Policy Research Working Paper 9557 Gravity Model–Based Export Potential An Application to Turkey Alen Mulabdic Pinar Yasar Macroeconomics, Trade and Investment Global Practice March 2021 Policy Research Working Paper 9557 Abstract This paper presents a framework to study countries’ export and Japan are important untapped destination markets, potentials. It uses a gravity model to develop measures of accounting for US$29 billion (16–17 percent of total export and trade policy potentials at the aggregate, bilateral, exports) of missing exports. Industry-level results suggest and industry levels. The methodology is applied to the case that Turkey has high export potential in the electronics and of Turkey. The analysis finds that Turkey was moderately chemical industries. under-exporting over 2010–17. The United States, China, This paper is a product of the Macroeconomics, Trade and Investment Global Practice. 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 amulabdic@worldbank.org and pyasar@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 Gravity Model–Based Export Potential: An Application to Turkey † Alen Mulabdic and Pinar Yasar* Keywords: Gravity Model, Export Potential, Turkey JEL Codes: F13, F14, F15 † This paper was written as a background paper for the Turkey Country Economic Memorandum: Leveraging Global Value Chains for Inclusive Growth. We are grateful to Sezai Ata, Cristina Constantinescu, Michael Ferrantino, David Knight, Habib Rab, Gonzalo Varela, and workshop participants at the World Bank in Ankara for helpful comments and suggestions. The paper also benefitted from inputs of experts from the Ministry of Trade, the Presidency of Strategy and Budget, the Ministry of Treasury and Finance, the Central Bank of the Republic of Turkey and the Ministry of Industry and Technology. Errors are our responsibility only. 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. * Alen Mulabdic, World Bank, 1818 H Street, Washington DC, USA, Email: amulabdic@worldbank.org; Pinar Yasar, World Bank, Ugur Mumcu Cad. No:88, Ankara, Turkey, Email: pyasar@worldbank.org. 1. Introduction The gravity model has become a workhorse tool for empirical analysis of international trade. The model has been widely used to estimate impact of geography and institutions on trade flows since the first application by Tinbergen (1962). More recent theoretical efforts, which started with Anderson (1979), have provided theoretical foundations for the gravity equation which can be derived from various theoretical models of international trade. These recent theoretical developments helped in the refinement of the original gravity equation which is now widely used to assess the effects of policy variables on trade flows as well as welfare (see Head and Mayer, 2014). This paper presents a framework to analyze countries’ export potentials using a gravity model. Our analysis incorporates recent developments of the empirical literature on the gravity model and proposes different measures of export potential that can be used to benchmark countries’ exports. In terms of empirical contribution, the paper addresses some of the issues of previous estimations of export potential. First, we estimate the gravity equation using the Poisson Pseudo Maximum Likelihood (PPML) estimator which corrects for the issues of zero trade flows and heteroskedasticity error terms (Silva and Tenreyro, 2006) which are typical of the log-linear form of the gravity equations used in the ordinary least squares (OLS) estimation. Second, in addition to the standard gravity variables, we proxy for multilateral resistance terms with measures of remoteness terms instead of only controlling for countries’ GDPs. The inclusion of remoteness terms has two implications: (i) it mitigates the bias outlined by Baldwin and Taglioni (2006), and (ii) it allows the model to predict aggregate under-exporting, which is not possible when fixed effects are included in a PPML model (Fally, 2012). We illustrate the framework analyzing Turkey’s export performance. 1 First, we use predicted and observed trade flows to construct an index of missing exports which varies between -100 and 100. 1 A number of studies estimate gravity equations to study Turkey’s export potential. For instance, Karagoz and Saray (2010) analyze trade between Turkey and Asia-Pacific countries and find that trade is below its 2 We find that Turkey is moderately under exporting relative to other countries. Countries with the highest export potential are Nepal, Burundi, and Rwanda. Second, we analyze bilateral trade flows to identify destination markets in which Turkey is under-exporting. We estimate that during the 2010-2017 period Turkey had on average 12 billion dollars of missing exports to the United States. This value is 30 percent higher than the current level of Turkish exports to the United States. Other important destination markets are China and Japan with 10 billion dollars (3 times higher than the current level of exports) and 7 billion dollars (14 times higher than the current level of exports), respectively. Third, industry-level results suggest that Turkey has a high export potential in electronics and chemical industries. Finally, we exploit data on the content of deep trade agreements to develop a measure of trade policy potential. First, we use a gravity model to assess the impact of deep trade agreements on Turkey’s trade. The results suggest that Turkey’s trade is more sensitive to deep agreements than that of other countries. Second, using the gravity estimates we investigate the possible impact on Turkey’s exports of deeper trade arrangements with its current preferential trade agreement (PTA) partners. Turkey could increase its exports from 1.8 percent to 9.2 percent by modernizing its existing trade agreements. To achieve this, Turkey would need to address in its existing trade agreements issues related to customs, export taxes, technical barriers to trade (TBT), sanitary and phytosanitary standards as well as more complex issues related to competition policy, intellectual property rights (IPR) protection, and movement of capital. The rest of the paper is organized as follows. The next section discusses the data and empirical strategy. Sections 3 and 4 present the aggregate and industry level results. The trade policy potential results are presented in Section 5. Concluding remarks follow. potential with Guinea, Peru, Myanmar, Mexico, the Lao People’s Democratic Republic, and Brunei. Ata (2012) estimates a gravity equation covering 67 countries and finds that Turkey under exports to 48 of 67 the countries in the sample. Finally, a World Bank (2014) report assesses the potential of services trade between Turkey and the EU by using a gravity model of trade in services. According to the results, Turkey is found to be under-trading services with nearly all EU member states, suggesting untapped potential to increase bilateral trade. 3 2. Data Sources and Empirical Strategy In this section, we provide details on the empirical methodology, data sources and measurement of the variables used in the empirical analysis. Our data on trade flows, reported imports, at the HS 6-digit (HS 1988/1992) are from the World Bank’s World Integrated Trade Solution (WITS). The data cover 105 countries across all geographic regions for the 2000-2018 period and cover 82 percent of world trade. 2 The sample is restricted to countries with population greater than 5 million in year 2000. Population data are from the World Development Indicators (WDI) database. In order to analyze industries that are of strategic interest to Turkey, we aggregate the HS 6-digit trade flows to obtain a new classification that covers 20 sectors. The Turkish Ministry of Trade identified 6 strategic sectors (in grey in Table 1) for which the goal is to expand Turkey’s global share. The sector “1-24 Food” covers all processed products in HS chapters 1-24, “28-38 Chemical” includes the sum of all 767 products in HS chapters 28-38, “84-85 Electronics” are all the electronics products under chapters 84-85, “84 Machinery” includes machinery and mechanical appliances in chapter 84, “85 Electrical” covers electrical machineries and equipment in chapter 85, and “86-89 Automotive” are all the products in chapters 86-89. 3 2 Specifications which include controls for capital stock per worker from the Penn World Tables 9.1 are restricted to the 2000-2017 period due to lack of data. 3 The concordance table between HS 1988/92 products and the new sectors is available upon request. 4 Table 1: Strategic Sectors in Grey Number of HS Sectors (HS Chapters) 6-digit Products 01-05 Animal 96 06-15 Vegetable 226 1-24 Food 381 16-24 Foodstuffs 26 25-27 Minerals 173 28-38 Chemicals 767 39-40 Plastic / Rubber 191 41-43 Hides, Skins 76 44-49 Wood 238 50-63 Textiles, Clothing 819 64-67 Footwear 55 68-71 Stone / Glass 196 72-83 Metals 597 84-85 Mach/Elec 23 84-85 Electronics 156 84 Machinery 438 85 Electrical 150 86-89 Automotive 133 90-97 Miscellaneous 389 98-99 Special 99 To empirically assess if Turkey is under exporting, we estimate a simple gravity model, which is widely used in the trade literature to assess the effects of trade policy changes on trade flows (Head and Mayer, 2014). As shown in Costinot and Rodríguez-Clare (2014), the following gravity equation emerges from different theoretical frameworks: 1− (1) = � � Π Ρ 1− (2) (Π )1− = �� � Ρ 1− 1− (3) �Ρ � = �� � Π 5 where is the bilateral trade flow from country to country , is country j’s total expenditure, = ∑ is country i’s income, is the elasticity of substitution among different product varieties, and is the bilateral trade costs between and . Π and Ρ are the outward and inward multilateral resistances that capture ’s and ’s market access respectively. Log-linearizing Equation (1) and assuming the equation holds in each year , we obtain the following gravity equation: ln ( ) = ln( ) + ln� � − ln() + (1 − ) ln� � − (1 − ) ln�Ρ � − (1 − )ln (Π ) (4) Exports , for 105 reporters, come from UN Comtrade. 4 As it is common in the gravity literature, we use exporter’s and importer’s nominal gross domestic products (GDPs), from the World Bank’s World Development Indicators (WDI), to proxy for total production ( ) and expenditure ( ). We also assume that bilateral trade costs are a function of the following observable variables: (1 − ) ln� � = 1 ln�1 + � + 2 + β3 ln( ) + 4 (5) + 5 + 6 are bilateral applied tariff duties, is an indicator variable that takes value of 1 if and have a trade agreement in year from Mario Larch’s Regional Trade Agreements Database from Egger and Larch (2008), is the geographical distance between and , is a variable that takes value of 1 for country-pairs that share a border, is a binary variable equal to 1 if and share the same language, and captures the presence of any colonial ties. Bilateral tariff duties are interpolated using data from the Market Access Map (MAcMap) database while all the other variables come from CEPII’s gravity database. 4 See Table A1 in the appendix for the descriptive statistics for the main variables used in the empirical analysis. 6 To control for the unobservable multilateral resistance terms defined in Equations (2) and (3), we construct “remoteness indexes” (Baier and Bergstrand, 2007; Wei, 1996). A popular alternative to this method requires the inclusion of exporter-year and importer-year fixed effects. These fixed effects account for multilateral resistance terms as well as any country specific time determinants of trade. However, in a PPML model fixed effects impose a perfect fit in terms of total exports and total imports for each country, which implies that countries’ total exports would be always perfectly predicted and never departing from their potential. The two indexes of remoteness are defined as the GDP weighted distance for exporters and importer : = � (6) � � = � (7) � � Finally, we include additional controls for factor endowments: natural resources and capital per worker (Chor, 2010; Romalis, 2004). First, to control for the presence of resource rich countries, we use data from the World Bank to construct variables equal to 1 if average rents from oil, coal, and mineral exceed 10 percent of GDP for the 2000-2018 period. Second, we follow Levchenko and Zhang (2014) and construct variables for capital stock per worker based on data from the Penn World Tables 9.1. In addition, we also include exporters’ and importers’ GDP per capita from the WDI to account for countries’ level of development, which can affect the composition and quality of imports and exports. 7 3. Is Turkey Under-Exporting Given Its Observable Characteristics? In this section we use estimates from the gravity model to benchmark Turkish exports. a. Gravity Model As it is standard in the recent trade literature, we use a PPML estimator to estimate the following gravity equation: = �1 ln�1 + � + 2 + 3 ln� � + 4 + 5 + 6 + 7 ln( ) + 8 ln� � + 9 ℎ (8) + 10 ℎ + 11 ln( ) + 12 ln( ) + 13 ln � � + 14 ln � � + 15 ln( ) + 16 ln� �� + Equation (8) is obtained by substituting (5), (6), and (7) in the exponential form of Equation (4), and by adding the error term . Additional controls include exporters’ and importers’ GDPs, per capita GDPs, dummy variables for resource rich economies and capital stock per worker ratios. Important features of the PPML estimator are that it accounts for the problems of zero trade flows, as not all countries trade with all the countries in the world, and issues related to the heteroskedasticity of trade data, which together can make the OLS estimates biased and inconsistent. Table 2 reports the PPML estimates from the gravity Equation (8). Results are in line with the trade gravity literature in terms of signs and magnitude of the coefficients. First, the results point to a significant effect of trade policy variables. RTAs are estimated to increase trade between 16 and 24 percent. The RTA coefficients are lower in Columns 2, 4 and 6 as tariff liberalizations are accounted for by the ( + 1) variable. Thus, in those specifications the RTA variable only captures reductions in non-tariff barriers. A one percent reduction in bilateral tariffs is estimated to increase trade between 1.6 and 2.5 percent. Distance is estimated to reduce bilateral trade, while sharing a border and speaking the same language has a positive impact on trade flows. The coefficients on the remoteness indexes suggest that larger and more remote countries trade more 8 intensively among themselves. Finally, more developed countries tend to export less while importers’ economic development and colonial ties are not statistically significant. Table 2: PPML Gravity Estimates (1) (2) (3) (4) (5) (6) VARIABLES Trade Trade Trade Trade Trade Trade RTA 0.211*** 0.171** 0.215*** 0.154** 0.209*** 0.148** (0.057) (0.066) (0.059) (0.067) (0.059) (0.067) ln(distance) -0.728*** -0.714*** -0.763*** -0.755*** -0.766*** -0.758*** (0.031) (0.034) (0.037) (0.038) (0.037) (0.038) Border 0.454*** 0.452*** 0.423*** 0.405*** 0.412*** 0.396*** (0.086) (0.086) (0.088) (0.087) (0.088) (0.087) Language 0.201** 0.194** 0.242*** 0.250*** 0.273*** 0.280*** (0.097) (0.096) (0.094) (0.093) (0.093) (0.093) Colony -0.124 -0.125 -0.112 -0.111 -0.135 -0.134 (0.106) (0.106) (0.100) (0.098) (0.102) (0.100) ln(GDP exp.) 0.849*** 0.852*** 0.892*** 0.899*** 0.889*** 0.896*** (0.021) (0.022) (0.034) (0.035) (0.034) (0.035) ln(GDP imp.) 0.824*** 0.818*** 0.828*** 0.836*** 0.830*** 0.837*** (0.030) (0.030) (0.032) (0.031) (0.031) (0.031) Mineral rich exp. 0.023*** 0.022*** 0.021*** 0.020*** 0.020*** 0.019*** (0.003) (0.003) (0.003) (0.003) (0.004) (0.004) Mineral rich imp. -0.010** -0.007* -0.009** -0.008** -0.011** -0.010** (0.005) (0.004) (0.004) (0.004) (0.004) (0.004) ln(Remoteness exp.) 0.548*** 0.555*** 0.516*** 0.522*** 0.559*** 0.566*** (0.086) (0.086) (0.090) (0.090) (0.100) (0.100) ln(Remoteness imp.) 0.554*** 0.545*** 0.533*** 0.504*** 0.651*** 0.615*** (0.098) (0.096) (0.096) (0.095) (0.114) (0.113) ln(GDP pc exp.) -0.126** -0.127** -0.177*** -0.183*** (0.050) (0.050) (0.049) (0.049) ln(GDP pc imp.) 0.005 -0.048 -0.189*** -0.228*** (0.032) (0.030) (0.054) (0.053) ln(K/L exp.) 0.074 0.081 (0.065) (0.064) ln(K/L imp.) 0.280*** 0.261*** (0.086) (0.083) ln(tariff+1) -1.591* -2.449*** -2.419*** (0.870) (0.891) (0.900) Observations 168,581 168,581 166,908 166,908 152,515 152,515 Robust standard errors, clustered at the country-pair level, are in parentheses. *** p<0.01, ** p<0.05, * p<0.1 9 b. Aggregate Results Does Turkey under export given its observable characteristics? To answer this question, we implement a two-step procedure based on estimates from Table 2. In a first step, we use the coefficients from Table 2 to predict bilateral trade flows based on countries’ observable characteristics. In a second step, we aggregate exports at the country level to calculate the aggregate index of missing exports which is defined as follows: ̂2 + ̂1 ln�1 + � + � = � ̂4 ̂3 ln� � + ̂6 + ̂5 + + ̂7 ln( ) + ̂8 ln� � + 9 ℎ + 10 ℎ + 11 ln( ) 1st Step (9) + 12 ln( ) + 13 ln � � + 14 ln � � + 15 ln( ) + 16 ln� �� �, − ∑ , ∑ 2nd Step , = � � ∗ 100 (10) �, + ∑ , ∑ Figure 1 presents the result from Equation (10). The index of missing exports varies between 100 and -100. The maximum value of the index is obtained when observed bilateral trade flows are equal to 0, but the model predicts positive exports; while the minimum value (i.e., -100) is obtained when the predicted value is equal to 0 and we observe positive export values. Panel A includes countries with an average index of missing exports greater than 10 percent over the 2010-2017 period, while Panel B shows the results for the other countries. The estimates suggest that Nepal, Burundi, and Rwanda have the highest export potential given their observable characteristics. The gravity model predicts these countries should be exporting at 10 least more than seven times what they exported. 5 At the other extreme, there are Malaysia, Vietnam, and South Africa that export more than what we would expect given their economic size, trade costs, geographic location, and other observable characteristics. Turkey is moderately under exporting. The index is equal to 21 which suggests that exports should have been 53 percent higher than what we observe in the data. Other countries in the same under-exporting range are the United Kingdom, the República Bolivariana de Venezuela, and Morocco. Figure 1: Country Level Index of Missing Exports Panel A 5 For a given export potential, we can calculate the relationship with respect to the observed trade flows using the following formula: 1+ �= 100 ∗ 1− 100 11 Panel B c. Index of Missing Exports by Destination Countries The aggregate results suggest that Turkey is exporting below its potential. To identify with which destination countries Turkey has the highest export potential, we take a closer look at differences between predicted and observed trade flows at the bilateral level. Also, we use the framework to understand if Turkey is under exporting to the target countries identified in the Export Master Plan. The Master Plan is based on macroeconomic data, market growth, and other factors, and it identified 17 countries which present the highest potential for Turkish exporters. 6 We use the predicted trade flows from Equation (9) and construct country-pair indicators of export potential as follows: 6 The 17 countries in the Turkish Export Master Plan are the United States, Brazil, China, Ethiopia, Morocco, South Africa, the Republic of Korea, India, Iraq, the United Kingdom, Japan, Kenya, Malaysia, Mexico, Uzbekistan, the Russian Federation and Chile. 12 � , − , , = � � ∗ 100 (11) � , + , � , are the predicted exports from Turkey to country in year and , are the where observed exports. Figure 2 presents the estimated trade potential for the 17 target countries identified in the Export Master Plan. Turkey is predicted to be over exporting to 4 out of the 17 countries in the plan. These countries are Ethiopia, Morocco, Uzbekistan, and the United Kingdom. One way to interpret these results is that expanding trade in these destinations could be difficult given their geography and size. However, the model does not account, for instance, for improvements in bilateral relations between Turkey and Uzbekistan and their strategic partnerships in the textile, metallurgical and automotive industries, which can increase trade in the future. Figure 2: Index of Missing Exports for Countries in the Export Master Plan There are other countries that could provide export opportunities for Turkish exporters. Figure 3 presents the results from Equation (11) for countries not included in the master plan. Turkey does not almost export to the Lao People’s Democratic Republic, which is ranked first in terms of 13 percent of missing exports. Other countries are Myanmar, a country with a big export potential, and Cambodia, a very competitive exporter according to the results in Figure 1. One issue with the export potential defined in Equation (11) is that the index fails to capture the size of the market defined in dollar terms. For instance, an index of missing exports equal to 99 for Lao PDR could be smaller in dollar terms than an index of 20 for Peru. To better understand the market size in destination countries, we take a closer look at deviations of observed trade flows from the predicted ones. That is to say, we look at the numerator of Equation (10). The dollar value figures of “missing” exports provide a better measure of opportunities and are more informative in case the government is planning activities aimed at promoting Turkish exporters in specific destinations (e.g., trade events), which might not be justified for smaller markets. Figure 3: Bilateral Index of Missing Exports (>20% only) 14 Figure 4 Panel A shows that ranking of countries changes dramatically when we look at missing exports. The United States is ranked first with around 12 billion dollars of missing exports, followed by China and Japan with 10 billion dollars and 7 billion dollars respectively. 7 Among the non-targeted countries, we find some important destination markets which have a relatively low index of missing exports but rank high in terms of market size. This is the case for Italy and France, ranked first and third, which together have missing exports of almost 8 billion dollars. Other important markets that are ranked high in terms of both the index and missing exports are Greece and Canada. The comparison between Panel A and Panel B indicates that the Turkish government identified the most important countries in terms market size and suggests that the list could have aimed at Canada, Indonesia, and Australia in case the aim is to diversify exports away the European Union. Table A2 shows that China is the top “over exporter” in Canada and Australia, while Thailand over exports around 4 billion dollars to Indonesia. Other important exporters in these three countries are Mexico, the Republic of Korea, and Malaysia. Germany is the only European country listed among the top five “over exporters” in Canada and Australia. 7 The average missing exports can produce inconsistent results when compared to the average index of missing exports. These inconsistencies are explained by the non-linear transformation used to construct the index of missing exports in which changes in missing exports can differ in magnitude from changes of the index. For instance, the Russian Federation has a positive average index of missing exports but negative average missing exports. In year 2017, missing exports declined by 32 percent while the index dropped by 48 percent. 15 Figure 4: Missing Exports (USD) Panel A: Target Countries Panel B: Other Countries (>200 USD mln) 16 4. Industry-Level Export Potential This section presents the industry level analysis. We extend the gravity model to the sectors described in Table 1. As it is standard in the gravity literature, we modify Equation (8) and estimate the following sector level gravity equation: = �1 ln�1 + � + 2 + 3 ln� � + 4 + 5 + 6 + 7 ln( ) + 8 ln� � + 9 ℎ + 10 ℎ + 11 ln( ) + 12 ln( ) (12) + 13 ln � � + 14 ln � � + 15 ln( ) + 16 ln� �� + where is the bilateral trade flow from country to country at time in sector . We estimate Equation (12) for each sector allowing each explanatory variable to have a sector specific impact. For instance, the impact on trade of speaking the same language, 5 , for the automotive industry may differ from the impact language has on trade in vegetables. Table 3 presents the industry level estimates from a PPML model. The estimates are qualitatively similar to the aggregate results, but there is heterogeneity in terms of magnitudes across industries. 8 The coefficient on the RTA indicator variable varies between 0.064, and not statistically significant, for electronics to 0.602, and highly statistically significant, for the automotive industry which translate to an increase in exports of 7 and 83 percent respectively. The results show that trade in the automotive industry is less sensitive with respect to distance, conditional on the fact that trade in the automotive is concentrated among neighboring countries – i.e., the coefficient on the border indicator variable is the largest for the automotive industry. Finally, the results suggest 8 The estimates for non-targeted industries are reported in Table A3 in the appendix. 17 that countries with relatively high capital per worker tend to export more electronics and products in the automotive industry. Table 3: Industry Level PPML Gravity Estimates Targeted Sectors (1) (2) (3) (4) (5) (6) Food Chemicals Machinery Electronics Electrical Automotive VARIABLES 1-24 28-38 84 84-85 85 86-89 RTA 0.438*** 0.108 0.137* 0.064 0.075 0.602*** (0.090) (0.084) (0.082) (0.160) (0.130) (0.134) ln(distance) -0.623*** -0.864*** -0.702*** -0.856*** -0.859*** -0.619*** (0.057) (0.039) (0.043) (0.111) (0.073) (0.060) Border 0.598*** 0.184** 0.383*** 0.003 0.487*** 0.620*** (0.107) (0.080) (0.110) (0.323) (0.177) (0.164) Language 0.293*** 0.424*** 0.017 0.411 0.016 -0.123 (0.087) (0.081) (0.088) (0.327) (0.228) (0.151) Colony 0.099 -0.079 0.112 -0.692*** -0.222 -0.282 (0.124) (0.103) (0.122) (0.202) (0.170) (0.182) ln(GDP exp.) 0.681*** 0.921*** 1.019*** 1.048*** 1.056*** 0.907*** (0.025) (0.023) (0.028) (0.080) (0.054) (0.038) ln(GDP imp.) 0.718*** 0.765*** 0.798*** 0.825*** 0.756*** 0.895*** (0.030) (0.023) (0.029) (0.102) (0.064) (0.045) ln(tariff+1) -1.838*** 3.120** -0.899 0.426 -0.030 -2.852** (0.369) (1.395) (1.295) (2.945) (1.225) (1.407) ln(GDP pc exp.) -0.226*** 0.254*** -0.101 -0.693*** -0.477*** -0.306*** (0.070) (0.066) (0.072) (0.115) (0.091) (0.099) ln(GDP pc imp.) -0.110* -0.202*** -0.126** -0.286** -0.267*** 0.029 (0.059) (0.049) (0.053) (0.140) (0.092) (0.093) Mineral rich imp. 0.175* -0.301*** 0.296*** -0.920*** -0.251* 0.516*** (0.103) (0.086) (0.099) (0.228) (0.133) (0.135) Mineral rich exp. -0.877*** -0.302** -2.219*** -4.080*** -2.761*** -2.350*** (0.175) (0.122) (0.277) (0.278) (0.323) (0.270) ln(Remoteness exp.) 0.719*** -0.491*** -0.075 1.704*** 0.525** 0.924*** (0.125) (0.137) (0.174) (0.283) (0.243) (0.263) ln(Remoteness imp.) 0.102 0.977*** 0.886*** 0.642** 0.953*** 0.632*** (0.129) (0.121) (0.120) (0.295) (0.222) (0.191) ln(K/L exp.) 0.278*** -0.062 0.277*** 0.347** 0.229* 0.815*** (0.079) (0.077) (0.100) (0.172) (0.137) (0.137) ln(K/L imp.) 0.188*** 0.277*** 0.075 0.490* 0.428** 0.061 (0.071) (0.062) (0.067) (0.297) (0.175) (0.105) Observations 152,515 152,515 152,515 152,515 152,515 152,515 Robust standard errors, clustered at the country-pair level, are in parentheses. *** p<0.01, ** p<0.05, * p<0.1 18 Similarly to the aggregate analysis, we use the estimates from Table 3 and Table A3 to calculate industry-level export potentials. We construct the industry level indexes of missing exports and calculate the difference between predicted and observed export flows to obtain the dollar figures for missing exports. Figure 5 presents the industry level export potential results. Panel A plots the index of missing exports for each sector outlined in Table 1, with the strategic industries in red (i.e., industries targeted in the Turkish Export Master Plan). The results suggest that Turkey under exports in all the targeted industries. Electronics, chemical, and processed food products are among the industries with the highest index of missing exports. These sectors, particularly electronics and chemical sectors have a significant share in total exports of Turkey (20 percent of total exports). Another important industry in terms export potential is footwear, while Turkey, given its observable characteristics, appears to be already exporting intensively in the textile, vegetables, and non-processed food (“16-24 foodstuffs”) industries. Other industries other than automotive in which Turkey is predicted to be under exporting are special products, and the miscellaneous category. These industries include exports of works of art, and military weapons which might be difficult to expand. In addition to having a high index of missing exports, the electronics and chemical industries are also important in terms market size (Figure 5 Panel B). Interestingly, the dollar figures show that the processed food industry has 6 billion dollars of missing exports, while it only ranks seventh in terms of the index. The high value of the index of the footwear industry translates into relatively small values of missing exports, equal to 1.2 billion dollars. 19 Figure 5: Industry Level Export Potential Panel A: Index of Missing Exports Panel B: Missing Exports (USD) 20 To understand the potential of insertion in global value chains (GVCs), we classify products according to archetypal GVCs using the conversion tables based on Ferrantino and Schmidt (2018). Ferrantino and Schmidt (2018) extend the mapping between intermediate and final goods developed by Sturgeon and Memedović (2011) to cover GVC trade in electronics, vehicles, machinery, electrical equipment, footwear, textile, and apparel. The original mapping uses the U.N. Statistical Division’s Broad Economic Categories (BEC) classification and the opinion of industry experts to classify products as belonging to one of three GVCs: apparel and footwear, electronics, and motor vehicles. Figure 6 presents the export potential results for different GVCs. Similarly to the industry-level results (Figure 5), we find that Turkey has a high export potential in electronics, footwear, and processed food. Electronics is the most important industry in terms of missing exports. 9 Turkey has a particularly low forward participation in electronics. 10 As exports of final electronics are also below potential, this industry could potentially benefit from an improved access to foreign intermediate inputs with lower local content requirements linked to subsidies. Finally, the GVC results show that Turkey is particularly competitive in the automotive and textile industries both in terms of intermediate and final goods exports. 9 The most important destinations in terms of missing exports for both final and intermediate electronics are the United States, China, and Italy. 10 In the context of archetypal GVCs, forward participation in GVCs are proxied by exports of intermediate products. While the intensity of backward linkages can be inferred by how much a country imports of intermediate goods. 21 Figure 6: Global Value Chains Export Potential Panel A: Index of Missing Exports Panel B: Missing Exports (USD) 22 5. Trade Policy Potential This section studies the impact of deep agreements on Turkey’s exports. We use a standard gravity model to assess the effect RTAs had on trade and then use the estimates from this analysis to evaluate the future of Turkish trade relations under different scenarios. As in Mulabdic et al. (2017) and Mattoo et al. (2017), we build a measure of depth of agreements between PTA members using information from the content of PTAs (Hofmann et al., 2019; WTO, 2011). The database covers all preferential trade agreements notified to the World Trade Organization (WTO) and in force up to December 2015 (i.e., 279 trade agreements). For each agreement, the data provide information on the coverage and legal enforceability of 52 policy areas. 11 Currently Turkey has 20 free trade agreements (FTAs) in force and there are 5 additional agreements under ratification process. According to the Ministry of Trade, Turkey has been engaged in the negotiation of new agreements as well as in negotiations aimed extending the scope of its current agreements. 12 Table 4 provides information on the content of Turkey’s agreements that were notified to the WTO before December 2015. The Content of Deep Trade Agreements database cover 18 trade agreements of which Turkey is a member country: 17 FTAs and the customs union with the EU. The most extensive agreement in terms of coverage of policy areas is the agreements with Morocco which covers 13 policy areas, while the agreements with the Arab Republic of Egypt and Jordan cover only 8 areas. To put these numbers in perspective, the Peru-Chile FTA includes 11 legally enforceable provisions, the United States-Korea Free Trade Agreement (KORUS FTA) signed in 2007 includes 15 provisions, and the EU, which comprises eight agreements—i.e., the Treaty of Rome and successive EU enlargements—cover 43 legally enforceable provisions. 11 See Table A4 in the appendix for the description of the 52 policy areas. 12 Turkey has been engaged in negations with Japan, Ukraine, Peru, Indonesia, Colombia, Ecuador, Mexico, Thailand, Pakistan, the Democratic Republic of Congo, Djibouti, Cameroon, Chad, the Seychelles, the Gulf Cooperation Council, Libya, and MERCOSUR. Available at: https://www.trade.gov.tr/free-trade- agreements (Accessed August 22, 2020). 23 Table 4: Content of Turkish Trade Agreements Bosnia and Korea, Palestinian Morocco Albania EFTA EU Herzegovina Chile Montenegro Republic of Authority 2006 2008 1992 1996 2003 2011 2010 2013 2005 FTA FTA FTA CU FTA FTA FTA FTA FTA FTA Industrial FTA Agriculture Customs Export Taxes Sanitary and phytosanitary (SPS) Technical barriers to trade (TBT) State trading enterprises (STE) Antidumping (AD) Countervailing measures (CVM) State Aid Public Procurement Trade in services agreement (GATS) Trade-related intellectual property rights (TRIPs) Competition Policy Environmental Laws Intellectual Property Rights (IPR) Labour Market Regulation Movement of Capital Agriculture Approximation of Legislation North Israel Mauritius Serbia Syria Georgia Macedonia Tunisia Egypt Jordan 1997 2013 2010 2007 2008 2000 2005 2007 2011 FTA FTA FTA FTA FTA FTA FTA FTA FTA FTA Industrial FTA Agriculture Customs Export Taxes Sanitary and phytosanitary (SPS) Technical barriers to trade (TBT) State trading enterprises (STE) Antidumping (AD) Countervailing measures (CVM) State Aid Public Procurement Trade in services agreement (GATS) Trade-related intellectual property rights (TRIPs) Competition Policy Environmental Laws Intellectual Property Rights (IPR) Labour Market Regulation Movement of Capital Agriculture Approximation of Legislation 24 We use a gravity model to analyze the impact of deep trade agreements on Turkey’s trade. As it is standard in the trade literature, we include importer‐time and exporter‐time fixed effects to account for country‐time specific determinants of trade (e.g., market size) as well as multilateral resistance terms (Anderson and Wincoop, 2004, 2003). We also construct intra-national trade flows using GDP data from the WDI as in Bergstrand et al. (2015). To control for all the time‐invariant determinants of trade costs (e.g., distance) and address endogeneity concerns in the formation of PTAs (Baier and Bergstrand, 2007), we include country‐pair fixed effects. Table 5 presents the PPML gravity estimates of depth using importer‐time, exporter‐time, and country-pair fixed effects. To ease interpretation, we transform the ℎ variable to be equal to 1 for the average depth in our sample. The results in column (1) suggest that signing an agreement with average depth increases bilateral trade by 18 percent. The impact of an average trade agreement is similar in magnitude to the estimated impact of a PTA in Table 2. In columns (2)-(3) we include interaction terms to allow for heterogeneous effects of deep PTAs for Turkey. The results suggest Turkey’s trade is more sensitive to deep agreements than other countries (column 2), especially its exports (column 3). Table 5: Deep Trade Agreements and Trade (1) (2) (3) VARIABLES Trade Trade Trade Depth 0.168*** 0.167*** 0.167*** (0.020) (0.020) (0.020) Depth*TUR exp 0.536*** (0.203) Depth*TUR imp 0.470** (0.191) Depth*TUR exp/imp 0.508*** (0.139) Observations 174,146 174,146 174,146 exporter-year FE YES YES YES importer-year FE YES YES YES exporter-importer FE YES YES YES Robust standard errors, clustered at the country-pair level, are in parentheses. *** p<0.01, ** p<0.05, * p<0.1 25 Finally, we use the estimates from Table 5 column (3) to investigate the possible impact on Turkey’s exports of future trade arrangements with its current PTA partners. Figure 7 presents the results for different scenarios in which Turkey deepens the scope of its current trade agreements. In the first case, we assume that increases the depth of its shallow agreements to be at least as deep as its average agreement. In the second scenario, we assume that Turkey updates all its agreements to cover as many areas as in its deepest agreement (i.e., 13 provisions as in the Turkey-Morocco FTA). Finally, the third scenario assumes that all the trade agreements are updated to cover all the provisions that have been included in at least one Turkish agreement (i.e., 18 provisions). We find that total exports increase under all three scenarios and that this increase is greater the higher the depth of the future arrangements. The increase in total exports ranges from 1.8 percent to 9.2 percent in the case Turkey modernizes all its trade agreements. To achieve this increase, Turkey would need to address in its trade agreements issues which are already partially addressed at the WTO such as customs, export taxes, technical barriers to trade (TBT), and sanitary and phytosanitary standards as well as new areas such as competition policy, intellectual property rights (IPR) protection, and movement of capital. There are two important caveats to be kept in mind. First, these calculations only account for the partial effects and do not have general equilibrium or welfare implications. 13 Second, the empirical model assumes that the marginal effect of an additional provision is the same regardless of what type of provision is included. 14 13 We do not account for possible re‐direction of exports to and from other countries (i.e. trade diversion and deflection), price adjustments or changes in wages. 14 It is possible that provisions on standards, investment or competition are likely to have a larger impact on trade than provisions that do not pertain to trade such as visa and asylum. 26 Figure 7: Trade Policy Potential 6. Conclusion In this paper, we present a framework to assess countries’ export potentials. We use a standard gravity model, which incorporates recent developments of the empirical trade literature, to construct export potential indicators at the aggregate, country-pair, and industry levels. To illustrate this framework, we analyze Turkey’s export potential. The results suggest Turkey had on average 12 billion dollars of missing exports to the United States over the 2010-2017 period. Other important destinations in terms of missing exports are China and Japan with 10 billion dollars and 7 billion dollars, respectively. Industry level results suggest that Turkey has a high export potential in electronics and chemical industries which are important in terms of both the index of missing exports and market size. The paper also investigates the potential impact of deep agreements on trade. We find that Turkey could achieve higher levels of exports through deeper trade arrangements with its current preferential trade agreement (PTA) partners. A modernization of Turkey’s existing trade 27 agreements has the potential of increasing its total exports between 1.8 to 9.2 percent. This would require Turkey to address in its trade agreements issues related to customs, export taxes, technical barriers to trade (TBT), sanitary and phytosanitary standards as well as more complex issues related to competition policy, intellectual property rights (IPR) protection, and movement of capital. 28 References Anderson, J.E., 1979. A theoretical foundation for the gravity equation. The American economic review 69, 106–116. Anderson, J.E., Wincoop, E. van, 2004. Trade Costs. Journal of Economic Literature 42, 691–751. Anderson, J.E., Wincoop, E. van, 2003. Gravity with Gravitas: A Solution to the Border Puzzle. American Economic Review 93, 170–192. Ata, S., 2012. Türkiye’nin İhracat Potansiyeli: Çekim modeli çerçevesinde bir inceleme. Presented at the International Conference on Eurasian Economies, 11–13. Baier, S.L., Bergstrand, J.H., 2007. Do free trade agreements actually increase members’ international trade? Journal of International Economics 71, 72–95. Baldwin, R., Taglioni, D., 2006. Gravity for Dummies and Dummies for Gravity Equations (NBER Working Paper No. 12516). National Bureau of Economic Research, Inc. Bergstrand, J.H., Larch, M., Yotov, Y.V., 2015. Economic integration agreements, border effects, and distance elasticities in the gravity equation. European Economic Review 78, 307–327. https://doi.org/10.1016/j.euroecorev.2015.06.003 Chor, D., 2010. Unpacking sources of comparative advantage: A quantitative approach. Journal of International Economics 82, 152–167. Costinot, A., Rodríguez-Clare, A., 2014. Chapter 4 - Trade Theory with Numbers: Quantifying the Consequences of Globalization, in: Gopinath, G., Helpman, E., Rogoff, K. (Eds.), Handbook of International Economics. Elsevier, pp. 197–261. https://doi.org/10.1016/B978-0-444-54314-1.00004-5 Egger, P., Larch, M., 2008. Interdependent preferential trade agreement memberships: An empirical analysis. Journal of International Economics 76, 384–399. https://doi.org/10.1016/j.jinteco.2008.08.003 Fally, T., 2012. Structural Gravity and Fixed Effect. Boulder, Colorado, University of Colorado Boulder, May. Ferrantino, M.J., Schmidt, G., 2018. Using gross trade data to map archetypal GVCs. The World Bank. Head, K., Mayer, T., 2014. Chapter 3 - Gravity Equations: Workhorse, Toolkit, and Cookbook, in: Gopinath, G., Helpman, E., Rogoff, K. (Eds.), Handbook of International Economics. Elsevier, 131–195. https://doi.org/10.1016/B978-0-444-54314-1.00003-3 Hofmann, C., Osnago, A., Ruta, M., 2019. The Content of Preferential Trade Agreements. World Trade Review 18, 365–398. https://doi.org/10.1017/S1474745618000071 Karagoz, K., Saray, M., 2010. Trade potential of Turkey with Asia-Pacific countries: Evidence from panel gravity model. Levchenko, A.A., Zhang, J., 2014. Ricardian productivity differences and the gains from trade. European Economic Review 65, 45–65. https://doi.org/10.1016/j.euroecorev.2013.10.006 Mattoo, A., Mulabdic, A., Ruta, M., 2017. Trade creation and trade diversion in deep agreements (No. 8206), Policy Research Working Paper Series. The World Bank. Mulabdic, A., Osnago, A., Ruta, M., 2017. Deep Integration and UK–EU Trade Relations, in: Campos, N.F., Coricelli, F. (Eds.), The Economics of UK-EU Relations: From the Treaty of Rome to the Vote for Brexit. Springer International Publishing, Cham, 253–282. https://doi.org/10.1007/978-3-319-55495-2_10 Romalis, J., 2004. Factor Proportions and the Structure of Commodity Trade. The American Economic Review 94, 67–97. 29 Silva, J.M.C.S., Tenreyro, S., 2006. The Log of Gravity. The Review of Economics and Statistics 88, 641–658. Sturgeon, T.J., Memedović, O., 2011. Mapping global value chains: Intermediate goods trade and structural change in the world economy. United Nations Industrial Development Organization. Tinbergen, J., 1962. Shaping the world economy; suggestions for an international economic policy. Wei, S.-J., 1996. Intra-National versus International Trade: How Stubborn are Nations in Global Integration? National Bureau of Economic Research Working Paper Series No. 5531. https://doi.org/10.3386/w5531 WTO, 2011. World Trade Report 2011- The WTO and Preferential Trade Agreements: From Co- existence to Coherence (Working Paper No. id:4335). eSocialSciences. 30 Appendix A: Additional Tables Table A1: Summary Statistics VARIABLES N Mean S.D. Min. Max. Exports (USD millions) 152,515 1,248.189 8,753.074 0.000 525,764.710 Border 152,515 0.031 0.174 0.000 1.000 Language 152,515 0.119 0.323 0.000 1.000 Mineral rich exp. 152,515 0.159 0.366 0.000 1.000 Mineral rich imp. 152,515 0.142 0.349 0.000 1.000 ln(distance) 152,515 8.696 0.763 5.081 9.886 ln(tariff+1) 152,515 0.059 0.057 0.000 0.455 ln(K/L exp.) 152,515 10.947 1.446 7.242 13.445 ln(K/L imp.) 152,515 11.062 1.433 7.283 13.445 RTA 152,515 0.283 0.451 0.000 1.000 ln(Remoteness exp.) 152,515 21.381 0.348 20.713 22.291 ln(Remoteness imp.) 152,515 21.388 0.349 20.713 22.291 ln(GDP exp.) 152,515 25.185 2.011 20.481 30.601 ln(GDP imp.) 152,515 25.373 1.975 20.481 30.601 ln(GDP pc exp.) 152,515 8.110 1.653 4.682 11.385 ln(GDP pc imp.) 152,515 8.262 1.646 4.682 11.385 Depth SLE 152,515 0.146 0.457 0.000 2.433 31 Table A2: Top "Over Exporters" in Markets with High Missing Exports for Turkey (USD millions) Top "Over Exporters" (USD millions) Turkey's missing Importer exports 1st 2nd 3rd 4th 5th 1 United States 12,220 China -265,943 Mexico -99,059 Germany -46,505 Canada -37,014 Japan -35,643 2 China 9,615 Korea, Rep. -68,766 Australia -66,046 Germany -56,274 Malaysia -36,202 Brazil -31,413 3 Japan 7,455 Australia -30,129 Saudi Arabia -29,034 Malaysia -18,150 Thailand -11,445 Vietnam -8,121 4 Italy 4,659 Germany -18,072 Netherlands -12,548 Belgium -10,089 Azerbaijan -6,425 Romania -2,650 Venezuela, 5 India 3,535 Saudi Arabia -21,323 Switzerland -21,070 Iraq -11,530 Nigeria -8,270 RB -8,080 Russian 6 Greece 3,466 Federation -2,692 Iraq -2,369 Netherlands -1,497 Kazakhstan -1,391 Korea, Rep. -1,048 7 France 3,194 China -15,186 Belgium -9,988 Netherlands -3,566 Kazakhstan -3,338 Tunisia -3,317 8 Korea, Rep. 1,902 Saudi Arabia -28,430 Australia -16,399 United States -11,442 Germany -11,028 Malaysia -7,463 9 Iraq 1,828 China -5,353 Korea, Rep. -2,110 Thailand -94 Ukraine -61 Argentina -58 10 Canada 1,818 China -21,070 Mexico -11,632 Germany -2,777 Korea, Rep. -1,916 Vietnam -1,647 11 Brazil 1,765 China -12,070 Germany -4,679 Korea, Rep. -4,530 Nigeria -3,058 Bolivia -1,413 12 Sweden 1,529 Germany -10,884 Denmark -8,830 Netherlands -7,391 Belgium -3,846 Finland -3,726 13 Austria 1,483 Germany -22,334 Czech Republic -2,458 Slovak Republic -1,794 Hungary -1,527 Switzerland -1,496 14 Indonesia 1,454 Thailand -4,270 Korea, Rep. -2,923 Saudi Arabia -2,604 Malaysia -1,186 Azerbaijan -1,034 15 Mexico 1,328 China -41,689 Korea, Rep. -10,267 Malaysia -5,538 Germany -4,976 Thailand -3,149 16 Australia 1,309 China -11,214 Thailand -7,518 Malaysia -6,351 Germany -5,552 Korea, Rep. -4,511 Russian 17 Netherlands 1,199 China -26,014 Belgium -12,600 United States -12,135 Federation -12,039 Malaysia -6,994 Russian 18 Finland 1,067 Federation -5,006 Sweden -3,886 Germany -2,739 Netherlands -1,993 Denmark -954 19 Denmark 946 Sweden -7,804 Netherlands -2,453 Belgium -412 Finland -274 Bangladesh -250 Russian 20 Hungary 938 Germany -17,681 Federation -3,889 Slovak Republic -3,827 Austria -3,433 Poland -2,998 Table A3: Industry Level PPML Gravity Estimates non-Targeted Sectors (1) (2) (3) (4) (5) (6) (7) (8) (9) (10) (11) (12) (13) (14) 39-40 41-43 50-63 68-71 01-05 06-15 16-24 25-27 Plastic / Hides 44-49 Textiles 64-67 Stone / 72-83 84-85 90-97 98-99 VARIABLES Animal Vegetable Foodstuffs Minerals Rubber Skins Wood Clothing Footwear Glass Metals Mach/Elec Miscellaneous Special RTA 0.021 0.366*** 0.063 0.243* 0.292*** -0.499*** 0.288** -0.009 -0.179 -0.033 0.360*** 0.868*** -0.049 0.352 (0.132) (0.115) (0.145) (0.131) (0.099) (0.136) (0.123) (0.089) (0.180) (0.260) (0.102) (0.145) (0.149) (0.227) ln(distance) -0.949*** -0.494*** -0.725*** -0.777*** -0.928*** -0.652*** -0.747*** -0.828*** -0.749*** -0.375*** -0.898*** -0.752*** -0.750*** -0.654*** (0.061) (0.081) (0.100) (0.087) (0.047) (0.081) (0.052) (0.055) (0.105) (0.129) (0.048) (0.068) (0.091) (0.138) Border 0.707*** 0.575*** 0.167 0.588*** 0.403*** 0.699** 0.795*** 0.275 0.552 0.124 0.490*** 0.585*** 0.336 0.802*** (0.178) (0.176) (0.255) (0.191) (0.111) (0.312) (0.145) (0.196) (0.348) (0.263) (0.086) (0.194) (0.208) (0.247) Language 0.304* 0.072 -0.126 0.441** 0.148 -0.063 0.294** 0.207 -0.456 1.243*** 0.076 -0.361 0.146 0.250 (0.178) (0.132) (0.154) (0.185) (0.091) (0.308) (0.147) (0.190) (0.428) (0.199) (0.088) (0.234) (0.182) (0.196) Colony -0.048 -0.125 -0.172 -0.009 -0.200 -0.655*** -0.029 -0.187 -0.499 -0.493* 0.140 0.003 -0.261* 0.945*** (0.163) (0.183) (0.225) (0.202) (0.149) (0.238) (0.166) (0.202) (0.333) (0.271) (0.133) (0.270) (0.153) (0.277) ln(GDP exp.) 0.682*** 0.661*** 0.552*** 0.677*** 0.923*** 1.126*** 0.680*** 0.974*** 1.170*** 0.690*** 0.871*** 0.847*** 1.054*** 0.650*** (0.032) (0.041) (0.050) (0.047) (0.035) (0.062) (0.045) (0.049) (0.115) (0.062) (0.032) (0.044) (0.060) (0.068) ln(GDP imp.) 0.820*** 0.857*** 0.627*** 0.952*** 0.820*** 0.647*** 0.810*** 0.762*** 0.810*** 0.698*** 0.806*** 0.882*** 0.855*** 1.131*** (0.044) (0.066) (0.051) (0.052) (0.036) (0.097) (0.048) (0.060) (0.131) (0.078) (0.029) (0.050) (0.073) (0.052) ln(tariff+1) -5.808*** -0.987 0.229 4.369 -2.835** -3.049 -4.652*** -2.592** -3.910* -1.927 -0.813 -3.276** 2.535 (1.163) (0.759) (0.479) (9.536) (1.297) (2.491) (1.002) (1.120) (2.026) (2.222) (1.236) (1.547) (1.834) ln(GDPpc exp.) -0.354*** -0.032 0.229 0.093 -0.490*** -0.649*** 0.046 -0.910*** -0.841*** 0.086 -0.319*** -0.559*** 0.002 0.106 (0.101) (0.090) (0.149) (0.106) (0.062) (0.104) (0.090) (0.083) (0.145) (0.225) (0.067) (0.129) (0.117) (0.227) ln(GDPpc imp.) -0.591*** -0.239*** -0.167 -0.443*** -0.353*** -0.443*** -0.193*** 0.001 0.176 -0.106 -0.323*** -0.277** 0.098 0.146 (0.096) (0.084) (0.118) (0.113) (0.063) (0.113) (0.070) (0.098) (0.142) (0.218) (0.054) (0.110) (0.103) (0.240) Mineral rich imp. -0.535*** 0.013 -0.025 -1.368*** -0.212** -1.354*** -0.042 -0.621*** 0.025 -1.239*** -0.118 -0.185 -0.535*** -0.061 (0.201) (0.134) (0.175) (0.254) (0.094) (0.250) (0.112) (0.171) (0.282) (0.206) (0.112) (0.176) (0.154) (0.521) Mineral rich exp. -0.194 -0.301* -1.041*** 2.515*** -0.650*** -2.097*** -0.143 -2.905*** -3.761*** -1.096*** 0.361** -2.962*** -2.919*** 0.512 (0.230) (0.180) (0.223) (0.131) (0.184) (0.261) (0.213) (0.155) (0.346) (0.251) (0.165) (0.436) (0.231) (0.461) ln(Remoteness exp.) 1.549*** 1.284*** -0.437 0.370 0.708*** -0.438 0.960*** 0.915*** 0.659* -0.445 0.753*** 1.142*** 0.430* 0.900** (0.207) (0.177) (0.279) (0.281) (0.146) (0.285) (0.189) (0.170) (0.345) (0.294) (0.136) (0.287) (0.224) (0.379) ln(Remoteness imp.) 0.248 -0.574*** 0.435* 0.837*** 0.614*** 1.575*** 0.501*** 0.232 0.480 0.334 0.468*** 0.920*** 0.790*** -0.422 (0.212) (0.178) (0.259) (0.263) (0.122) (0.371) (0.145) (0.227) (0.492) (0.298) (0.124) (0.206) (0.231) (0.299) ln(K/L exp.) 0.074 -0.118 -0.600*** -0.060 0.644*** 0.103 0.065 0.213* -0.130 -0.040 0.284*** 0.708*** -0.350* 0.166 (0.111) (0.098) (0.167) (0.100) (0.096) (0.179) (0.101) (0.109) (0.240) (0.209) (0.079) (0.165) (0.181) (0.231) ln(K/L imp.) 1.318*** 0.191* 0.561*** 0.497*** 0.231*** 0.951*** 0.120 0.204 0.553** 0.281 0.283*** 0.517*** 0.268 -0.336 (0.146) (0.102) (0.164) (0.145) (0.082) (0.268) (0.100) (0.181) (0.277) (0.229) (0.075) (0.136) (0.189) (0.324) Observations 152,515 152,515 152,515 152,515 152,515 152,515 152,515 152,515 152,515 152,515 152,515 152,515 152,515 152,515 Robust standard errors, clustered at the country-pair level, are in parentheses. *** p<0.01, ** p<0.05, * p<0.1 33 Table A4: Description of the 52 provisions in the Content of Deep Trade Agreements Database WTO-plus areas FTA Industrial Tariff liberalization on industrial goods; elimination of non-tariff measures FTA Agriculture Tariff liberalization on agriculture goods; elimination of non-tariff measures Customs Provision of information; publication on the Internet of new laws and regulations; training Export Taxes Elimination of export taxes SPS Affirmation of rights and obligations under the WTO Agreement on SPS; harmonization of SPS measures TBT Affirmation of rights and obligations under WTO Agreement on TBT; provision of information; harmonization of regulations; mutual recognition agreements STE Establishment or maintenance of an independent competition authority; nondiscrimination regarding production and marketing condition; provision of information; affirmation of Art XVII GATT provision AD Retention of Antidumping rights and obligations under the WTO Agreement (Art. VI GATT). CVM Retention of Countervailing measures rights and obligations under the WTO Agreement (Art VI GATT) State Aid Assessment of anticompetitive behaviour; annual reporting on the value and distribution of state aid given; provision of information Public Progressive liberalisation; national treatment and/or non-discrimination principle; publication of laws and Procurement regulations on the Internet; specification of public procurement regime TRIMs Provisions concerning requirements for local content and export performance of FDI GATS Liberalisation of trade in services TRIPs Harmonisation of standards; enforcement; national treatment, most-favoured nation treatment WTO-X areas Anti-Corruption Regulations concerning criminal offence measures in matters affecting international trade and investment Competition Maintenance of measures to proscribe anticompetitive business conduct; harmonisation of competition laws; Policy establishment or maintenance of an independent competition authority Environmental Development of environmental standards; enforcement of national environmental laws; establishment of Laws sanctions for violation of environmental laws; publications of laws and regulation IPR Accession to international treaties not referenced in the TRIPs Agreement Investment Information exchange; Development of legal frameworks; Harmonisation and simplification of procedures; National treatment; establishment of mechanism for the settlement of disputes Labour Market Regulation of the national labour market; affirmation of International Labour Organization (ILO) commitments; Regulation enforcement Movement of Liberalisation of capital movement; prohibition of new restrictions Capital Consumer Harmonisation of consumer protection laws; exchange of information and experts; training Protection Data Protection Exchange of information and experts; joint projects Agriculture Technical assistance to conduct modernisation projects; exchange of information Approximation Application of EC legislation in national legislation of Legislation Audio Visual Promotion of the industry; encouragement of co-production Civil Protection Implementation of harmonised rules Innovation Participation in framework programmes; promotion of technology transfers Policies Cultural Promotion of joint initiatives and local culture Cooperation Economic Exchange of ideas and opinions; joint studies Policy Dialogue Education and Measures to improve the general level of education Training Energy Exchange of information; technology transfer; joint studies Financial Set of rules guiding the granting and administration of financial assistance Assistance Health Monitoring of diseases; development of health information systems; exchange of information Human Rights Respect for human rights Illegal Conclusion of re-admission agreements; prevention and control of illegal immigration Immigration Illicit Drugs Treatment and rehabilitation of drug addicts; joint projects on prevention of consumption; reduction of drug supply; information exchange Industrial Assistance in conducting modernisation projects; facilitation and access to credit to finance Cooperation Information Exchange of information; dissemination of new technologies; training Society Mining Exchange of information and experience; development of joint initiatives Money Harmonisation of standards; technical and administrative assistance Laundering Nuclear Safety Development of laws and regulations; supervision of the transportation of radioactive materials Political Convergence of the parties’ positions on international issues Dialogue Public Technical assistance; exchange of information; joint projects; Training Administration Regional Promotion of regional cooperation; technical assistance programmes Cooperation Research and Joint research projects; exchange of researchers; development of public-private partnership Technology SME Technical assistance; facilitation of the access to finance Social Matters Coordination of social security systems; non-discrimination regarding working conditions Statistics Harmonisation and/or development of statistical methods; training Taxation Assistance in conducting fiscal system reforms Terrorism Exchange of information and experience; joint research and studies Visa and Exchange of information; drafting legislation; training Asylum Source: WTO (2011). 35