WPS7953 Policy Research Working Paper 7953 Firms’ Export Decisions Demand Trumps Financial Shocks Francesca de Nicola Shawn W. Tan Trade and Competitiveness Global Practice Group January 2017 Policy Research Working Paper 7953 Abstract This paper studies the relationship between access to credit, (extensive margin) and on the firm’s share of foreign sales demand shocks, and export market adjustments using firm- (intensive margin). Foreign shocks to demand only affect the level panel survey data for 24 economies in the Eastern firm’s share of foreign sales. Conversely, the role of financial Europe and Central Asian region. The study finds that constraints on either the extensive or the intensive margin is domestic shocks to demand have a significant influence more nuanced. The results are robust to various specifications on the firm’s decision to participate in international markets of financial constraints and different estimation methods. This paper is a product of the Trade and Competitiveness 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 fdenicola@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 Firms’ Export Decisions: Demand Trumps Financial Shocks Francesca de Nicola Shawn W. Tan The World Bank The World Bank JEL Classification: D92, F14, F36. Keywords: Heterogeneous Firms, Export Margins, Credit Constraints. 1 Introduction What drives international trade? What hampers its growth? Academics and policy makers devote a lot of attention to these questions. Multiple factors have been under the lens but typically different causes have been examined separately. In this paper, we investigate the role of financial and real sector elements on trade dynamics using firm-level data from 24 countries in the Eastern Europe and Central Asia (ECA) region. We document the relative importance of adjustment at the extensive (number of exporters) and intensive (quantity of exports) margins during the period from 2002 to 2013. We then analyze the relative importance of credit as opposed to demand shocks in the firm’s decision to export and the amount of exports in a formal econometric setting. We find that demand shocks tend to significantly affect the decision to export as well as the volumes exported. Conversely, credit constraints appear to play a more limited role. These results tend to be robust to a wide range of specifications. However, there are differences in the relative importance of these elements when we consider the type of export margin (extensive and intensive) as well as the time period (before or after the crisis). Domestic demand shocks matter across all specifications as they influence firms’ decision to export and the amount of exports. Conversely, foreign demand shocks significantly influence the intensive but not the extensive margin, and their role becomes more pronounced when we distinguish between pre- and post-crisis periods. Finally, credit constraints only play a role in the intensive margins and mainly when we do not take into account the role of the financial crisis. The firm’s subjective perception that financial constraints are a hurdle for business appears to matter more than the objective inability to secure a loan after applying. The evidence is robust to controlling for possible endogeneity. The paper contributes to two branches in the literature examining the roles of finance and demand shocks in international trade. The literature has mostly examined both roles separately and finds that they are important to trade. Our paper will examine both roles together and finds that demand shocks may be more important to international trade. The paper is closely related to Nguyen and Qian (2014), that examines the roles of demand shocks and credit constraints on firms in the ECA countries. Focusing only on the financial crisis period (specifically 2009), the authors show that demand shock is the main factor for a worsening in firm performance (sales, capacity, and employment) is the decline in demand for the firm’s goods. They highlight the role demand shocks play in the financial crisis over credit constraints. Our paper extends this discussion by examining the roles demand shocks and credit constraints have on firm export behavior and how the relative importance of these roles changes over a longer time period that includes the financial crisis. Various studies document the importance of finance for international trade transactions: fi- nancial institutions not only provide trade finance through payment insurance and guarantees to exporters, they are also key suppliers of services such as the evaluation of counterparty default risk. Trade finance is a powerful catalyzer of international trade, Auboin (2009) estimates that about 90% of international trade transactions rely on some form of trade finance. The global financial crisis has heightened the interest in the impact of credit crunches on trade. 2 The role of finance may have been magnified since borrowing may become harder during times of stress. Recent evidence suggests that the trade-finance nexus seems to be one explanation for the trade contractions during both the Japanese slump and the recent U.S. financial crisis. Amiti and Weinstein (2011) show that during the Japanese crisis of the late 1990s and early 2000s, the lack of financing accounted for nearly one-third of the plunge in exports. Similarly, using monthly U.S. import data, Chor and Manova (2012) find that countries with tighter credit availability during the crisis exported less to the United States. Moreover, exports to the United States contracted more in sectors that are more heavily dependent on external financing. International evidence is in line with the findings from the U.S. Paravisini, Rappoport, Schnabl, and Wolfenzon (2015) match Peruvian customs and firm-level bank credit data to accurately measure the exogenous credit shocks, and estimate that a 10% contraction in credit supply from poorly performing banks translates into a 2.3% fall in the export volumes. There is limited understanding of the channels through which credit constraints may affect or be affected by export status. Several studies based on Melitz (2003) show that more productive firms export as they can afford the fixed costs of exporting. But less attention has been dedicated to the nature of the fixed costs. Indeed many studies assume perfect capital markets and rule out the existence of credit to finance either fixed or variable costs. Two noteworthy exceptions are Chaney (2016) and Manova (2013) that introduce firm-level liquidity into a trade model of heterogeneous firms and link country-specific financial development to export status. The logic underlying these models is that in the presence of binding credit constraints, a high-productivity firm may still be unable to finance export costs and operate internationally, but there may be low-productivity firms that are financially unconstrained and able to export. The relationship between productivity and access to credit may alter the sharp prediction of the standard heterogeneous firms model so that credit constraint may lower the (weighted) average productivity of the industry. A fundamental difference in these models is whether credit is used to finance the fixed or variable costs of exporting. Chaney (2016) conjectures that firms use the cash flows from domestic sales to finance the fixed costs of entering foreign markets. More productive firms achieving higher profits are more likely to reinvest earnings. Once firms have enough liquidity to pay the fixed cost of exporting, they are able to finance the variable costs of expanding the scale of production with internal funds or even by external borrowing. In contrast, Manova (2013) proposes a different framework to analyze the role of credit constraints, assuming that credit affects exporting status because also the variable production costs need to be financed with external capital. In her model of heterogeneous firms, industries differ in their dependence on external finance and level of asset tangibility. By exploiting the variation in the degree of financial development across countries, she uncovers evidence that credit constraints are a significant determinant of international trade patterns. She shows that firms in financially-vulnerable sectors in financially-developed countries gain the most from the liberalization process. Once these firms become exporters, they start trading greater volumes and at the same time export a wider variety of products. Alternative interpretations of the relationship among productivity, liquidity constraints, and 3 export decisions have been suggested. Causality may run in the opposite direction, where firms overcome financial constraints by diversifying their sales markets and choose to become exporters to overcome liquidity constraints.1 Trade policies may also play a role.2 In parallel, a limited but growing literature on the role of demand shocks on trade flows is developing. When faced with demand shocks, firms may seek to spread their sales between do- mestic and foreign markets to smooth their cash flows. Indeed, Campa and Shaver (2002) suggest that Spanish manufacturing firms may decide to export to foreign markets to smooth their cash flow and thus exploit the imperfect correlation between the destination country and the Spanish business cycles. Behrens, Corcos, and Mion (2013) use micro data to identify the fall in the de- mand for tradable goods as the main driver for trade decline in Belgium. More attenuated effects are documented by Eaton, Kortum, Neiman, and Romalis (2011) whose counterfactual analysis shows that non-durables demand shocks lead to 18% fall in world trade. Claessens, Tong, and Wei (2012) take a different approach and study the impact of the crisis on cross-country balance sheets of non-financial firms. They find that the crisis hampered more firms with greater sensitivity to the business cycle and trade. Conversely, financing openness (measured as total international assets plus liabilities over GDP) played a more limited role. The recent crisis has highlighted the significant role of real sector shocks on trade dynamics, as suggested by Eaton, Kortum, Neiman, and Romalis (2011). The paper proceeds in three sections. We discuss the multiple data sources used in Section 2. We then outline the empirical strategy and present the results in Subsection 3.1 and 3.2. We conclude in Section 4. 2 Data We base our analysis on information from three main data sources. First, the Business Environment and Enterprise Performance Surveys (BEEPSs) provide information about firms’ characteristics. Second, the World Development Indicators (WDI) and Global Development Finance (GDF) contain the macroeconomic variables to correct for inflation and exchange rate fluctuations, and capture the macroeconomic conditions faced by firms. Third, the National Accounts Estimates of Main Aggregates and Trade Policy Information System (TPIS) data allow us to estimate demand shocks. The World Bank jointly with the European Bank for Reconstruction and Development admin- istered the BEEP Surveys for four years (2002, 2005, 2009 and 2013) in 27 Eastern European and Central Asia countries.3 The firms interviewed were selected to form a representative sample of the 1 In a study of Spanish manufacturing firms, Campa and Shaver (2002) suggest that firms may decide to export to foreign markets to smooth their cash flow and thus exploit the imperfect correlation between the destination country and the Spanish business cycles. 2 Pavcnik (2002) concludes that Chilean manufacturing plants in import-competing sectors underwent large productiv- ity improvements during the massive trade liberalization of the late 1970s. Similarly, examining a period of significant changes in Colombian trade policy across industries between 1977 and 1991, Fernandes (2007) finds that tariff liber- alization boosts plant productivity and the effect appears to be particularly strong if the firm is larger and operating in a less competitive industry. 3 Specifically the surveys are fielded in Albania, Armenia, Azerbaijan, Belarus, Bosnia and Herzegovina, Bulgaria, Croa- 4 population of non-agricultural firms within each country.4 The BEEPSs contain both qualitative and quantitative information about the firms and the business environment where they operate. We exploit the sections of the surveys that focus on the business perceptions about the most relevant obstacles to investment, the relative importance of different constraints to boosting employment and productivity, and the effects of the country-specific environment on firms’ international com- petitiveness. To ensure comparability across countries and over time, we convert in US dollar at constant prices the values of the variables of interest from the BEEPS expressed in local currencies. The WDI and GDF provide exchange rate time series and estimates of the GDP deflator used to convert the nominal time series in real terms. The use of the GDP deflator implies that the inflation rate has been constant across sectors in a given country and during a specific year. The use of sector- specific price indexes for each ISIC sector would be preferable but these series are not available for the countries we study. We account for country-specific factors such as general economic conditions with GDP (constant 2000 US$), likelihood to obtain credit with the number of commercial bank branches (per 100,000 adults), domestic credit provided by banking sector (% of GDP), and the extra-cost of borrowing (lending rate minus deposit rate, %). The United Nations Statistics Division (UNSD) manages the National Accounts Estimates of Main Aggregates that report the gross value added at constant prices (in USD) by country and sector for all the years and countries in the BEEPS. The TPIS data set reports bilateral export volumes and we can then estimate the demand shocks based on the strategy proposed by Coulibaly, Sapriza, and Zlate (2012). The authors construct an index of demand as a function of the firms’ exports-to-sales ratio, the sector-specific exposure to various foreign destinations, and real GDP growth across destinations as a proxy for the change in demand. We provide details regarding the construction of these indices in Section 3.1. 3 Empirical Analysis Firms in our sample constitute a representative sample of non-agricultural firms in 27 ECA coun- tries. On average, they have been operative for 15 years and are middle-sized with less than 100 employees (Table 1). They tend to be liquidity constrained based on both the subjective and ob- jective measures, and operate in a relatively volatile macroeconomic environment as indicated by the fact that domestic GDP fluctuates from −4.5% up to +25%. Only 26% of the firms in the sample export to foreign markets, and they tend to make relatively small shares of their total sales abroad (11%). The panel nature of the data also provides important insights from the cross-country and time variations. For example, in Kazakhstan firms’ exports amount to only 3% of total sales, while in tia, Czech Republic, Estonia, Former Yugoslavian Republic of Macedonia (FYRM), Georgia, Hungary, Kazakhstan, Kyrgyz Republic, Latvia, Moldova, Montenegro, Poland, Romania, Russian Federation, Serbia, Slovak Republic, Turkey and Ukraine. 4 In the 2002 and 2005, the sample is self-weighted, while in 2009 and 2013 firms are sampled randomly with replacement according to sector of activity, firm size, and geographical location strata. 5 Slovenia they account for 22%. In order to explore these differences, in Figure 1 we plot the changes in the extensive and intensive margins of exports within each country during the 2002-2013 period. On the y-axis we plot the changes from 2002 and 2009, while on the x-axis we plot the changes between 2009 and 2013. Observing scatter dots concentrated in the first (third) quadrants is a sign of a boom (bust) in exports that continued during the entire sample period. Conversely, scatter dots in the second (fourth) quarter indicate that exports fell (rose) in 2013-2009 after an initial increase (decline) in 2002-2009. We find that many ECA countries are in the fourth quadrant: there was a decline in the extensive (share of exporters) and intensive (share of exports) components of the changes in trade in 2002-2009, and an increase in 2009-2013. During the period 2002-2009, the share of exporting firms and exports declinedby almost 15% and 30% respectively across most ECA countries. Macedonia, FYR experienced the largest fall in the share of exports and in the number of exporters. Some of these countries increased the intensive and extensive margins in 2009-2013. Conversely, the share of exporters and exports in Turkey increased over both periods, and substantially during the second period. 3.1 Empirical methodology We test the relative importance of demand shocks and credit constraints of firm i in country j at time t on the ability to export (extensive margin) in Equation 1 and the relative extent of exports (intensive margin) in Equation 2 by estimating the following equations: P rob(Exportijt > 0) = αCreditConstraintijt + βDomesticDemandShockjt +γF oreignDemandShock−jt + δXijt + ijt (1) Exportijt = αCreditConstraintit + βDomesticDemandShockjt +γF oreignDemandShock−jt + δXijt + υijt (2) The dependent variables (P rob(Exportijt > 0) and Exportijt ) are constructed from the BEEPS survey question: “What percentage of your firm’s sales are sold domestically, exported directly, or indirectly through a distributor?” Firms that report a positive value for either direct or indirect exports are identified as exporters, with the dummy variable P rob(Exportijt > 0) taking a value of one. The sum of direct and indirect exports yields the value of total export as a percentage of total sales used to estimate Equation 2. For robustness checks, we use alternative definitions of the dependent variables. First we consider excluding exports sold indirectly through a distributor. Second, we measure the dependent variable in Equation 2 as the total value of exports, allowing for a distinction between small and large firms.5 There is an objective and a subjective measure of the variable CreditConstraintijt . The sub- 5 Total value of exports is not directly recorded in the BEEPs survey. It is calculated by multiplying the share of exports in total sales and the value of total sales. While export value is a preferable dependent variable to the share of exports in total sales, there are inconsistencies in the way total sales are recorded across the countries and time periods. We cannot dismiss the potentially considerable measurement error and related bias. Hence, we use this measure only for robustness checks. 6 jective measure is based on the self-reported difficulty of financing for the operation and growth of the business. In 2002 and 2005 respondents are separately asked to assess the degree of difficulty in access to (“e.g., collateral required or financing not available from banks”) and cost of financing (“e.g., interest rates and charges”), ranging from 1 (“No obstacle”) to 4 (“Major obstacle”). In 2009 and 2013, these two questions are merged into one and respondents are asked about the concerns due to access to financing. Also the scale of the importance of the constraint is different and varies from 1 (“No obstacle”) to 5 (“Very severe obstacle”). We harmonize the scales across years and construct the variable as follows:     1, if the access to and the cost of financing are not an obstacle;   2, if either the access to or the cost of financing are a minor obstacle; CreditConstraintijt =    3, if either the access to or the cost of financing are a moderate obstacle;   4, if both access and cost of financing are a major obstacle. The subjective measure of CreditConstraint may raise questions about consistency as it is self-reported and suffers from individual bias. We thus compare the subjective measure with an objective firm-level indicator. A firm is objectively deemed credit unconstrained if one the following conditions applies: (1) it did not apply for a line of credit or loan because it “did not need a loan, the establishment has sufficient capital”, or (2) it applied for any new loans or new lines of credit and was not rejected.6 The subjective and objective measures of being capital constrained are significantly correlated. This evidence is consistent with Hallward-Driemeier and Aterido (2009) and Gorodnichenko and Schnitzer (2010) that relate subjective financing constraints to the amount of credit issued to the private sector, expressed as a percentage of GDP. However, the objective measure of credit constraint may suffer from endogeneity bias. Firms may be credit unconstrained because banks extend them credit (supply-side channel) or because they do not demand credit as demand falls (demand-side channel). To disentangle these two chan- nels we propose two instruments for the objective measure: the share of domestic credit provided by the banking sector (as a percentage of GDP), and the share of foreign banks out of the total number of banks in each country at time t and t − 1. Both instruments capture the health of the financial market and are correlated to whether or not a firm may receive credit but exogenous to the firm’s export decisions. Credit constraints may have an important impact on firm exports in ECA countries, especially given the large presence of foreign banks in the region. Based on data from Claessens and van Horen (2012), Figure 2 shows that the ECA region stands out in terms of presence of foreign banks. As European banks are the largest sources of FDI in the banking sector of ECA countries, this initial wave of transmission of credit shocks produced immediate effects on the credit markets of the countries we study.7 There is indeed growing evidence that banks transmit credit shocks 6 Alternative possible answers to question (1) are “Application procedures for loans or lines of credit are complex”, “Interest rates are not favorable”, “Collateral requirements are too high”, “Size of loan and maturity are insufficient”, “It is necessary to make informal payments to get bank loans”, “Did not think it would be approved”. 7 Giannetti and Laeven (2012) show that the collapse of the global market for syndicated loans during financial crisis can in part be explained by a flight home effect whereby lenders rebalance their loan portfolios in favor of domestic 7 internationally, quickly and powerfully, and in particular that the volatility of foreign lending is higher than that of domestic lending. Lending by bank subsidiaries appears to be more procyclical and “crunched” during the recent crisis.8 More relevant for this study is the evidence from Popov and Udell (2012). Focusing on the Eastern European countries in our data set, these authors show that firms were more credit constrained if they were dealing with banks that experienced a decline in equity and Tier 1 capital, as well as losses on financial assets. Riskier firms as well as firms with fewer tangible assets were more affected by positive or negative shocks suffered by banks. We account for both foreign and domestic shocks to demand, DomesticDemandShockjt and F oreignDemandShock−jt , in our estimation framework. Following Coulibaly, Sapriza, and Zlate (2012), foreign demand shocks are captured as the percentage change in GDP in the destination country weighted by the shares of the exports from country of firm i at time t − 1 to each des- tination country.9 Domestic demand shocks are computed as the percentage change in domestic GDP, consistent with the measure of foreign demand shocks. We also use the firm-specific level of utilization of facilities and manpower (“capacity”) as alternative measures of demand shocks for robustness checks. The measures are not included in the baseline regressions as they are not available for all time periods: less than 40% of the firms reported their capacity level in the last two rounds of the surveys. The vector Xijt contains the additional firm-level characteristics accounting for the heterogene- ` la ity among firms. Because of data limitations, we cannot calculate total factor productivity a Olley and Pakes (1996), but define the labor productivity as the log of the ratio of total sales to employment, a widely used alternative in the trade literature. Even though basic, this proxy is preferable to more elaborate measures of productivity in the absence of input or output prices, as in our case. We account for the age of the firm, computed as the difference between the time of the survey and the year the firm began operating in a given country. We also control for the size of the firm, including a binary variable that takes value one if the firm has more than one but less than 19 employees. Industry fixed effects are also added: the firms in the sample belong to the real estate (8%), manufacturing (27%), construction (13%), retail (40%), hotel and restaurant (5%), and transportation (8%) industries. The estimation of Equation 1 will be done through a linear probability model (LPM). While the LPM tends to generate biased and inconsistent estimates, as discussed in Horrace and Oaxaca (2006), it is not a large concern given the structure of our data. Irrespective of the set of covariates borrowers with the home bias of lenders’ loan origination increasing by approximately 20% if the bank’s home country experiences a banking crisis. The larger contraction in lending during the financial crisis by foreign-owned banks relative to locally-funded domestic banks is documented also by Ongena, Peydr´ o, and Van Horen (2015) that focus specifically on countries from the ECA region and Turkey. 8 For example, using U.S. data from the recent financial crisis, Cetorelli and Goldberg (2012) show that foreign banks pulled significant funding from their U.S. branches during the Great Recession with sizable effects on their U.S. lending. They estimate that the average-sized branch experienced a 12% net internal fund withdrawal and that, on average, for each dollar of funds transferred internally to the parent, branches decreased lending supply by about 40 to 50 cents. Similar findings are also documented in Coulibaly, Sapriza, and Zlate (2012). 9 The BEEPS data do not track the firm-specific export destinations. We use aggregate trade data from the World Integrated Trade Solutions (WITS) database. 8 used, at least 96% of predicted probabilities lie between 0 and 1, indicating that the estimates for only less than 4% of the sample may be biased or inconsistent.10 Even if one focuses on this 4% of observations, Wooldridge (2010) notes that “[...] if the main purpose is to estimate the partial effect of [the independent variable] on the response probability, averaged across the distribution of [the independent variable], then the fact that some predicted values are outside the unit interval may not be very important.” The computational tractability of the LPM also leads us to favor it over the nonlinear alternatives. The alternative of relying on Probit models would lead to biased estimates when including fixed effects because of incidental parameter problems (Greene (2004)). The benefits of accounting for industry and year fixed effects outweigh the costs of relying on LPM, in our view. Our sample includes periods prior, during and after the recent global crisis, covering the span between the 2002 and 2013 surveys. We explore whether the role of demand and credit constraints evolved over time interacting the main demand and credit shocks with time dummies. The dummy Crisis? takes value one for the years from 2009 to 2013. 3.2 Results We test the role and relative importance of demand and credit constraints on the extensive (Equa- tion 1) and intensive (Equation 2) margins of exports. We first conduct the analysis on the full sample and then focus on the crisis period in order to assess the robustness of the results to a period of stress. We find the usual predictions for export participation of firms: larger, more productive and (at least partially) foreign-owned firms are more likely to become exporters (Table 2). Negative domestic demand shocks can significantly increase the probability to export. Controlling for the heterogeneity in firm characteristics that can affect the firm’s ability to pay for the costs of ex- porting, a 1% decrease in domestic GDP increases the likelihood of selling abroad by about 0.5%. Production capacity is difficult to change in the short run and when local demand drops, firms are likely to look for alternative markets to sell their products to remain profitable. Thus, as the do- mestic economy worsens, firms branch out of their national borders in search of potential markets. The importance of domestic demand shocks is unaffected by the crisis (Table 3). The results hold also after controlling for the firm-specific level of utilization of facilities and manpower (Table A.1, A.2). On the other hand, foreign demand shocks play a more limited role when we consider the full sample. This result suggests that there are other factors that are influencing the choice of export market besides the demand at the destination country.11 However, when we examine the role of foreign shocks before and after the crisis, they are responsible for as much as 1.8% increase in the probability to export during the non-crisis period. Credit constraints do not significantly influence 10 Values outside of this range vary at most from -0.04 to 1.08, considering all predicted probabilities across models. 11 One factor may be the role of business and social networks in providing information and contacts in the destination market. Firms are more likely to export to markets where they have common links or relationship (Rauch (2001), Combes, Lafourcade, and Mayer (2005) and Chaney (2014)). 9 the choice to export (Tables 2 and 3). The results are robust to the use of both the subjective and objective measures of credit constraint and estimating directly or indirectly through instrumental variables. Restricting the focus only to those firms that export directly leads to qualitatively similar results. Next, we explore the respective role and importance of demand and financial shocks on the intensive margin measured by the share of exports in total sales (Table 4). Factors that play a marginal role in the decision to export, such as credit constraints and foreign demand shocks, take the center stage in determining the amount a firm exports. An increase in credit constraints will decrease the firm’s share of exports in total sales, while an increase in foreign shock will increase the share of exports. The results in Table 4 indicate that the subjective financial constraints have the largest impact on a firm’s exports. A one standard deviation reduction in subjective credit constraints corresponds to a 0.13 standard deviation increase in share of sales exported. Similar reductions in objective credit constraints and domestic demand shocks correspond to a 0.09 and 0.01 standard deviation raise in the share exports respectively.12 The negative relationship between domestic demand shocks and the share of exports in the firm’s sales suggests that firms substitute between local and foreign sales as their production capacities are fixed. Ceteris paribus as local demand conditions improve, the amount of exports decreases. Interestingly, the role of credit constraints becomes marginal when we distinguish between the crisis and pre-crisis period (Table 5). Similar findings emerge when we account for the level of capacity exploited (Table A.3, A.4).13 In all these cases, irrespective of specification chosen (foreign or domestic) demand shocks matter and appear with the expected sign. A positive shock abroad or a negative one at home increases the share of sales sold outside of the national borders. We plot the average marginal effects of credit constraints and demand shocks for easier com- parison across them in Figures 3 and 4. The vertical line represents the y-axis passing through the origin, while the horizontal lines indicate the 95% confidence interval around each coefficient capture by the dot. Demand shocks play a significant role in determining how much a firm exports, with foreign shocks playing a more prominent role. Interestingly, the perception to be financially constraints reduces less the volume of exports during the crisis, when access to finance is likely to be tighter. Finally, we test the robustness of the results by controlling for possible sample selection bias (Table 6). The lack of significance of the inverse Mill’s ratio (IMR) points to the lack of selection bias. The IMR is obtained assuming that (i) customs and trade regulations, (ii) labor regulations, or (iii) transportation as the main obstacle to business, or (iv) being (at least partially) foreign owned affects the probability of being an exporter, but not the share of exports in total sales. 12 The p-values of the Endogeneity test indicates that we cannot reject the null hypothesis that “Unable to obtain a loan?” can be treated as exogenous. 13 We obtain similar findings using the firm’s share of sales directly exported. The results are available upon request from the authors. 10 4 Conclusions Both demand shocks and credit constraints can affect the extensive and intensive margins of ex- ports. Demand shocks are important for both margins, but it depends on the source of the shocks. Domestic demand shocks are important for both the firm’s export participation and the share of exports in the firm’s sales. Foreign demand shocks are only relevant for the intensive margin and become more prominent when we distinguish between crisis and non-crisis periods. Similarly, credit constraints are only relevant for the intensive margin but only when we do not account for the financial crisis. The results suggest the relative importance of demand shocks over credit constraints for ex- porting firms, which remains over the financial crisis period. They confirm the findings in Nguyen and Qian (2014), where demand shocks are the main factor for poor firm performance over the financial crisis, and show that these results apply also to a period of no financial stress. Finally, before concluding, two points are worth highlighting. First, the evidence does not imply that credit constraints are not important for exports, but that they play a less prominent role compared to demand shocks. Second, the results presented are based on the evidence from countries in Eastern Europe and Central Asia. The (relative) importance of demand and credit shocks may be different in other developing countries that may suffer more acutely from underdeveloped financial markets. 11 5 Figure Figure 1: Trade dynamics: with linear prediction line 10 TUR TUR 5 LVA SRB POL SRB CZE 0 BLR 0 CZE KGZ Changes 2009-2002 Changes 2009-2002 POL KAZ LVA UKR HRV ARM HUN KGZ BLR -10 ARM -5 ALB GEO UKR GEO HUN AZE KAZ BGR AZE BGR ROM ROM MDA HRV ALB MDA -20 -10 MKD MKD -30 -15 -10 0 10 20 30 -5 0 5 10 15 Changes 2013-2009 Changes 2013-2009 (a) Share of exporters (b) Share of export Figure 2: Foreign banks in ECA % foreign banks South Asia % foreign bank assets Middle East and North Africa East Asia and Pacific Other High Income OECD Latin America and the Caribbean Europe and Central Asia Sub-Saharan Africa 0 20 40 60 12 Figure 3: The choice of how much to export: marginal effects -1.884 -3.566 Subj. fin. constr. Obj. fin. constr. 0.127 0.129 Foreign shock Foreign shock -0.045 -0.045 Domestic shock Domestic shock -4.000 -3.000 -2.000 -1.000 0.000 -8.000 -6.000 -4.000 -2.000 0.000 (a) OLS (b) OLS -19.942 Obj. fin. constr. 0.135 Foreign shock -0.042 Domestic shock -100.000 -50.000 0.000 50.000 (c) IV 13 Figure 4: The choice of how much to export: before and after the crisis, marginal effects -4.414 -6.305 Subj. fin. constr. Obj. fin. constr. 2.581 2.852 Subj. fin. constr. X Crisis? Obj. fin. constr. X Crisis? 0.176 0.160 Foreign shock Foreign shock -0.051 -0.034 Foreign shock X Crisis? Foreign shock X Crisis? -0.045 -0.051 Domestic shock Domestic shock 0.001 0.006 Domestic shock X Crisis? Domestic shock X Crisis? -6.000 -4.000 -2.000 0.000 2.000 4.000 -10.000 -5.000 0.000 5.000 10.000 (a) OLS (b) OLS 159.984 Obj. fin. constr. -190.938 Obj. fin. constr. X Crisis? -0.105 Foreign shock 0.243 Foreign shock X Crisis? -0.194 Domestic shock 0.154 Domestic shock X Crisis? -400.000 -200.000 0.000 200.000 400.000 (c) IV 14 6 Table Table 1: Summary statistics N Mean SD 95% C.I. Exporter(1=Yes) 23342 .258 0.008 .243 .273 Direct exporter(1=Yes) 23341 .21 0.01 .19 .22 Share of Exported Sales 23342 11 0.41 9.8 11 Share of Sales Exported Directly 23341 7.9 0.35 7.2 8.6 Subj. financial constraint 23342 3.1 0.02 3 3.1 Unable to obtain a loan? (Yes=1) 23342 .58 0.01 .56 .59 Foreign shock (%∆ GDP) 23342 .82 0.03 .76 .88 Domestic shock (%∆ GDP) 23342 2.8 0.08 2.6 2.9 Foreign shock (%∆ GDP × %Export) 23342 8.4 0.71 7 9.8 Domestic shock (%∆ GDP × (100-%Export)) 23342 249 7.68 234 264 Log(Labor Productivity) 23342 2.9 0.03 2.9 3 Firm Age 23342 15 0.17 15 16 Small Firm? (Yes=1) 23342 .81 0.01 .8 .82 Foreign Ownership? (1=Yes) 23342 .081 0.00 .072 .09 Domestic credit by banks (% of GDP) 23342 48 0.23 48 48 Share number of foreign banks 23342 57 0.39 56 58 Share number of foreign banks at t−1 23342 56 0.40 56 57 Capacity 15204 75 0.54 74 76 Inventory 10144 35 1.29 32 37 15 Table 2: The decision to become an exporter OLS OLS OLS OLS OLS IV IV Subj. financial -0.005 -0.004 constraint (0.008) (0.008) Unable to obtain a 0.021 0.022 0.450 0.205 loan? (Yes=1) (0.015) (0.015) (0.299) (0.423) Foreign shock 0.007 0.007 0.007 0.004 (%∆ GDP) (0.005) (0.005) (0.005) (0.009) Domestic shock -0.005** -0.005** -0.005** -0.006* (%∆ GDP) (0.002) (0.002) (0.002) (0.003) Log(Labor 0.039*** 0.040*** 0.040*** 0.039*** 0.039*** 0.048*** 0.042*** Productivity) (0.005) (0.005) (0.005) (0.005) (0.005) (0.008) (0.008) Firm Age 0.001* 0.001 0.001 0.001* 0.001* 0.001 0.001 (0.001) (0.001) (0.001) (0.001) (0.001) (0.001) (0.001) Small Firm? (Yes=1) -0.120*** -0.120*** -0.118*** -0.120*** -0.118*** -0.079** -0.102*** (0.017) (0.017) (0.017) (0.017) (0.017) (0.033) (0.039) Foreign Ownership? 0.218*** 0.216*** 0.217*** 0.218*** 0.219*** 0.227*** 0.224*** (1=Yes) (0.027) (0.027) (0.027) (0.027) (0.027) (0.031) (0.030) Industry F.E. Yes Yes Yes Yes Yes Yes Yes Year F.E. Yes Yes Yes Yes Yes Yes Yes Mean dep. var. 0.269 0.269 0.269 0.269 0.269 0.269 0.269 F stat 57 64 Kleibergen-Paap LM test 0.008 0.107 Hansen J stat 0.099 0.000 Endog. test 0.033 0.093 N 23342 23342 23342 23342 23342 23342 23342 The dependent variable is the P rob(Exportit > 0). Robust standard errors are reported in brackets. The symbols ***,**,* represent significance at the 1%, 5%, and 10% level. Under the null hypothesis of the Kleibergen-Paap LM test, the structural equation is underidentified, that is instruments are not relevant. Under the null hypothesis of the Hansen J test, the instruments are valid. The null of the endogeneity test is that the endogenous regressor(s) can be treated as exogenous. 16 Table 3: The decision to become an exporter: before and after the crisis OLS OLS OLS OLS OLS IV IV Crisis period? -0.016 0.063 0.085 -0.041 -0.020 -0.183 0.186 (1=Yes) (0.083) (0.094) (0.094) (0.085) (0.082) (0.408) (0.352) Subj. financial 0.028 0.028 constraint (0.028) (0.028) Subj. financial -0.034 -0.033 constraint × Crisis? (0.029) (0.029) Unable to obtain a 0.014 0.013 0.190 0.536 loan? (Yes=1) (0.038) (0.038) (0.581) (0.451) Unable to obtain a 0.007 0.009 0.256 -0.304 loan? (Yes=1) × Crisis? (0.041) (0.041) (0.693) (0.661) Foreign shock 0.016*** 0.017*** 0.016*** 0.007 (%∆ GDP) (0.006) (0.006) (0.005) (0.009) Foreign shock -0.009 -0.010 -0.010 -0.004 × Crisis? (0.007) (0.007) (0.007) (0.014) Domestic shock -0.007*** -0.007*** -0.007*** -0.006*** (%∆ GDP) (0.001) (0.001) (0.001) (0.002) Domestic shock 0.002 0.002 0.002 0.000 × Crisis? (0.003) (0.003) (0.003) (0.004) Log(Labor 0.039*** 0.040*** 0.039*** 0.040*** 0.039*** 0.048*** 0.042*** Productivity) (0.005) (0.005) (0.005) (0.005) (0.005) (0.008) (0.008) Firm Age 0.001* 0.001 0.001 0.001 0.001 0.001 0.001 (0.001) (0.001) (0.001) (0.001) (0.001) (0.001) (0.001) Small Firm? (Yes=1) -0.120*** -0.120*** -0.120*** -0.118*** -0.118*** -0.079** -0.099** (0.017) (0.017) (0.017) (0.017) (0.017) (0.032) (0.039) Foreign Ownership? 0.218*** 0.216*** 0.218*** 0.217*** 0.219*** 0.226*** 0.226*** (1=Yes) (0.027) (0.027) (0.027) (0.027) (0.027) (0.031) (0.030) Industry F.E. Yes Yes Yes Yes Yes Yes Yes Year F.E. Yes Yes Yes Yes Yes Yes Yes Mean dep. var. 0.269 0.269 0.269 0.269 0.269 0.269 0.269 F stat 50 53 Kleibergen-Paap LM test 0.037 0.196 Hansen J stat 0.022 0.000 Endog. test 0.000 0.000 N 23342 23342 23342 23342 23342 23342 23342 The dependent variable is the P rob(Exportit > 0). Robust standard errors are reported in brackets. The symbols ***,**,* represent significance at the 1%, 5%, and 10% level. Under the null hypothesis of the Kleibergen-Paap LM test, the structural equation is underidentified, that is instruments are not relevant. Under the null hypothesis of the Hansen J test, the instruments are valid. The null of the endogeneity test is that the endogenous regressor(s) can be treated as exogenous. 17 Table 4: The choice of how much to export OLS OLS OLS OLS OLS IV IV Subj. financial -2.659** -1.884* constraint (1.190) (1.059) Unable to obtain a -3.777* -3.566* -82.438** -19.942 loan? (Yes=1) (2.153) (1.956) (39.184) (30.848) Foreign shock 0.127*** 0.127*** 0.129*** 0.135*** (%∆ GDP × %Export) (0.026) (0.025) (0.025) (0.024) Domestic shock -0.045*** -0.045*** -0.045*** -0.042*** (%∆ GDP × (100-%Export)) (0.004) (0.004) (0.004) (0.006) Log(Labor -2.847*** -2.869*** -2.364*** -2.429*** -2.452*** -5.413*** -2.858*** Productivity) (0.677) (0.686) (0.627) (0.632) (0.637) (1.701) (1.081) Firm Age -0.241*** -0.241*** -0.125** -0.122** -0.120** -0.130 -0.098 (0.056) (0.056) (0.055) (0.055) (0.055) (0.105) (0.076) Small Firm? (Yes=1) -4.414** -4.603** -3.141* -3.144* -3.322* -8.324** -4.153* (2.000) (2.023) (1.837) (1.832) (1.836) (3.656) (2.339) Foreign Ownership? 10.042*** 10.077*** 10.197*** 10.043*** 10.027*** 6.253 9.245*** (1=Yes) (2.485) (2.471) (2.362) (2.364) (2.334) (3.818) (2.694) Industry F.E. Yes Yes Yes Yes Yes Yes Yes Year F.E. Yes Yes Yes Yes Yes Yes Yes Mean dep. var. 41.796 41.796 41.796 41.796 41.796 41.796 41.796 F stat 5 27 Kleibergen-Paap LM test 0.081 0.260 Hansen J stat 0.138 0.007 Endog. test 0.005 0.586 N 6275 6275 6275 6275 6275 6275 6275 The dependent variable is the Share(ExportedSalesit ). Robust standard errors are reported in brackets. The symbols ***,**,* represent significance at the 1%, 5%, and 10% level. Under the null hypothesis of the Kleibergen-Paap LM test, the structural equation is underidentified, that is instruments are not relevant. Under the null hypothesis of the Hansen J test, the instruments are valid. The null of the endogeneity test is that the endogenous regressor(s) can be treated as exogenous. 18 Table 5: The choice of how much to export: before and after the crisis OLS OLS OLS OLS OLS IV IV Crisis period? 15.817*** -16.874 18.451*** -16.260 16.571*** 47.058 28.722 (1=Yes) (4.190) (19.947) (5.974) (12.695) (4.081) (31.907) (21.548) Subj. financial -4.595 -1.056 constraint (4.979) (1.140) Subj. financial 1.990 -0.840 constraint × Crisis? (5.127) (1.569) Unable to obtain a -11.479 -2.017 8.960 7.650 loan? (Yes=1) (7.634) (1.542) (28.602) (11.337) Unable to obtain a 7.880 -1.514 -86.938* -19.250 loan? (Yes=1) × Crisis? (7.956) (2.498) (50.341) (33.585) Foreign shock 0.199*** 0.198*** 0.197*** 0.202*** (%∆ GDP × %Export) (0.010) (0.010) (0.010) (0.014) Foreign shock -0.074*** -0.074*** -0.071*** -0.073** (%∆ GDP × %Export) × (0.028) (0.027) (0.027) (0.029) Crisis? Domestic shock -0.037*** -0.037*** -0.037*** -0.038*** (%∆ GDP × (100-%Export)) (0.002) (0.002) (0.002) (0.002) Domestic shock -0.007 -0.007 -0.007 -0.005 %∆ GDP) × Crisis? (0.005) (0.005) (0.005) (0.008) Log(Labor -2.364*** -2.864*** -2.428*** -2.881*** -2.449*** -5.280*** -2.643** Productivity) (0.635) (0.677) (0.640) (0.686) (0.645) (1.604) (1.058) Firm Age -0.126** -0.242*** -0.123** -0.241*** -0.121** -0.140 -0.111 (0.055) (0.056) (0.055) (0.056) (0.055) (0.100) (0.074) Small Firm? (Yes=1) -3.175* -4.331** -3.180* -4.527** -3.353* -7.822** -3.724 (1.842) (2.001) (1.836) (2.022) (1.842) (3.469) (2.269) Foreign Ownership? 10.122*** 10.050*** 9.972*** 10.092*** 9.958*** 6.592* 9.591*** (1=Yes) (2.361) (2.485) (2.364) (2.471) (2.334) (3.623) (2.641) Industry F.E. Yes Yes Yes Yes Yes Yes Yes Year F.E. Yes Yes Yes Yes Yes Yes Yes Mean dep. var. 41.796 41.796 41.796 41.796 41.796 41.796 41.796 F stat 5 76 Kleibergen-Paap LM test 0.050 0.113 Hansen J stat 0.036 0.000 Endog. test 0.430 0.000 N 6275 6275 6275 6275 6275 6275 6275 The dependent variable is the Share(ExportedSalesit ). Robust standard errors are reported in brackets. The symbols ***,**,* represent significance at the 1%, 5%, and 10% level. Under the null hypothesis of the Kleibergen-Paap LM test, the structural equation is underidentified, that is instruments are not relevant. Under the null hypothesis of the Hansen J test, the instruments are valid. The null of the endogeneity test is that the endogenous regressor(s) can be treated as exogenous. 19 Table 6: The choice of how much to export: is there a sample selection problem? OLS OLS OLS OLS OLS IV IV Subj. financial -1.809 -0.875 constraint (1.183) (0.982) Unable to obtain a -2.924 -2.343 -60.005* 13.174 loan? (Yes=1) (2.143) (1.952) (34.856) (36.712) Foreign shock 0.148*** 0.148*** 0.149*** 0.144*** (%∆ GDP × %Export) (0.021) (0.021) (0.021) (0.023) Domestic shock -0.045*** -0.045*** -0.045*** -0.047*** (%∆ GDP × (100-%Export)) (0.005) (0.005) (0.005) (0.008) Inverse Mill’s Ratio 1.258 2.223 3.282 2.531 2.843 -9.212 5.755 (8.626) (8.569) (7.664) (7.740) (7.680) (16.166) (9.336) Log(Labor -2.525*** -2.492*** -1.799** -1.882** -1.891** -5.171** -1.278 Productivity) (0.926) (0.932) (0.774) (0.780) (0.791) (2.356) (1.591) Firm Age -0.229*** -0.224*** -0.094 -0.098 -0.096 -0.246** -0.088 (0.074) (0.073) (0.073) (0.073) (0.072) (0.097) (0.071) Small Firm? (Yes=1) -7.032** -7.542** -5.602* -5.337 -5.586* -6.588 -5.692* (3.480) (3.450) (3.252) (3.283) (3.259) (5.171) (3.261) Foreign Ownership? 9.760** 10.167** 11.035*** 10.621** 10.694** 1.439 12.955** (1=Yes) (4.498) (4.454) (4.207) (4.231) (4.165) (8.763) (6.271) Industry F.E. Yes Yes Yes Yes Yes Yes Yes Year F.E. Yes Yes Yes Yes Yes Yes Yes Mean dep. var. 41.866 41.866 41.866 41.866 41.866 41.866 41.866 F stat 6 33 Kleibergen-Paap LM test 0.111 0.393 Hansen J stat 0.048 0.017 Endog. test 0.068 0.918 N 5977 5977 5977 5977 5977 5977 5977 The dependent variable is the Share(ExportedSalesit ). Robust standard errors are reported in brackets. The symbols ***,**,* represent significance at the 1%, 5%, and 10% level. Under the null hypothesis of the Kleibergen-Paap LM test, the structural equation is underidentified, that is instruments are not relevant. Under the null hypothesis of the Hansen J test, the instruments are valid. The null of the endogeneity test is that the endogenous regressor(s) can be treated as exogenous. 20 A Appendix: Robustness checks Table A.1: The decision to become an exporter OLS OLS OLS OLS OLS IV IV Subj. financial -0.015 -0.014 constraint (0.013) (0.013) Unable to obtain a 0.013 0.020 -1.337* -1.391*** loan? (Yes=1) (0.024) (0.024) (0.706) (0.504) Foreign shock -0.007 -0.007 -0.007 0.019 (%∆ GDP) (0.009) (0.009) (0.009) (0.018) Domestic shock -0.011*** -0.010*** -0.011*** -0.000 (%∆ GDP) (0.003) (0.003) (0.004) (0.007) Capacity -0.000 -0.000 -0.000 -0.000 -0.000 -0.003* -0.003** (0.001) (0.001) (0.001) (0.001) (0.001) (0.002) (0.001) Log(Labor 0.034*** 0.040*** 0.040*** 0.034*** 0.034*** 0.003 0.005 Productivity) (0.008) (0.008) (0.008) (0.008) (0.008) (0.024) (0.018) Firm Age 0.002*** 0.002*** 0.002*** 0.002*** 0.002*** 0.003** 0.003*** (0.001) (0.001) (0.001) (0.001) (0.001) (0.001) (0.001) Small Firm? (Yes=1) -0.272*** -0.272*** -0.272*** -0.271*** -0.270*** -0.386*** -0.389*** (0.024) (0.024) (0.024) (0.024) (0.024) (0.074) (0.060) Foreign Ownership? 0.201*** 0.197*** 0.200*** 0.199*** 0.203*** 0.124 0.123* (1=Yes) (0.034) (0.035) (0.035) (0.034) (0.034) (0.076) (0.073) Industry F.E. Yes Yes Yes Yes Yes Yes Yes Year F.E. Yes Yes Yes Yes Yes Yes Yes Mean dep. var. 0.326 0.326 0.326 0.326 0.326 0.326 0.326 F stat 18 15 Kleibergen-Paap LM test 0.146 0.013 Hansen J stat 0.563 0.561 Endog. test 0.001 0.000 N 15204 15204 15204 15204 15204 15204 15204 The dependent variable is the P rob(Exportit > 0). Robust standard errors are reported in brackets. The symbols ***,**,* represent significance at the 1%, 5%, and 10% level. Under the null hypothesis of the Kleibergen-Paap LM test, the structural equation is underidentified, that is instruments are not relevant. Under the null hypothesis of the Hansen J test, the instruments are valid. The null of the endogeneity test is that the endogenous regressor(s) can be treated as exogenous. 21 Table A.2: The decision to become an exporter: before and after the crisis OLS OLS OLS OLS OLS IV IV Crisis period? 0.056 0.074 0.096 0.007 0.035 0.143 0.916* (1=Yes) (0.102) (0.148) (0.149) (0.115) (0.115) (0.703) (0.470) Subj. financial -0.001 -0.002 constraint (0.021) (0.021) Subj. financial -0.015 -0.012 constraint × Crisis? (0.025) (0.025) Unable to obtain a -0.007 -0.007 -1.422* -0.427 loan? (Yes=1) (0.033) (0.034) (0.757) (0.344) Unable to obtain a 0.021 0.029 0.075 -1.085* loan? (Yes=1) × Crisis? (0.042) (0.042) (0.884) (0.623) Foreign shock 0.009 0.009 0.009 0.004 (%∆ GDP) (0.006) (0.006) (0.006) (0.012) Foreign shock -0.016 -0.016 -0.016 0.019 × Crisis? (0.011) (0.011) (0.011) (0.025) Domestic shock -0.006*** -0.006*** -0.006*** -0.007*** (%∆ GDP) (0.001) (0.001) (0.001) (0.002) Domestic shock -0.005 -0.005 -0.006 0.008 × Crisis? (0.004) (0.004) (0.004) (0.008) Capacity -0.000 -0.000 -0.000 -0.001 -0.000 -0.002* -0.001 (0.000) (0.000) (0.000) (0.000) (0.000) (0.001) (0.001) Capacity × 0.000 0.000 0.000 0.000 0.000 -0.001 -0.003 Crisis? (0.001) (0.001) (0.001) (0.001) (0.001) (0.002) (0.002) Log(Labor 0.034*** 0.040*** 0.034*** 0.040*** 0.035*** 0.003 0.004 Productivity) (0.008) (0.008) (0.008) (0.008) (0.008) (0.024) (0.019) Firm Age 0.002*** 0.002*** 0.002*** 0.002*** 0.002*** 0.003** 0.003** (0.001) (0.001) (0.001) (0.001) (0.001) (0.001) (0.001) Small Firm? (Yes=1) -0.272*** -0.272*** -0.271*** -0.271*** -0.270*** -0.387*** -0.400*** (0.024) (0.024) (0.024) (0.024) (0.024) (0.075) (0.067) Foreign Ownership? 0.201*** 0.197*** 0.199*** 0.200*** 0.202*** 0.123 0.120 (1=Yes) (0.034) (0.035) (0.034) (0.035) (0.034) (0.075) (0.078) Industry F.E. Yes Yes Yes Yes Yes Yes Yes Year F.E. Yes Yes Yes Yes Yes Yes Yes Mean dep. var. 0.326 0.326 0.326 0.326 0.326 0.326 0.326 F stat 14 16 Kleibergen-Paap LM test 0.012 0.000 Hansen J stat 0.039 0.000 Endog. test 0.032 0.205 N 15204 15204 15204 15204 15204 15204 15204 The dependent variable is the P rob(Exportit > 0). Robust standard errors are reported in brackets. The symbols ***,**,* represent significance at the 1%, 5%, and 10% level. Under the null hypothesis of the Kleibergen-Paap LM test, the structural equation is underidentified, that is instruments are not relevant. Under the null hypothesis of the Hansen J test, the instruments are valid. The null of the endogeneity test is that the endogenous regressor(s) can be treated as exogenous. 22 Table A.3: The choice of how much to export OLS OLS OLS OLS OLS IV IV Subj. financial -1.206 0.267 constraint (1.310) (1.118) Unable to obtain a -4.291* -1.435 -17.575 -17.962 loan? (Yes=1) (2.334) (2.058) (23.814) (17.023) Foreign shock 0.114*** 0.114*** 0.114*** 0.115*** (%∆ GDP × %Export) (0.021) (0.021) (0.021) (0.021) Domestic shock -0.058*** -0.058*** -0.058*** -0.053*** (%∆ GDP × (100-%Export)) (0.005) (0.005) (0.005) (0.007) Capacity 0.086 0.081 0.083* 0.084* 0.080 0.055 0.050 (0.058) (0.057) (0.049) (0.049) (0.049) (0.071) (0.057) Log(Labor -3.380*** -3.484*** -2.666*** -2.663*** -2.698*** -3.830*** -3.067*** Productivity) (0.792) (0.785) (0.671) (0.673) (0.670) (1.067) (0.817) Firm Age -0.256*** -0.249*** -0.125* -0.124* -0.123* -0.230*** -0.104 (0.066) (0.065) (0.065) (0.065) (0.065) (0.075) (0.072) Small Firm? (Yes=1) -10.818*** -11.195*** -8.781*** -8.801*** -8.874*** -12.025*** -9.954*** (2.350) (2.314) (2.071) (2.070) (2.064) (2.772) (2.459) Foreign Ownership? 14.775*** 14.659*** 15.199*** 15.224*** 15.134*** 13.977*** 14.381*** (1=Yes) (2.941) (2.909) (2.855) (2.854) (2.857) (3.118) (2.904) Industry F.E. Yes Yes Yes Yes Yes Yes Yes Year F.E. Yes Yes Yes Yes Yes Yes Yes Mean dep. var. 43.293 43.293 43.293 43.293 43.293 43.293 43.293 F stat 13 34 Kleibergen-Paap LM test 0.044 0.022 Hansen J stat 0.000 0.001 Endog. test 0.493 0.081 N 4958 4958 4958 4958 4958 4958 4958 The dependent variable is the Share(ExportedSalesit ). Robust standard errors are reported in brackets. The symbols ***,**,* represent significance at the 1%, 5%, and 10% level. Under the null hypothesis of the Kleibergen-Paap LM test, the structural equation is underidentified, that is instruments are not relevant. Under the null hypothesis of the Hansen J test, the instruments are valid. The null of the endogeneity test is that the endogenous regressor(s) can be treated as exogenous. 23 Table A.4: The choice of how much to export: before and after the crisis OLS OLS OLS OLS OLS IV IV Crisis period? 16.129** -17.683 14.907 -23.251 15.714** -40.104 25.639 (1=Yes) (6.648) (22.975) (9.398) (14.931) (6.845) (34.221) (17.344) Subj. financial 0.300 -0.105 constraint (3.789) (1.144) Subj. financial -1.455 0.376 constraint × Crisis? (4.008) (1.605) Unable to obtain a -6.569 -1.985 -73.944** 12.357 loan? (Yes=1) (6.475) (1.931) (35.157) (10.117) Unable to obtain a 2.477 0.667 56.265 -13.257 loan? (Yes=1) × Crisis? (6.912) (2.887) (42.760) (23.487) Foreign shock 0.195*** 0.195*** 0.195*** 0.196*** (%∆ GDP × %Export) (0.011) (0.011) (0.011) (0.015) Foreign shock -0.087*** -0.086*** -0.086*** -0.087*** (%∆ GDP × %Export) × (0.025) (0.025) (0.025) (0.027) Crisis? Domestic shock -0.037*** -0.037*** -0.037*** -0.039*** (%∆ GDP × (100-%Export)) (0.002) (0.002) (0.002) (0.003) Domestic shock -0.022*** -0.022*** -0.021*** -0.019** %∆ GDP) × Crisis? (0.005) (0.005) (0.005) (0.009) Capacity 0.058** 0.116* 0.058** 0.106* 0.056** 0.002 0.064** (0.024) (0.068) (0.023) (0.063) (0.023) (0.102) (0.031) Capacity × 0.025 -0.029 0.027 -0.025 0.025 0.052 0.018 Crisis? (0.056) (0.089) (0.056) (0.085) (0.055) (0.128) (0.074) Log(Labor -2.686*** -3.442*** -2.683*** -3.546*** -2.715*** -3.947*** -2.781*** Productivity) (0.681) (0.788) (0.682) (0.782) (0.679) (0.982) (0.787) Firm Age -0.125* -0.258*** -0.125* -0.252*** -0.124* -0.226*** -0.125* (0.064) (0.066) (0.064) (0.065) (0.064) (0.074) (0.071) Small Firm? (Yes=1) -8.815*** -10.603*** -8.838*** -10.983*** -8.906*** -11.664*** -8.573*** (2.075) (2.342) (2.074) (2.308) (2.069) (2.754) (2.513) Foreign Ownership? 15.045*** 14.746*** 15.066*** 14.618*** 14.986*** 13.922*** 15.145*** (1=Yes) (2.832) (2.945) (2.831) (2.915) (2.834) (3.102) (2.889) Industry F.E. Yes Yes Yes Yes Yes Yes Yes Year F.E. Yes Yes Yes Yes Yes No No Mean dep. var. 43.293 43.293 43.293 43.293 43.293 43.293 43.293 F stat 12 71 Kleibergen-Paap LM test 0.091 0.149 Hansen J stat 0.000 0.015 Endog. test 0.425 0.000 N 4958 4958 4958 4958 4958 4958 4958 The dependent variable is the Share(ExportedSalesit ). Robust standard errors are reported in brackets. The symbols ***,**,* represent significance at the 1%, 5%, and 10% level. Under the null hypothesis of the Kleibergen-Paap LM test, the structural equation is underidentified, that is instruments are not relevant. Under the null hypothesis of the Hansen J test, the instruments are valid. The null of the endogeneity test is that the endogenous regressor(s) can be treated as exogenous. 24 References Amiti, M., and D.E. Weinstein (2011): “Exports and financial shocks,” The Quarterly Journal of Economics, 126(4), 1841–1877. Auboin, Marc (2009): “Boosting the availability of trade finance in the current crisis: Background analysis for a substantial G20 package,” CEPR Policy Insight, 35. Behrens, Kristian, Gregory Corcos, and Giordano Mion (2013): “Trade crisis? What trade crisis?,” Review of Economics and Statistics, 95(2), 702–709. e M., and J. Myles Shaver (2002): “Exporting and Capital Investment: On the Strategic Campa, Jos´ Behavior of Exporters,” IESE Business School, University of Navarra, Discussion Paper. Cetorelli, Nicola, and Linda S Goldberg (2012): “Follow the Money: Quantifying Domestic Effects of Foreign Bank Shocks in the Great Recession,” The American Economic Review, 102(3), 213– 218. Chaney, Thomas (2014): “The network structure of international trade,” The American Economic Review, 104(11), 3600–3634. (2016): “Liquidity constrained exporters,” Journal of Economic Dynamics and Control. Chor, Davin, and Kalina Manova (2012): “Off the Cliff and Back: Credit Conditions and Inter- national Trade during the Global Financial Crisis,” Journal of International Economics, 87(1), 117–133. Claessens, Stijn, Hui Tong, and Shang-Jin Wei (2012): “From the financial crisis to the real econ- omy: Using firm-level data to identify transmission channels,” Journal of International Eco- nomics, 88(2), 375–387. Claessens, S., and N. van Horen (2012): “Foreign Banks: Trends, Impact and Financial Stability,” International Monetary Fund Working Paper WP/12/10. Combes, Pierre-Philippe, Miren Lafourcade, and Thierry Mayer (2005): “The trade-creating effects of business and social networks: evidence from France,” Journal of international Economics, 66(1), 1–29. Coulibaly, Brahima, Horacio Sapriza, and Andrei Zlate (2012): “Financial frictions, trade credit, and the 2008-09 global financial crisis,” . Eaton, Jonathan, Samuel Kortum, Brent Neiman, and John Romalis (2011): “Trade and the global recession,” Discussion paper, NBER Working Paper No. 16666. Fernandes, Ana M (2007): “Trade policy, trade volumes and plant-level productivity in Colombian manufacturing industries,” Journal of International Economics, 71(1), 52–71. 25 Giannetti, Mariassunta, and Luc Laeven (2012): “The Flight Home Effect: Evidence from the Syndicated Loan Market During Financial Crises ?,” 104(1), 23–43. Gorodnichenko, Y., and M. Schnitzer (2010): “Financial constraints and innovation: Why poor countries don’t catch up,” Discussion paper, National Bureau of Economic Research. Greene, William (2004): “The behaviour of the maximum likelihood estimator of limited dependent variable models in the presence of fixed effects,” The Econometrics Journal, 7(1), 98–119. Hallward-Driemeier, M., and R. Aterido (2009): “Comparing Apples with Apples: How to Make (More) Sense in Subjective Ranking of Constraints to Business,” Policy Research Working Paper Series 5054, The World Bank. Horrace, William C, and Ronald L Oaxaca (2006): “Results on the bias and inconsistency of ordinary least squares for the linear probability model,” Economics Letters, 90(3), 321–327. Manova, Kalina (2013): “Credit constraints, heterogeneous firms, and international trade,” The Review of Economic Studies, 80(2), 711–744. Melitz, Marc J. (2003): “The Impact of Trade on Intra-Industry Reallocations and Aggregate Industry Productivity,” Econometrica, 71(6), 1695–1725. Nguyen, Ha, and Rong Qian (2014): “Demand collapse or credit crunch to firms? Evidence from the World Bank’s financial crisis survey in Eastern Europe,” Journal of International Money and Finance, 47, 125–144. Olley, GS, and A. Pakes (1996): “The dynamics of productivity in the telecommunications equip- ment industry,” Econometrica, 64(6), 1263–1297. e-Luis Peydr´ Ongena, Steven, Jos´ o, and Neeltje Van Horen (2015): “Shocks Abroad, Pain at Home? Bank-Firm-Level Evidence on the International Transmission of Financial Shocks,” IMF Eco- nomic Review, 63(4), 698–750. Paravisini, Daniel, Veronica Rappoport, Philipp Schnabl, and Daniel Wolfenzon (2015): “Dissecting the effect of credit supply on trade: Evidence from matched credit-export data,” The Review of Economic Studies, 82(1), 333–359. Pavcnik, N (2002): “Trade Liberalization, Exit, and Productivity Improvements: Evidence from Chilean Plants,” Review of economic studies, 69(1), 245–276. Popov, A., and G.F. Udell (2012): “Cross-border banking, credit access, and the financial crisis,” Journal of International Economics, 87(1), 147–161. Rauch, James E (2001): “Business and social networks in international trade,” Journal of economic literature, pp. 1177–1203. Wooldridge, Jeffrey M (2010): Econometric analysis of cross section and panel data. MIT press. 26