unconnected firms happen to be unproductive as well, the negative impacts of misallocation on aggregate productivity is certainly not as large. Drawing on the seminal work of Restuccia and Rogerson (2008), a growing number of studies have quantified the costs generated by resource misallocation. Hsieh and Klenow (2009), in their landmark study, examine the quantitative effect of resource misallocation. The basic underlying assumption in their paper is that if resource misallocation is completely removed, the marginal products of labor and capital for all firms should be equalized. Therefore any unequal marginal products of production factors are due to resource misallocation. With this assumption, they estimate that if the problem of resource allocation in China and India is improved, meaning, if capital and labor are hypothetically reallocated to equalize 4   marginal products to the extent observed in the United States, manufacturing TFP can increase 30% to 50% in China and 40% to 60% in India. Subsequent research following the methodology of Hsieh and Klenow (2009) confirms the quantitative importance of misallocations for several countries. Examples are Camacho and Conover (2010) for Colombia, Busso et al. (2013) for Latin American countries, and Kalemli-Ozcan and Sorensen (2012) and Cirera et al. (2015) for African countries. 4. Framework In this section, we present the HK framework which constitutes the theoretical background to measurement of within sector misallocation. Consider an economy with many sectors, denoted s. A final output Y is produced in each country using a Cobb-Douglas production technology: ∏ , (1) where is the value added share of sector s, and ∑ 1. Each sector's output is the aggregate of the individual firm's output , using the CES technology: ∑ , (2) where is the differentiated product by firm i in sector s. Each firm produces a differentiated product with the standard Cobb-Douglas production function: , (3) where stands for firm-specific productivity, and are the firm's capital and labor respectively, and is the industry-specific capital share. Note that the assumption in this framework is that firms in the same narrowly-defined sector (i.e. 4-digit NACE) have the same production function. Each establishment maximizes current profits: 1 1 , (4) where is the firm's value added (which is the firm's revenue minus the cost of intermediate inputs), w and R are the cost of one unit of labor and capital respectively. The term denotes firm-specific output distortions that reduce firms' revenues. Many factors could contribute to output distortions, ranging from transportation costs to harassment from authorities. These factors could reduce output for a given set of input. The firm-specific "capital" distortions, which raise the cost of capital (relative to labor), is denoted 5   with . Credit market imperfections (such as differential access to finance) and labor market frictions could contribute to different "capital" distortions across firms. From Hsieh and Klenow (2009), we distinguish between two productivity measures, one expressed in physical units (TFPQ) and the other in monetary value (TFPR): (5) (6) It is important to note that it is normal that TFPQ differs across firms, different firms may have different productivity levels. However, in this framework, if there were no distortions, TFPR should be equalized across firms in the same industry. This is because of the assumption in the model that firms are monopolistically competitive. Without distortions, low productivity firms have less resources and produce less. Since their product is relatively scarce, they can charge a higher price, , which equates across firms i. In other words, in the absence of distortions, more capital and labor should be allocated to firms with higher TFPQ to the point where their higher output results in a lower price and the exact same TFPR as smaller firms. As a consequence, any dispersion of TFPR across firms within an industry is an indication of distortions. A firm with higher TFPR than the sector average is more "taxed", meaning, it suffers more obstacles, than other firms. To empirically implement the HK framework, we follow HK and choose the elasticity of substitution 3, R=10 (assuming the real interest rate=5% and the depreciation rate of 5%). Follow HK, we use capital share, ,and labor share, 1 , from the U.S. manufacturing sectors. The underlying assumption is that capital and labor shares from sectors in the U.S. represent the least distorted environment. Any deviation of capital-labor share from the U.S.'s level is an indication of distortions. The output and capital wedges can be computed as follows: 1 (7) 1 (8) Note that is firm i's wage bill; is the firm's value added. Both values are available in the census data. To understand the intuition of equation (8), we rewrite it as: 1 (9) 6   Note that is the labor-capital ratio in the undistorted (U.S.) environment. If firm i's actual labor capital ratio is higher than the undistorted labor capital ratio, this indicates that the firms face difficulties accessing capital (relative to hiring labor), and as a result, use less capital than the optimal level. This is equivalent to stating the firm has a positive capital wedge . Armed with and , HK shows that TFPR can be calculated as: (10) Equation (10) implies that in the absence of distortions (i.e. = 0 and =0), TFPR is the same for all firms “i" within a sector “s”. Using this equation, one can induce that a firm with higher and/or higher also has a higher TFPR. In addition, the industry level is: (11) ∑ ∑ Note that when there are no distortions (i.e. = 0 and =0) for all i, the right hand side of (11) equals the right hand side of (10), which means that TFPR are equalized for all i. Firm i's productivity can be calculated as: (12) and the efficient industry's productivity level (when all marginal products are equalized) is: ∑ (13) From (10) to (13), we can calculate the ratio of the actual TFP in the economy to the efficient level of TFP: ∏ ∑ (14) We calculate the ratio of actual TFP to the efficient level of TFP and then aggregate this ratio across sectors using the Cobb-Douglas aggregator. 7   5. Data We use the Annual Industry and Services Statistics (AISS) to carry out the empirical exercise. AISS is a survey data set conducted annually by the Turkish Statistics Institute (TURKSTAT) since 2003. AISS covers all firms which employ more than 20 workers and draws a sample from firms employing 20 or fewer workers. Our analysis is based on the entire coverage period of 2003 – 2014 with a particular focus on the manufacturing sector. We exclude firms employing 20 or fewer workers, because only a subset of these firms are included in the data. We also observe some inconsistencies in sample weights of drawn firms in some years.4 Therefore, we keep our focus on the firms that have more than 20 workers given that these firms constitute the whole population instead of a sample. The data set covers a large set of variables including investment, sales, energy expenditures, material expenditures, number of employees, ownership type, location, and industry. The number of employees is available at the gender and paid/unpaid breakdown. Location is provided at the NUTS3 level (province), and industry classification is available in NACE Rev2 at 4-digit level.5 Investment expenditures are reported in three categories, namely, computer and programing, machinery and equipment, and buildings and structure. We calculate a firm level capital series by using the disaggregated investment series and corresponding depreciation rates. Following Taymaz and Yilmaz (2009), depreciation rates of 5%, 10%, and 30% are used for building and structure, machinery and equipment, and computer and programing, respectively, to construct initial capital stock and to apply the perpetual inventory method. For the firms that report non-zero investment at their initial year, we calculate capital stock by dividing the firms’ average investment with the depreciation rate of investment. For the firms that report zero investment at their initial year, it is assumed that they cannot be producing without capital. Therefore, the initial capital stock is calculated at the year that they report positive investment and this amount is iterated back to the beginning year by dividing capital stock by the value (1 - each year. After calculating a capital stock series for building and structure, machinery and equipment, and computer and programming, these series are aggregated to form the capital stock series of the firm. All the monetary variables are reported in Turkish Lira with current prices. We normalize the input expenditures with the corresponding 3-digit deflators. The firm level output is deflated by 3-digit output price deflators.                                                              4  For instance, the sample weights are assigned to be 1 for all sampled firms in some years.  5  NUTS stands for “Nomenclature des Unités Territoriales Statistiques”.  8   Our empirical exercise also requires sector-level capital/labor ratios in the US, which are obtained from NBER manufacturing database. Since US sectors are coded according to the SIC (Standard Industrial Classification), we implement the necessary conversion between SIC and NACE while merging the US capital/labor ratios with our firm-level data. In the data cleaning process we dropped observations at a number of different steps, ending up with a smaller data set than the original one. The original data set has 1,275,049 observations for the entire coverage period of 2003-2014. In the first step, we drop non-manufacturing firms because our focus is on manufacturing sector firms.6 This step reduced the number of observations in our data set significantly to 352,150. In the second step, we restrict our data set only to the firms that have more than 20 workers because of the reasons explained above. This step further decreased the size of our data set to 222,393 firms. In the third step, we drop the firms that operate in a given year but have 0 capital stock. We assume that firms cannot operate without a capital stock and consider these observations as inconsistencies. This brought the number of observations down to 211,450. In the fourth step, in order to obtain the capital and labor shares from the United States, we had to match 4-digit NACE Rev2 codes with the relevant 4-digit SIC codes. We dropped the industries that did not have a close counterpart. Moreover, we also dropped industries that have labor share values greater than 1. These reductions decreased the size to 196,155. In the final step, we drop 4-digit industries that have fewer than 10 firms in a given year. Otherwise, analyzing the allocative efficiency in a 4-digit industry that has fewer than 10 firms would not have made much sense. Also, we eliminated any inconsistent observations that were left in the data set. In addition, following Hsieh and Klenow (2009), we trimmed the 1% tails of log ⁄ and log ⁄ across industries before calculating the gains from our hypothetical liberalization. The final data set has 181,052 observations.                                                              6 In addition, the calculation of the service sector’s TFP is challenging. Since it is more problematic to assign the traditional Cobb-Douglas production function to the service sector than to the manufacturing sector, the estimated TFP for the service sector with a Cobb-Douglas production function would have larger measurement errors. 9   Table 1: Size of the Dataset st nd 1 Step 2 Step 3rd Step 4th Step 5th Step 6th Step Original Manufacturing More Than After Capital After Final Year Dataset Firms 20 Workers Stock Conversion Dataset 2003 67,516 10,750 8,021 7,809 7,246 6,796 2004 71,973 13,157 10,182 9,811 9,100 8,438 2005 63,304 16,654 14,172 13,561 12,596 11,729 2006 85,016 17,446 15,302 14,738 13,675 12,811 2007 83,963 17,269 15,095 14,547 13,491 12,567 2008 82,662 31,134 17,793 16,860 15,466 14,264 2009 99,921 35,059 16,061 15,506 14,354 13,147 2010 106,714 33,890 21,218 20,305 18,792 17,190 2011 138,013 41,194 23,755 22,636 21,029 19,249 2012 147,916 43,281 25,856 24,578 22,875 21,115 2013 168,618 47,000 26,791 25,208 23,435 21,598 2014 159,433 45,316 28,147 25,891 24,096 22,148 Total 1,275,049 352,150 222,393 211,450 196,155 181,052 Source: TURKSTAT and authors’ calculations. 6. Results 6.1 Measuring distortions Table 2 presents the benchmark results of the paper. The first column shows the year of the data coverage. The second column displays the number of firms in each year. Note that only firms with more than 20 workers are included in our data set. The third column shows the potential TFP gains if TFPR are equalized among firms within a sector (i.e. if the resource misallocation problem is completely removed). The fourth column shows the potential TFP gains if allocative efficiency in Turkey improves to the U.S. level (thus misallocation reduces to the U.S. level). The columns five through ten show the statistics for the distributions of TFPQ, log , and of TFPR, log ⁄ , over time. The final two columns show the standard deviations of capital and output wedges. 10   Table 2: Baseline Results TFP Gains TFPQ TFPR Wedge Year Firms Equalizin Moving to Capital Output 1997 U.S. SD 75-25 90-10 SD 75-25 90-10 g TFPR SD SD Efficiency 2003 6,796 123.1 56.1 1.46 2.10 3.87 0.97 1.28 2.51 1.63 0.97 2004 8,438 89.2 32.4 1.28 1.81 3.38 0.85 1.12 2.20 1.62 0.88 2005 11,729 86.1 30.2 1.20 1.67 3.12 0.81 1.04 2.05 1.60 0.84 2006 12,811 79.5 25.6 1.14 1.61 2.98 0.76 1.01 1.93 1.57 0.86 2007 12,567 77.3 24.1 1.13 1.57 2.96 0.75 0.97 1.88 1.56 0.89 2008 14,264 77.1 23.9 1.13 1.56 2.96 0.74 0.96 1.90 1.58 0.90 2009 13,147 76.7 23.6 1.15 1.56 2.99 0.76 0.95 1.90 1.55 0.93 2010 17,190 70.2 19.1 1.10 1.46 2.83 0.74 0.93 1.88 1.57 0.94 2011 19,249 73.2 21.2 1.11 1.48 2.88 0.74 0.95 1.89 1.57 0.95 2012 21,115 69.7 18.8 1.08 1.44 2.78 0.73 0.92 1.85 1.58 0.94 2013 21,598 69.1 18.4 1.09 1.47 2.80 0.74 0.93 1.87 1.58 0.91 2014 22,148 78.0 24.5 1.11 1.50 2.85 0.76 0.96 1.88 1.59 0.90 Source: TURKSTAT and authors’ calculations. The main take-away message of the paper is shown in columns 3 and 4. They display hypothetical aggregate TFP gains from removing misallocation over the years. Larger hypothetical gains imply that the resource misallocation across firms within a sector is more pronounced. The TFP gain of 78% in 2014 means that if resources are allocated efficiently across firms within a sector, i.e. more productive firms having more resources and less productive firms having less resources, Turkey’s manufacturing TFP would increase by 78% (Table 2, column 3). Over time, Turkey has reduced the within-sector resource misallocation quite significantly. Potential TFP gains had been declining over time, from 123% in 2003 to 69% in 2013. However the speed of the improvement had markedly slowed down after 2007, except in 2010 when Turkey’s economy bounced back from the global financial crisis and grew by around 9%. Moreover, the most recent data suggests that the earlier trend reversed in 2014, and resource misallocation worsened in Turkey’s manufacturing industry. In fact, the potential TFP gains increased from 69% in 2013 to 78% in 2014. Following Hsieh and Klenow (2009), we measure how much aggregate manufacturing TFP in Turkey could increase if capital and labor were reallocated to equalize marginal products across firms within each four- digit sector to the extent observed in the United States. The United States is a critical benchmark. Comparing countries with the United States has an advantage. To the extent the measurement errors of capital stock and marginal productivity are present to a similar extent in both the U.S. and Turkey data, the comparison would account for such errors. The TFP gain of 24.5% in 2014 implies that if the resource 11   misallocation problem is at the U.S. level in 1997, Turkey’s manufacturing TFP would increase by 24.5%. Similarly, the speed of allocative efficiency improvement started to slow down around 2007, and the trend has reversed in 2014. Compared to the 1997 U.S. benchmark, Turkey’s allocative efficiency improved by 8.3% from 2004 to 2007, or about 2.7% per year. A determined implementation of second-generation reforms over the next few years could produce similar improvements in Turkey, assuming a moderate speed of convergence to ‘U.S efficiency’. This would constitute an important boost to manufacturing productivity growth, which nearly stagnated in recent years. Consistent with the results on potential gains, TFPR dispersion declined significantly until 2007, suggesting decreased firm-level distortions. The standard deviation of TFPR dropped from 0.97 in 2003 to 0.75 in 2007. However, the TFPR dispersion had remained mostly unchanged after 2007 with its standard deviation staying around 0.74 before increasing to 0.76 in 2014. Although the improvement in TFPR dispersion tracked by standard deviation had stopped after 2007, it had continued at the edges of the distribution, albeit at a very slow pace. Nonetheless, the increase in TFPR dispersion in 2014 is also reflected in the tails of the distribution. Figure 1 plots the distribution of log TFPR, log ⁄ , as a measure of firm- specific distortions in Turkey. The plot has been demeaned in order to show distance from the sector- specific average. By plotting this measure of distortions over time, the most outstanding observation we have is that TFPR distribution in 2004 is more dispersed than those in other years. Moreover, the improvement came from the left tail, which contains the firms that have low TFPR. In other words, some of the firms lost their implicit subsides over time, leading to the left tail moving inwards towards the mean. Table 2 shows that the dispersion of TFPQ also improved rapidly between 2003 and 2007, slowed substantially after 2007, and worsened in 2014. Figure 2 plots the distribution of log TFPQ, log . There is clearly more TFPQ dispersion in 2004 than 2014. The TFPQ in 2004 has a thicker left tail than in 2014. On the other hand, there is almost no difference in the right tail between 2004 and 2014, implying that the improvement in TFPQ dispersion overtime has come from the unproductive firms. Although this exercise does not say anything about the dynamics of the improvement, results imply that either unproductive firms converged closer to the sector averages, or exit the market, or combination of both effects, leading to a thinner left tail and less misallocation. The elimination of firm specific distortions over time could have driven unproductive firms out of the market or could have 12   downsized unproductive firms given that, on average, unproductive firms receive implicit subsidies while productive firms are subject to implicit tax.7 Figure 1: TFPR Dispersion over time Figure 2: TFPQ Dispersion over time                                                              7 The relation between TFPR and TFPQ is discussed in more details in section 4.1.2, where results suggest strong positive correlation between TFPR and TFPQ. 13   These numbers clearly suggest that within-sector misallocation decreased over time in Turkey, more rapidly in the early 2000s. Interestingly, the timing coincides with Turkey’s reforms in early 2000, some of which might have helped with reducing resource misallocation. Turkey’s Transitions Report (World Bank, 2014) describes the early 2000 reforms in the finance sectors, and competition-based regulations and institutions, in details: “The deep crisis of 2001 provided an opportunity to introduce wholesale changes to Turkey`s governance model and to overcome the inconsistencies between macroeconomic measures and microeconomic regulations and to get rid of incentives which had plagued repeated stabilization efforts throughout late 1980s and 1990s. After the banking crisis in 2001, the country embarked on a concerted path of structural reforms. Key structural reform measures focused on the five main areas involving state-business relations, enterprise, finance, infrastructure, cities, and public finance management. Prior to 2001, political forces in Turkey, like in many other countries before, used and shaped financial sector policies and the operation of the financial system for political gain and influence, to the detriment of economic performance and welfare. The basic aim of the reforms was to develop a credible policy framework that would provide the proper incentives for the financial system to thrive and benefit the population in general. The approach followed, based on greater transparency and accountability in policymaking, increased sector competition and strong regulatory and legal steps to limit moral hazard, appears to have worked in fostering a stable, more efficient, and trustworthy financial sector. The financial sector reforms focused on four key areas: the enhancement of the Central Bank independence, the restricting and partial privatization of state-owned banks, the establishment of an independent regulator and the prompt resolution of failed private banks, the transparent recapitalization of the core of the banking system. Taken as a whole all these reforms contributed to a major change in the way of doing banking in Turkey. Another important development was macroeconomic stabilization. Fiscal prudence and sustained declines in inflation and real interest rates led to a major restructuring in the balance sheets of the banking system and increased credit to the private sector. Turkey`s perceived country risk also declined significantly so that Turkey could benefit from in the increase in global capital inflows that were in abundance during 2000s. This liquidity provided the financial resources for private sector growth. Together macroeconomic stabilization at home and the “Great Moderation” in the world economy enhanced access to finance and lowered barriers for new firm entry and expansion. At the same time, Turkey benefited from the gradual emergence of “European Union (EU)-anchored” competition-based regulations and institutions. The adoption of generally pro-market trade, financial and regulatory policies- together with infrastructure and urbanization policies that supported domestic market integration have created the basis for productive 14   resource reallocations. As a result of these multifaceted regulatory reforms, Turkey progressively reduced the gap to the frontier in the World Bank`s Doing Business indicators.” However, the reform momentum slowed significantly after 2007 (World Bank, 2014). Indicators from various data sources (such as Doing Business) also confirm the slowdown in institutional and business climate improvements after 2007. Interestingly, the timing of the slowdown also matches the slowdown in resource allocation improvement after 2007. How is Turkey compared to other countries? To answer this question, we put these numbers in a cross- country context to get a sense of the economic magnitude of these numbers. Table 3 compares the latest available year in each country: Turkey in 2014, United States in 1997, India in 1994, and China in 2005.8 There is more TFPQ dispersion in Turkey than in the United States and China, but less than India. The ratio of 75th to 25th percentiles of TFPQ in the latest year are 5.0 in India, 4.5 in Turkey, 3.6 in China and 3.2 in United States.9 Table 3 also provides TFPR dispersion statistics for the same group of country-years. The TFPR dispersion in Turkey is significantly more than that in the United States, while is only slightly more than those in China and India. These numbers suggest that distortions are greater in Turkey than in United States, China and India. The ratio of 75th to 25th percentiles of TFPR in the latest year are 2.6 in Turkey, 2.3 in China, 2.2 in India, and 1.7 in United States. Table 3: Dispersion of TFPR and TFPQ U.S. (1997) Turkey (2014) India (1994) China (2005) TFPR TFPQ TFPR TFPQ TFPR TFPQ TFPR TFPQ S.D 0.49 0.84 0.76 1.11 0.67 1.23 0.63 0.95 75-25 0.53 1.17 0.96 1.50 0.81 1.60 0.82 1.28 90-10 1.19 2.18 1.88 2.85 1.60 3.11 1.59 2.44 N 194,669 194,669 22,148 22,148 41,006 41,006 211,304 211,304 Source: Authors’ calculations, HK (2009). What is the potential for TFP growth from a reduction in misallocation? We calculate “efficient” output to compare it with the actual output level to measure the hypothetical gains from removing the within-sector misallocation of resources (Table 4). Aggregate TFP gains, from full liberalization by equalizing marginal revenue products across the existing set of firms in each 4-digit industry, are around 78.0% in Turkey in 2014, compared to 127.5% in India in 1994 and 86.6% in China in 2005. Overall, Turkey’s misallocation                                                              8 We rely on Hsieh and Klenow (2009) for the numbers of United States, China and India. 9 Exponentials of the corresponding numbers in Table 2. 15   level in 2014 is smaller than China’s in 2005 and India’s in 1994. In 2005, however, it was at the same level as China in the same year. How do we explain the different results regarding TFPR dispersion and TFP gains between Turkey, India and China? The answer is with TFPQ dispersion. If country A’s TFPQ is more dispersed than country B’s, country A’s TFPR could be more dispersed than country B’s, even if the levels of misallocation in the two countries are similar. Applying the same argument to Turkey, China, and India, since Turkey’s TFPQ is more dispersed than that in China, Turkey’s TFPR can be more dispersed and at the same time Turkey’s misallocation is less severe than China’s. Another explanation is with industry weights. It can be seen from equation (14), when calculating TFP gains, we take industry weights into account. Thus, one can consider TFPR dispersion as the measure of unweighted misallocation, while do TFP gains as the measure of weighted misallocation. The results show that TFPR dispersion is higher in Turkey, but TFP gain is lower in Turkey. This is most likely because the sectors that have high TFPR dispersion have lower weight in GDP in Turkey, making TFP gains from reallocation of resources within sectors smaller compared to peer countries. Table 4: TFP Gains From Equalizing TFPR Within Industries US Turkey Turkey India China 1997 2014 2005 1994 2005 Full Liberalization 42.9 78.0 86.1 127.5 86.6 Moving to U.S. Efficiency - 24.5 30.2 59.2 30.6 Source: Authors’ calculations, HK (2009). 6.2 Distortion and Productivity In the absence of frictions, more capital and labor should be allocated to firms with higher TFPQ to the point where their higher output results in lower prices and the exact same TFPRs as those of smaller firms. Hence, TFPR would not vary across firms within an industry unless firms face distortions. However, in reality, capital and labor distortions engender TFPR dispersion within industries, as shown in Figure 1. Distortions would be particularly harmful if they are positively correlated with firm’s physical productivity (i.e. they are higher for more productive firms). To see the relationship between productivity and distortions, Figure 3 plots TFPQ against TFPR. In the frictionless world, all firms would fall along the zero log ⁄ line. Along this undistorted equilibrium line firms would differ only on their physical productivity, TFPQ. However, figure 3 shows that TFPR is strongly increasing in TFPQ in Turkey, suggesting that more productive firms face larger distortions. In other words, the figure implies that high productivity firms are subject to higher implicit taxes that keep these firms smaller than their optimal levels. 16   Similarly, low productivity firms receive implicit subsidies that enable these firms to expand and lower the firms’ marginal products. Figure 3: Productivity vs Distortions Distortions, log (TFPR) Physical Productivity, log (TFPQ) Figure 4: Productivity vs Capital Wedge Figure 5: Productivity vs Output Wedge Output Wedge Capital Wedge Physical Productivity, log (TFPQ) Physical Productivity, log (TFPQ) To better understand the sources of distortions, we decompose the overall distortion into the “capital” wedge, log 1 , and the “output” wedge, log . Figure 4 shows that there is no systematic relation between the capital wedge and productivity level. At almost every productivity level, capital wedge dispersion is quite high, indicating that at every given level of productivity, some firms have an easier access to capital markets while some firms have difficulties. However, the capital wedge does not 17   systematically increase as firm level productivity increases. This suggests that capital distortions are not the main drivers of resource misallocation in Turkey. Figure 5 shows that output wedges are monotonically increasing in productivity. This suggests that compared to a frictionless equilibrium, productive firms are subject to a larger output wedge, causing them to produce less than their optimal output, while unproductive firms receive an implicit output subsidy and produce beyond their optimal level, resulting in an inefficient allocation of resources and thus lower aggregate TFP. Figure 4 and Figure 5 jointly suggest that the answers to resource misallocation in Turkey lies in the output markets, not in the factor markets. Thus, policy measures that focus on output markets and eliminate distortions can reduce misallocation and bolster aggregate TFP in Turkey. 6.3 Distortions and Firm Size The relationship between distortions and size is an important dimension to examine to understand the costs of misallocation. Size is measured in terms of value added of the firms. Figure 6 plots the relationship between firm size and productivity distribution of firms with respect to their sector averages, while Figure 7 plots the relationship between firm size and distortions. It is clear that there is a strong positive relationship between firm size and productivity, suggesting that larger firms are on average more productive than their smaller counterparts. Similarly, large firms have higher TFPR compared to the sector averages, whereas small firms have smaller TFPR than the sector averages. This implies that small firms are less productive, but receive implicit subsidies that grow them larger than otherwise they would. Moreover, on average, medium, and especially large, firms are more productive than their competitors, but they face greater implicit taxes that raise their marginal products and constrain their growth. As a result, these more productive, large firms remain smaller than they would otherwise. Overall, Figures 6 and 7 together imply that small, unproductive firms operate at the expense of large, productive firms, as a result of idiosyncratic distortions, leading to a significant misallocation of resources within industries. In the absence of distortions, small firms would cut their production and increase their prices, ending up with higher TFPR, whereas medium and large firms would expand their production and lower their prices, ending up with lower TFPR. Large, more productive firms would expand, while small less productive firms would downsize until TFPR across all firms equalized within industries. 18   Figure 6: Productivity vs Firm Size Figure 7: Distortions vs Firm Size Physical Productivity,  Distortions, log (TFPR) log (TFPQ) Size (Value Added) Size (Value Added) 6.4 Misallocation by Industry The results presented so far are for the manufacturing sector as a whole, weighted by the industry value added shares. By focusing on the aggregate outcome we might obscure important differences across industries. The finding will provide evidence for policymakers to focus on certain industries to address resource misallocation. Hence, it is instructive to investigate to what extent distortions vary across sectors. We have grouped industries that are closely related under broader categories in order to reduce the number of industries. The food sector includes manufacturing of food products, beverages and tobacco products (NACE rev. 2 2-digit sector codes 10-12). The textiles sector includes manufacturing of textiles and wearing apparel (NACE rev. 2 2-digit sector codes 13-14). The chemicals sector includes manufacturing of chemicals and chemical products, rubber and plastic products (NACE rev. 2 2-digit sector codes 20 and 22). The metals sector includes manufacturing of basic metals and of fabricated metal products, except machinery and equipment (NACE rev. 2 2-digit sector codes 24 and 25). The transport sector includes manufacturing of motor vehicles, trailers and semi-trailers, and manufacturing of other transport equipment (NACE rev. 2 2-digit sector codes 29 and 30). Finally, the electronics sector includes manufacture of computer, electronic and optical products, and manufacturing of electrical equipment (NACE rev. 2 2-digit sector codes 26 and 27). The remaining sectors are stand-alone sectors, and do not include any other 2-digit sector. Table 5 shows the number of firms in each industry in each year. 19   Table 5: Number of Firms in Each Industry Year Elect. Machinery Metals Furniture Chemicals Text. Trans. Food Leather 2003 237 367 587 196 571 2,829 268 584 96 2004 316 501 686 267 729 3,475 370 653 117 2005 388 772 1,065 426 1,025 4,605 577 856 186 2006 390 888 1,213 506 1,152 4,770 637 924 220 2007 426 938 1,169 535 1,151 4,511 673 874 222 2008 476 1,206 1,269 667 1,285 4,628 706 1,511 248 2009 486 1,026 1,296 612 1,234 4,027 724 1,415 216 2010 609 1,340 1,665 945 1,688 5,032 809 1,934 343 2011 664 1,497 2,028 1,120 1,856 5,637 793 2,117 398 2012 720 1,682 2,329 1,231 1,968 6,120 833 2,340 488 2013 771 1,759 2,438 1,257 2,009 6,252 865 2,231 505 2014 812 1,874 2,551 1,364 2,087 6,125 879 2,209 519 Source: TURKSTAT and authors’ calculations. Table 6 presents the potential TFP gains if resource misallocation is removed in each of these stand-alone industries. Table 6 shows that potential TFP gains are above 80% in the Furniture and Chemicals sectors, and around 90% in the Textiles, Transportation, Food, and Leather sectors in 2014. Thus, the room for TFP growth by removing distortions and improving allocative efficiency is large in these sectors. Conversely, the potential TFP gains are about 50% in Machinery and Metals, implying that reforms can also yield meaningful TFP growth in these sectors. The best performing sector is Electronics, where potential TFP gains are close to 0% in 2014. Also note that misallocation in Electronics was already low in 2003 in comparison to the other sectors. In 2011 and 2013, the potential TFP gains were negative in the Electronics sector, implying that equalizing TFPR within 4-digit industries would lead to lower aggregate TFP in this sector. This may be due to the relationship between distortions and productivity level. In section 6.2, we showed that TFPR is positively correlated with TFPQ in the manufacturing sector in Turkey, meaning that distortions favor less productive firms and punish more productive firms. However, if TFPR is negatively correlated with TFPQ, in other words, if distortions punish less productive firms and favor more productive firms, potential TFP gains can be negative. A closer look at the electronics sector in Turkey and a deeper analysis of market structure, infrastructure, rules, and regulations in this sector can provide valuable lessons for other sectors and policy makers. 20   Table 6: TFP Gains by Industries Year Elect. Machinery Metals Furniture Chemicals Text. Trans. Food Leather 2003 41 90 141 80 119 147 89 165 370 2004 48 47 60 84 104 110 70 146 133 2005 21 68 67 54 82 103 84 131 128 2006 57 58 48 71 67 78 82 123 117 2007 11 57 30 73 63 77 99 124 68 2008 34 60 66 86 96 72 106 73 75 2009 28 64 75 55 100 64 99 73 89 2010 27 52 57 75 76 75 65 75 70 2011 -4 55 54 63 80 86 63 64 78 2012 7 41 53 75 75 84 79 63 80 2013 -2 46 71 94 73 79 105 71 81 2014 7 47 52 80 86 88 89 92 94 Source: TURKSTAT and authors’ calculations. 6.5 Misallocation by Technology Intensity of Sectors In this section, we separately implement the same exercise by constraining the sample to individual technology classes to shed light on whether distortions and misallocation vary with the technology intensity. Misallocation would be particularly harmful if it systematically distorts resources in the more innovative and dynamic sectors. In this section, we rely on Eurostat technology intensity classifications that are based on NACE Rev.2. The low technology class includes the following 2-digit sectors: food products, beverages, tobacco products, textiles, wearing apparel, leather and related products, wood and of products of wood and cork, except furniture, paper and paper products, printing and reproduction of recorded media, manufacturing of furniture, and other manufacturing.10 The medium-low technology class includes coke and refined petroleum products, rubber and plastic products, non-metallic mineral products, basic metals, fabricated metal products, except machinery and equipment, and repair and installation of machinery and equipment.11 The medium-high technology class includes manufacturing of chemicals and chemical products, electrical equipment, machinery and equipment n.e.c, vehicles, trailers and semi-trailers, and other transportation equipment.12 The high technology class includes manufacturing of basic pharmaceutical products and                                                              10 NACE rev. 2 2-digit sector codes 10-18, 31, and 32. 11 NACE rev. 2 2-digit sector codes 19, 22-25, and 33. 12 NACE rev. 2 2-digit sector codes 20, 27-30. 21   13 pharmaceutical preparations, computer, electronic, and optical products. Table 7 shows the number of firms in each technology class in each year. Table 7: TFP Gains by Technology Classes Low Tech Medium-Low Medium-High Tech High Tech Tech Year Firms TFP Firms TFP Firms TFP Gains Firms TFP Gains Gains Gains 2003 4,153 161 1,564 121 1,007 69 79 535 2004 5,081 119 1,888 63 1,364 47 111 185 2005 6,896 104 2,769 71 1,949 58 123 142 2006 7,333 89 3,205 57 2,165 66 125 135 2007 7,049 84 3,140 49 2,239 59 146 276 2008 8,073 83 3,411 54 2,597 68 195 221 2009 7,234 62 3,270 60 2,464 77 175 209 2010 9,549 75 4,363 47 3,068 62 213 216 2011 10,662 74 5,116 47 3,263 66 211 252 2012 11,727 75 5,622 44 3,578 62 211 231 2013 11,869 81 5,783 56 3,745 58 213 175 2014 11,913 89 6,076 57 3,934 61 229 215 Source: TURKSTAT and authors’ calculations. There is no systematic relationship between the technology intensity of sectors and misallocation, according to Table 7. However, the high tech class exhibits the highest frictions in Turkey, suggesting that the most innovative sectors are highly distorted.14 Medium-low and medium-high tech classes are the best performing technology classes with respect to the allocative efficiency. The potential TFP gains are around 60% in these two technology classes in 2014. Although the potential TFP gains in these classes are lower than the remaining technology classes, the potential gains are still substantial. The potential gains are around 90% in low technology class in 2014, suggesting that aggregate TFP in this class can almost double if TFPR equalizes within 4-digit industries. The aggregate TFP in high tech can more than double if distortions are eliminated in this technology class. The Turkish government has various subsidy and incentive programs for firms in high technology sectors in order to achieve its aspirations of increasing value added and technology content of production. Although there are eligibility criteria for these subsidies, programs do not control for productivity levels of firms. Perhaps industrial subsidies have supported less productive firms, and might have caused a larger misallocation in high technology class, compared to other technology classes. If this assertion is true, the                                                              13 NACE rev. 2 2-digit sector codes 21 and 26. 14 The results should be taken with caution because of the small sample size for high tech firms. 22   availability of various and large incentive programs for high tech firms, which do not control for firm productivity, could explain larger dispersion in this class. However, we leave the identification of the reasons behind differences in technology classes for future research. Table 8: TFP Gains From Full Liberalization Year Low Tech Medium-Low Tech Medium-High Tech High Tech 2003 161 121 69 535 2004 119 63 47 185 2005 104 71 58 142 2006 89 57 66 135 2007 84 49 59 276 2008 83 54 68 221 2009 62 60 77 209 2010 75 47 62 216 2011 74 47 66 252 2012 75 44 62 231 2013 81 56 58 175 2014 89 57 61 215 Source: TURKSTAT and authors’ calculations. 6.6 Misallocation by Region In this section, we separately implement the same exercise by constraining the sample to individual regions to examine the variation of distortions and misallocation among different regions. Unfortunately, a regional identifier is not available for 2013 and 2014, thus we restrict our regional analysis to the time frame of 2003-12. Following Gonenc et al. (2012) and World Bank (2014), we group the regions of Turkey at the NUTS2 level into 3 sub-regions. We identify those sub-regions that had relatively low value added per capita in 2004 and experienced high job growth between 2004 and 2012 as “Tigers”. Sub-regions that had high value added in 2004 are belonging to the “West”, representing provinces that have traditionally acted as industrial growth centers. The rest of the sub-regions are identified as the “Others”. 23   Figure 8: Regional Classification West, Anatolian Tigers, Others Source: World Bank (2014) Table 8 suggests that the misallocation within Tigers is smaller than West and Others. This is an interesting result because it indicates that business environment in Tigers treat firms relatively equally compared to a more developed region like West. Table 8 shows the hypothetical gains from equalizing TFPR within sectors for each region. The reallocation of resources from less productive firms to more productive firms within industries could increase TFP by about 81% in Others, 70% in West, and 46% in Tigers in 2012. The potential TFP gains are large in each region, thus new reform momentum can yield significant improvements in aggregate productivity in every region. Especially in West, where firms are disproportionately located, elimination of distortions could lead to a substantial rise in TFP not only in West, but also in aggregate manufacturing in Turkey, given that West has the largest weight among regions. 24   Table 9: TFP Gains by Regions Anatolian Tigers West Others Year Firms TFP Firms TFP Firms TFP Gains Gains Gains 2003 864 146 5,405 117 113 142 2004 1,156 83 6,704 85 131 141 2005 1,743 68 9,308 89 171 137 2006 2,032 62 10,041 77 232 71 2007 1,969 56 9,806 76 248 77 2008 2,361 56 11,000 78 328 37 2009 2,073 48 10,196 80 305 36 2010 3,051 54 13,076 71 526 68 2011 3,594 44 14,500 72 646 63 2012 4,145 46 15,737 70 783 81 Source: TURKSTAT and authors’ calculations. 7. Conclusion This paper uses Hsieh and Klenow’s (2009) framework to quantitatively examine resource misallocation within narrow industries in Turkey. We find that Turkey’s resource misallocation is substantial. The hypothetical gain of moving to ‘U.S. efficiency’ is 24.5% of manufacturing TFP in 2014. The availability of long panel data allows us to examine resource misallocation over time and by disaggregated sectors. Allocative efficiency improved since 2003, but began to slow down since 2007. Moreover, in 2014, the latest year of observation, the resource misallocation in Turkey worsened, reversing the earlier trend. The finding suggests there is substantial room to improve TFP by reducing misallocation within sectors. Particularly, we have identified the four sectors where misallocation is most pronounced and gains are around 90%: textile, transport, food, and leather. A first direction for further research would be to more closely examine drivers of reduction in misallocation in the past twelve years. Identifying causal relationships between changes in the regulatory environment or policies, and improvement in misallocation, could shed light on this important phenomena and could be useful for policy makers to design policies to address the misallocation issue. A second, and related, avenue for future research is to look at cross-sectional differences in misallocation patterns among regions, industries, and technology classes, and to identify causes of such differences. One option is to analyze how regional and industrial subsidies affect differences in misallocation patterns across regions and industries. This analysis could help design policies to reduce regional gaps and boost aggregate productivity growth in the manufacturing sector. 25   References Busso, M., Madrigal, L. and C. Pages (2013). Productivity and resource misallocation in Latin America. The BE Journal of Macroeconomics, 13(1):903-932. Camacho, A. and Conover, E. (2010). Misallocation and productivity in Colombia’s manufacturing industries. IADB Research Department Publications 4654 Cirera, Xavier, Roberto N. Fattal Jaef and Hibret B. Maemir (2015) Taxing the good? Distortions, Misallocation, and Productivity in Sub-Saharan Africa, mimeo Foster, L., Haltiwanger, J. and C. Syverson (2008). Reallocation, Firm Turnover, and Efficiency: Selection and Productivity or Profitability. American Economic Review, 98 (2008), 394-425. Gonenc, R., Rohn, O., Koen, V. and S. Saygili (2012). Structural Reforms to Boost Turkey`s Long-Term Growth. OECD Economics Department Working Papers, 987. Hsieh, C.-T. and Klenow, P. J. (2009). Misallocation and manufacturing TFP in China and India. The Quarterly Journal of Economics, 124(4):1403-1448. Kalemli-Ozcan, S. and Sorensen, B. E. (2012). Misallocation, property rights, and access to finance: Evidence from within and across Africa. NBER Working paper Working Paper 18030 Restuccia, D. and Rogerson, R. (2008). Policy distortions and aggregate productivity with heterogeneous establishments. Review of Economic Dynamics, 11(4):707-720. Ryzhenkov, M. (2016). Resource misallocation and manufacturing productivity: The case of Ukraine. Journal of Comparative Economics, Vol 44 (1). Taymaz, Erol and Yilmaz, Kamil. (2009). Foreign Direct Investment and Productivity Spillovers: Identifying Linkages through Product-based Measures. Unpublished manuscript. World Bank (2014). Turkey’s Transitions. Washington, D.C.: World Bank Group. 26