Policy Research Working Paper 10403 Export Survival The Role of Banks and Stock Markets Melise Jaud Madina Kukenova Martin Strieborny Macroeconomics, Trade and Investment Global Practice April 2023 Policy Research Working Paper 10403 Abstract Banks and stock markets play distinct roles in helping of working capital. And the trade credit can act as a sub- exporters survive in foreign markets, conditional on the stitute for external financing only from banks and only in specific financial needs of exported products. Stock markets the presence of well-established export links. These results rather than banks help exporters who lack easily collat- on product-level export survival provide new insights into eralizable tangible assets. Active rather than large stock the transmission process from finance to the real economy. markets promote exports of products requiring high levels 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 mjaud@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 Export Survival: The Role of Banks and Stock Markets∗ Melise Jaud† Madina Kukenova‡ Martin Strieborny§ JEL classification: G10, G21, F14. Keywords: banks versus stock markets, transmission from finance to real economy, finance and export survival ∗ The previous version of this paper circulated under the title ”Financial dependence and intensive margin of trade”. We would like to thank Tibor Besedeˇ s, Xin Huang, Jean Imbs, Beata Javorcik, Rafael Lalive, Luis Santos-Pinto, Pascal St-Amour and the seminar and conference participants at Univer- sity of Lausanne, Lund University and the 37th AFFI conference for very useful comments and sug- gestions. Melise Jaud appreciates the financial support from the Swiss National Science Foundation. Madina Kukenova appreciates the grant received from the Swiss School of Higher Education. Mar- tin Strieborny appreciates the financial support from the Handelsbanken Research Foundation. The most recent version of this paper including an accompanying Online Appendix can be downloaded from www.martinstrieborny.com † World Bank; E-Mail: mjaud@worldbank.org ‡ Swiss School of Higher Education; E-Mail: madina.kukenova@sshe.ch § University of Glasgow, Adam Smith Business School, Room 469, Main Building, University Avenue, Glasgow, G12 8QQ, United Kingdom. E-Mail: martin.strieborny@glasgow.ac.uk 1 Introduction Banks and stock markets operate in fundamentally different ways, with stock markets providing a public platform that directly connects investors with firms while banks serve as an intermediary between savers and borrowers. Such structural differences have inspired an extensive literature on the distinct roles of banks and stock markets in the real economy u¸ (e.g., Demirg¨ c–Kunt and Maksimovic 1998, 2002; Levine and Zervos 1998; Beck and Levine 2002; Levine 2002; Beck and Levine 2004; Shen and Lee 2006; Song and Thakor u¸ 2010; Demirg¨ c–Kunt, Feyen, and Levine 2013; Langfield and Pagano 2016). However, the evidence arising from this research is often inconclusive, especially when it comes to the stock markets (see, e.g., Zingales 2015). This paper examines how domestic banks and stock markets help exporters survive in foreign product markets, moving beyond the traditional focus of existing literature on domestic output. Financial constraints are even more important in international trade than in domestic production due to additional costs like cross-border transport, customs clearance, conforming with foreign product market regulations, etc. The bilateral character of the data in international trade also allows us to exploit the variation in the strength of domestic banks and stock markets across exporting countries while at the same time holding constant the product market structure as well as the legal, regulatory and institutional environment by focusing on a single destination market (the United States). Last but not least, the highly disaggregated product-level data on international trade are available and directly comparable across many different countries at all stages of banking and stock market development. The focus of our analysis is the different financing needs of exported products that can be addressed by well-developed financial markets and institutions. Certain products rely on extensive investment in physical capital (Rajan and Zingales 1998), high levels of working capital (Raddatz 2006), or a significant use of intangible assets that are less suitable to serve as collateral (Braun 2003; Claessens and Laeven 2003). Two additional transmission channels from finance to the real economy relate to the firm-to-firm trade credit as an alternative source of external finance (Fisman and Love 2003) and to the al- 1 locative role of the financial sector (Fisman and Love 2007). We compute these different financial needs for narrowly defined industrial sectors and match them with the disag- gregated product-level data on international trade. Given that the explored transmission channels might operate differently in the short term and long term, we examine export survival at time horizons ranging from one to ten years. We also allow for the possibility that the size and the activity of domestic stock markets play different roles in shaping export survival at the product level. Our three main empirical findings are as follows. First, well-developed stock markets rather than strong banks are key in promoting export survival of products from industries with a high share of intangible assets. This result supports the notion that stock markets have an advantage over banks when it comes to promoting innovative sectors of the economy that face increasing returns to scale and a pronounced technological change (see, e.g., Allen 1993; Aghion et al. 2004; Brown, Martinsson, and Petersen 2013, 2017; Hsu, Tian, and Xu 2014). Second, active stock markets play the most important role for the export survival of products with high liquidity needs, possibly reflecting the role active stock markets have in monitoring the efficiency of working capital management (see, e.g., Gill and Biger 2013; Ben-Nasr 2016; Zeidan and Shapir 2017). By contrast, large stock markets are often dominated by banks in promoting these products, especially when it comes to long-term export survival. Third, trade credit operates as a substitute for external finance from banks but not from stock markets. In addition, this substitution works only for long-term export survival, suggesting that the channel can successfully operate only alongside well-established export links where exporters and their foreign customers already enjoy the mutual trust necessary for an extensive use of trade credit. Overall, our results suggest that banks, large stock markets, and active stock markets all play different roles in shaping product-level export survival, depending on the exam- ined transmission channel and the time horizon under consideration. This relates our work to three distinct strands of the existing literature. First, the paper is related to the existing literature examining the distinct roles of banks and stock markets. This line of research has investigated the importance of banks 2 u¸ and stock markets for output growth (e.g., Demirg¨ c–Kunt and Maksimovic 1998; Levine and Zervos 1998; Beck and Levine 2004; Shen and Lee 2006), the economic and financial consequences of having a bank-based or a market-based financial system (e.g., Beck and u¸ Levine 2002; Demirg¨ c–Kunt and Maksimovic 2002; Levine 2002; Langfield and Pagano 2016; Bats and Houben 2020), as well as the evolving roles of banks and stock markets during the process of economic development (e.g., Tadesse 2002; Song and Thakor 2010; u¸ Demirg¨ c–Kunt, Feyen, and Levine 2013). We focus on specific transmission channels from banks and stock markets, utilizing disaggregated data on international trade that are available and directly comparable across different countries. One of the channels we examine also relates our work to the recent literature on banks and intangible capital (Dell’Ariccia et al. 2020; Beck et al. 2021) while our findings on large versus active stock markets complement the recent research on different characteristics of stock markets (Bae, Bailey, and Kang 2021). Second, the paper connects to the strand of literature looking at the role of finance in international trade (e.g., Kletzer and Bardhan 1987; Beck 2002; Greenaway, Guariglia, ericourt 2010; Amiti and Weinstein 2011; and Kneller 2007; Manova 2008; Berman and H´ Ju and Wei 2011; Levchenko, Lewis, and Tesar 2011; Minetti and Zhu 2011; Bricongne uls 2015; et al. 2012; Chor and Manova 2012; Becker, Chen, and Greenberg 2013; Muˆ Paravisini et al. 2015). Several papers in this area look at how the financial sector shapes export entry, export exit or export volume alongside some of the transmission channels that we also explore (e.g., Beck 2003; Svaleryd and Vlachos 2005; Hur, Raj, and Riyanto 2006; Manova 2013; Fan, Lai, and Li 2015; Crino and Ogliari 2017).1 We focus on the relative importance of banks and stock markets while also allowing for distinct roles for active versus large stock markets. In terms of the outcome variables, we explore both short-term and long-term export dynamics by looking at the export survival across 1 Rather than investigating these transmission channels, several other papers look at the role of banks and stock markets in international trade from different angles. Cho et al. (2019) develop a model link- ing bank and bond financing with firm size distribution, gains from trade, and the real exchange rate. Amissah et al. (2021) explore theoretically and empirically the bi-directional relationship between financial infrastructure and comparative advantage of countries. 3 different time horizons.2 Finally, we examine two additional transmission channels from finance to the real economy that have so far attracted little attention in the finance-trade literature. s and Prusa Third, the paper relates to the trade survival literature (e.g., Besedeˇ s and 2006a, 2006b; Nitsch 2009; Brenton, Saborowski, and von Uexkull 2010; Besedeˇ ek´ Prusa 2011; B´ ozy 2012; Hess and Persson 2012; Besedeˇ es and Murak¨ s 2013; Cadot et al. 2013; Albornoz, Fanelli, and Hallak 2016; Araujo, Mion, and Ornelas 2016). A few papers explore particular aspects of the relationship between finance and export survival. Jaud, Kukenova, and Strieborny (2018) document the disciplining impact of banks on exporting managers, Jaud, Kukenova, and Strieborny (2015, 2019) investigate s, the importance of financial frictions in the context of agricultural exports, and Besedeˇ Kim, and Lugovskyy (2014) use the survival analysis to compute a measure of project risk.3 However, the trade survival literature has so far not systematically looked at the distinct roles of banks and stock markets or the different transmission channels from finance to export survival at different time horizons. The rest of the paper is organized as follows. Section 2 outlines our empirical ap- proach. Section 3 describes the construction of our dataset. Section 4 discusses the results about the relative importance of banks and stock markets alongside our three main transmission channels. Section 5 explores if and how these results change when we distinguish between large and active stock markets. Section 6 reports the results for two additional transmission channels and other tests. Section 7 concludes and suggests directions for further research. 2 The finance-trade literature often focuses on the extensive margin of international trade and measures export success in terms of export entry. However, most exports cease shortly after they start without leaving a significant mark on the long-term export performance of a country (see, e.g., Besedeˇ s and Prusa 2006a, 2006b; B´ es and Murak¨ ek´ ozy 2012; Albornoz, Fanelli, and Hallak 2016). Export survival is also an important component of the intensive margin of international trade that often sets apart the successful developing countries from the unsuccessful ones (Besedeˇs and Prusa 2011). 3 Bridges and Guariglia (2008) and G¨ org and Spaliara (2014) look at how export status affects the link between financial factors and firm survival but do not examine the link between finance and trade. Jaud, Kukenova, and Strieborny (2021) examine export survival but focus on the role of foreign investors rather than the domestic financial system. 4 2 Empirical Approach We examine the importance of well-developed banks and stock markets for export success at different time horizons through the lenses of five well-established transmission channels from finance to the real economy. Subsection 2.1 describes in more detail the transmission channels, while Subsection 2.2 discusses our estimation strategy. 2.1 Transmission Channels The transmission channels examined in this paper have been originally developed in papers exploring the impact of finance on economic growth (Rajan and Zingales 1998; Braun 2003; Claessens and Laeven 2003; Fisman and Love 2003, 2007; Raddatz 2006). As outlined below, these channels might be even more important in the context of finance and international trade. Investment needs. Firms in certain industries do not generate sufficient cash flows to maintain the necessary level of physical capital from internal financial sources alone. Typical technological reasons for high investment needs include the large scale of a typ- ical investment project or a long gestation period (time span between the start of an investment project and the start of actual production making use of this project) in a given industry. Consequently, the firms in these industries need to rely on providers of external finance like banks and stock markets to fund a significant part of their invest- ment needs (Rajan and Zingales 1998). This would be even more true in the case of exporting that requires substantial additional investments due to adjusting of products to different consumer preferences, satisfying regulatory requirements in foreign countries, establishing and maintaining distribution networks abroad, etc. Liquidity needs. Firms in certain industries require high levels of liquid funds (working capital) to maintain their operations. Typical technological reasons for high liquidity needs are a long production process or the necessity to hold a significant amount of inventories in a given industry. A smooth operating performance in these industries would therefore generally require external funding provided by banks or stock markets (Raddatz 2006). Again, exporting activities magnify liquidity needs, because the time-intensive cross-border shipping, customs clearance, and product distribution in a foreign country 5 further increase the time lag between paying for the purchased inputs and receiving payments for the sold products. Asset intangibility. Firms in certain industries utilize a high share of intangible assets in their production process. The examples of such assets include intellectual property, client lists, blueprints and building designs, brand recognition, human and organizational capital, etc. Intangible assets are less suitable to serve as collateral because they are less durable and can be more easily stolen by competitors or employees. Intangible assets also have a limited liquidation value, both because they are often industry-specific or even firm-specific, and because the management can more easily hide them from the providers of external finance in case of the firm’s default. The lower collateral and liquidation value of intangible assets present an important obstacle for obtaining the necessary external finance (Shleifer and Vishny 1992; Rampini and Viswanathan 2010; Falato et al. 2020). Importantly, it presents a bigger problem in financially underdeveloped countries charac- terized by less efficient monitoring and screening of borrowers and worse protection of the rights of investors and debtholders. Firms operating in financially underdeveloped coun- tries therefore usually need to possess easily collateralizable tangible assets like plants and machinery in order to obtain the necessary external financing. Consequently, indus- tries relying for technological reasons on a high share of intangible assets would benefit disproportionately more from being located in a financially developed country (Braun 2003; Claessens and Laeven 2003). Given the higher need for external finance in cross- border trade, this would arguably be even more the case when it comes to the exporting activities of such industries. Trade credit. Borrowing from business partners in the form of trade credit can serve as a substitute for external financing by banks and stock markets. Therefore, firms in indus- tries that rely on trade credit financing should benefit comparatively less from the devel- opment of formal financial institutions and markets (Fisman and Love 2003). However, this alternative financing channel often relies on well-established trust between buyers and suppliers of given products. When it comes to production growth of industries dependent on trade credit, the effect is therefore more visible in the growth of already established 6 firms rather than in entry of new firms (Fisman and Love 2003). Well-established trust between providers and recipients of trade credit is arguably even more important in in- ternational trade where higher risk and longer time spans between production and final delivery put an additional burden on the business partner who provides the trade credit. One would therefore expect trade credit to matter especially for the long-term export survival of products already established in the foreign markets and to matter rather less for the immediate survival of new products that have just entered these markets. Global growth opportunities. Fisman and Love (2007) argue that a well-developed financial system increases an economy’s resource allocation towards firms and sectors that have better growth opportunities. They suggest this allocative transmission channel as an alternative to the channel of investment needs by Rajan and Zingales (1998) and demonstrate that global growth opportunities perform better than investment needs in explaining the impact of the overall financial development on industrial growth. We ex- amine how these results translate into the context of international trade while considering different time horizons of export survival and separate measures for the strength of banks and stock markets. 2.2 Estimation Strategy and Empirical Specifications An empirical model inspired by Araujo, Mion, and Ornelas (2016) provides us with a unified framework to explore export survival at different time horizons, allowing for the possibility that the examined transmission channels operate differently in the short term and long term. In contrast to the non-linear estimators like logit or probit, this linear probability model also facilitates the inclusion of full sets of diverse fixed effects that are an indispensable part of the identification strategy based on Rajan and Zingales (1998). This difference-in-difference estimation strategy employs proxies of financial vulnerability at the level of industrial sectors that capture the exogenous financing needs driven by technological characteristics of different industrial sectors including the scale of a typical investment project, length of the production process, or typical amount of inventories (see, 7 e.g., Rajan and Zingales 1998, Braun and Larrain 2005; Raddatz 2006).4 This represents the main advantage of these sectoral measures over firm-level financial variables. The data observed at the firm level are by their very nature a result of both the external financial constraints facing the firm and the endogenous financial decisions of the firm itself, making causal inferences difficult. Combining the trade survival framework of Araujo, Mion, and Ornelas (2016) with the difference-in-differences identification strategy inspired by Rajan and Zingales (1998) yields the following empirical model: l/t Skic0 = β1 F inancialV ulnerabilityi ∗ Banksc,t0 +β2 F inancialV ulnerabilityi ∗ StockM arketsc,t0 (1) +controlskic,t0 φ + δk + δc∗t0 + εkic,t0 , where the dependent variable is the probability of export survival of product k from industrial sector (industry) i exported by country c to the United States. The export survival probability is measured l years (l = 1, 5, 10) after the beginning of export spell (t0 ).5 Our main variables of interest are the interaction terms of financial vulnerability with banking and stock market development (F inancialV ulnerabilityi ∗ Banksc,t0 , Financial Vulnerability i *StockMarkets c,t0 ). The direct effects of the individual components of these interaction terms (F inancialV ulnerabilityi , Banksc,t0 , StockM arketsc,t0 ) are absorbed by the product and country-time fixed effects (δk , δc∗t0 ). We use different measures of financial vulnerability to explore different transmission channels from finance to export survival: investment needs, liquidity needs, and asset intangibility. Equation 1 thus comprises nine different specifications that look at the relative importance of banks and stock markets for export survival at three different time 4 Other finance papers using the difference-in-differences identification strategy of Rajan and Zingales (1998) include Cetorelli and Gambera (2001), Braun and Larrain (2005), Cetorelli and Strahan (2006), Kroszner, Laeven and Klingebiel (2007), Gupta and Yuan (2009), Strieborny and Kukenova (2016) and many others. Several papers have used similar methodology outside the survival framework when looking at financial or non-financial determinants of international trade. The former include Beck (200), Svaleryd and Vlachos (2005), Hur, Raj, and Riyanto (2006), Manova (2013), Fan, Lai, and Li (2015), Crino and Ogliari (2017), while the latter include Romalis (2004), Levchenko (2007) or Nunn (2007). 5 In this survival framework, we thus estimate a linear probability model for various lengths of export spells. See Araujo, Mion, and Ornelas (2016) for a more detailed and more technical discussion. 8 horizons alongside three different transmission channels. A positive coefficient β1 and/or β2 would suggest that well-developed banks and/or stock markets particularly promote export survival of products from industries that suffer from a given source of financial vulnerability. As an example, let us pick from the nine variants of Equation 1 the one looking at l=10/t0 export survival after 10 years (Skic ) and focusing on the high share of intangible and thus not easily collateralizable assets in the production process as the source of financial vulnerability.6 In this specification, an insignificant coefficient β1 for the interaction term AssetIntangibilityi ∗ Banksc,t0 and a positive and significant coefficient β2 for the inter- action term AssetIntangibilityi ∗ StockM arketsc,t0 would have the following economic interpretation: It is only well-developed stock markets but not well-developed banks that promote the long-term export survival of financially vulnerable products from industries that for technological reasons rely on a high share of intangible assets. In some estimations, we also control for two additional transmission channels from finance to the real economy that are related to the firm-to-firm trade credit as an alter- native source of external finance (Fisman and Love 2003) and to the allocative role of the financial sector (Fisman and Love 2007). All regressions are estimated by OLS with robust standard errors clustered at the country-time level (c ∗ t0 ), where country c refers to the exporting country and time t0 refers to the beginning of a given export spell. All estimations contain a full set of product-level fixed effects (δk ) and fixed effects at the country-time level (δc∗t0 ). The product-level fixed effects control for all product characteristics that could affect the probability of export survival (e.g., large weight or volume of a product increasing the shipping costs). The product fixed effects also absorb the direct effects of all industry characteristics entering the interaction terms included in the regressions (investment needs, liquidity needs, asset intangibility, physical and hu- man capital intensity). The country-time fixed effects (i.e., the interacted fixed effects of the exporting country c and the year t0 when a given export spell started) control for 6 l=10/t This variant of Equation 1 thus writes: Skic 0 = β1 AssetIntangibilityi ∗ Banksc,t0 + β2 AssetIntangibilityi ∗ StockM arketsc,t0 + controlskic,t0 φ + δk + δc∗t0 + εkic,t0 9 all time-varying characteristics of the exporting countries that could affect the chances for the subsequent export survival. Consequently, they absorb the direct effects of those time-varying country characteristics that enter our interaction terms (various dimensions of bank and stock market development, GDP per capita, real exchange rate, endowments of the exporting country with physical and human capital). The country-time effects also control for all time-invariant country characteristics potentially affecting export perfor- mance (country size, access to the sea, etc.) and time-varying conditions in the world market (global business cycle, technological progress reducing shipping costs, etc.).7 Our focus on a single destination market allows for a cleaner identification strategy that relies on a more parsimonious set of fixed effects when controlling for potential sources of the omitted variable bias. In our setting, the included (exporting)country-time fixed effects also control for time-varying origin-destination characteristics like political conflicts or deepening economic interdependence between the exporting and the importing country. This alleviates the need for including complex origin-destination-time fixed effects into our regressions. Similarly, the included product effects also control for various observable and unobservable destination-product characteristics. Any remaining omitted variable bias affecting the estimates for β1 and β2 would need to work simultaneously via time-varying country characteristic correlated with our measures of bank and stock market development and via industry characteristic correlated with our measures of financial vulnerability. To take care of this possibility, we include in all regressions the interactions of financial vulnerability measures with both GDP per capita and real exchange rate of the exporting country, as well as the interactions of physical (human) capital intensity in a given industry with physical (human) capital endowment in a given country. The row vector of control variables (controlskic,t0 ) also includes product-level control variables that vary across countries and time: initial export, total export, number of suppliers, and a multiple spell dummy. 7 See also the discussion of the fixed effects in the context of the applied difference-in-differences me- thodology in Raddatz (2006, p. 682-683) or Cetorelli and Strahan (2006, p. 442-443). 10 3 Data Our unit of analysis is an export spell defined as the length (in years) that a country c exports a particular good k to the United States without interruption. There can be multiple spells for a given country-product pair if a country ceases and then re-starts exporting the same product to the US destination market. All time-varying explanatory variables are measured at the beginning of a given export spell. Table 10 in the Appendix shows summary statistics for the main variables used in our analysis. When controlling for all variables of interest, our final database consists of 252,147 export spells of 3,300 HS 6-digit products exported from 71 countries to the United States over the period 1995-2011. Online Appendix A provides the list of exporting countries in our sample. 3.1 Financial Vulnerability The standard empirical proxies capturing the different financial vulnerabilities have been developed in a series of papers that follow different industrial classifications at different levels of aggregation. Rajan and Zingales (1998) compute investment needs for 36 different sectors at the level of 3-digit and 4-digit ISIC classification, Claessens and Laeven (2003) compute asset intangibility for 20 sectors following the SIC 2-digit classification, and Raddatz (2006) computes liquidity needs for 70 sectors at the ISIC 4-digit level. Fisman and Love (2003, 2007) follow the definition of 36 ISIC sectors used by Rajan and Zingales (1998) when computing their measures of trade credit dependence and global growth opportunities. We recompute these five measures at a significantly more disaggregated level of 192 industrial sectors (120 4-digit and 72 3-digit sectors according to the SIC classification). After merging these measures with our trade data and other control variables, the final sample comprises 174 industrial sectors (108 4-digit and 66 3-digit SIC sectors). Defining financial vulnerability across narrowly defined sectors not only allows for a more precise measurement, but it also represents a better fit with the highly disaggregated data on 11 international trade described in Subsection 3.3.8 To ensure the comparability of different measures capturing different aspects of financial vulnerability, we compute the measures in a unified way and normalize them to be between zero and one.9 Following Rajan and Zingales (1998), we compute Investment needs as the difference between capital expenditures (Compustat item 128) and cash flow (Compustat item 110), divided by capital expenditures.10 This proxy captures the share of the necessary invest- ment into physical capital that cannot be financed from internally generated funds and therefore requires external financing. The reliance on large US firms covered in Compu- stat in the computation assures that the resulting share reflects the investment that a representative firm in the industry truly requires but cannot finance itself. For smaller US firms and most other firms around the world, the observed use of external financing represents an equilibrium outcome between the firm’s demand and the available supply from local banks and stock markets. By contrast, large publicly traded firms in the US face only minimal frictions in accessing external finance so that the observed amount of their external funding represents a good measure of their actual financing needs (Rajan and Zingales 1998). Following Raddatz (2006), we compute Liquidity needs as the median ratio of total inventories to sales (i.e. dividing Compustat item 3 by Compustat item 12). A higher 8 The use of more broadly defined sectors in the original papers fits with the prevailing aggregate focus of the finance-growth literature at that time. Existing trade literature has mostly followed these broader industries when using the difference-in-difference identification strategy adopted in this paper. More recently, both finance and trade papers using disaggregated data have started to recompute some of these measures for narrowly defined industrial sectors. Crin` o and Ogliari (2017) recompute investment needs and asset tangibility for 273 SIC industries in their study on financial frictions and product quality in international trade. Tong and Wei (2011) recompute investment needs and liquidity needs for around 250 SIC industries when investigating the impact of financial crisis on the firm-level stock market returns. 9 In the first step, we compute median value of a given measure for each US firm in Compustat over the period 1989-2006. In the second step, we look at the computed measures for all firms in a given industry and take the value of the median firm to represent the industry, following the approach of Raddatz (2006). Other approaches used in the original papers include taking mean rather than median in the first step or eliminating the two-steps procedure altogether and take an average value across all firm-year observations in a given industry in just one step. More fundamentally, none of the measures in the original papers is normalized in order to allow for better comparability with other measures. This is only natural given that each of these papers focuses primarily on one particular transmission channel and the measure of financial vulnerability associated with it. 10 Compustat reports data for different firms and years in different formats. For formats 1/2/3, we compute the cash flow as the sum of following variables: funds from operations (operating income + depreciation), decreases in inventories, decreases in receivables, and increases in payables. For format 7, we use the variable cash flow directly provided in Compustat. 12 value for this ratio means a lower share of investment in inventories that can be financed by ongoing revenue, suggesting a higher level of liquidity needs that have to be financed externally. The reason for focusing on inventories rather than the overall investment in working capital is to capture the technologically driven exogenous part of liquidity needs. The completion of final goods take longer in certain industries, requiring higher inventory values during the production process. This technological link is arguably weaker for other parts of liquid assets like cash (Raddatz 2006, p. 685). Following Claessens and Laeven (2003), we measure Asset intangibility as the ra- tio of the net value of intangible assets (Compustat item 33) to the net fixed assets (Compustat item 8).11 In Compustat, the intangibles include the value of blueprints or building designs, patents, copyrights, trademarks, franchises, organization costs, client lists, computer software patent costs, licenses, and goodwill. The ratio thus measures for a representative firm in a given industry the relative importance of assets that are less suitable to serve as collateral due to their lack of physical manifestation as opposed to the easily collateralizable fixed assets. The proxies for our two additional transmission channels also rely on Compustat data on large US firms. Following Fisman and Love (2003), we compute Trade credit depen- dency as the ratio of accounts payable (Compustat item 70) to total assets (Compustat item 6). The ratio represents the share of total assets that is financed by trade credit, capturing the ability of a representative firm in a given industry to rely on informal credit from its business partners rather than on formal financing from financial markets and intermediaries. We measure Global growth opportunities by the actual growth in real sales for the representative firm in a given industry in the US. This proxy is based on the argument outlined in Fisman and Love (2007) that large publicly traded US firms respond optimally to worldwide industry-specific shocks to growth opportunities, making the actual growth rate of these firms a good proxy for these growth opportunities. Table 1 shows that all five measures are only weakly and often negatively correlated with each other. The measures thus seem to represent genuinely distinct transmission 11 The net fixed assets equal gross property, plan and equipment (Compustat item 7) minus accumulated depreciation, depletion and amortization (Compustat item 196). 13 Table 1: Correlations among the measures of financial vulnerability Investment needs Liquidity needs Asset intangibility Growth opp. Trade credit Investment needs 1 Liquidity needs -0.0497 1 Asset intangibility -0.1822 -0.1346 1 Growth opportunities 0.2399 0.0512 -0.1042 1 Trade credit -0.0391 -0.0693 -0.1704 0.0256 1 The table reports correlations among five measures of financial vulnerability described in Subsection 3.1. All measures are computed at the level of the 4-digit and 3-digit sectors according to the SIC classification. channels from finance to the real economy. A closer look at the exposure of individual industries to the three main sources of financial vulnerability further reinforces this point. For example, the production process in the cigarettes industry (SIC code 2111) is highly dependent on intangible assets. At the same time, the cigarettes industry generates high levels of cash flow, which makes it less dependent on external funding for its investment needs. Another example involves the two industries representing publishing and printing of newspapers (SIC code 2711) and periodicals (SIC code 2721) that face low liquidity needs but use high levels of intangible assets in their production process. Online Appendix B provides more details by listing 15 industries with both the lowest and the highest levels of investment needs, liquidity needs and asset intangibility.12 Given that all five variables are measured at the industry level, their direct effects are captured in our regressions by the included product fixed effects (δk ). Consequently, the variables described in this subsection enter the regressions explicitly only as part of interaction terms with the measures capturing the strength of financial institutions and markets that are described in the next subsection. 3.2 Financial Institutions and Markets Our main measure capturing the strength of financial institutions in the exporting country is called Banks. This standard measure of a well-developed banking sector represents the ratio between credit provided by deposit-taking banks to the private sector and GDP of a given country. In additional tests, we also use two alternative measures looking at the broader role played by financial institutions in the real economy. Bank assets represent the value of the claims held by deposit-taking banks vis-a-vis the domestic real 12 As we normalize all measures to be between zero and one, the value for the least (most) exposed industry for every measure is zero (one). 14 non-financial sector, again normalized by the GDP level in a given country. Total credit represents the ratio of all credit to the private sector over GDP. The variable thus includes credit provided both by deposit-taking banks and by other financial institutions. Our main measure of the depth of financial markets in the exporting country is called Stock markets. It represents the ratio between stock market capitalization (the market value of all listed stocks) and GDP of a given country, and it is a standard measure of well-developed stock markets. Stock market turnover (the ratio between the value of stock market transactions and the stock market capitalization) serves as a proxy for an active and liquid (rather than large) stock market. In additional tests, we also use Stock market value traded (the value of stock market transactions divided by GDP). The data are from the Global Financial Development Database that builds on the ˇ ak et al. earlier work by Beck et al. (2000) and has been substantially extended by Cih´ (2012). The database contains various indicators of financial development across countries and over time and is regularly updated. All measures representing financial institutions and markets vary over time. We mea- sure them at the beginning of a given export spell in order to capture the impact of the predetermined financial environment on the subsequent export performance. The direct effects of these time-varying country-level financial variables are thus captured in our regressions by the included country-time fixed effects (δc∗t0 ), with time referring to the beginning of a given export spell (t0 ). Consequently, the variables described in this subsection enter the regressions explicitly only as part of interaction terms with the industry-level measures of financial vulnerability. 3.3 Product-Related Trade Variables The export survival rate in the US market and all trade-related control variables at the product level (initial export, total export, number of suppliers, a multiple spell dummy) are computed using the BACI dataset developed by the CEPII and described in Gaulier and Zignago (2010). The dataset provides harmonized bilateral trade flows for more than 5,000 HS 6-digit products and 143 countries. Export flows are reported annually 15 in values (thousands of US dollars) and quantities.13 Trade data at the HS 6-digit level allows us to account for export failures at a detailed product level which would otherwise be undetected when using more aggregated industry data. In our estimations, we focus on the manufacturing exports to the United States during the 1995-2011 period.14 The three time-varying variables (initial export, total export, number of suppliers) are measured at the beginning of the export spell. The terms ‘given year’ or ‘time’ thus refer to the initial year of a given export spell. Initial export is measured at the product- country-time level and represents the export value of a given product exported by a given country to the US market in the initial year of exporting. The variable captures the levels of mutual trust and expectations about the potential future exports that exist among the trade partners at the beginning of a new trade relationship. Total export is measured at the product-country-time level and represents the overall value of the country’s exports of a given product to the world market in a given year. It serves as a proxy for the size and overall export performance of a given exporting country in that given product. Number of suppliers is measured at the product-time level and counts the number of countries that export a given product in a given year to the United States. The variable can represent both the level of competition and the potential market size for a given product in the US destination market. A dummy variable called Multiple spell is measured at the product-country level. It equals one if a given country exports a given product to the United States during two or more separate export spells, identifying instances when the country at some point stops exporting a product but later re-enters the US market with the same product again. The inclusion of this variable accounts for the possibility that repeated exits and re-entries into exporting of the same product to the same destination might affect this product’s chances for export survival at various time horizons. 13 We do not need export values for measuring the export survival rate but we do use them for con- structing some of the trade-related control variables. 14 We start our sample period in 1995 due to the high number of missing values before 1994. In particular, we are using BACI in HS classification from 1992 that covers the period 1994-2011. As the survival analysis relies on the length of export spells, we cannot use the data from the initial year. This leaves us with the data for 1995-2011. We finish our sample period in 2011 to match the time period for which we have data for all our industry-level variables. 16 3.4 Other Control Variables We include in our regressions also several country-level and industry-level variables from the real economy. The real GDP per capita is reported in constant 2005 US dollars. Real exchange rate is computed as the nominal effective exchange rate (a measure of the value of a currency against a weighted average of several foreign currencies) divided by a price deflator or index of costs. The source for both variables is the World Development Indicators of the World Bank. We also control for countries’ endowments with physical and human capital, relying on data from the Penn World Table (version 8.1). The stock of physical capital per worker in a given country is constructed according to the perpet- ual inventory method. Human capital per worker is calculated from the average years of schooling in a given country using attainment data. These time-varying economic charac- teristics of exporting countries are measured at the beginning of a given export spell. We also include two time-invariant industry-level characteristics - physical capital intensity and human capital intensity. Both variables are from Braun (2003) and are measured at the ISIC 3-digit level. The direct effects of these real-economy variables are absorbed by the country-time and product fixed effects included in our empirical specifications. However, these control variables do explicitly enter the regressions as parts of various interaction terms. 4 Banks vs Stock Markets This section relies on standard measures of well-developed banks and stock markets when examining their relative importance for export survival at different time horizons. Sub- section 4.1 provides preliminary graphical evidence, both by capturing the overall impact of banks and stock markets on export survival, and by further exploring the transmission channels related to high investment needs, high liquidity needs, and asset intangibility. Subsection 4.2 reports full results from additional regressions focusing on these three channels from finance to export survival (Equation 1). 17 4.1 Graphical Evidence Figure 1a and Figure 1b report the Kaplan-Meier survival functions that capture the probability of export spells (continuous exporting of a given product from a given country to the US market) to survive after year 1,2, etc. The survival probability in the first year is one by default given the annual frequency of the data. Figure 1a compares the average survival in the US market for products exported from countries with a well- developed versus underdeveloped domestic banking sector. The solid line captures the export survival for products from countries at the 25th percentile of banking development while the dashed line captures the export survival for products from countries at the 75th percentile of banking development. Analogously, Figure 1b compares average export survival for products from countries at the 25th percentile versus the 75th percentile of the stock market development. In both figures, the dashed line is located above the solid line, suggesting that products have better survival chances in the US market if they are exported from countries with a strong domestic banking sector or a well-developed stock market. Figure 1: Export survival and financial development (a) Banking development (b) Stock market development 1.00 1.00 0.75 0.75 Survival probability Survival probability 0.50 0.50 0.25 0.25 0.00 0.00 0 5 10 15 20 0 5 10 15 20 Years elapsed Years elapsed Low High Low High Note: The solid line (denoted ‘Low’) captures the export survival for products from countries at the 25th percentile of banking (stock market) development. The dashed line (denoted ‘High’) captures the export survival for products from countries at the 75th percentile of banking (stock market) development. Next, we explore the specific transmission channels that could explain the reduced- form relationship between finance and export survival from Figure 1a and Figure 1b. Figure 2 examines these transmission channels in the context of strong banks, and Fig- 18 ure 3 examines them in the context of well-developed stock markets. In particular, we regress the export survival at the product level after year 1,2,...,10 on the interaction terms between three measures of financial vulnerability (investment needs, liquidity needs, asset intangibility) and stock market/banking development. We then plot the coefficient esti- mates of these interaction terms together with the bounds of their 90 per cent confidence interval. Figure 2 and Figure 3 thus visually summarize a series of regression results about the impact of banks and stock markets on export survival. Unlike the uncondi- tional export survival analyzed in Figure 1a and Figure 1b, the regressions summarized in Figure 2 and Figure 3 also control for fixed effects at the (exporting country)*time and product level as well as for several trade-related control variables (initial export, total export, number of suppliers, multiple spell dummy). Figure 2: Export survival and banks: The transmission channels Note: Each of the three graphs summarizes ten regressions representing one transmission channel from banking development to export survival at ten different time horizons. We regress the export survival at the product level after year 1,2,...,10 on the interaction term between a given measure of financial vulnerability (investment needs, liquidity needs, asset intangibility) and the banking development in the exporting country. The horizontal axis in the graphs represents the ten different time horizons of export survival, the solid line connects the ten coefficient estimates for the corresponding interaction term, and the shaded area represents the 90 per cent confidence interval for these point estimates. While the graphs focus on the visual summary of the coefficient estimates for the main interaction terms, the underlying regressions control also for (exporting country)*time and product fixed effects, as well as for several trade-related control variables (initial export, total export, number of suppliers, multiple spell dummy). Robust standard errors are clustered at the (exporting country)*time level. The results in Figure 2 and Figure 3 suggest that all three transmission channels play an important role in driving the reduced-form relationship from Figure 1a and Figure 1b. The effects are generally significant, ranging from the very short-term export survival of 19 one year up to the long-term survival of ten years after the beginning of an export spell. The exceptions are the insignificant effects of both banks and stock markets for exports from industries with high investment needs in the shorter term as well as the insignificant effects of stock markets on exports from industries with high liquidity needs in the longer term.15 Figure 3: Export survival and stock markets: The transmission channels Note: Each of the three graphs summarizes ten regressions representing one transmission channel from stock market development to export survival at ten different time horizons. We regress the export survival at the product level after year 1,2,...,10 on the interaction term between a given measure of financial vulnerability (investment needs, liquidity needs, asset intangibility) and the stock market development in the exporting country. The horizontal axis in the graphs represents the ten different time horizons of export survival, the solid line connects the ten coefficient estimates for the corresponding interaction term, and the shaded area represents the 90 per cent confidence interval for these point estimates. While the graphs focus on the visual summary of the coefficient estimates for the main interaction terms, the underlying regressions control also for (exporting country)*time and product fixed effects, as well as for several trade-related control variables (initial export, total export, number of suppliers, multiple spell dummy). Robust standard errors are clustered at the (exporting country)*time level. 15 One should read the graphs in the following way. If for a given year 1,2,...,10, the whole confidence interval is above zero, then the estimated coefficient for the corresponding main interaction term is positive and statistically significant for that year. For example, in the first graph of Figure 2, the whole confidence interval is above zero in year 10, meaning that the coefficient of the interaction term between investment needs and banking development is positive and significant in a regression examining the ten-year export survival at the product level. This suggest that banking development over-proportionately promotes export survival of products with high investment needs at the horizon of ten years. In year 1 on the same graph, the coefficient line is above zero but the confidence interval spreads both above and below zero. The estimated coefficient is thus positive but not statistically different from zero, suggesting that banking development does not play a particular role in promoting one-year export survival of products with high investment needs. The transmission mechanism tested in the first graph of Figure 2 thus seems to play an important role for the export survival in the long term but not in the short term. 20 4.2 Main Channels This subsection provides a further analysis of the three transmission channels from well- developed financial markets and institutions to the export survival at the product level. Additionally to the fixed effects and trade-related control variables included in regressions summarized by Figure 2 and Figure 3, the regressions reported both in this subsection and in the remainder of the paper control also for non-financial channels related to the Heckscher-Ohlin forces of comparative advantage in international trade, the economic development, the business cycle fluctuations, and changes in the real exchange rate. As we now report extensive regression tables rather than summarizing graphs, we do not provide all results for the export survival after 1,2,...,10 years but focus on the export survival at the one, five, and ten year time horizon. Table 2 looks at the transmission channel related to products from industries with high investment needs. In columns (1) to (3) of Table 2, we look whether a well-developed banking system particularly promotes such products, taking as a dependent variable the export survival after one, five, and ten years. In columns (4) to (6) of Table 2, we focus on the importance of stock markets, again looking at the product-level export survival after one, five, and ten years. A positive and significant coefficient for the main interaction term (investment needs interacted with banks in the first three columns, investment needs interacted with stock markets in the last three columns) would confirm that providing external finance for investment into physical capital is an important transmission channel from finance to export survival. We also control for the non-financial transmission chan- nels related to economic development and currency value, by interacting our measure of investment needs with both GDP per capita and real exchange rate. Furthermore, we include other control variables that could affect the probability of export survival at the product level - initial export, total export, number of suppliers, multiple spell dummy, and interactions of physical (human) capital intensity at the industry level with physical (human) capital endowment at the country level. The last two interaction terms are based on the standard Heckscher-Ohlin theory of international trade and control for the importance of countries’ factor endowments in shaping the international trade flows. 21 Table 2: Banks, stock markets, and investment needs (1) (2) (3) (4) (5) (6) Dep. var.: Survival 1y 5y 10y 1y 5y 10y investment needs × banks -0.011 0.034 0.056b (0.047) (0.033) (0.027) investment needs × stock markets 0.034 0.062b 0.057a (0.040) (0.026) (0.020) investment needs × GDPpc -0.004 -0.005 -0.010 0.006 -0.002 0.002 (0.018) (0.012) (0.010) (0.017) (0.011) (0.009) investment needs × real exchange rate -0.642 -0.396 -0.407c -1.933b -0.703 -0.420 (0.464) (0.258) (0.215) (0.874) (0.638) (0.528) initial export 0.020a 0.016a 0.012a 0.020a 0.016a 0.012a (0.002) (0.002) (0.001) (0.002) (0.002) (0.001) total export 0.041a 0.032a 0.024a 0.041a 0.032a 0.024a (0.001) (0.001) (0.001) (0.001) (0.001) (0.001) number of suppliers 0.003a -0.001a -0.004a 0.003a -0.001b -0.004a (0.000) (0.000) (0.000) (0.000) (0.001) (0.000) multiple spell -0.042a -0.268a -0.351a -0.044a -0.274a -0.360a (0.008) (0.015) (0.021) (0.008) (0.016) (0.022) phys. cap. intensity × physical capital -0.139b -0.166a -0.100b -0.047 -0.080 -0.023 (0.067) (0.047) (0.040) (0.079) (0.053) (0.043) hum. cap. intensity × human capital 0.274a 0.265a 0.221a 0.279a 0.289a 0.244a (0.027) (0.021) (0.019) (0.031) (0.024) (0.021) Observations 252,147 252,147 252,147 243,509 243,509 243,509 R-squared 0.263 0.546 0.692 0.254 0.547 0.695 Country-Time FE yes yes yes yes yes yes Product FE yes yes yes yes yes yes Dependent variable is the probability of export survival of product k from industrial sector (industry) i exported by country c to the USA. Export survival probability is measured l years after the beginning of export spell, with l = 1 in columns (1) and (4), l = 5 in columns (2) and (5), and l = 10 in columns (3) and (6). The regressions are estimated by OLS and contain a full set of fixed effects at the product level and the (exporting country)*time level, with time referring to the beginning of a given export spell. Investment needs is defined at the industry level i and represents the difference between capital expenditures and cash flow, divided by capital expenditures. Banks represents the ratio between credit from deposit-taking banks to the private sector and GDP in a given exporting country c. Stock markets is the ratio between stock market capitalization and GDP of a given exporting country c. Other variables entering regressions directly or as a part of interaction terms include GDPpc (GDP per capita in country c reported in constant 2005 US dollars), real exchange rate (nominal effective exchange rate divided by price deflator or index of costs of country c), initial export (export value of a product k exported by country c to the US in the initial year of exporting), total export (value of all exports from country c to the world market), number of suppliers (number of countries exporting product k to the US), multiple spell (dummy variable that equals one if country c exports product k to the US during more than one export spell), and interaction terms between physical and human capital endowments of country c and the corresponding capital intensities at the industry level. All time-varying explanatory variables are measured at the beginning of the export spell. Robust standard errors are clustered at the (exporting country)*time level, with time referring to the beginning of a given export spell. a , b , c denote statistical significance at the 1%, 5%, and 10% level, respectively. The results reported in Table 2 suggest that stock markets play a more robust role than banks when it comes to facilitating exports of products with high investment needs. While banks have a significant role for the export survival only at the longest horizon of ten years (third column), stock markets promote export survival of these products at horizons of both five and ten years (fifth and sixth column). As for the non-financial channels, economic development seems to play no particular role in promoting exports of products with high investment needs as documented by the insignificant results for the interaction 22 term of investment needs and GDP per capita. The interaction term of investment needs and real exchange rate is also insignificant in four out of six specifications.16 Table 3 explores the transmission channel related to high liquidity needs of products from industries that require large amount of working capital. Again, we explore separately the role of banks (first three columns) and stock markets (last three columns), looking at export survival over the horizons of one, five, and ten years as our dependent variable. The results suggest that strong banks play a somewhat more important role than deep stock markets in promoting export survival of products with high liquidity needs, especially in the longer term. In the case of the immediate survival over a one-year horizon, both banks and stock markets promote exports of products with high liquidity needs as attested by positive and significant main interaction terms both in column (1) and in column (4) of Table 3. At the time horizon of five years, the significance level for the main interaction term is higher for banks in column (2) than for stock markets in column (5) of Table 3. And when it comes to the very long-term survival of ten years reported in columns (3) and (6) of Table 3, only banks seem to play a significant role. As for the non-financial channels, the interaction term including real exchange rate is insignificant in four out of six specifications, similarly to the previous table. However, the interaction term of liquidity needs and GDP per capita is positive and significant in all six columns of Table 3, suggesting that a higher level of economic development over-proportionately benefits the exports of products that require external finance to fund their working capital. Table 4 examines the transmission channel linked to products from industries with a high share of intangible assets. When it comes to promoting exports of such products, it is only deep stock markets and not strong banks that matter. The interaction term between asset intangibility and banks is insignificant in the first three columns, while the interaction term of asset intangibility with stock markets is positive and highly signifi- cant in the last three columns of Table 4. As for the non-financial channels, economic development does not seem to play a particular role in promoting exports from industries 16 These insignificant results do not imply that economic development and the real exchange rate do not have any impact on export performance in general. The results for the two interaction terms in our difference-in-difference framework merely suggest that this impact does not vary between products with high versus low investment needs. 23 Table 3: Banks, stock markets, and liquidity needs (1) (2) (3) (4) (5) (6) Dep. var.: Survival 1y 5y 10y 1y 5y 10y liquidity needs × banks 0.103a 0.082a 0.050b (0.039) (0.029) (0.022) liquidity needs × stock markets 0.127a 0.052b -0.009 (0.026) (0.024) (0.020) liquidity needs × GDPpc 0.059a 0.035a 0.028a 0.061a 0.043a 0.038a (0.013) (0.012) (0.010) (0.013) (0.011) (0.009) liquidity needs × real exchange rate -1.246b -0.362 -0.638 -1.392c -0.317 -0.415 (0.596) (0.504) (0.394) (0.722) (0.586) (0.470) initial export 0.020a 0.016a 0.012a 0.020a 0.016a 0.012a (0.002) (0.002) (0.001) (0.002) (0.002) (0.001) total export 0.041a 0.032a 0.024a 0.041a 0.032a 0.024a (0.001) (0.001) (0.001) (0.001) (0.001) (0.001) number of suppliers 0.003a -0.001a -0.004a 0.003a -0.001b -0.004a (0.000) (0.000) (0.000) (0.000) (0.001) (0.000) multiple spell -0.042a -0.268a -0.352a -0.044a -0.274a -0.360a (0.008) (0.015) (0.021) (0.008) (0.016) (0.022) phys. cap. intensity × physical capital -0.056 -0.110b -0.060 0.027 -0.028 0.016 (0.069) (0.050) (0.042) (0.081) (0.059) (0.046) hum. cap. intensity × human capital 0.265a 0.259a 0.218a 0.273a 0.285a 0.241a (0.027) (0.021) (0.019) (0.031) (0.024) (0.022) Observations 252,147 252,147 252,147 243,509 243,509 243,509 R-squared 0.264 0.546 0.692 0.254 0.547 0.695 Country-Time FE yes yes yes yes yes yes Product FE yes yes yes yes yes yes Dependent variable is the probability of export survival of product k from industrial sector (industry) i exported by country c to the USA. Export survival probability is measured l years after the beginning of export spell, with l = 1 in columns (1) and (4), l = 5 in columns (2) and (5), and l = 10 in columns (3) and (6). The regressions are estimated by OLS and contain a full set of fixed effects at the product level and the (exporting country)*time level, with time referring to the beginning of a given export spell. Liquidity needs represents the median ratio of total inventories to sales in industry i. Other variables are described in Table 2. All time-varying explanatory variables are measured at the beginning of the export spell. Robust standard errors are clustered at the (exporting country)*time level, with time referring to the beginning of a given export spell. a , b , c denote statistical significance at the 1%, 5%, and 10% level, respectively. characterized by a high share of intangible assets, as the corresponding interaction term is marginally significant only in two out of six specifications. By contrast, the interaction term of asset intangibility and real exchange rate is negative and highly significant in all six columns of Table 4. This suggest that a strong domestic currency disproportionately hurts exports of those products whose manufacturing process heavily relies on intangible assets. A comparison of the results from Table 4 with Figure 2 and Figure 3 demonstrates the importance of controlling for alternative determinants of export survival like economic development, exchange rate or factor endowments. According to Figure 2 and Figure 3, both strong banks and deep stock markets promote exports of products from industries with a high share of intangible assets. Stock markets maintain this beneficial impact also 24 after controlling for the differential impact of GDP per capita, real exchange rate, and countries’ endowments with physical and human capital in columns (4) to (6) of Table 4. By contrast, the positive impact of banks on export survival of products from industries with a high share of intangible assets completely disappears when we control for the same additional variables in columns (1) to (3) of Table 4. Table 5 runs a direct horse race between banks and stock markets for each of the three main transmission channels, allowing for a further scrutiny of the results reported in the previous tables. Columns (1) to (3) of Table 5 correspond to Table 2, looking at products requiring external finance for investment into physical capital. Columns (4) to (6) of Table 5 correspond to Table 3, focusing on products requiring external finance for working capital. Columns (7) to (9) of Table 5 correspond to Table 4, looking at products whose manufacturing process relies on intangible assets. Like in the previous three tables, the dependent variable is probability of export survival after one, five, and ten years, but the main interaction terms including banks and stock markets enter the regression simul- taneously rather than separately. We also control for the same non-financial transmission channels and additional variables as in Table 2, Table 3, and Table 4. The results in Table 5 are in accordance with results from the previous three tables. Stock markets seem to play a more important role in promoting exports with high invest- ment needs (first to third columns of Table 5) while banks seem to be more important for export survival of products with high liquidity needs (fourth to sixth columns of Table 5). These two sets of results apply especially when it comes to export survival at longer hori- zons. And it is only deep stock markets and not strong banks that promote export survival of products from industries with a high share of intangible assets, independently on the examined time horizon (seventh to ninth columns of Table 5). The results for non-financial channels are also in accordance with the previous tables. Economic development disproportionately promotes export survival for products with high liquidity needs (fourth to sixth columns of Table 5), and a strong domestic currency disproportionately hurts export performance of products from industries with a high share of intangible assets (seventh to ninth columns of Table 5). 25 Table 4: Banks, stock markets, and asset intangibility (1) (2) (3) (4) (5) (6) Dep. var.: Survival 1y 5y 10y 1y 5y 10y asset intangibility × banks 0.067 0.065 0.036 (0.056) (0.043) (0.032) asset intangibility × stock markets 0.213a 0.181a 0.087a (0.039) (0.030) (0.028) asset intangibility × GDPpc 0.017 0.025c 0.015 0.014 0.022c 0.011 (0.018) (0.015) (0.013) (0.017) (0.013) (0.011) asset intangibility × real exchange rate -2.114a -2.055a -1.534a -2.829a -3.340a -2.298a (0.759) (0.728) (0.558) (1.049) (0.775) (0.614) initial export 0.020a 0.016a 0.012a 0.020a 0.016a 0.012a (0.002) (0.002) (0.001) (0.002) (0.002) (0.001) total export 0.041a 0.032a 0.024a 0.041a 0.032a 0.024a (0.001) (0.001) (0.001) (0.001) (0.001) (0.001) number of suppliers 0.003a -0.001a -0.004a 0.003a -0.001b -0.004a (0.000) (0.000) (0.000) (0.000) (0.001) (0.000) multiple spell -0.042a -0.268a -0.351a -0.044a -0.274a -0.360a (0.008) (0.015) (0.021) (0.008) (0.016) (0.022) phys. cap. intensity × physical capital -0.123c -0.146a -0.089b -0.033 -0.063 -0.016 (0.067) (0.046) (0.039) (0.079) (0.054) (0.043) hum. cap. intensity × human capital 0.268a 0.258a 0.217a 0.274a 0.282a 0.241a (0.027) (0.021) (0.019) (0.031) (0.024) (0.021) Observations 252,147 252,147 252,147 243,509 243,509 243,509 R-squared 0.263 0.546 0.692 0.254 0.547 0.695 Country-Time FE yes yes yes yes yes yes Product FE yes yes yes yes yes yes Dependent variable is the probability of export survival of product k from industrial sector (industry) i exported by country c to the USA. Export survival probability is measured l years after the beginning of export spell, with l = 1 in columns (1) and (4), l = 5 in columns (2) and (5), and l = 10 in columns (3) and (6). The regressions are estimated by OLS and contain a full set of fixed effects at the product level and the (exporting country)*time level, with time referring to the beginning of a given export spell. Asset intangibility is the ratio of the net value of intangible assets to the net fixed assets in industry i. Other variables are described in Table 2. All time-varying explanatory variables are measured at the beginning of the export spell. Robust standard errors are clustered at the (exporting country)*time level, with time referring to the beginning of a given export spell. a , b , c denote statistical significance at the 1%, 5%, and 10% level, respectively. For space reasons, we do not report in the remaining sections the coefficients for following control variables: initial export, total export, number of suppliers, multiple spell dummy, and interaction terms between physical and human capital endowments of country c and the corresponding capital intensities at the industry level. 5 Size vs Activity of Stock Markets The results reported in the previous section revealed several patterns regarding the im- portance of banks and stock markets for export survival at the product level. This section explores if and how these patterns change when we distinguish between large and active stock markets. In Table 6, we replace the traditional measure of deep stock markets - the normalized stock market capitalization (the market value of all listed shares divided by GDP) with 26 Table 5: The three main channels: Banks versus stock markets (1) (2) (3) (4) (5) (6) (7) (8) (9) Dep. var.: Survival 1y 5y 10y 1y 5y 10y 1y 5y 10y investment needs × banks -0.003 0.006 0.046c (0.052) (0.035) (0.028) investment needs × stock markets 0.032 0.064b 0.044b (0.044) (0.027) (0.020) investment needs × GDPpc 0.006 -0.003 -0.008 (0.019) (0.012) (0.010) investment needs × real exchange rate -1.944b -0.744 -0.646 (0.920) (0.657) (0.551) liquidity needs × banks 0.029 0.068b 0.063a (0.041) (0.032) (0.023) liquidity needs × stock markets 0.112a 0.029 -0.026 (0.028) (0.026) (0.020) liquidity needs × GDPpc 0.058a 0.036a 0.030a (0.014) (0.013) (0.010) liquidity needs × real exchange rate -1.456c -0.630 -0.668 (0.758) (0.637) (0.485) asset intangibility × banks -0.031 -0.026 -0.009 (0.062) (0.045) (0.035) asset intangibility × stock markets 0.224a 0.187a 0.091a (0.043) (0.032) (0.031) asset intangibility × GDPpc 0.015 0.025 0.012 (0.019) (0.016) (0.014) asset intangibility × real exchange rate -2.631b -3.129a -2.184a (1.115) (0.829) (0.662) initial export 0.019a 0.016a 0.012a 0.019a 0.016a 0.012a 0.019a 0.016a 0.012a (0.002) (0.002) (0.001) (0.002) (0.002) (0.001) (0.002) (0.002) (0.001) total export 0.041a 0.032a 0.024a 0.041a 0.033a 0.024a 0.041a 0.032a 0.024a (0.001) (0.001) (0.001) (0.001) (0.001) (0.001) (0.001) (0.001) (0.001) number of suppliers 0.003a -0.001b -0.004a 0.003a -0.001b -0.004a 0.003a -0.001b -0.004a (0.000) (0.001) (0.000) (0.000) (0.001) (0.000) (0.000) (0.001) (0.000) multiple spell -0.045a -0.276a -0.363a -0.046a -0.276a -0.363a -0.045a -0.276a -0.363a (0.008) (0.016) (0.022) (0.008) (0.016) (0.022) (0.008) (0.016) (0.022) phys. cap. intensity × physical capital -0.056 -0.099c -0.039 0.020 -0.044 0.003 -0.044 -0.082 -0.031 (0.080) (0.055) (0.044) (0.082) (0.060) (0.048) (0.079) (0.055) (0.044) hum. cap. intensity × human capital 0.284a 0.292a 0.248a 0.278a 0.287a 0.244a 0.279a 0.285a 0.244a (0.032) (0.025) (0.022) (0.032) (0.025) (0.022) (0.032) (0.024) (0.022) Observations 235,294 235,294 235,294 235,294 235,294 235,294 235,294 235,294 235,294 R-squared 0.259 0.548 0.698 0.259 0.548 0.699 0.259 0.548 0.699 Country-Time FE yes yes yes yes yes yes yes yes yes Product FE yes yes yes yes yes yes yes yes yes Dependent variable is the probability of export survival of product k from industrial sector (industry) i exported by country c to the USA. Export survival probability is measured l years after the beginning of export spell, with l = 1 in columns (1), (4) and (7), l = 5 in columns (2), (5) and (8), and l = 10 in columns (3), (6) and (9). The regressions are estimated by OLS and contain a full set of fixed effects at the product level and the (exporting country)*time level, with time referring to the beginning of a given export spell. All variables are described in Table 2, Table 3, and Table 4. All time-varying explanatory variables are measured at the beginning of the export spell. Robust standard errors are clustered at the (exporting country)*time level, with time referring to the beginning of a given export spell. a , b , c denote statistical significance at the 1%, 5%, and 10% level, respectively. the stock market turnover (the value of stock market transactions relative to the market value of all listed shares). While stock market capitalization measures the size of the stock market relative to the size of the economy, the stock market turnover captures liquidity or activity of a stock market in a given country. Otherwise, the specifications correspond to Table 5. Columns (1) to (3) of Table 6 explore the impact of active stock markets on the export survival of products with high investment needs. The interaction term including 27 Table 6: The three main channels: The role of active stock markets (1) (2) (3) (4) (5) (6) (7) (8) (9) Dep. var.: Survival 1y 5y 10y 1y 5y 10y 1y 5y 10y investment needs × banks 0.006 0.036 0.065b (0.050) (0.035) (0.028) investment needs × stock market turnover 0.043 0.015 0.025 (0.037) (0.028) (0.021) investment needs × GDPpc 0.009 -0.004 -0.008 (0.019) (0.013) (0.011) investment needs × real exchange rate -2.004b -0.684 -0.592 (0.925) (0.669) (0.562) liquidity needs × banks 0.065 0.061b 0.031 (0.040) (0.029) (0.023) liquidity needs × stock market turnover 0.090a 0.066a 0.043a (0.026) (0.021) (0.015) liquidity needs × GDPpc 0.056a 0.035a 0.030a (0.014) (0.013) (0.010) liquidity needs × real exchange rate -1.161 -0.384 -0.498 (0.763) (0.620) (0.479) asset intangibility × banks 0.076 0.060 0.030 (0.061) (0.043) (0.031) asset intangibility × stock market turnover 0.099a 0.079a 0.045b (0.036) (0.028) (0.018) asset intangibility × GDPpc 0.008 0.021 0.010 (0.019) (0.016) (0.013) asset intangibility × real exchange rate -2.126c -2.716a -1.967a (1.160) (0.859) (0.660) Observations 232,766 232,766 232,766 232,766 232,766 232,766 232,766 232,766 232,766 R-squared 0.259 0.549 0.700 0.260 0.549 0.701 0.259 0.549 0.701 Country-Time FE yes yes yes yes yes yes yes yes yes Product FE yes yes yes yes yes yes yes yes yes Full set of controls included yes yes yes yes yes yes yes yes yes Dependent variable is the probability of export survival of product k from industrial sector (industry) i exported by country c to the USA. Export survival probability is measured l years after the beginning of export spell, with l = 1 in columns (1), (4) and (7), l = 5 in columns (2), (5) and (8), and l = 10 in columns (3), (6) and (9). The regressions are estimated by OLS and contain a full set of fixed effects at the product level and the (exporting country)*time level, with time referring to the beginning of a given export spell. Stock market turnover is measured as the ratio between the value of stock market transactions and the stock market capitalization. Other variables are described in Table 2, Table 3, and Table 4. The full set of controls also includes initial export, total export, number of suppliers, multiple spell, and interaction terms between physical and human capital endowments of country c and the corresponding capital intensities at the industry level. All time-varying explanatory variables are measured at the beginning of the export spell. Robust standard errors are clustered at the (exporting country)*time level, with time referring to the beginning of a given export spell. a b c , , denote statistical significance at the 1%, 5%, and 10% level, respectively. stock market turnover is insignificant across all time dimensions. This contrasts with the previous results for the size of stock markets that did promote export survival of these products at the longer term horizons of five and ten years (second and third columns of Table 5). While active stock markets do affect the real economy in other contexts (e.g., Levine and Zervos 1998, Manova 2008), our results suggest that they do not play a significant role in promoting the export survival of products with high investment needs. Columns (4) to (6) of Table 6 focus on the role played by active stock markets in the export survival of products with high liquidity needs. In the previous specifications looking at large stock markets, banks dominated stock market capitalization in promoting the longer term export survival for these products (fifth and sixth columns of Table 5). The results are very different when looking at the activity of the stock market instead. 28 The interaction term of liquidity needs and stock market turnover is positive and highly significant across all time horizons (fourth to sixth columns of Table 6), while strong banks seem to have a significant effect only at the time horizon of five years (fifth column of Table 6). It is thus active rather than large stock markets that play the crucial role in promoting export survival of products with high liquidity needs. Finally, the last three columns of Table 6 examine the importance of active stock markets for the export survival of products whose manufacturing process requires a high share of intangible assets. Here the results are qualitatively the same as in the last three columns of Table 5. Whether we look at the size or the activity dimension, it is well- developed stock markets rather than banks that improve the export performance of these products across all time horizons. Comparing the results in Table 6 with Table 5 suggest that the relative impact of banks and stock markets often depends on the distinction between large and active stock markets. In the remainder of this section, we further explore this issue by running an alter- native horse race. Instead of looking at the relative importance of banks and large/active stock markets within individual transmission channels, we examine the relative impor- tance of the three main transmission channels for the overall impact of banks, large stock markets, and active stock markets. Besides the main interaction terms, each specification in Table 7 also controls for interaction terms of the three measures of financial vulnerabil- ity with GDP per capita and with real exchange rate (coefficients for these six additional interaction terms not reported due to space reasons). In columns (1) to (3) of Table 7, we interact each of the main proxies of financial vulnerability with our measure of banking development and allow all three interaction terms to enter the regressions simultaneously. The results suggest that strong banks shape the export survival especially through the transmission channel of liquidity needs. When it comes to the long-term export survival at the time horizon of ten years (third column), well-developed banks also help products with high investment needs. Independent of the time horizon of export survival, banks do not seem to play any role in promoting products whose manufacturing process requires a high share of intangible assets. 29 Table 7: An alternative horse race across transmission channels (1) (2) (3) (4) (5) (6) (7) (8) (9) Dep. var.: Survival 1y 5y 10y 1y 5y 10y 1y 5y 10y investment needs × banks 0.002 0.044 0.064b (0.047) (0.033) (0.027) liquidity needs × banks 0.103a 0.083a 0.052b (0.039) (0.029) (0.022) asset intangibility × banks 0.063 0.065 0.038 (0.056) (0.043) (0.032) investment needs × stock markets 0.054 0.076a 0.062a (0.040) (0.026) (0.019) liquidity needs × stock markets 0.128a 0.053b -0.007 (0.026) (0.024) (0.020) asset intangibility × stock markets 0.215a 0.184a 0.090a (0.039) (0.030) (0.028) investment needs × stock market turnover 0.055 0.021 0.035c (0.036) (0.027) (0.019) liquidity needs × stock market turnover 0.097a 0.070a 0.046a (0.025) (0.021) (0.015) asset intangibility × stock market turnover 0.111a 0.083a 0.047a (0.034) (0.027) (0.017) Observations 252,147 252,147 252,147 243,509 243,509 243,509 240,981 240,981 240,981 R-squared 0.264 0.546 0.692 0.255 0.547 0.695 0.255 0.549 0.697 Country-Time FE Yes Yes Yes Yes Yes Yes Yes Yes Yes Product FE Yes Yes Yes Yes Yes Yes Yes Yes Yes Full set of controls included Yes Yes Yes Yes Yes Yes Yes Yes Yes Dependent variable is the probability of export survival of product k from industrial sector (industry) i exported by country c to the USA. Export survival probability is measured l years after the beginning of export spell, with l = 1 in columns (1), (4) and (7), l = 5 in columns (2), (5) and (8), and l = 10 in columns (3), (6) and (9). The regressions are estimated by OLS and contain a full set of fixed effects at the product level and the (exporting country)*time level, with time referring to the beginning of a given export spell. Each specification controls for interaction terms of the three measures of financial vulnerability with GDP per capita and with real exchange rate. All variables are described in Table 2, Table 3, and Table 4. The full set of controls also includes initial export, total export, number of suppliers, multiple spell, and interaction terms between physical and human capital endowments of country c and the corresponding capital intensities at the industry level. All time-varying explanatory variables are measured at the beginning of the export spell. Robust standard errors are clustered at the (exporting country)*time level, with time referring to the beginning of a given export spell. a , b , c denote statistical significance at the 1%, 5%, and 10% level, respectively. In columns (4) to (6) of Table 7, we interact the three financial vulnerability proxies with the size dimension of stock market development. At all the time horizons, the posi- tive impact of large stock markets on export survival manifests itself especially through alleviating the financial vulnerability arising from the presence of intangible assets. Large stock markets also promote products with high investment needs, especially when it comes to export survival at longer time horizons. When it comes to products with high liquidity needs, the size of stock markets matters in particular for the short-term export survival. In columns (7) to (10) of Table 7, we interact our proxies of financial vulnerability with the activity of a given stock market. Similarly to large stock markets explored in the previous three columns, the active stock markets play an important role in promoting the export survival of products whose manufacturing process requires a high share of intan- gible assets. However, the other two transmission channels operate differently in the case 30 of active stock markets. Across all time horizons, alleviating the financial vulnerability of products with high liquidity needs represents an important transmission channel from active stock markets to export survival. At the same time, stock market activity does not play a substantial role in promoting exports of products with high investment needs. Overall, the results of the horse race across transmission channels in Table 7 are consistent with the previous findings from horse races between banks and large stock markets in Table 5 and between banks and active stock markets in Table 6. 6 Further Tests 6.1 Additional Channels This subsection extends the baseline empirical model to include the transmission channels of trade credit and global growth opportunities. Besides examining the robustness of our previous results, the exploration of these two channels in the context of export survival is worthwhile in its own right. Table 8 takes as a point of departure Table 5 in Subsection 4.2, adding the interaction terms of trade credit dependency with both banks and stock markets. These interaction terms allow us to examine if trade credit can act as a substitute for external finance provided by financial intermediaries and markets. According to this hypothesis, firms that have access to trade credit from their business partners suffer relatively less in countries with underdeveloped financial systems. Such firms would therefore also benefit somewhat less than other firms from the process of financial development. Consequently, we would expect negative coefficients for these additional interaction terms as well-developed banks and stock markets would be comparatively less useful in promoting export survival of products from industries that can rely on trade credit. Note that in our difference-in- differences framework, the negative coefficients do not imply that well-developed financial markets and institutions actually hurt the export survival of these products. The products from industries that can rely on trade credit simply do not benefit from a well-developed financial system quite as much as products from industries that lack access to trade credit as an alternative form of financing. The results in Table 8 show that the distinction between banks and stock markets as 31 well as the time horizon of export survival both matter for this additional transmission channel. The interaction term of trade credit dependency with banks has the expected negative sign and is highly significant for the long-term export survival at the horizons of both five years (second, fifth, and eighth columns of Table 8) and ten years (third, sixth, and ninth columns of Table 8). By contrast, the interaction term of trade credit dependency with stock markets is never significant in these specifications. When it comes to the short-term survival at the one-year horizon (first, fourth, and seventh columns of Table 8), the interaction terms of trade credit with both banks and stock markets are insignificant. Table 8: Additional channel from finance to export survival: Trade credit (1) (2) (3) (4) (5) (6) (7) (8) (9) Dep. var.: Survival 1y 5y 10y 1y 5y 10y 1y 5y 10y investment needs × banks -0.004 0.002 0.042 (0.052) (0.035) (0.028) investment needs × stock markets 0.032 0.065b 0.045b (0.044) (0.027) (0.020) investment needs × GDPpc 0.007 -0.003 -0.007 (0.019) (0.012) (0.010) investment needs × real exchange rate -1.949b -0.755 -0.657 (0.920) (0.658) (0.552) liquidity needs × banks 0.029 0.064b 0.059b (0.041) (0.032) (0.023) liquidity needs × stock markets 0.112a 0.030 -0.025 (0.028) (0.026) (0.019) liquidity needs × GDPpc 0.058a 0.035a 0.030a (0.014) (0.013) (0.010) liquidity needs × real exchange rate -1.458c -0.635 -0.673 (0.758) (0.637) (0.486) asset intangibility × banks -0.029 -0.019 -0.002 (0.063) (0.045) (0.036) asset intangibility × stock markets 0.226a 0.185a 0.089a (0.042) (0.032) (0.031) asset intangibility × GDPpc 0.015 0.025 0.012 (0.019) (0.016) (0.014) asset intangibility × real exchange rate -2.647b -3.148a -2.201a (1.115) (0.828) (0.662) trade credit dependency × banks -0.014 -0.050a -0.051a -0.007 -0.044a -0.046a -0.011 -0.048a -0.050a (0.023) (0.016) (0.012) (0.023) (0.016) (0.012) (0.023) (0.016) (0.013) trade credit dependency × stock markets -0.007 0.015 0.015 -0.003 0.015 0.014 -0.015 0.008 0.012 (0.020) (0.015) (0.011) (0.020) (0.015) (0.011) (0.019) (0.015) (0.011) Observations 235,294 235,294 235,294 235,294 235,294 235,294 235,294 235,294 235,294 R-squared 0.259 0.548 0.699 0.259 0.548 0.699 0.259 0.548 0.699 Country-Time FE yes yes yes yes yes yes yes yes yes Product FE yes yes yes yes yes yes yes yes yes Full set of controls included yes yes yes yes yes yes yes yes yes Dependent variable is the probability of export survival of product k from industrial sector (industry) i exported by country c to the USA. Export survival probability is measured l years after the beginning of export spell, with l = 1 in columns (1), (4) and (7), l = 5 in columns (2), (5) and (8), and l = 10 in columns (3), (6) and (9). The regressions are estimated by OLS and contain a full set of fixed effects at the product level and the (exporting country)*time level, with time referring to the beginning of a given export spell. Trade credit dependency is the ratio of accounts payable to total assets in industry i. Other variables are described in Table 2, Table 3, and Table 4. The full set of controls also includes initial export, total export, number of suppliers, multiple spell, and interaction terms between physical and human capital endowments of country c and the corresponding capital intensities at the industry level. All time-varying explanatory variables are measured at the beginning of the export spell. Robust standard errors are clustered at the (exporting country)*time level, with time referring to the beginning of a given export spell. a , b , c denote statistical significance at the 1%, 5%, and 10% level, respectively. 32 Trade credit can thus act as a substitute only for external financing by banks and not by stock markets. Furthermore, this transmission channel matters only for long- term export survival. In other words, trade credit only promotes exports of products that are already well-established in the destination market, while not being particularly beneficial for products whose exports started only recently. These results are consistent with the argument by Fisman and Love (2003) that firms first need to establish a positive reputation before being able to rely on trade credit from business partners as a substitute for a formal bank credit. Regarding our three main transmission channels from finance to export survival, Ta- ble 8 confirms the findings from Table 5. Deep stock markets promote long-term export survival of products with high investment needs (columns (1) to (3) of Table 8). Strong banks play a more important role in promoting export survival of products with high liquidity needs over longer time horizons, while deep stock markets are decisive for the one-year survival of these products (columns (4) to (6) of Table 8). And stock markets dominate banks across all time horizons when it comes to improving the export perfor- mance of products from industries with a high share of intangible assets (columns (7) to (9) of Table 8). Table 9 adds to the nine specifications from Table 5 the interaction terms of growth opportunities with banks and with stock markets. Fisman and Love (2007) find that in the context of promoting industrial growth, the interaction term of financial development with growth opportunities dominates the interaction term of financial development with investment needs. In the context of export performance, it turns out to be the export survival at longer time horizons where the inclusion of the channel related to growth opportunities affects the results for the channel related to investment needs. When looking at the short-term survival of one year in column (1) of Table 9, neither banks nor stock markets seem to particularly promote products with high investment needs, in accordance with the previous findings in Table 5. However, when it comes to the long-term export survival at the horizon of five and ten years in columns (2) and (3) of Table 9, it is banks rather stock markets that promote export performance of products with high investment 33 needs. This is the opposite result compared to the findings in Table 5. Part of the explanation might lie in a possible collinearity problem between various interaction terms. The fact that the interaction terms of growth opportunities with banks and stock markets are both highly significant but their regression coefficients have opposite signs throughout Table 9 would also support this interpretation.17 Table 9: Additional channel from finance to export survival: Growth opportunities (1) (2) (3) (4) (5) (6) (7) (8) (9) Dep. var.: Survival 1y 5y 10y 1y 5y 10y 1y 5y 10y investment needs × banks 0.045 0.064c 0.090a (0.053) (0.034) (0.028) investment needs × stock markets -0.003 0.036 0.024 (0.047) (0.027) (0.021) investment needs × GDPpc 0.005 -0.005 -0.009 (0.019) (0.012) (0.011) investment needs × real exchange rate -1.945b -0.743 -0.645 (0.920) (0.656) (0.550) liquidity needs × banks 0.022 0.059c 0.056b (0.042) (0.032) (0.023) liquidity needs × stock markets 0.120a 0.035 -0.022 (0.029) (0.027) (0.019) liquidity needs × GDPpc 0.058a 0.035a 0.030a (0.014) (0.013) (0.010) liquidity needs × real exchange rate -1.452c -0.629 -0.667 (0.759) (0.637) (0.484) asset intangibility × banks -0.020 -0.013 0.000 (0.062) (0.045) (0.035) asset intangibility × stock markets 0.217a 0.181a 0.086a (0.042) (0.032) (0.031) asset intangibility × GDPpc 0.013 0.023 0.011 (0.019) (0.016) (0.014) asset intangibility × real exchange rate -2.625b -3.149a -2.199a (1.114) (0.828) (0.661) growth opportunities × banks -0.065a -0.079a -0.060a -0.060a -0.072a -0.051a -0.062a -0.075a -0.054a (0.021) (0.016) (0.012) (0.021) (0.016) (0.012) (0.020) (0.015) (0.012) growth opportunities × stock markets 0.049a 0.038a 0.027a 0.052a 0.041a 0.027a 0.044a 0.036a 0.026a (0.017) (0.012) (0.010) (0.016) (0.012) (0.009) (0.015) (0.012) (0.010) Observations 235,294 235,294 235,294 235,294 235,294 235,294 235,294 235,294 235,294 R-squared 0.259 0.548 0.699 0.259 0.548 0.699 0.259 0.548 0.699 Country-Time FE yes yes yes yes yes yes yes yes yes Product FE yes yes yes yes yes yes yes yes yes Full set of controls included yes yes yes yes yes yes yes yes yes Dependent variable is the probability of export survival of product k from industrial sector (industry) i exported by country c to the USA. Export survival probability is measured l years after the beginning of export spell, with l = 1 in columns (1), (4) and (7), l = 5 in columns (2), (5) and (8), and l = 10 in columns (3), (6) and (9). The regressions are estimated by OLS and contain a full set of fixed effects at the product level and the (exporting country)*time level, with time referring to the beginning of a given export spell. Growth opportunities is the growth in real sales for the representative firm in industry i in the US. Other variables are described in Table 2, Table 3, and Table 4. The full set of controls also includes initial export, total export, number of suppliers, multiple spell, and interaction terms between physical and human capital endowments of country c and the corresponding capital intensities at the industry level. All time-varying explanatory variables are measured at the beginning of the export spell. Robust standard errors are clustered at the (exporting country)*time level, with time referring to the beginning of a given export spell. a , b , c denote statistical significance at the 1%, 5%, and 10% level, respectively. The results for the other two main transmission channels from finance to export survival are not substantially affected by the inclusion of the channel related to global growth opportunities. 17 Fisman and Love (2007) do not encounter this problem as they interact growth opportunities only with a general measure of financial development, which is the sum of the two proxies capturing the strength of the banks and the depth of stock markets in a given country. 34 6.2 Stock Market Value Traded and Broader Measures of Bank Development This subsection looks at three alternative measures of well-developed financial markets and institutions. First, we use the normalized stock market value traded (the value of stock market transactions divided by GDP) that lies somewhere in-between our measures for size and activity of stock markets.18 In another set of additional tests, we replace the traditional measure of a well-developed banking sector - credit provided by deposit-taking banks to the private sector divided by GDP - by two broader proxies for importance of financial institutions in the real economy. One proxy looks at a broader measure of claims banks have vis-a-vis the rest of the economy: the ratio of bank assets to GDP. The other proxy looks beyond traditional banks at a broader group of financial institutions: the total credit by both banks and other financial institutions divided by GDP. The results for these alternative measures broadly confirm the results from Section 4. The only qualitative difference is the impact of stock market value traded on products with high investment needs. Compared to stock market capitalization, the stock market value traded seems to matter for export survival of these products at shorter time horizons. For space reasons, we report the detailed results for the stock market value traded and the two alternative bank measures in Online Appendix C. 7 Conclusions We examine the transmission process from financial markets and institutions in exporting countries to short-term and long-term export survival at the product level. Our results suggest that banks, active stock markets, and large stock markets all play distinct roles in helping exported products to survive in the highly competitive US destination market. Large and active stock markets promote both short-term and long-term export sur- vival of products from industries with a high share of intangible assets. Strong banks do not seem to play any significant role in this regard. Given the association of intangible assets with innovative activities and a pronounced technological change, our findings pro- 18 Note that multiplying the stock market capitalization (the market value of all listed shares divided by GDP) with the stock market turnover (the value of stock market transactions relative to the market value of all listed shares) yields the stock market value traded. For a further discussion of various dimensions and measures of banking and stock market development see, e.g., Levine and Zervos (1998), Beck, Demirg¨ c–Kunt, and Levine (2000, 2010), Manova (2008), Strieborny and Kukenova (2016). u¸ 35 vide further support for the notion that stock markets are more important than banks in promoting R&D investment (Aghion et al. 2004; Brown, Martinssson, and Petersen 2013), encouraging innovation (Hsu, Tian, and Xu 2014), and helping industries that face increasing returns to scale and rapid technological change (Allen 1993; Brown, Martins- son, and Petersen 2017). Our product-level evidence on export survival also complements the recent research on banks and intangible capital. Dell’Ariccia et al. (2020) show that banks decrease their commercial lending if local firms increasingly use intangible capital. Beck et al. (2021) find that liquidity created by banks promotes only tangible invest- ment and consequently does not benefit those countries that specialize in industries using intangible assets. When it comes to promoting export survival of products with high liquidity needs, it is specifically the active but not the large stock markets that matter. This result might be related to the roles played by corporate governance and shareholders’ monitoring in improving the efficiency of working capital management (see, e.g., Gill and Biger 2013, Ben-Nasr 2016, Zeidan and Shapir 2017). Arguably, it is the activity rather than size of the stock market that strengthens the ability of shareholders to monitor the firms’ management. In particular, active and liquid stock markets make it easier for shareholders to put the management under pressure by selling or threatening to sell the shares of the firm (see, e.g., Edmans and Manso 2011 or Edmans and Holderness 2017). In accordance with the traditional role of the banking system in covering the working capital needs of firms, well-developed banks also promote export survival of products with high liquidity needs, especially at longer time horizons.19 Our empirical analysis also reveals the necessity of long-term export links if trade credit among business partners is to serve as a viable source of export financing. Based on our results, informal credit between upstream and downstream firms cannot substitute for well-developed financial markets and institutions in promoting immediate survival of 19 Besides the traditional route of bank loans, banks can support firms’ working capital management also by specialized instruments like letters of credit. These instruments are particularly important in international trade and at the same time do not directly enter our proxies for banking development. The results in this paper might therefore represent only a lower bound for the importance of a well- developed banking system for the export survival of products with high liquidity needs. 36 newly exported products. The ability to rely on trade credit matters only for long-term export survival, helping products that are already well-established in the destination market. Trade credit also seems to serve as a substitute for external financing from banks rather than from stock markets, in accordance with the previous research focusing on industrial output (e.g., Fisman and Love 2003). No clear pattern emerges in the export survival of products with high investment needs. This is especially the case when we control for the alternative channel of growth opportunities. These results complement previous research on production and stock re- turns. Fisman and Love (2007) provide evidence that growth opportunities dominate investment needs in the transmission process from financial development to industrial growth. Tong and Wei (2011) find less robust results for the investment channel in the context of the transmission from the 2007-2009 financial crisis to stock market returns at the firm level, especially when compared to the channel of liquidity needs. Our results on product-level export survival suggest three possible directions for fur- ther research on the distinct roles played by financial markets and institutions in inter- national trade. First, it seems that the specific source of products’ financial needs affects the relative importance of banks versus stock markets in promoting export success. It matters whether given products need to fund their investment needs related to physi- cal capital, to fund their liquidity needs related to working capital, or to overcome the focus of more rudimentary financial systems on collateralizable tangible assets. Second, the relative importance of banks versus stock markets can vary even for the same group of products, depending on the specific time horizon used for measuring export success. Third, the distinction between the size and the activity of stock markets matters sub- stantially. Further research on these three topics could improve our understanding of the transmission process from finance to the real economy. References Aghion, P., S. Bond, A. Klemm, and I. Marinescu, 2004, “Technology and Financial Structure: Are Innovative Firms Different?” Journal of the European Economic Association 2(2-3), 277–288. Albornoz, F., S. Fanelli, and J.C. Hallak, 2016, “Survival in Export Markets,” Journal of International Economics 102, 262–281. 37 Allen, F., 1993, “Stock Markets and Resource Allocation,” in C. Mayer and X. Vives, (eds.): Capital Markets and Financial Intermediation, Cambridge University Press, Cambridge, U.K., 81-116. Amissah, E., S. Bougheas, F. Defever, and R. Falvey, 2021, “Financial System Architecture and the Patterns of International Trade,” European Economic Review 136, article 103751. Amiti, M., and D.E. Weinstein, 2011, “Exports and Financial Shocks,” Quarterly Journal of Economics 126(4), 1841–1877. Araujo, L., G. Mion, and E. Ornelas, 2016, “Institutions and Export Dynamics,” Journal of International Economics 98, 2–20. Bae, K.-H., W. Bailey, and J. Kang, 2021, “Why is Stock Market Concentration Bad for the Economy?” Journal of Financial Economics 140, 436–459. Bats, J.V., and A.C.F.J.. Houben, 2020, “Bank-Based versus Market-Based Financing: Impli- cations for Systemic Risk,” Journal of Banking and Finance 114, article 105776 Beck, T., 2002, “Financial Development and International Trade: Is There a Link?” Journal of International Economics 57, 107–131. Beck, T., 2003, “Financial Dependence and International Trade,” Review of International Eco- nomics, 11(2), 296-316. Beck, T., A.Demirg¨ c–Kunt, and R. Levine, 2000, “A New Database on Financial Development u¸ and Structure,” World Bank Economic Review 14, 597-605. Beck, T., A. Demirg¨ c–Kunt, and R. Levine, 2010, “Financial Institutions and Markets across u¸ Countries and over Time: The Updated Financial Development and Structure Database,” World Bank Economic Review 24, 77-92. Beck, T., R. D¨ ottling, T. Lambert, and M. van Dijk, 2021, “Liquidity Creation, Investment, and Growth,” Working Paper. Beck, T., and R. Levine, 2002, “Industry Growth and Capital Allocation: Does Having a Market– or Bank–Based System Matter?” Journal of Financial Economics 64, 147–180. Beck, T., and R. Levine, 2004, “Stock Markets, Banks, and Growth: Panel Evidence,” Journal of Banking and Finance 28, 423–442. Becker, B., J. Chen, and D. Greenberg, 2013, “Financial Development, Fixed Costs, and International Trade,” Review of Corporate Finance Studies 2(1), 1–28. Bekaert, G., C.R. Harvey, C. Lundblad, and S. Siegel, 2007, “Global Growth Opportunities and Market Integration,” Journal of Finance 62(3), 1081-1137. ek´ B´ es, G., and B. Murak¨ ozy, 2012, “Temporary Trade and Heterogeneous Firms,” Journal of International Economics 87, 232–246. Ben-Nasr, H., 2016, “State and Foreign Ownership and the Value of Working Capital Manage- ment,” Journal of Corporate Finance 41, 217–240. Berman, N., and J. H´ ericourt, 2010, “Financial Factors and the Margins of Trade: Evidence from Cross-Country Firm-Level Data,” Journal of Development Economics 93(2), 206-217. Besedeˇ s, T., 2013, “The Role of NAFTA and Returns to Scale in Export Duration,” CESifo Economic Studies 59(2), 306-336. Besedeˇ s, T., and T.J. Prusa, 2006a, “Ins, Outs, and the Duration of Trade,” Canadian Journal of Economics 39(1), 266-295. Besedeˇ s, T., and T.J. Prusa, 2006b, “Product Differentiation and Duration of US Import Trade,” Journal of International Economics 70, 339-358. Besedeˇ s, T., and T.J. Prusa, 2011, “The Role of Extensive and Intensive Margins and Export Growth,” Journal of Development Economics 96, 371-379. Besedeˇ s, T., B.-C. Kim, and V. Lugovskyy, 2014, “Export Growth and Credit Constraints,” European Economic Review 70, 350-370. Braun, M., 2003, “Financial Contractibility and Asset Hardness,” Working Paper. Braun, M., and B. Larrain, 2005, “Finance and the Business Cycle: International, Inter- Industry Evidence,” Journal of Finance 60(3), 1097-1128. 38 Brenton, P., C. Saborowski, and E. von Uexkull, 2010, “What Explains the Low Survival Rate of Developing Country Export Flows?” World Bank Economic Review 24(3), 474–499. Bricongne, J.-C., L. Fontagn´ e, G. Gaulier, D. Taglioni, and V. Vigard, 2012, “Firms and the Global Crisis: French Exports in the Turmoil,” Journal of International Economics 87, 134–146. Bridges, S., and A. Guariglia, 2008, “Financial Constraints, Global Engagement, and Firm Survival in the United Kingdom: Evidence from Micro Data,” Scottish Journal of Political Economy 55(4), 444–464. Brown, J.R., G. Martinsson, and B.C. Petersen, 2013, “Law, Stock Markets, and Innovation,” Journal of Finance 68(4), 1517–1549. Brown, J.R., G. Martinsson, and B.C. Petersen, 2017, “Stock Markets, Credit Markets, and Technology-Led Growth,” Journal of Financial Intermediation 32, 45–59. Cadot, O., L. Iacovone, M.D. Pierola, and F. Rauch, 2013, “Success and Failure of African Exporters,” Journal of Development Economics 101, 284–296. Cetorelli, N., and M. Gambera, 2001, “Banking Market Structure, Financial Dependence and Growth: International Evidence from Industry Data,” Journal of Finance 56(2), 617-648. Cetorelli, N., and P.E. Strahan, 2006, “Finance as a Barrier to Entry: Bank Competition and Industry Structure in Local U.S. Markets,” Journal of Finance 61(1), 437-461. Cho, I., S. Contessi, K.N. Russ, and D. Valderrama, 2019, “Financial Choice and International Trade,” Journal of Economic Behavior and Organization 157, 297–319. Chor, D., and K. Manova, 2012, “Off the Cliff and Back? Credit Conditions and International Trade during the Global Financial Crisis,” Journal of International Economics 87, 117-133. ˇ ak, M., A. Demirg¨ Cih´ c-Kunt, E. Feyen, and R. Levine, 2012, “Benchmarking Financial Sys- u¸ tems around the World,” World Bank Policy Research Working Paper No. 6175. Claessens, S., and L. Laeven, 2003, “Financial Development, Property Rights, and Growth,” Journal of Finance 58(6), 2401–2436. Crin`o, R., and L. Ogliari, 2017, “Financial Imperfections, Product Quality, and International Trade,” Journal of International Economics 104, 63-84. Dell’Ariccia, G., D. Kadyrzhanova, C. Minoiu, and L. Ratnovski, 2020, “Bank Lending in the Knowledge Economy,” Review of Financial Studies, forthcoming. Demirg¨ c–Kunt, A., E. Feyen, and R. Levine, 2013, “The Evolving Importance of Banks and u¸ Securities Markets,” World Bank Economic Review 27(3), 476–490. Demirg¨ c–Kunt, A., and V. Maksimovic, 1998, “Law, Finance, and Firm Growth,” Journal of u¸ Finance 53(6), 2107–2137. Demirg¨ c–Kunt, A., and V. Maksimovic, 2002, “Funding Growth in Bank-Based and Market- u¸ Based Financial Systems: Evidence from Firm-Level Data,” Journal of Financial Eco- nomics 65, 337–363. Edmans, A., and C.G. Holderness, 2017, “Blockholders: A Survey of Theory and Evidence,” in B. Hermalin and M. Weisbach (eds): Handbook of the Economics of Corporate Gov- ernance, North Holland, Amsterdam, Netherlands, 541-636. Falato, A., D. Kadyrzhanova, J. Sim, and R. Steri, 2020, “Rising Intangible Capital, Shrinking Debt Capacity, and the US Corporate Savings Glut,” Journal of Finance, forthcoming. Fan, H., E. L.-C. Lai, and Y.A. Li, 2015, “Credit Constraints, Quality, and Export Prices: Theory and Evidence from China,” Journal of Comparative Economics 43, 390–416. Edmans, A., and G. Manso, 2011, “Governance Through Trading and Intervention: A Theory of Multiple Blockholders,” Review of Financial Studies 24(7), 2395–2428. Fisman, R., and I. Love, 2003, “Trade Credit, Financial Intermediary Development, and In- dustry Growth,” Journal of Finance 58(1), 353–374. Fisman, R., and I. Love, 2007, “Financial Dependence and Growth Revisited,” Journal of the European Economic Association 5(2-3), 470–479. Gaulier, G., and S. Zignago, 2010, “BACI: International Trade Database at the Product Level. 39 The 1994-2007 Version,” CEPII Working Paper No. 2010-23. Gill, A.S., and N. Biger, 2013, “The Impact of Corporate Governance on Working Capital Management Efficiency of American Manufacturing Firms,” Managerial Finance 39(2), 116–132. G¨org, H., and M.-E. Spaliara, 2014, “Financial Health, Exports and Firm Survival: Evidence from UK and French Firms,” Economica 81, 419-444. Greenaway, D., A. Guariglia, and R. Kneller, 2007, “Financial Actors and Exporting Decisions,” Journal of International Economics 73, 377-395. Gupta, N., and K. Yuan, 2009, “On the Growth Effect of Stock Market Liberalizations,” Review of Financial Studies 22(11), 4715-4752. Hess, W., and M. Persson, 2012, “The Duration of Trade Revisited: Continuous-Time vs. Discrete-Time Hazards,” Empirical Economics 43, 1083–1107. Hur, J., M. Raj, and Y.E. Riyanto, 2006, “Finance and Trade: A Cross-Country Empirical Analysis on the Impact of Financial Development and Asset Tangibility on International Trade,” World Development 34(10), 1728–1741. Hsu, P.-H., X. Tian, and Y. Xu, 2014, “Financial Development and Innovation: Cross-country Evidence,” Journal of Financial Economics 112, 116–135. Jaud, M., M. Kukenova, and M. Strieborny, 2015, “Financial Development and Sustainable Exports: Evidence from Firm-product Data,” World Economy 38(7), 1090-1114. Jaud, M., M. Kukenova, and M. Strieborny, 2018, “Finance, Comparative Advantage, and Resource Allocation,” Review of Finance 22(3), 1011-1061. Jaud, M., M. Kukenova, and M. Strieborny, 2019, “Finance and Export Survival: The Case of the MENA Region and Sub-Saharan Africa,” in Tibor Besedeˇ s and Volker Nitsch (eds.): Disrupted Economics Relationships, MIT Press, Cambridge, USA. Jaud, M., M. Kukenova, and M. Strieborny, 2021, “Stock Market Liberalizations and Export Dynamics,” Working Paper. Ju, J., and S.-J. Wei, 2011, “When Is Quality of Financial System a Source of Comparative Advantage?” Journal of International Economics 84, 178–187. Kletzer, K., and P. Bardhan, 1987, “Credit Markets and Patterns of International Trade,” Journal of Development Economics 27, 57-70. Kroszner, R.S., L. Laeven, and D. Klingebiel, 2007, “Banking Crises, Financial Dependence, and Growth,” Journal of Financial Economics 84, 187-228. Langfield, S., and M. Pagano, 2016, “Bank Bias in Europe: Effects on Systemic Risk and Growth,” Economic Policy 31(85), 51–106. Levchenko, A.A., 2007, “Institutional Quality and International Trade,” Review of Economic Studies 74(3), 791-819. Levchenko, A.A., L.T. Lewis, and L.L. Tesar, 2011, “The Role of Trade Finance in the U.S. Trade Collapse: A Skeptic’s View,” in J.-P. Chauffour and M. Malouche (eds.): Trade Finance During the Great Trade Collapse, World Bank, Washington DC, USA, 133–147. Levine, R., 2002, “Bank-Based or Market-Based Financial Systems: Which is Better?” Journal of Financial Intermediation 11, 398-428. Levine, R., and S. Zervos, 1998, “Stock Markets, Banks, and Economic Growth,” American Economic Review 88(3), 537-558. Manova, K., 2008, “Credit Constraints, Equity Market Liberalizations and International Trade,” Journal of International Economics 76, 33-47. Manova, K., 2013, “Credit Constraints, Heterogeneous Firms, and International Trade,” Re- view of Economic Studies 80, 711–744. Minetti, R., and S.C. Zhu, 2011, “Credit Constraints and Firm Export: Microeconomic Evi- dence from Italy,” Journal of International Economics 83, 109-125. Muˆuls, M., 2015, “Exporters, Importers and Credit Constraints,” Journal of International Economics 95, 333–343. 40 Nitsch, V., 2009, “Die Another Day: Duration in German Import Trade,” Review of World Economics 145(1), 133–154. Nunn, N., 2007, “Relationship-Specificity, Incomplete Contracts, and the Pattern of Trade,” Quarterly Journal of Economics 122(2), 569-600. Paravisini, D., V. Rappoport, P. Schnabl, and D. Wolfenzon, 2015, “Dissecting the Effect of Credit Supply on Trade: Evidence from Matched Credit–Export Data,” Review of Economic Studies 82, 333–359. Raddatz, C., 2006, “Liquidity Needs and Vulnerability to Financial Underdevelopment,” Jour- nal of Financial Economics 80(3), 677-722. Rajan, R.G., and L. Zingales, 1998, “Financial Dependence and Growth,” American Economic Review 88(3), 559-586. Rampini, A.A., and S. Viswanathan, 2010, “Collateral, Risk Management, and the Distribution of Debt Capacity,” Journal of Finance 65(6), 2293-2322. Romalis, J., 2004, “Factor Proportions and the Structure of Commodity Trade,” American Economic Review 94(1), 67-97. Shen, C.-H., and C.-C. Lee, 2006, “Same Financial Development yet Different Economic Growth - Why?” Journal of Money, Credit, and Banking 38(7), 1907-1944. Shleifer, A., and R.W. Vishny, 1992, “Liquidation Values and Debt Capacity: A Market Equi- librium Approach,” Journal of Finance 47(4), 1343-1366. Song, F., and A.V. Thakor, 2010, “Financial System Architecture and the Co-Evolution of Banks and Capital Markets,” Economic Journal 120, 1021-1055. Strieborny, M., and M. Kukenova, 2016, “Investment in Relationship-Specific Assets: Does Finance Matter?” Review of Finance 20(4), 1487-1515. Svaleryd, H., and J. Vlachos, 2005, “Financial Markets, the Pattern of Industrial Specializa- tion and Comparative Advantage: Evidence from OECD Countries,” European Economic Review 49, 113–144. Tadesse, S., 2002, “Financial Architecture and Economic Performance: International Evi- dence,” Journal of Financial Intermediation 11, 429–454. Tong, H., and S.-J. Wei, 2011, “The Composition Matters: Capital Inflows and Liquidity Crunch During a Global Economic Crisis,” Review of Financial Studies 24(6), 2023-2052. Zeidan, R., and O.M. Shapir, 2017, “Cash Conversion Cycle and Value-enhancing Operations: Theory and Evidence for Free Lunch,” Journal of Corporate Finance 45, 203–219. Zingales, L., 2015, “Presidential Address: Does Finance Benefit Society?” Journal of Finance 70(4), 1327–1363. APPENDIX 41 Table 10: Summary statistics Variable Obs. Mean Std. Dev. Q1 Median Q3 investment needs 252147 0.39 0.05 0.37 0.38 0.40 liquidity needs 252147 0.25 0.08 0.21 0.26 0.30 asset intangibility 252147 0.05 0.05 0.02 0.04 0.07 trade credit dependency 252147 0.40 0.11 0.34 0.39 0.46 global growth opportunities 252147 0.41 0.16 0.29 0.40 0.52 investment needs x banks 252147 0.28 0.19 0.12 0.25 0.38 investment needs x stock markets 235294 0.22 0.21 0.07 0.15 0.30 investment needs x GDPpc 252147 3.60 0.69 3.18 3.65 3.99 investment needs x real exchange rate 252147 -0.87 0.12 -0.90 -0.86 -0.82 investment needs x bank assets 252082 0.31 0.20 0.15 0.27 0.42 investment needs x total credit 251588 0.28 0.20 0.12 0.26 0.39 investment needs x stock market value traded 233886 0.13 0.18 0.02 0.06 0.16 investment needs x stock market turnover 232766 0.20 0.20 0.07 0.16 0.28 liquidity needs x banks 252147 0.18 0.14 0.07 0.15 0.25 liquidity needs x stock markets 235294 0.14 0.14 0.04 0.09 0.19 liquidity needs x GDPpc 252147 2.31 0.79 1.82 2.28 2.75 liquidity needs x real exchange rate 252147 -0.56 0.17 -0.67 -0.57 -0.46 liquidity needs x bank assets 252082 0.20 0.15 0.09 0.16 0.27 liquidity needs x total credit 251588 0.18 0.14 0.07 0.15 0.25 liquidity needs x stock market value traded 233886 0.08 0.12 0.01 0.04 0.10 liquidity needs x stock market turnover 232766 0.13 0.14 0.04 0.10 0.17 asset intangibility x banks 252147 0.03 0.05 0.01 0.02 0.05 asset intangibility x stock markets 235294 0.03 0.05 0.00 0.01 0.03 asset intangibility x GDPpc 252147 0.47 0.49 0.14 0.32 0.63 asset intangibility x real exchange rate 252147 -0.11 0.12 -0.16 -0.08 -0.04 asset intangibility x bank assets 252082 0.04 0.05 0.01 0.02 0.05 asset intangibility x total credit 251588 0.04 0.05 0.01 0.02 0.05 asset intangibility x stock market value traded 233886 0.02 0.03 0.00 0.00 0.02 asset intangibility x stock market turnover 232766 0.03 0.05 0.00 0.01 0.03 trade credit dependency x banks 252147 0.28 0.21 0.12 0.24 0.39 trade credit dependency x stock markets 235294 0.22 0.22 0.07 0.15 0.30 global growth opportunities x banks 252147 0.29 0.23 0.11 0.23 0.40 global growth opportunities x stock markets 235294 0.23 0.24 0.06 0.14 0.31 banks 252147 0.71 0.47 0.31 0.65 0.98 stock markets 235294 0.56 0.52 0.17 0.37 0.76 GDPpc (log) 252147 9.24 1.28 8.38 9.52 10.34 real exchange rate 252147 -2.23 0.02 -2.24 -2.23 -2.22 bank assets 252082 0.78 0.51 0.39 0.69 1.08 total credit 251588 0.73 0.50 0.30 0.68 1.00 stock market value traded 233886 0.33 0.45 0.04 0.15 0.40 stock market turnover 232766 0.52 0.51 0.17 0.41 0.73 initial export 252147 2.95 2.42 1.18 2.48 4.23 total export 252147 6.27 2.69 4.40 6.31 8.13 number of suppliers 252147 37.99 19.36 24.00 35.00 48.00 multiple spell 252147 0.73 0.44 0.00 1.00 1.00 physical capital intensity 252147 0.06 0.02 0.05 0.06 0.07 human capital intensity 252147 0.98 0.23 0.81 1.06 1.13 physical capital intensity x physical capital 252147 0.74 0.27 0.62 0.71 0.86 human capital intensity x human capital 252147 2.17 0.57 1.73 2.24 2.62 physical capital 252147 11.46 0.91 10.94 11.68 12.17 human capital 252147 2.20 0.25 2.08 2.27 2.38 We measure all time-varying explanatory variables at the beginning of a given export spell. The time dimension in our dataset is therefore reduced to the initial year of a given export spell - t0 . The variables GDP pcc,t0 , initial exportck,t0 , total exportck,t0 are taken in log terms. 42