TY - JOUR
AU - Strieborny, Martin
AB - Abstract Can financial institutions and markets enhance the discipline imposed by competitive product markets and thus improve resource allocation in the real economy? We address this question in the context of international trade, using disaggregated product-level data from seventy-one countries exporting to the USA. We show that exported products exit the US market sooner if they stand far away from the exporting country’s comparative advantage. This pattern is stronger when the exporting country has a well-developed banking system, but it is unaffected by the depth of stock markets. These results are in accordance with theories stressing the disciplining role of debt and monitoring abilities of banks. 1. Introduction Can well-functioning financial institutions and markets enhance the discipline provided by product markets and help avoid waste of scarce resources? The question is important because product markets often represent a rather slow-acting disciplining force (Jensen, 1993). And on the conceptual grounds, the answer seems to be affirmative. According to Levine (2005), creditors and shareholders can often improve the resource allocation by inducing managers to maximize firm value. The immediate threats of bankruptcy by creditors (Grossman and Hart, 1982) or takeover through the stock markets (Scharfstein, 1988; Stein, 1988) seem to be particularly strong devices enabling financial agents to enhance the ultimate but slow-acting disciplining impact of product markets on firms’ behavior.1 However, it can be at times challenging to distinguish empirically between the disciplining effects of financial and product markets. The quality of financial markets and institutions and the quality of product markets are namely often correlated both across countries and across geographic and administrative units within countries. For example, the overall legal framework and government regulations affect the potential efficiency of the disciplining pressures imposed by both financial agents and product markets. Our paper attempts to disentangle the disciplining effects imposed by well-developed financial institutions and markets from that of the product markets in the context of international trade. Our identification strategy consists of a difference-in-difference approach using data on more than 4,500 different products exported to the United States (USA) by seventy-one countries at different stages of financial development. The impact of financial markets and institutions is captured by the varying strengths of banks, stock, and bond markets across countries exporting to the USA. The disciplining effect of product markets is captured by the varying level of expected competitive pressures faced by a given product in the US destination market. Our measure of these product market pressures is motivated by the Heckscher–Ohlin theory of comparative advantage in international trade and captures the extent of discrepancy between each product’s revealed factor intensity (RFI) and each country’s factor endowment. Intuitively, exported products that extensively use factors of production (e.g., physical or human capital) that are scarce and therefore expensive in the exporting country will face much stronger headwinds in the highly competitive US destination market. There are several advantages in using the international trade framework for examining the disciplining role of finance and product markets. First, we can reasonably assume orthogonality between the competitive pressures in the US product markets and the varying levels of strength of financial markets and institutions across countries exporting their products to the USA. The bilateral character of the international trade data thus allows us to better distinguish the impact of financial markets and institutions that are measured at the exporting country level from the impact of product markets in the US destination market. Second, the international trade data are highly disaggregated, internationally comparable, and available for a large set of different countries and products. Table A.I in Appendix A provides some examples of different product categories to convey a better idea about the level of disaggregation in our data. This allows us to control for product heterogeneity at a sufficient level of disaggregation while at the same time minimizing the problems of comparability and external validity that arise in the context of existing national data classification on domestic production. In particular, there are no readily available firm-level data on domestic production that would be both internationally comparable and available for a large number of countries. Third, there is the issue of reverse causality between financial factors and outcomes in the real economy. For example, Greenaway, Guariglia, and Kneller (2007) find no evidence that firms with a better ex-ante financial health are more likely to enter foreign markets. They do, however, find strong evidence that firms’ financial health improves once they start exporting. This result poses serious challenge for studies examining whether better developed financial institutions and markets promote exports of financially vulnerable firms.2 The utilization of product data can help to address this issue. In particular, the lack of product’s congruence with the comparative advantage of the exporting country (distance to comparative advantage — DCAck,tz) is a technological characteristic capturing the extent to which the product’s manufacturing process uses up production factors that do not correspond to the factor endowment of a given economy. Presumably, neither the various political factors affecting export performance nor the export performance itself will alter the capital or labor intensity of individual products. Our empirical results confirm that the larger the distance between the exported product’s RFI and the exporting country’s factor endowment, the sooner the product exits the US destination market. Our measure of distance to comparative advantage thus seems to be a good proxy for competitive pressures facing a given product. Highly competitive product markets in the USA indeed force out exporters who rely mostly on scarce and expensive production factors from the exporting country. Although not the main point of our paper, this result is certainly interesting on its own and warrants further research. Crucially, the products whose factor intensity puts them at a comparative disadvantage exit the US market even faster if the exporting country has a well-developed banking system. Strong banks in the country of origin thus seem to reinforce the disciplining impact of a competitive destination market on the international trade flows. Interestingly, we do not find the same reinforcing effect in the case of deep and liquid stock markets in the exporting countries. These findings are robust to a battery of sensitivity checks, including the use of various proxies for the strength of banks, stock, and bond markets and alternative ways of measuring distance to comparative advantage. We thus see the main contribution of our paper in providing new evidence about the unique importance of banks in enhancing the disciplining impact of product markets and improving resource allocation in the real economy. In that sense, the paper contributes to three strands of literature examining the impact of finance on real economy. First, our finding that well-developed banks but not deep stock markets can push country’s exports toward products congruent with its comparative advantage provides empirical support for theories stressing the disciplining role of debt (e.g., Stiglitz, 1974; Grossman and Hart, 1982; Jensen, 1986; Harris and Raviv, 1990; Stulz, 1990; Hart and Moore, 1995; Aghion, Dewatripont, and Rey, 1999) and the superior monitoring abilities of banks (e.g., Stiglitz and Weiss, 1983; Diamond, 1984; Fama, 1985). In Section 2, we present a simple model of our own that illustrates how a well-developed banking sector can enhance the incentive impact of product market competition in the context of international trade. One could also see our results in the light of an idea introduced by Allen (1993), building upon the conceptual framework of Diamond (1984). Allen (1993) argues that banks will have advantage over stock markets in monitoring firms when there is a broad consensus on what the firms should do and the main issue is to ensure that these actions are indeed taken by the firms’ management. Arguably, this is exactly the situation when it comes to the question whether firms should focus on exporting products that correspond to the comparative advantage of the country of origin, in particular if their destination market is the highly competitive US market. Second, an influential strand of finance literature has examined the role of financial development in improving the allocation of physical capital (e.g., Wurgler, 2000; Beck and Levine, 2002; Love, 2003). In our paper, we utilize the framework of international trade to complement this scholarly work in two dimensions. For one thing, we are able to examine the impact of finance on resource allocation in a much more disaggregated setting at the level of individual products rather than industries while still having the benefit of using data comparable across a wide range of countries. Moreover, using trade data and the concept of RFI allows us to compare products’ factor intensities with countries’ factor endowments and thus examine the disciplining role of finance in a somewhat broader context of the allocation of production factors that include physical capital, human capital, and natural resources.3 The focus on the overall resource allocation relates our paper to Fisman and Love (2004) who develop two indirect tests for examining the role of finance in resource allocation based on the idea that industry global shocks are correlated across countries and that countries at similar levels of per-capita income should experience demand-driven similarities in their industrial growth rates. Another important paper in this area is Bekaert et al. (2007) which stresses the importance of financial openness rather than domestic financial development for the capital allocation. We do not examine the issue of financial openness in our framework as it is covered elsewhere (Jaud, Kukenova, and Strieborny, 2016). Third, the paper contributes to the growing body of scholarly work on the effects of finance on international trade. This literature has provided empirical support for the importance of well-functioning financial environment for countries’ and firms’ successful export performance in two ways. It has showed that domestic financial development improves export performance, particularly for industries dependent on external finance and firms facing credit constraints (e.g., Beck, 2002, 2003; Ju and Wei, 2005; Greenaway, Guariglia, and Kneller, 2007; Manova, 2008; Jaud, Kukenova, and Strieborny, 2009; Berman and Héricourt, 2010; Becker, Chen, and Greenberg, 2013; Manova, 2013; Besedes, Kim, and Lugovskyy, 2014; Jaud, Kukenova, and Strieborny, 2015). A related strand of papers has explored the negative impact of disturbances in the financial system on trade performance, especially during the recent global financial crisis (e.g., Bricongne et al., 2010; Amiti and Weinstein, 2011; Levchenko, Lewis, and Tesar, 2011; Chor and Manova, 2012; Paravisini et al., 2015). Our paper complements these findings by showing that a well-developed banking system does not provide non-discriminate support for all kinds of exports. Instead, we provide evidence suggesting that banks do specifically promote exports on the “right side” of the comparative advantage. While providing the necessary funds for the often expensive exporting activities, banks in the countries of origin seem also to reinforce the discipline imposed by the product markets in the destination country. The rest of this paper is structured as follows. Section 2 outlines a simple theoretical framework while the remaining sections comprise the empirical analysis that constitutes the main focus of the paper. Sections 3 and 4 discuss the data and our empirical strategy, respectively. Section 5 reports our main results and Section 6 offers a series of robustness checks. Section 7 concludes. 2. An Illustrative Model To provide additional motivation for our empirical approach, this section presents a simple model centering around an agent who maximizes perks associated with her position as export manager while taking into account the competitive pressures in the destination market and the probability of obtaining export financing in her country.4 Like the rest of our paper, the model deliberately focuses on the long-term export survival rather than export entry. The slow-acting disciplining impact of competitive product markets might namely not affect the export entry much. In the model, we thus assume a mid-level manager who is allocated an existing export project and can initiate its termination but is not involved in decisions about entry into exporting. 2.1 Assumptions and Timing of the Model We consider a three period model where a risk-neutral manager takes a decision at time t = 0 in order to maximize her expected utility in periods t = 1 and t = 2.5  t=0 At time t = 0, the manager is allocated by her supervisor an ongoing project that represents export of a specific product (we thus use the terms project and product interchangeably). The project is one of two types: efficient (E) with probability λ∈[0,1] or inefficient (I) with probability (1−λ). Efficient projects are exports of products that correspond to the comparative advantage of the exporting country, while inefficient ones do not. The extent of the export project’s inefficiency is captured by the exported product’s distance to the comparative advantage, d∈[0,1). Defining di, i={E, I} as the distance to comparative advantage of a type i project thus implies dE=0 and 0<dI<1. The manager observes the type of the project and decides between pursuing the originally assigned project or initiating its termination and a reassignment to another ongoing project. The manager knows that a reassigned project will again be efficient (E) with probability λ∈[0,1] or inefficient (I) with probability (1−λ). An efficiency-maximizing manager would always initiate termination of a type I project. However, the manager in our model maximizes her own export-driven perks (X), with both efficient and inefficient projects generating the same level of these perks.6 Moreover, initiating termination of an ongoing export project imposes costs on the manager that we assume to be equal to a fraction of her perks (kX, with 0<k<1). We can interpret k as an additional effort on the part of manager (convincing the manager’s superior about the necessity of altering firm’s export portfolio, dealing with corporate procedures associated with terminating an ongoing project, etc.), which disturbs the managerial “quiet life” (Bertrand and Mullainathan, 2003). Alternatively, k can be seen as the probability that manager fails to achieve reassignment to another export project and loses her perks.   t=1 In the first period (t = 1), the manager has to secure external funding from a bank for either the originally assigned export project or the project to which she was reassigned.7 With probability φ the bank funding is secured, the export project remains active, and the manager receives her perks X in the first period. With probability (1−φ) the bank refuses, export project gets terminated, and the manager does not receive any perks. The level of financial development (f) in the manager’s country affects the probability of obtaining bank funding φ in two ways. A more developed banking system implies a better availability of external finance to the real economy, which ceteris paribus increases the probability of any given export project to receive bank funding (∂φ(.)∂f>0). At the same time, banks in more financially developed countries possess better monitoring and screening skills. This implies a disproportionately lower probability of obtaining bank funding for exports characterized by large distance to the comparative advantage d in countries at a high level of financial development f (∂2φ(.)∂d∂f<0).8  t=2 In the second period (t = 2), that can be seen as a proxy for “long term,” the project that secured bank funding either succeeds (with probability μ) or fails (with probability of 1−μ) in a competitive destination market. In the former case, the manager receives X also in the second period. In the latter case, she receives in the second period a fraction of her perks lX, with 0<l<1. This fraction can be thought of as, for example, a wage premium in a future job associated with having prior “exporting experience” on her CV (Mion and Opromolla, 2011). The probability of project’s success in a competitive destination market μ depends on the efficiency of the project. Formally, let μi={μI,μE} be the success probability of a project of type i (i∈I, E) in a competitive destination market, with 0≤μI<μE≤1. All projects corresponding to comparative advantage are assumed to share the same probability of export success μE.9 The probability of export success for the inefficient projects decreases in the extent of the inefficiency, that is, in the distance to comparative advantage (∂μI(.)∂d<0). 2.2 Managerial Decision The manager knows all the parameters of the model (d,f,l,k,μE,λ). She is also aware of the level of competitive pressures faced by the inefficient projects in the destination market (∂μI(.)∂d) as well as of the availability of external finance and monitoring skills of the banks in her country and their impact on the probability of securing bank funding (∂φ(.)∂f, ∂2φ(.)∂d∂f). Taking as given these external constraints in her pursuit of maximizing export perks, the manager decides at time t = 0 whether to initiate termination of an inefficient project.10,11 The utility of the manager if she continues (C) with an inefficient export project writes:   UC=φ(d,f){X+μI(d)X+[1−μI(d)]lX}, (1) where φ(d,f) is the probability of obtaining bank funding for an inefficient project and the expression in curled brackets represents the expected managerial perks in the first and second period. Manager’s utility in case of initiating termination (T) of an inefficient project and pursuing reassignment to an efficient one writes:   UT=λφ(0,f){X+μEX+[1−μE]lX}+[1−λ]φ(d,f){X+μI(d)X+[1−μI(d)]lX}−kX, (2) where φ(0,f) is the probability of obtaining bank funding for an efficient project with zero distance to comparative advantage and λ is the probability of being reassigned to such project after initiating the termination of the originally allocated export project. The first set of curled brackets captures the expected perks if the manager succeeds in being reassigned to an efficient project while the second set of curled brackets represents the perks in case of reassignment to an inefficient project. The manager maximizes her own utility so that: Umanager=Max{UT,UC}. She will thus initiate termination of a project that does not correspond to the comparative advantage if and only if12:   UT−UC≥0 (3)  λφ(0,f){X+μEX+[1−μE]lX}−λφ(d,f){X+μI(d)X+[1−μI(d)]lX}−kX≥0 (4)  φ(0,f){1+μE+[1−μE]l}−φ(d,f){1+μI(d)+[1−μI(d)]l}≥kλ (5) The right-hand side (RHS) of Equation (5) represents the cost of terminating an inefficient export project from the point of view of self-interested manager. This cost is increasing in the share of perks lost by initiating the termination of an inefficient project (k) and decreasing in the probability of being reassigned to an efficient project (λ). The left-hand side (LHS) of Equation (5) represents the benefit for the manager from terminating an inefficient project. This benefit increases with the exported product’s distance to the comparative advantage d. Formally, the derivative of the LHS with respect to d is positive:   −∂φ(d,f)∂d︸≤0{1+μI(d)+[1−μI(d)]l}−φ(d,f)∂μI(d)∂d︸<0[1−l]>0.  (6) The intuition is as follows. The further away the project is from the comparative advantage, the less likely it will survive the product market competition in the destination country (∂μI(d)∂d<0), depriving the manager of the [1−l] fraction of her perks in the second period for a project that has received external funding (which happens with probability φ). This mechanism captured by the second term determines also the sign of the whole expression, under the plausible assumption that the extent of inefficiency of export projects does not increase the probability of obtaining external funding (∂φ(d,f)∂d≤0). The incentive impact of product market competition in the destination country is further reinforced by the banking development in the country of origin. Formally, the second derivative of the LHS of Equation (5) with respect to distance to comparative advantage d and financial development f is also positive:   −∂2φ(d,f)∂d∂f︸<0{1+μI(d)+[1−μI(d)]l}−[1−l]∂φ(d,f)∂f︸>0∂μI(d)∂d︸<0>0.  (7) The intuition for this result is as follows. The disciplining impact of a competitive destination market on export success and the related managerial perks is less relevant if projects fail indiscriminately due to the lack of bank funding. The financial development increases the probability of export projects to receive such funding (the external finance availability channel, ∂φ(d,f)∂f>0) so that the ultimate success or failure of an export project is increasingly determined by its alignment with comparative advantage, reinforcing the incentive effect of the product market competition (∂μI(d)∂d<0). The overall effect is even stronger if a more developed banking sector possess better monitoring and screening skills so that the additional funds get channeled disproportionately toward export projects closer to the comparative advantage of the exporting country (the bank monitoring channel, ∂2φ(d,f)∂d∂f<0).13 Equations (6) and (7) capture the idea that product market competition in the destination country provides incentives for perks-maximizing managers to initiate termination of inefficient export projects and that this incentive effect of product market competition is amplified in exporting countries with a well-developed banking sector. The model can thus provide a simple motivation for the empirical part of our paper that tests the following two hypotheses: H1: The probability of terminating an ongoing export spell of a given product into a competitive destination market increases with the distance of this product to the comparative advantage of the exporting country. H2: This disciplining impact of distance to comparative advantage is magnified by the level of banking development in the exporting country. The phrasing in both the model and in hypothesis H2 stresses the disciplining impact of the banking sector. Stock and bond markets could be in principle also a source of financing for export projects and thus reinforce the disciplining impact of the product markets via the external finance availability channel (∂φ(d,f)∂f>0).14 Making the standard assumption about superior monitoring abilities of banks [e.g., based on arguments by Diamond (1984) or Fama (1985)], the presence of the monitoring channel (∂2φ(d,f)∂d∂f<0) in Equation (7) should nevertheless lead to more robust results for the banking sector as opposed to the stock and bond markets. Our empirical results using various proxies for the strength of banks, stock, and bond markets in the exporting countries will confirm this conjecture. 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 defined at the HS-6 digit level to the USA without interruption. Let z(ck) denote an export spell of product k exported from country c to the USA. Furthermore, let tz be the spell initiation year, that is, the first year in which the export spell is active. As the initiation year is spell-specific, indexing observations by initiation year eliminates the need to index them by spell. Note that there can be multiple spells for a given country-product pair.15 The unit of observation is thus an (exporting country) * product * (initiation year) triplet. All time-varying explanatory variables are measured at the time tz. 3.1 Financial Variables We use several variables measuring the strength of financial markets and institutions in the exporting country. As our main proxy for the strength of the banking sector, we use the ratio of the bank credit to private sector over the country’s GDP (BCc,tz). The ratio of the stock market capitalization over the GDP (StMc,tz) captures the depth of stock markets in the exporting country. Both variables are the standard proxies for the strength of banks and stock markets in a given country (see, e.g., Strieborny and Kukenova, 2016). We also use various alternative measures to test the robustness of our results. In the case of banks, these measures include bank assets (BAc,tz), total credit (TCc,tz), non-bank credit (NBCc,tz), and credit to government (GCc,tz). Bank assets are claims on domestic real non-financial sector by deposit money banks as a share of GDP, total credit is the ratio of private credit by deposit money banks and other financial institutions over GDP, non-bank credit is the credit extended by other financial institutions than banks as a share of GDP (i.e., the difference between values for total credit and bank credit), credit to government is the credit extended to the general government sector, including central, state, and local governments and social security funds. In the case of stock markets, the alternative measures are stock market value traded (StMVTc,tz) and stock market turnover (StMTc,tz). Stock market value traded is the value of stock market transactions relative to GDP and stock market turnover is the ratio of the value of total shares traded to market capitalization. We use three proxies for the strength and development of bond markets. Domestic private bonds (DPBc,tz) are the ratio of outstanding domestic private debt securities over GDP while the proxy called international private bonds (IPBc,tz) represents the ratio of outstanding international private debt securities over GDP. The third proxy, non-financial international bonds (NFIBc,tz), captures the ratio of international corporate bonds issued by non-financial firms to GDP. The data are from the Global Financial Development Database that builds on the earlier work by Beck, Demirgüç-Kunt, and Levine (2000) and has been substantially extended by Cihak et al. (2012). The database contains various indicators of financial development across countries and over time and is regularly updated. The exception is the variable non-financial international bonds (NFIBc,tz) that comes from the statistics compiled by the Bank for International Settlements (BIS). All financial variables measured at the exporting country level enter the regressions both directly (BCc,tz, StMc,tz, etc.) and interacted with our product-country measure of distance to comparative advantage (BCc,tz*DCAck,tz, StMc,tz*DCAck,tz, etc.). 3.2 Distance to Comparative Advantage — Conceptual Basis According to a long-standing idea within the Heckscher–Ohlin theoretical framework, international trade in multiple commodities can be explained by a “product chain of comparative advantage” that ranks the products in order of their factor intensities (Jones, 1956–57; Bhagwati, 1972). In the context of multiple products and multiple countries, countries are arranged along the product chain in accordance with their relative factor endowments, with each country exporting the products within its own segment and importing all the others (Deardorff, 1979). Each country thus specializes in exporting products whose relative factor intensity broadly corresponds to the relative factor endowment of this country.16 In the Heckscher–Ohlin framework, the “distance to comparative advantage” thus captures the extent of discrepancy between each product’s RFI and each country’s factor endowment. This approach should be seen as complementary to recent literature on the alternative Ricardian theory that stresses technology rather than factor endowments as a possible source of comparative advantage in international trade (see, e.g., Dornbusch, Fischer, and Samuelson, 1977; Eaton and Kortum, 2002; Costinot, 2009; Costinot, Donaldson, and Komunjer, 2012).17 Our focus on factor endowments as the main source of comparative advantage is a pragmatic decision motivated by data availability at a highly disaggregated level in a broad cross-country setting.18 Moreover, Morrow (2010) shows that omitting Ricardian forces do not bias tests of the Heckscher–Ohlin model while ignoring Heckscher–Ohlin forces can lead to biased tests of the Ricardian model.19 3.3 Distance to Comparative Advantage — Measurement The specific measure of distance to comparative advantage (DCA) used in our paper has been developed by Cadot, Carrère, and Strauss-Kahn (2011). It is computed at the six-digit level of the HS classification and compares in a given year the RFI of a given product with the factor endowment of a given country. A large distance to comparative advantage thus identifies products that represent inefficient exports as they use up scarce rather than abundant resources in a given economy. Specifically, the formula for the Euclidean distance of product k from the comparative advantage of country c, in the initiation year tz of export spell z, writes:   DCAck,tz=std(κc,tz−κ^k,tz)2+std(hc,tz−h^k,tz)2, (8) where κc,tz and hc,tz are the endowments in physical and human capital of country c, and κ^k,tz and h^k,tz are the corresponding revealed factor intensities of product k, all in log terms and measured in the initiation year tz of the export spell z. In one of the robustness tests, we will extend the two-factor framework of distance to comparative advantage and consider also the role of natural resources by adding arable land per worker as a third production factor. Since κ and h are measured in different units, we use normalized differences between product factor intensities and country factor endowments with mean 0 and standard deviation 1 (“std” in the formula above stands for “standardized”). This ensures that physical and human capital are given equal weights in the overall distance formula.20 The data on national factor endowments are from Cadot, Tumurchudur, and Shirotori (2009). In a given year, the stock of physical capital per worker in a given country (κ) is constructed according to the perpetual inventory method. Human capital per worker (h) is calculated from the average years of schooling in a given country using attainment data. The product revealed factor intensities are from Cadot, Tumurchudur, and Shirotori (2009). They are calculated as weighted averages of the factor endowments of the countries exporting a given product, following the methodology introduced by Hausmann, Hwang, and Rodrik (2007). Specifically, the revealed physical capital intensity of product k in year tz is calculated as:   κ^k,tz=∑cωck,tzκc,tz, (9) where κc,tz is country c’s endowment in physical capital at time tz, and the weights are given by ωck,tz=Xck,tz/Xc,tz∑cXck,tz/Xc,tz, with X denoting export values. These weights correspond to the revealed comparative advantage of country c in product k. The numerator in the formula for ωck,tz measures the importance of product k in the overall exports of country c. The denominator aggregates the export shares of product k across all countries. Weighting countries’ factor endowments by the revealed comparative advantage rather than by the simple export shares allows us to account for distortions due to differences in countries’ size [Hausmann, Hwang, and Rodrik (2007) and Cadot, Tumurchudur, and Shirotori (2009) provide numerical examples].21 Similarly, the revealed human capital intensity of product k is calculated as:   h^k,tz=∑cωck,tzhc,tz, (10) where hc,tz is endowment in human capital of country c at time tz. 3.4 Other Variables Product-level 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, 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 six-digit products and 143 countries. Export flows are reported annually in values (thousands of US dollars) and quantities.22 Trade data at the HS six-digit level is particularly fitting for survival analysis. It 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 USA from the 1995 to 2005 period. We start our sample period in 1995 due to the high number of missing values before 1994.23 We finish our sample period in 2005 in order to avoid possible spurious correlations due to the recent global financial crisis that affected both our empirical proxies for the strength of banks, bond, and stock markets and the functioning of international trade in manufacturing goods. We plan to examine the issue of interactions between financial shocks and crises and product market discipline in later research. Macroeconomic variables: Real exchange rate is 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 is the World Development Indicators of the World Bank. The annual data for the GDP per capita measured in log terms (GDPpcc,tz) are taken from the World Development Indicator Report 2006 and are reported in constant 2000 US dollars. Both macroeconomic variables measured at the exporting country level enter the regressions directly (RERc,tz, GDPpcc,tz) as well as interacted with our measure of distance to comparative advantage (RERc,tz*DCAck,tz, GDPpcc,tz*DCAck,tz). Industry-level variables: Let j denote an industry. External finance dependence at the level of ISIC four-digit sectors comes from Raddatz (2006) and is based on financial data about US firms from Compustat. In particular, the dependence on external finance of industry j (ExFj) is defined as capital expenditures minus cash flow from operations, divided by capital expenditures, for the median firm in that industry. Human capital intensity (HumIntj) and physical capital intensity (CapIntj) for ISIC three-digit sectors come from Braun (2003) and have already been used in trade literature (Manova, 2008, 2013).24,25 All industry characteristics (ExFj, CapIntj, HumIntj) are computed solely from the US data and thus do not vary across exporting countries. The US market is large, diversified, well-functioning, and comparatively frictionless. Industry characteristics based on the US data can thus be interpreted as exogenous technological characteristics that are not driven by various market imperfections prevalent in many countries. This idea comes back to the seminal paper of Rajan and Zingales (1998).26 Table I shows descriptive statistics for the main variables used in our analysis. When controlling for all variables of interest, our final database consists of 191,078 export spells of 4,562 HS six-digit products from seventy-one exporting countries to the USA over the period 1995–2005. Appendix B gives the list of countries in our sample. Table I. Summary statistics: data Let j denote an industry (ISIC three-digit or four-digit), k denote product (HS six-digit). The time dimension in our dataset is reduced to the initiation year. All time-varying variables take value in time tz. The variables GDPpcc,tz, initial_exportck,tz, total_exportck,tz are taken in log terms. Variable  Obs.  Mean  Standard deviation  Min  Max  Distance to comparative advantage (DCAck,tz)  191,078  1.18  0.79  0  6.64  Banks (BCc,tz)  191,078  0.61  0.46  0.03  2.18  Stock markets (StMc,tz)  175,481  0.50  0.46  0.01  3.03  GDP per capita (GDPpcc,tz)  191,078  9.16  0.94  6.29  10.47  Dependence on external finance (ExFj)  191,078  0.25  0.29  –0.55  1.26  Banks×dist. to comp. adv. (BCc,tz*DCAck,tz)  191,078  0.69  0.75  0  12.09  Stock markets×dist. to comp. advantage (StMc,tz*DCAck,tz)  175,481  0.54  0.70  0  14.71  GDPpc×dist. to comp. advantage (GDPpcc,tz*DCAck,tz)  191,078  10.65  6.73  0.01  69.00  Banks×dep. on ext. finance (BCc,tz*ExFj)  191,078  0.15  0.25  –1.19  2.74  Stock markets×dep. on ext. finance (StMc,tz*ExFj)  175,481  0.12  0.22  –1.65  3.82  Bank assets (BAc,tz)  191,078  0.70  0.47  0.04  2.48  Total credit (TCc,tz)  191,078  0.63  0.44  0.03  2.28  Non-bank credit (NBCc,tz)  191,078  0.02  0.19  –1.19  0.83  Credit to government (GCc,tz)  54,329  51.91  24.66  6.90  143.80  Stock market value traded (StMVTc,tz)  173,368  0.27  0.34  0.00  2.00  Stock market turnover (StMTc,tz)  172,200  0.56  0.60  0.00  5.12  Domestic private bonds to GDP (DPBc,tz)  127,260  0.23  0.22  0.00  1.42  International private bonds to GDP (IPBc,tz)  158,291  0.09  0.09  0.00  0.92  Non-financial international bonds to GDP (NFIBc,tz)  168,113  0.02  0.03  0.00  0.16  Real exchange rate (RERc,tz)  129,642  –2.23  0.02  –2.31  –2.10  Average export growth  191,078  0.02  0.44  –7.25  5.23  Physical capital (κc,tz)  191,078  10.50  1.17  6.35  12.09  Human capital (hc,tz)  191,078  7.24  2.56  0.69  11.89  Physical capital intensity (CapIntj)  191,078  0.06  0.02  0.02  0.20  Human capital intensity (HumIntj)  191,078  0.95  0.24  0.50  1.66  Physical capital×phys. cap. intensity (κc,tz*CapIntj)  191,078  0.67  0.27  0.12  2.36  Human capital×hum. cap. intensity (hc,tz*HumIntj)  191,078  6.90  3.10  0.34  19.67  Initial export (initial_exportck,tz)  191,078  10.12  2.64  0.00  23.72  Total export (total_exportck,tz)  191,078  13.00  2.83  0.00  24.07  Number of suppliers (NSuppliersk,tz)  191,078  37.07  19.09  1.00  136  Multiple spell (multiplespellck)  191,078  0.58  0.49  0.00  1.00  Alternative measures of distance to comparative advantage  Absolute distance  191,078  1.60  1.12  0.00  9.20  Euclidean distance, including land per worker  191,078  1.54  0.80  0.02  7.26  Absolute distance, including land per worker  191,078  2.39  1.28  0.03  12.18  Euclidean distance, excluding origin country  190,490  1.18  0.79  0.00  6.58  Absolute distance, excluding origin country  190,490  1.60  1.11  0.01  9.12  Euclidean distance, excluding USA as destination country  191,078  1.18  0.79  0.00  6.64  Absolute distance, excluding USA as destination country  191,078  1.60  1.12  0.00  9.19  Variable  Obs.  Mean  Standard deviation  Min  Max  Distance to comparative advantage (DCAck,tz)  191,078  1.18  0.79  0  6.64  Banks (BCc,tz)  191,078  0.61  0.46  0.03  2.18  Stock markets (StMc,tz)  175,481  0.50  0.46  0.01  3.03  GDP per capita (GDPpcc,tz)  191,078  9.16  0.94  6.29  10.47  Dependence on external finance (ExFj)  191,078  0.25  0.29  –0.55  1.26  Banks×dist. to comp. adv. (BCc,tz*DCAck,tz)  191,078  0.69  0.75  0  12.09  Stock markets×dist. to comp. advantage (StMc,tz*DCAck,tz)  175,481  0.54  0.70  0  14.71  GDPpc×dist. to comp. advantage (GDPpcc,tz*DCAck,tz)  191,078  10.65  6.73  0.01  69.00  Banks×dep. on ext. finance (BCc,tz*ExFj)  191,078  0.15  0.25  –1.19  2.74  Stock markets×dep. on ext. finance (StMc,tz*ExFj)  175,481  0.12  0.22  –1.65  3.82  Bank assets (BAc,tz)  191,078  0.70  0.47  0.04  2.48  Total credit (TCc,tz)  191,078  0.63  0.44  0.03  2.28  Non-bank credit (NBCc,tz)  191,078  0.02  0.19  –1.19  0.83  Credit to government (GCc,tz)  54,329  51.91  24.66  6.90  143.80  Stock market value traded (StMVTc,tz)  173,368  0.27  0.34  0.00  2.00  Stock market turnover (StMTc,tz)  172,200  0.56  0.60  0.00  5.12  Domestic private bonds to GDP (DPBc,tz)  127,260  0.23  0.22  0.00  1.42  International private bonds to GDP (IPBc,tz)  158,291  0.09  0.09  0.00  0.92  Non-financial international bonds to GDP (NFIBc,tz)  168,113  0.02  0.03  0.00  0.16  Real exchange rate (RERc,tz)  129,642  –2.23  0.02  –2.31  –2.10  Average export growth  191,078  0.02  0.44  –7.25  5.23  Physical capital (κc,tz)  191,078  10.50  1.17  6.35  12.09  Human capital (hc,tz)  191,078  7.24  2.56  0.69  11.89  Physical capital intensity (CapIntj)  191,078  0.06  0.02  0.02  0.20  Human capital intensity (HumIntj)  191,078  0.95  0.24  0.50  1.66  Physical capital×phys. cap. intensity (κc,tz*CapIntj)  191,078  0.67  0.27  0.12  2.36  Human capital×hum. cap. intensity (hc,tz*HumIntj)  191,078  6.90  3.10  0.34  19.67  Initial export (initial_exportck,tz)  191,078  10.12  2.64  0.00  23.72  Total export (total_exportck,tz)  191,078  13.00  2.83  0.00  24.07  Number of suppliers (NSuppliersk,tz)  191,078  37.07  19.09  1.00  136  Multiple spell (multiplespellck)  191,078  0.58  0.49  0.00  1.00  Alternative measures of distance to comparative advantage  Absolute distance  191,078  1.60  1.12  0.00  9.20  Euclidean distance, including land per worker  191,078  1.54  0.80  0.02  7.26  Absolute distance, including land per worker  191,078  2.39  1.28  0.03  12.18  Euclidean distance, excluding origin country  190,490  1.18  0.79  0.00  6.58  Absolute distance, excluding origin country  190,490  1.60  1.11  0.01  9.12  Euclidean distance, excluding USA as destination country  191,078  1.18  0.79  0.00  6.64  Absolute distance, excluding USA as destination country  191,078  1.60  1.12  0.00  9.19  Table I. Summary statistics: data Let j denote an industry (ISIC three-digit or four-digit), k denote product (HS six-digit). The time dimension in our dataset is reduced to the initiation year. All time-varying variables take value in time tz. The variables GDPpcc,tz, initial_exportck,tz, total_exportck,tz are taken in log terms. Variable  Obs.  Mean  Standard deviation  Min  Max  Distance to comparative advantage (DCAck,tz)  191,078  1.18  0.79  0  6.64  Banks (BCc,tz)  191,078  0.61  0.46  0.03  2.18  Stock markets (StMc,tz)  175,481  0.50  0.46  0.01  3.03  GDP per capita (GDPpcc,tz)  191,078  9.16  0.94  6.29  10.47  Dependence on external finance (ExFj)  191,078  0.25  0.29  –0.55  1.26  Banks×dist. to comp. adv. (BCc,tz*DCAck,tz)  191,078  0.69  0.75  0  12.09  Stock markets×dist. to comp. advantage (StMc,tz*DCAck,tz)  175,481  0.54  0.70  0  14.71  GDPpc×dist. to comp. advantage (GDPpcc,tz*DCAck,tz)  191,078  10.65  6.73  0.01  69.00  Banks×dep. on ext. finance (BCc,tz*ExFj)  191,078  0.15  0.25  –1.19  2.74  Stock markets×dep. on ext. finance (StMc,tz*ExFj)  175,481  0.12  0.22  –1.65  3.82  Bank assets (BAc,tz)  191,078  0.70  0.47  0.04  2.48  Total credit (TCc,tz)  191,078  0.63  0.44  0.03  2.28  Non-bank credit (NBCc,tz)  191,078  0.02  0.19  –1.19  0.83  Credit to government (GCc,tz)  54,329  51.91  24.66  6.90  143.80  Stock market value traded (StMVTc,tz)  173,368  0.27  0.34  0.00  2.00  Stock market turnover (StMTc,tz)  172,200  0.56  0.60  0.00  5.12  Domestic private bonds to GDP (DPBc,tz)  127,260  0.23  0.22  0.00  1.42  International private bonds to GDP (IPBc,tz)  158,291  0.09  0.09  0.00  0.92  Non-financial international bonds to GDP (NFIBc,tz)  168,113  0.02  0.03  0.00  0.16  Real exchange rate (RERc,tz)  129,642  –2.23  0.02  –2.31  –2.10  Average export growth  191,078  0.02  0.44  –7.25  5.23  Physical capital (κc,tz)  191,078  10.50  1.17  6.35  12.09  Human capital (hc,tz)  191,078  7.24  2.56  0.69  11.89  Physical capital intensity (CapIntj)  191,078  0.06  0.02  0.02  0.20  Human capital intensity (HumIntj)  191,078  0.95  0.24  0.50  1.66  Physical capital×phys. cap. intensity (κc,tz*CapIntj)  191,078  0.67  0.27  0.12  2.36  Human capital×hum. cap. intensity (hc,tz*HumIntj)  191,078  6.90  3.10  0.34  19.67  Initial export (initial_exportck,tz)  191,078  10.12  2.64  0.00  23.72  Total export (total_exportck,tz)  191,078  13.00  2.83  0.00  24.07  Number of suppliers (NSuppliersk,tz)  191,078  37.07  19.09  1.00  136  Multiple spell (multiplespellck)  191,078  0.58  0.49  0.00  1.00  Alternative measures of distance to comparative advantage  Absolute distance  191,078  1.60  1.12  0.00  9.20  Euclidean distance, including land per worker  191,078  1.54  0.80  0.02  7.26  Absolute distance, including land per worker  191,078  2.39  1.28  0.03  12.18  Euclidean distance, excluding origin country  190,490  1.18  0.79  0.00  6.58  Absolute distance, excluding origin country  190,490  1.60  1.11  0.01  9.12  Euclidean distance, excluding USA as destination country  191,078  1.18  0.79  0.00  6.64  Absolute distance, excluding USA as destination country  191,078  1.60  1.12  0.00  9.19  Variable  Obs.  Mean  Standard deviation  Min  Max  Distance to comparative advantage (DCAck,tz)  191,078  1.18  0.79  0  6.64  Banks (BCc,tz)  191,078  0.61  0.46  0.03  2.18  Stock markets (StMc,tz)  175,481  0.50  0.46  0.01  3.03  GDP per capita (GDPpcc,tz)  191,078  9.16  0.94  6.29  10.47  Dependence on external finance (ExFj)  191,078  0.25  0.29  –0.55  1.26  Banks×dist. to comp. adv. (BCc,tz*DCAck,tz)  191,078  0.69  0.75  0  12.09  Stock markets×dist. to comp. advantage (StMc,tz*DCAck,tz)  175,481  0.54  0.70  0  14.71  GDPpc×dist. to comp. advantage (GDPpcc,tz*DCAck,tz)  191,078  10.65  6.73  0.01  69.00  Banks×dep. on ext. finance (BCc,tz*ExFj)  191,078  0.15  0.25  –1.19  2.74  Stock markets×dep. on ext. finance (StMc,tz*ExFj)  175,481  0.12  0.22  –1.65  3.82  Bank assets (BAc,tz)  191,078  0.70  0.47  0.04  2.48  Total credit (TCc,tz)  191,078  0.63  0.44  0.03  2.28  Non-bank credit (NBCc,tz)  191,078  0.02  0.19  –1.19  0.83  Credit to government (GCc,tz)  54,329  51.91  24.66  6.90  143.80  Stock market value traded (StMVTc,tz)  173,368  0.27  0.34  0.00  2.00  Stock market turnover (StMTc,tz)  172,200  0.56  0.60  0.00  5.12  Domestic private bonds to GDP (DPBc,tz)  127,260  0.23  0.22  0.00  1.42  International private bonds to GDP (IPBc,tz)  158,291  0.09  0.09  0.00  0.92  Non-financial international bonds to GDP (NFIBc,tz)  168,113  0.02  0.03  0.00  0.16  Real exchange rate (RERc,tz)  129,642  –2.23  0.02  –2.31  –2.10  Average export growth  191,078  0.02  0.44  –7.25  5.23  Physical capital (κc,tz)  191,078  10.50  1.17  6.35  12.09  Human capital (hc,tz)  191,078  7.24  2.56  0.69  11.89  Physical capital intensity (CapIntj)  191,078  0.06  0.02  0.02  0.20  Human capital intensity (HumIntj)  191,078  0.95  0.24  0.50  1.66  Physical capital×phys. cap. intensity (κc,tz*CapIntj)  191,078  0.67  0.27  0.12  2.36  Human capital×hum. cap. intensity (hc,tz*HumIntj)  191,078  6.90  3.10  0.34  19.67  Initial export (initial_exportck,tz)  191,078  10.12  2.64  0.00  23.72  Total export (total_exportck,tz)  191,078  13.00  2.83  0.00  24.07  Number of suppliers (NSuppliersk,tz)  191,078  37.07  19.09  1.00  136  Multiple spell (multiplespellck)  191,078  0.58  0.49  0.00  1.00  Alternative measures of distance to comparative advantage  Absolute distance  191,078  1.60  1.12  0.00  9.20  Euclidean distance, including land per worker  191,078  1.54  0.80  0.02  7.26  Absolute distance, including land per worker  191,078  2.39  1.28  0.03  12.18  Euclidean distance, excluding origin country  190,490  1.18  0.79  0.00  6.58  Absolute distance, excluding origin country  190,490  1.60  1.11  0.01  9.12  Euclidean distance, excluding USA as destination country  191,078  1.18  0.79  0.00  6.64  Absolute distance, excluding USA as destination country  191,078  1.60  1.12  0.00  9.19  Table II provides some preliminary evidence on the relationship between export survival, finance, and comparative advantage. Products that are closer to the comparative advantage of the exporting countries seem to face better survival odds in the US destination market. The average length of export spells with initial value of distance to comparative advantage DCAck,tz below the 25th percentile is 5.05 years while the average length for export spells with initial value of DCAck,tz above the 75th percentile is 4.00 years. The corresponding median values are 3 and 2 years, respectively. A strong domestic banking sector in the exporting country (initial value of BCc,tz above the 75th percentile versus below the 25th percentile) also seems to improve products’ survival chances on the US market (average spell lengths of 5.60 years versus 3.57 years and median spell lengths of 4 years versus 2 years). Well-developed stock markets seem to play quantitatively a somewhat lesser role (5.13 years versus 4.10 years in case of average spell lengths and 3 years versus 2 years in case of median spell lengths). However, these prima-facie differences do not control for export spells heterogeneity and are subject to potential confounding influences. Next section thus turns to a systematic exploration of the disciplining effect of product markets and finance on countries’ export composition within a formal econometric framework. Table II. Summary statistics: export spells The time dimension in our dataset is reduced to the initiation year tz of export spell z. All time-varying variables are measured in the year tz. DCAck,tz is distance to comparative advantage measured as the Euclidean distance of the RFI of product k to the factor endowment of country c. BCc,tz is the proxy for banking development in country c measured as bank credit over GDP. StMc,tz is the proxy for stock markets development in country c measured as stock market capitalization over GDP.   Number of spells  Mean spell length  Median spell length  Full sample  191,078  4.79  3  DCAck,tz < 25th percentile  47,769  5.05  3  DCAck,tz > 75th percentile  47,769  4.00  2  BCc,tz< 25th percentile  47,733  3.57  2  BCc,tz > 75th percentile  47,735  5.60  4  StMc,tz < 25th percentile  43,766  4.10  2  StMc,tz > 75th percentile  43,729  5.13  3    Number of spells  Mean spell length  Median spell length  Full sample  191,078  4.79  3  DCAck,tz < 25th percentile  47,769  5.05  3  DCAck,tz > 75th percentile  47,769  4.00  2  BCc,tz< 25th percentile  47,733  3.57  2  BCc,tz > 75th percentile  47,735  5.60  4  StMc,tz < 25th percentile  43,766  4.10  2  StMc,tz > 75th percentile  43,729  5.13  3  Table II. Summary statistics: export spells The time dimension in our dataset is reduced to the initiation year tz of export spell z. All time-varying variables are measured in the year tz. DCAck,tz is distance to comparative advantage measured as the Euclidean distance of the RFI of product k to the factor endowment of country c. BCc,tz is the proxy for banking development in country c measured as bank credit over GDP. StMc,tz is the proxy for stock markets development in country c measured as stock market capitalization over GDP.   Number of spells  Mean spell length  Median spell length  Full sample  191,078  4.79  3  DCAck,tz < 25th percentile  47,769  5.05  3  DCAck,tz > 75th percentile  47,769  4.00  2  BCc,tz< 25th percentile  47,733  3.57  2  BCc,tz > 75th percentile  47,735  5.60  4  StMc,tz < 25th percentile  43,766  4.10  2  StMc,tz > 75th percentile  43,729  5.13  3    Number of spells  Mean spell length  Median spell length  Full sample  191,078  4.79  3  DCAck,tz < 25th percentile  47,769  5.05  3  DCAck,tz > 75th percentile  47,769  4.00  2  BCc,tz< 25th percentile  47,733  3.57  2  BCc,tz > 75th percentile  47,735  5.60  4  StMc,tz < 25th percentile  43,766  4.10  2  StMc,tz > 75th percentile  43,729  5.13  3  4. Empirical Strategy The difference-in-difference approach in this paper relies on exploiting the varying strength of the competitive pressures a given product faces in the US destination market and the varying strength of financial markets and institutions in the countries exporting these products to the USA. Our focus on a single destination market filters out potential confounding influences arising from the destination markets heterogeneity. The use of bilateral trade data thus allows us a cleaner distinction between the impact of financial development measured at the level of exporting countries and the competitive pressures in the US product market.27 We choose the USA as our preferred destination market because it represents the largest among the highly competitive product markets of the developed countries.28 We will perform robustness tests with alternative destination markets. We are interested in the long-term survival of the export spells and rely on a formal duration (survival) analysis, where the dependent variable is the hazard rate (hazard function) hck(t) — the probability that a given product k exported by country c exits the US market at time t, conditional on its survival until that time (i.e., conditional on the fact that country c exported the given product k to US continuously for t years, counted from the initiation year tz of the export spell z).29,30 This econometric framework thus captures the probability of products’ long-term exporting failure, which makes it particularly suitable to maintain our focus on long-term impact of product markets and financial institutions. It is also in accordance with our simple model that centers around an export manager and her decision whether to initiate the termination of an ongoing export project. We rely on the widely used continuous time proportional hazard model proposed by Cox (1972), henceforth Cox model [see Online Appendix A for further technical details and Kiefer (1988) for a detailed description of duration analysis]. In the Cox model, the hazard function varies from zero (meaning certain survival at time t) to infinity (meaning certain failure at time t) and is defined as a product of two multiplicative terms. The first term is a baseline (underlying) hazard function h0 that is the function of time alone. The second term is an exponential function containing a set of independent variables that shift the baseline hazard function. An important advantage of the semiparametric Cox model is the fact that the shape of the baseline hazard function is left unspecified so that no distributional assumptions are required. Such assumptions could possibly lead to misspecifications and biased estimates. 4.1 The Role of Banks We start by examining the allocative and selective role of banks. We ask whether these external debtholders enhance the disciplining effect of product markets by pushing the composition of manufacturing exports toward goods congruent with the comparative advantage of the exporting country. Specifically, we regress the hazard rate hck(t) of a country-product export spell to the USA on the following independent variables: the Euclidean distance between the RFI of product k and the factor endowment of exporting country c at the initiation of the export spell at time tz (distance to comparative advantage — DCAck,tz), an interaction term between the ratio of the bank credit over GDP in exporting country c and our measure of distance to comparative advantage at time tz (BCc,tz*DCAck,tz), the bank credit over GDP in exporting country c (BCc,tz), a vector of control variables (x), as well as a full set of exporting-country fixed effects (δc) and initiation-year fixed effects (δtz). Note that the time dimension in our dataset is reduced to the initiation years and time-varying variables are thus measured at the time tz. The unit of analysis is an observation at the (exporting country) * product * (initiation year) level. Our estimation equation focusing on the disciplining impact of banks and product markets thus writes:   hck(t)=h0,j(t) exp  [β1DCAck,tz+β2BCc,tz*DCAck,tz+γBCc,tz+xck,tzφ+δc+δtz+εck,tz], (11) where εck,tz is the error term. Equation (11) is estimated using partial maximum likelihood and standard errors are clustered at the (exporting country) * (initiation year) level (c*tz). The relationship between the covariates and the hazard rate is log-linear, allowing parameters to be interpreted as semi-elasticities. Our two main variables of interest are the proxy for distance of the exported product to the comparative advantage of the exporting country (DCAck,tz) and the interaction term of this proxy with the strength of the banking sector in the exporting country (BCc,tz*DCAck,tz). The parameter β1 captures the disciplining effect of the competitive US market. In particular, the variable DCAck,tz captures the strength of competitive pressures that a given product faces in the US destination market dependent on how optimally it uses the available production factors endowment of the exporting country. A positive coefficient β1 would confirm that products using a high share of scarce rather than abundant production factors of the exporting country indeed face a higher hazard rate in the US market. The parameter β2 captures the additional role of banks in reinforcing this pattern. A positive coefficient β2 would suggest that a strong banking system pushes the export composition toward the comparative advantage of a given country even before the competition in the US market sets in.31 Positive coefficients β1 and β2 would also be in accordance with hypotheses H1 and H2 derived in the context of our simple model in Section 2. In some regressions, we also use other banking variables that measure various dimensions of the banking system in the country of origin — bank assets, total credit, non-bank credit, and credit to government. Similarly to our main banking proxy, these measures also enter the regressions both directly (BAc,tz, TCc,tz, NBCc,tz, GCc,tz) and interacted with the distance of the exported product to the comparative advantage of the exporting country (BAc,tz*DCAck,tz, TCc,tz*DCAck,tz, NBCc,tz*DCAck,tz, GCc,tz*DCAck,tz). We control for heterogeneity across manufacturing sectors by fitting a stratified Cox model at the industry level j. This allows the shape of the baseline hazard function, h0,j(t), to vary across 118 ISIC four-digit industries and adds more flexibility to the model. We also address the omitted variable bias by the use of fixed effects. In the Cox model, the inclusion of fixed effects results in a shift of the baseline hazard function. The exporting-country fixed effects (δc) control for a wide array of observable and unobservable characteristics of exporting countries that might affect the chances of their products to survive in the US market. These factors include physical and cultural proximity to the USA, common border, common language, etc. The initiation-year fixed effects (δtz) control for the possibility that the initial conditions in the first year of export spell (tz) might influence the products’ chances for subsequent survival in the US market. The vector x includes the GDP per capita of the exporting country c taken in log terms (GDPpcc,tz), country c’s endowments in physical and human capital (κc,tz,hc,tz) and their interactions with the corresponding capital intensities at the industry level (κc,tz*CapIntj,hc,tz*HumIntj), as well as several product-related control variables. We include the log value of export of product k from country c to the US market in the initial year of the trade relationship tz (initial_exportck,tz). This reflects the level of confidence importers in the USA have in the reliability of their trading partners. Additionally, we include the total log export value of product k from country c to all countries in the initial year of the trade relationship (total_exportck,tz). This variable captures the experience the exporting country has in supplying the world market with product k. We further control for the varying degree of competition in the US product markets by incorporating the number of countries exporting a given product k to the USA in year tz (NSuppliersk,tz). We also account for trade relationships that experience multiple spells within the same (exporting country) * product pair by including a dummy variable (multiple_spellck) marking all spells within such “on-and-off” trade relationships.32 In some specifications, the vector x also includes GDP per capita of the exporting country c interacted with the distance to comparative advantage (GDPc,tz*DCAck,tz), real exchange rate (RERc,tz), and real exchange rate interacted with the distance to comparative advantage (RERc,tz*DCAck,tz). Finally, we re-estimate Equation (11), stratifying by product k, allowing the baseline function to vary across individual products rather than just across industries. The estimating equation then becomes:   hck(t)=h0,k(t) exp  [β1DCAck,tz+β2BCc,tz*DCAck,tz+γBCc,tz+xck,tzφ+δc+δtz+εck,tz]. (12) This specification allows for a different baseline hazard function, h0,k(t), for every of the 4,562 analyzed products from the HS six-digit classification. In terms of controlling for product heterogeneity, the specification is thus even more stringent than would be a one including product fixed effects. Fixed effects merely shift the baseline hazard function while stratification allows different products to have different underlying hazard functions (see Online Appendix B for a more technical treatment of this issue). In Online Appendix C.1, we allow the underlying hazard rate to differ even for the same product k if the export spell started in different year tz. 4.2 The Role of Stock Markets As the next step in our analysis, we examine whether well-developed stock markets also enhance the disciplining effect of product markets in the context of international trade. We re-estimate Equation (11), replacing the variable measuring the strength of the banking sector (BCc,tz) with the ratio of the stock market capitalization over the GDP in the exporting country c measured in the initiation year of the export spell tz (StMc,tz). This variable captures the depth and liquidity of stock markets in the exporting country. The estimating equation focusing on the role of stock markets in enhancing the product market discipline thus writes:   hck(t)=h0,j(t) exp  [β1DCAck,tz+β2StMc,tz*DCAck,tz+γStMc,tz+xck,tzφ+δc+δtz+εck,tz], (13) where εck,tz is the error term. As previously, Equation (13) is estimated using partial maximum likelihood and standard errors are clustered at the (exporting country) * (initiation year) level (c*tz). We also include exporting-country and initiation-year fixed effects (δc,δtz) and stratify the estimation by 118 ISIC four-digit industry categories j. The parameter β1 captures the disciplining effect of the competitive US market and the parameter β2 captures the additional effect of well-developed stock markets in the exporting country. In some regressions, we use stock market value traded and stock market turnover as alternative measures of the depth and liquidity of stock markets in the exporting country. Like with other financial variables, these measures enter the regressions both directly (StMVTc,tz, StMTc,tz) and interacted with the distance of the exported product to the comparative advantage of the exporting country (StMVTc,tz*DCAck,tz, StMTc,tz*DCAck,tz). The non-financial control variables are identical to the ones in Equation (11). Similarly to the regressions focusing on the banking sector, also in the case of stock markets we re-estimate Equation (13), stratifying at the product level k and thus allowing the baseline hazard function h0 to vary across the 4,562 products according to the HS six-digit classification:   hck(t)=h0,k(t) exp  [β1DCAck,tz+β2StMc,tz*DCAck,tz+γStMc,tz+xck,tzφ+δc+δtz+εck,tz]. (14) 4.3 Estimation Issues The previous subsections described our general empirical strategy, including our handling of the omitted variables bias. We discussed the use of control variables, employment of the full sets of exporting-country and initiation-year fixed effects, and stratification of the estimated equation at the industry and product level. However, estimating the effect of financial institutions and markets on export survival raises some further identification issues, like endogeneity or unobserved heterogeneity. In this section, we describe these issues and our handling of them. First, financial development may be endogenous to export performance (Do and Levchenko, 2007). In particular, our regressors measuring the strength of banks and the depth of stock markets in exporting countries could be subject to the endogeneity bias if credit and/or stock markets in the economy expand in anticipation of future export booms. Fortunately, this bias is less of an issue in our survival framework where the dependent variable is the hazard rate of highly disaggregated trade relationships and not the annual volume of export. Additionally, we measure all explanatory variables, including our bank and stock market proxies, in the initiation years of these individual trade relationships (tz). Second, there are some issues associated with using the Cox model in the context of export survival.33 The Cox model is based on the assumption that the explanatory variables influencing survival have a proportionate impact on the baseline hazard function. A change in covariates would thus shift the hazard function by the same factor in any period. Brenton, Saborowski, and von Uexkull (2010) provide evidence that this restrictive assumption typically does not hold in export-duration samples. The violation of the proportional hazards assumption may be due to unobserved heterogeneity that causes the estimated parameters to depend on the duration time. Alternatively, it may be that the effect of the explanatory variables is intrinsically non-proportional. Our Cox model is stratified to account for unobserved time-invariant heterogeneity at the industry or product level. This implies that the baseline hazard rate varies across industries and products. However, covariate effects are still assumed to be constant across industry or product strata.34 In order to account for unobserved heterogeneity and for the possibility of intrinsically non-proportional effects of explanatory variables, we will follow Araujo, Mion, and Ornelas (2014) and estimate a variant of our main specification using a linear probability model with fixed effects in one of the robustness checks reported in Online Appendix C.3. Finally, because all our financial variables (BCc,tz, StMc,tz, etc.) vary at the country-time level, where time tz stands for the year when the export spell has been initiated, we report in all tables robust standard errors clustered at the country*(initiation year) level. This procedure avoids biasing the standard errors downward for our financial development variables whose value is constant across product-level observations within the country-time clusters (Moulton, 1986, 1990; Cameron and Trivedi, 2005). We will cluster standard errors also at the country level (Bertrand, Duflo, and Mullainathan, 2004) and use the two-way clustering for standard errors (Cameron, Gelbach, and Miller, 2006) as additional robustness checks in Online Appendix C.1. 5. Empirical Results 5.1 The Role of Banks In Table III, we investigate the interplay between disciplinary pressures from product markets and banks toward exporting patterns congruent with the idea of comparative advantage. Table III reports estimates using Equation (11) in Columns (1)–(3) and Equation (12) in Columns (4)–(6). The dependent variable is the probability of exiting the US market (hazard rate, in the terminology of survival analysis) for product k exported from country c. All regressions control for fixed effects at the level of exporting countries and initiation years of the export spells. Table III. Banks and comparative advantage The dependent variable is the hazard rate on the US destination market for an export spell of product k from country c that started at time tz. All regressions are estimated using the Cox Proportional Hazard model (Cox, 1972) and control for country- and time-fixed effects. Estimations also allow for different baseline hazard rate across industries (Columns 1–3) and across products (Columns 4–6) by defining industry j and product k as strata variables. Our first variable of interest is the Euclidean distance of the RFI of product k to the factor endowment of exporting country c (distance to comparative advantage — DCAck,tz). Our second variable of interest is the interaction between the distance measure and banking development in country c proxied by bank credit over GDP (banks×distance to comparative advantage — BCc,tz*DCAck,tz). The control variables include the direct effect of bank credit over GDP (banks — BCc,tz) as well as GDP per capita of country c (GDPpcc,tz), countries’ endowments of physical capital (κc,tz) and human capital (hc,tz), corresponding physical and human capital intensities at the industry level (CapIntj, HumIntj), initial export value to the USA (initial_exportck,tz), total export value to the world market (total_exportck,tz), number of countries exporting product k to the USA (NSuppliersk,tz), and a dummy variable taking value one if the spell is a higher order spell (multiplespellck). All time-varying explanatory variables are measured in the initial year of the export spell tz. We report coefficients and not hazard ratios. Robust standard errors clustered at country*time (c*tz) level are in parentheses. *,**,*** denote statistical significance at the 10%, 5%, and 1% level, respectively. Dep. Var.: hazard rate  (1)  (2)  (3)  (4)  (5)  (6)  Distance to comparative advantage  0.126***  0.102***  0.076***  0.145***  0.109***  0.080***  (0.007)  (0.013)  (0.012)  (0.008)  (0.014)  (0.013)  Banks×distance to comparative advantage    0.041**  0.043**    0.068***  0.064***    (0.016)  (0.017)    (0.018)  (0.019)  Banks    –0.035  –0.039    –0.081**  –0.082*      (0.036)  (0.040)    (0.039)  (0.043)  GDPpc  0.141  0.235**  0.287**  0.207**  0.312***  0.332***    (0.099)  (0.103)  (0.115)  (0.096)  (0.100)  (0.112)  Physical capital      –0.053      0.035        (0.111)      (0.107)  Human capital      0.085      0.105        (0.065)      (0.066)  Physical capital× phys. cap. intensity      1.031***      1.011***      (0.165)      (0.167)  Human capital× hum. cap. intensity      –0.091***      –0.098***      (0.009)      (0.010)  Initial export  –0.084***  –0.084***  –0.085***  –0.091***  –0.091***  –0.093***    (0.005)  (0.005)  (0.005)  (0.005)  (0.005)  (0.005)  Total export  –0.104***  –0.105***  –0.104***  –0.121***  –0.121***  –0.120***    (0.003)  (0.003)  (0.003)  (0.003)  (0.003)  (0.003)  Number of suppliers  –0.020***  –0.020***  –0.020***  0.007***  0.006***  0.005***    (0.001)  (0.001)  (0.001)  (0.001)  (0.001)  (0.001)  Multiple spell  1.024***  1.013***  1.045***  1.026***  1.014***  1.033***    (0.090)  (0.091)  (0.094)  (0.083)  (0.084)  (0.084)  Exporting country FE  Yes  Yes  Yes  Yes  Yes  Yes  Year FE  Yes  Yes  Yes  Yes  Yes  Yes  ISIC4 industry strata  Yes  Yes  Yes        HS6 product strata        Yes  Yes  Yes  Observations  220,041  211,643  191,078  220,041  211,643  191,078  Dep. Var.: hazard rate  (1)  (2)  (3)  (4)  (5)  (6)  Distance to comparative advantage  0.126***  0.102***  0.076***  0.145***  0.109***  0.080***  (0.007)  (0.013)  (0.012)  (0.008)  (0.014)  (0.013)  Banks×distance to comparative advantage    0.041**  0.043**    0.068***  0.064***    (0.016)  (0.017)    (0.018)  (0.019)  Banks    –0.035  –0.039    –0.081**  –0.082*      (0.036)  (0.040)    (0.039)  (0.043)  GDPpc  0.141  0.235**  0.287**  0.207**  0.312***  0.332***    (0.099)  (0.103)  (0.115)  (0.096)  (0.100)  (0.112)  Physical capital      –0.053      0.035        (0.111)      (0.107)  Human capital      0.085      0.105        (0.065)      (0.066)  Physical capital× phys. cap. intensity      1.031***      1.011***      (0.165)      (0.167)  Human capital× hum. cap. intensity      –0.091***      –0.098***      (0.009)      (0.010)  Initial export  –0.084***  –0.084***  –0.085***  –0.091***  –0.091***  –0.093***    (0.005)  (0.005)  (0.005)  (0.005)  (0.005)  (0.005)  Total export  –0.104***  –0.105***  –0.104***  –0.121***  –0.121***  –0.120***    (0.003)  (0.003)  (0.003)  (0.003)  (0.003)  (0.003)  Number of suppliers  –0.020***  –0.020***  –0.020***  0.007***  0.006***  0.005***    (0.001)  (0.001)  (0.001)  (0.001)  (0.001)  (0.001)  Multiple spell  1.024***  1.013***  1.045***  1.026***  1.014***  1.033***    (0.090)  (0.091)  (0.094)  (0.083)  (0.084)  (0.084)  Exporting country FE  Yes  Yes  Yes  Yes  Yes  Yes  Year FE  Yes  Yes  Yes  Yes  Yes  Yes  ISIC4 industry strata  Yes  Yes  Yes        HS6 product strata        Yes  Yes  Yes  Observations  220,041  211,643  191,078  220,041  211,643  191,078  Table III. Banks and comparative advantage The dependent variable is the hazard rate on the US destination market for an export spell of product k from country c that started at time tz. All regressions are estimated using the Cox Proportional Hazard model (Cox, 1972) and control for country- and time-fixed effects. Estimations also allow for different baseline hazard rate across industries (Columns 1–3) and across products (Columns 4–6) by defining industry j and product k as strata variables. Our first variable of interest is the Euclidean distance of the RFI of product k to the factor endowment of exporting country c (distance to comparative advantage — DCAck,tz). Our second variable of interest is the interaction between the distance measure and banking development in country c proxied by bank credit over GDP (banks×distance to comparative advantage — BCc,tz*DCAck,tz). The control variables include the direct effect of bank credit over GDP (banks — BCc,tz) as well as GDP per capita of country c (GDPpcc,tz), countries’ endowments of physical capital (κc,tz) and human capital (hc,tz), corresponding physical and human capital intensities at the industry level (CapIntj, HumIntj), initial export value to the USA (initial_exportck,tz), total export value to the world market (total_exportck,tz), number of countries exporting product k to the USA (NSuppliersk,tz), and a dummy variable taking value one if the spell is a higher order spell (multiplespellck). All time-varying explanatory variables are measured in the initial year of the export spell tz. We report coefficients and not hazard ratios. Robust standard errors clustered at country*time (c*tz) level are in parentheses. *,**,*** denote statistical significance at the 10%, 5%, and 1% level, respectively. Dep. Var.: hazard rate  (1)  (2)  (3)  (4)  (5)  (6)  Distance to comparative advantage  0.126***  0.102***  0.076***  0.145***  0.109***  0.080***  (0.007)  (0.013)  (0.012)  (0.008)  (0.014)  (0.013)  Banks×distance to comparative advantage    0.041**  0.043**    0.068***  0.064***    (0.016)  (0.017)    (0.018)  (0.019)  Banks    –0.035  –0.039    –0.081**  –0.082*      (0.036)  (0.040)    (0.039)  (0.043)  GDPpc  0.141  0.235**  0.287**  0.207**  0.312***  0.332***    (0.099)  (0.103)  (0.115)  (0.096)  (0.100)  (0.112)  Physical capital      –0.053      0.035        (0.111)      (0.107)  Human capital      0.085      0.105        (0.065)      (0.066)  Physical capital× phys. cap. intensity      1.031***      1.011***      (0.165)      (0.167)  Human capital× hum. cap. intensity      –0.091***      –0.098***      (0.009)      (0.010)  Initial export  –0.084***  –0.084***  –0.085***  –0.091***  –0.091***  –0.093***    (0.005)  (0.005)  (0.005)  (0.005)  (0.005)  (0.005)  Total export  –0.104***  –0.105***  –0.104***  –0.121***  –0.121***  –0.120***    (0.003)  (0.003)  (0.003)  (0.003)  (0.003)  (0.003)  Number of suppliers  –0.020***  –0.020***  –0.020***  0.007***  0.006***  0.005***    (0.001)  (0.001)  (0.001)  (0.001)  (0.001)  (0.001)  Multiple spell  1.024***  1.013***  1.045***  1.026***  1.014***  1.033***    (0.090)  (0.091)  (0.094)  (0.083)  (0.084)  (0.084)  Exporting country FE  Yes  Yes  Yes  Yes  Yes  Yes  Year FE  Yes  Yes  Yes  Yes  Yes  Yes  ISIC4 industry strata  Yes  Yes  Yes        HS6 product strata        Yes  Yes  Yes  Observations  220,041  211,643  191,078  220,041  211,643  191,078  Dep. Var.: hazard rate  (1)  (2)  (3)  (4)  (5)  (6)  Distance to comparative advantage  0.126***  0.102***  0.076***  0.145***  0.109***  0.080***  (0.007)  (0.013)  (0.012)  (0.008)  (0.014)  (0.013)  Banks×distance to comparative advantage    0.041**  0.043**    0.068***  0.064***    (0.016)  (0.017)    (0.018)  (0.019)  Banks    –0.035  –0.039    –0.081**  –0.082*      (0.036)  (0.040)    (0.039)  (0.043)  GDPpc  0.141  0.235**  0.287**  0.207**  0.312***  0.332***    (0.099)  (0.103)  (0.115)  (0.096)  (0.100)  (0.112)  Physical capital      –0.053      0.035        (0.111)      (0.107)  Human capital      0.085      0.105        (0.065)      (0.066)  Physical capital× phys. cap. intensity      1.031***      1.011***      (0.165)      (0.167)  Human capital× hum. cap. intensity      –0.091***      –0.098***      (0.009)      (0.010)  Initial export  –0.084***  –0.084***  –0.085***  –0.091***  –0.091***  –0.093***    (0.005)  (0.005)  (0.005)  (0.005)  (0.005)  (0.005)  Total export  –0.104***  –0.105***  –0.104***  –0.121***  –0.121***  –0.120***    (0.003)  (0.003)  (0.003)  (0.003)  (0.003)  (0.003)  Number of suppliers  –0.020***  –0.020***  –0.020***  0.007***  0.006***  0.005***    (0.001)  (0.001)  (0.001)  (0.001)  (0.001)  (0.001)  Multiple spell  1.024***  1.013***  1.045***  1.026***  1.014***  1.033***    (0.090)  (0.091)  (0.094)  (0.083)  (0.084)  (0.084)  Exporting country FE  Yes  Yes  Yes  Yes  Yes  Yes  Year FE  Yes  Yes  Yes  Yes  Yes  Yes  ISIC4 industry strata  Yes  Yes  Yes        HS6 product strata        Yes  Yes  Yes  Observations  220,041  211,643  191,078  220,041  211,643  191,078  Columns (1)–(3) of Table III allow for different baseline hazard rate across industries by defining industry as the strata variable (Equation 11). The first column focuses on the disciplining impact of product markets. The positive and significant coefficient for the distance to comparative advantage (DCAck,tz) confirms the importance of competitive product markets in enforcing the optimal allocation of resources. Products with factor intensity far away from the endowment of the exporting country face a significantly higher probability of failure in the US destination market. In terms of control variables, higher initial values of exports both to the US and to all destinations (initial_exportck,tz,total_exportck,tz) decrease the hazard rate. Intuitively, products survive longer on the US market when the importers are willing to accept a higher initial shipment and when the exporting country has experience with exporting a given product to other markets as well. The coefficient for the multiple spell dummy (multiplespellck) is positive and significant, suggesting a higher risk of failure for products that repeatedly exit and re-enter the US market. The last product-related variable is the number of suppliers (NSuppliersk,tz), which has a negative impact on the hazard rate. This result is rather counter-intuitive as a higher number of countries exporting a given product should increase the competition in the destination market. A possible explanation could be the positive network effect among exporters (Koenig, 2009; Cassey and Schmeiser, 2010; Cadot et al., 2013). The effect of the GDP per capita of the exporting country (GDPpcc,tz) has no significant effect in this specification. The second column in Table III examines whether domestic banks provide an additional check on inefficient exporting. The regressors now also include the ratio of bank credit over the GDP in the exporting country (banks — BCc,tz) and an interaction term between this variable and the distance of exported product to the exporting country’s comparative advantage (banks × distance to comparative advantage — BCc,tz*DCAck,tz). A well-developed banking sector should in general help the exporters survive on foreign markets. Domestic banks (BCc,tz) indeed lower the hazard rate, but this direct effect is not statistically significant. In contrast, the interaction term between banks and distance to comparative advantage (BCc,tz*DCAck,tz) has a positive and statistically significant effect on the hazard rate. The coefficient for the distance to comparative advantage measure (DCAck,tz) also remains highly significant. Joint interpretation of positive estimates for both coefficients (DCAck,tz, BCc,tz*DCAck,tz) suggests that banks push exporters to abandon products that are already facing an uphill battle on the US destination market due to the suboptimal use of the domestic factor endowments. With regard to our control variables, the GDP per capita of the exporting country (GDPpcc,tz) has a positive and significant effect on the hazard rate. While this result might appear counter-intuitive at first sight, two features of our estimation strategy provide an explanation. First, we control for exporting-country fixed effects in all regressions. The effect of GDPpcc,tz is thus identified solely from variations within countries over time. These variations emerge both from growth trend and from business cycle fluctuations. Second, we measure all time-varying regressors in the first year of an export spell (tz). Economically, the positive estimated coefficient for GDPpcc,tz would then imply that exports initiated at the peak of a business cycle in the country of origin face higher risk of failure. Possible reasons for this effect include over-confidence of exporters during a boom or difficulties to maintain the costly presence in the US destination market once the business climate at home deteriorates. In the third column of Table III, we add interaction terms between the exporting countries’ factor endowments and the corresponding factor intensities at the sectoral level (physical capital × physical capital intensity — κc,tz*CapIntj, human capital × human capital intensity — hc,tz*HumIntj). These interacted variables control for the possibility that products from industries extensively using physical or human capital survive longer in the US destination market if the exporting country is abundant in such a capital. When adding these interaction terms, we also control for the direct effect of exporting countries’ endowments in physical capital (κc,tz) and human capital (hc,tz)35 while the direct effect of the corresponding factor intensities (CapIntj, HumIntj) is captured by the industry strata effects.36 Our main interaction term capturing the disciplining effects of banks (banks × distance to comparative advantage — BCc,tz*DCAck,tz) maintains a positive and statistically significant coefficient. Similarly, the direct effect of distance to comparative advantage (DCAck,tz) still translates into a higher hazard rate of exports, confirming the disciplining impact of the competitive US market. The interaction term between human capital and human capital intensity (hc,tz*HumIntj) has the expected negative sign while the direct effects of endowments in physical and human capital (κc,tz, hc,tz) are insignificant. The interaction between physical capital and physical capital intensity (κc,tz*CapIntj) has a positive sign, suggesting that products of capital-intensive industries coming from capital-abundant countries face a higher risk of exit from the US destination market. This rather counter-intuitive result is similar to Manova (2008), who finds a negative effect of this interaction term on export volume. In Columns (4)–(6) of Table III, we explore the disciplining effects of the US destination market and banking system in the country of origin within a more stringent econometric specification. We replicate the specifications in Columns (1)–(3) but apply the stratification according to HS six-digit products (Equation 12) rather than ISIC four-digit industries. Concerning our main focus on the interplay between disciplining forces of product markets and banks, the results confirm in qualitative terms those obtained in Columns (1)–(3). Both distance to the comparative advantage and the interaction of this variable with the strength of banking system in the country of origin retain a positive and significant effect on products’ probability of exit from the US market. Furthermore, both the point estimate and the significance level of our main interaction term (banks × distance to comparative advantage — BCc,tz*DCAck,tz) increase after stratification at the product rather than just industry level. Accounting for unobserved heterogeneity at the product level also yields two changes regarding our control variables. First, the number of suppliers (NSuppliersk,tz) — measuring the number of countries exporting the given product to the USA and thus capturing the strength of import competition on the US product markets — now has the expected positive sign, increasing the products’ hazard rate. Second, the expected negative direct effect of banks (BCc,tz) on the hazard rate is now statistically significant in the fifth and sixth column where it is included. Column (6) of Table III represents our preferred specification. Unless stated otherwise, later estimations take this specification as the reference point. Overall, the results from Table III suggest a statistically significant disciplining effect of both product markets and banks when it comes to terminating exports to the USA that do not optimally use the available factors of production in the countries of origin. Before moving to further empirical analysis, let us consider the economic significance of our results. One way to get a sense of the economic magnitude regarding our results is as follows. Let us assume that Turkey would reach the average level of banking development among the OECD countries, that is, the value of the variable BCc,tz would increase from 0.16 to 0.84. Equation (12) and estimated coefficients from Column (6) of Table III then imply that the hazard rate for any given product exported from Turkey to the USA would change by β2ΔBCc,tz*DCAck,tz+γΔBCc,tz=ΔBCc,tz*[β2DCAck,tz+γ]=(0.84−0.16)*[0.064*DCAck,tz−0.082]. In case of a product that is far away from the comparative advantage of Turkey like “Equipment for automatically developing photographic films” (DCAck,tz=1.90, in the top 25th percentile), the hazard rate of exports would then increase by approximately 2.7%: (0.84−0.16)*[0.064*1.90−0.082]=0.027. In case of a product well aligned with Turkish comparative advantage like “Footwear with leather” (DCAck,tz=0.71, in the bottom 25th percentile), the hazard rate would actually decrease by approximately 2.5%: (0.84−0.16)*[0.064*0.71−0.082]=−0.025. In case of Turkey, a better developed banking sector could thus increase the difference in export hazard rates between products in the top and bottom 25th percentile of distance to comparative advantage by 5.2%. This is a large effect, also when compared with the direct disciplining impact of product markets. Taking the example of the same two Turkish exports but abstracting from the impact of banks and considering only the effect of distance to comparative advantage would namely suggest that the product “Equipment for automatically developing photographic films” faces approximately a 17% higher hazard rate on the US destination market than the product “Footwear with leather”.37,38 In Table IV, we further examine the disciplining role banks have on exports not corresponding to the notion of comparative advantage. In Columns (1) and (2), we look at alternative proxies of the strength of the banking sector in the exporting country. In Columns (3)–(6), the focus is on the disciplining role of traditional banks versus the impact of various other financial institutions. Table IV. Alternative measures of banks and comparative advantage The dependent variable is the hazard rate on the US destination market for an export spell of product k from country c that started at time tz. All regressions are estimated using the Cox Proportional Hazard model (Cox, 1972) and control for country- and time-fixed effects and allow for different baseline hazard rate across products by defining product k as strata variable. Bank assets are claims on domestic real non-financial sector by deposit money banks as a share of GDP, total credit is the ratio of private credit by deposit money banks and other financial institutions over GDP, non-bank credit is the credit extended by other financial institutions than banks as a share of GDP (i.e., the difference between values for total and banks credit), credit to government is the credit extended to the general government sector, including central, state and local governments, and social security funds. Other variables are defined in Table III. Full set of controls corresponds to Column (6) of Table III. We report coefficients and not hazard ratios. Robust standard errors clustered at country*time (c*tz) level are in parentheses. *,**,*** denote statistical significance at the 10%, 5%, and 1% level, respectively. Dep. Var.: hazard rate  (1)  (2)  (3)  (4)  (5)  (6)  (7)  Distance to comparative advantage  0.078***  0.083***  0.120***  0.081***  0.086***  0.047*  0.001  (0.014)  (0.014)  (0.008)  (0.014)  (0.020)  (0.025)  (0.022)  Banks×distance to comparative advantage        0.064***    0.055***  0.053**        (0.019)    (0.021)  (0.021)  Bank assets×distance to comp. advantage  0.051***              (0.019)              Total credit×distance to comp. advantage    0.047**              (0.020)            Non-bank credit×dist. to comp. advantage      –0.053  –0.005            (0.062)  (0.057)        Credit to government× dist. to comp. adv.          –0.001**  –0.001***            (0.000)  (0.000)    Banks        –0.046    –0.211**  –0.147          (0.043)    (0.096)  (0.093)  Bank assets  –0.113*                (0.061)              Total credit    0.022                (0.042)            Non-bank credit      0.147*  0.104              (0.088)  (0.086)        Credit to government          –0.001  –0.002              (0.002)  (0.002)    Full set of controls  Yes  Yes  Yes  Yes  Yes  Yes  Yes  Exporting country FE  Yes  Yes  Yes  Yes  Yes  Yes  Yes  Year FE  Yes  Yes  Yes  Yes  Yes  Yes  Yes  HS6 product strata  Yes  Yes  Yes  Yes  Yes  Yes  Yes  Observations  196,627  197,450  191,078  191,078  54,329  54,329  54,329  Dep. Var.: hazard rate  (1)  (2)  (3)  (4)  (5)  (6)  (7)  Distance to comparative advantage  0.078***  0.083***  0.120***  0.081***  0.086***  0.047*  0.001  (0.014)  (0.014)  (0.008)  (0.014)  (0.020)  (0.025)  (0.022)  Banks×distance to comparative advantage        0.064***    0.055***  0.053**        (0.019)    (0.021)  (0.021)  Bank assets×distance to comp. advantage  0.051***              (0.019)              Total credit×distance to comp. advantage    0.047**              (0.020)            Non-bank credit×dist. to comp. advantage      –0.053  –0.005            (0.062)  (0.057)        Credit to government× dist. to comp. adv.          –0.001**  –0.001***            (0.000)  (0.000)    Banks        –0.046    –0.211**  –0.147          (0.043)    (0.096)  (0.093)  Bank assets  –0.113*                (0.061)              Total credit    0.022                (0.042)            Non-bank credit      0.147*  0.104              (0.088)  (0.086)        Credit to government          –0.001  –0.002              (0.002)  (0.002)    Full set of controls  Yes  Yes  Yes  Yes  Yes  Yes  Yes  Exporting country FE  Yes  Yes  Yes  Yes  Yes  Yes  Yes  Year FE  Yes  Yes  Yes  Yes  Yes  Yes  Yes  HS6 product strata  Yes  Yes  Yes  Yes  Yes  Yes  Yes  Observations  196,627  197,450  191,078  191,078  54,329  54,329  54,329  Table IV. Alternative measures of banks and comparative advantage The dependent variable is the hazard rate on the US destination market for an export spell of product k from country c that started at time tz. All regressions are estimated using the Cox Proportional Hazard model (Cox, 1972) and control for country- and time-fixed effects and allow for different baseline hazard rate across products by defining product k as strata variable. Bank assets are claims on domestic real non-financial sector by deposit money banks as a share of GDP, total credit is the ratio of private credit by deposit money banks and other financial institutions over GDP, non-bank credit is the credit extended by other financial institutions than banks as a share of GDP (i.e., the difference between values for total and banks credit), credit to government is the credit extended to the general government sector, including central, state and local governments, and social security funds. Other variables are defined in Table III. Full set of controls corresponds to Column (6) of Table III. We report coefficients and not hazard ratios. Robust standard errors clustered at country*time (c*tz) level are in parentheses. *,**,*** denote statistical significance at the 10%, 5%, and 1% level, respectively. Dep. Var.: hazard rate  (1)  (2)  (3)  (4)  (5)  (6)  (7)  Distance to comparative advantage  0.078***  0.083***  0.120***  0.081***  0.086***  0.047*  0.001  (0.014)  (0.014)  (0.008)  (0.014)  (0.020)  (0.025)  (0.022)  Banks×distance to comparative advantage        0.064***    0.055***  0.053**        (0.019)    (0.021)  (0.021)  Bank assets×distance to comp. advantage  0.051***              (0.019)              Total credit×distance to comp. advantage    0.047**              (0.020)            Non-bank credit×dist. to comp. advantage      –0.053  –0.005            (0.062)  (0.057)        Credit to government× dist. to comp. adv.          –0.001**  –0.001***            (0.000)  (0.000)    Banks        –0.046    –0.211**  –0.147          (0.043)    (0.096)  (0.093)  Bank assets  –0.113*                (0.061)              Total credit    0.022                (0.042)            Non-bank credit      0.147*  0.104              (0.088)  (0.086)        Credit to government          –0.001  –0.002              (0.002)  (0.002)    Full set of controls  Yes  Yes  Yes  Yes  Yes  Yes  Yes  Exporting country FE  Yes  Yes  Yes  Yes  Yes  Yes  Yes  Year FE  Yes  Yes  Yes  Yes  Yes  Yes  Yes  HS6 product strata  Yes  Yes  Yes  Yes  Yes  Yes  Yes  Observations  196,627  197,450  191,078  191,078  54,329  54,329  54,329  Dep. Var.: hazard rate  (1)  (2)  (3)  (4)  (5)  (6)  (7)  Distance to comparative advantage  0.078***  0.083***  0.120***  0.081***  0.086***  0.047*  0.001  (0.014)  (0.014)  (0.008)  (0.014)  (0.020)  (0.025)  (0.022)  Banks×distance to comparative advantage        0.064***    0.055***  0.053**        (0.019)    (0.021)  (0.021)  Bank assets×distance to comp. advantage  0.051***              (0.019)              Total credit×distance to comp. advantage    0.047**              (0.020)            Non-bank credit×dist. to comp. advantage      –0.053  –0.005            (0.062)  (0.057)        Credit to government× dist. to comp. adv.          –0.001**  –0.001***            (0.000)  (0.000)    Banks        –0.046    –0.211**  –0.147          (0.043)    (0.096)  (0.093)  Bank assets  –0.113*                (0.061)              Total credit    0.022                (0.042)            Non-bank credit      0.147*  0.104              (0.088)  (0.086)        Credit to government          –0.001  –0.002              (0.002)  (0.002)    Full set of controls  Yes  Yes  Yes  Yes  Yes  Yes  Yes  Exporting country FE  Yes  Yes  Yes  Yes  Yes  Yes  Yes  Year FE  Yes  Yes  Yes  Yes  Yes  Yes  Yes  HS6 product strata  Yes  Yes  Yes  Yes  Yes  Yes  Yes  Observations  196,627  197,450  191,078  191,078  54,329  54,329  54,329  Column (1) of Table IV reports the results of our preferred specification from Column (6) in Table III, replacing the credit by banks to private real sector (BCc,tz) with the overall bank assets (BAc,tz) — both normalized by GDP of a given country. The main interaction term in the first column of Table IV is thus bank assets × distance to comparative advantage (BAc,tz*DCAck,tz). In Column (2), we replace the bank credit (BCc,tz) with the total credit by both banks and other financial institutions (TCc,tz) — again, normalized by GDP. The main interaction term thus becomes total credit × distance to comparative advantage (TCc,tz*DCAck,tz). Compared with Table III, Column (1) of Table IV looks at a broader measure of the claims banks have on the rest of the economy rather than at extended credit alone while Column (2) looks at credit extended by a broader set of financial institutions rather than by the deposit-taking money banks alone. Our main results from Table III are robust to using these broader measures of the importance of financial institutions for the real economy. Comparing Column (2) from Table IV to Column (6) from Table III, it seems that the results for a broader set of financial institutions are somewhat weaker than when looking at traditional banks alone. In particular, the main interaction term in Column (2) of Table IV (total credit × distance to comparative advantage — TCc,tz*DCAck,tz) has a lower level of significance, suggesting a possibly lower disciplining effect of other financial institutions on exporting activities of the real sector. This could be explained by the presence of state banks and other governmental financial institutions in many countries in our sample. These institution do not rely on deposit taking but receive a bulk of their funding directly from government. Arguably, such institutions face lower pressure to invest their funds efficiently and are thus less suitable as a disciplining device toward their own borrowers. We examine this hypothesis more directly in Column (3) of Table IV, where our main financial proxy is the difference between total credit by all banks and financial institutions and the credit extended by deposit-taking money banks, capturing the credit extended by other financial institutions (non-bank credit — NBCc,tz). The non-significant interaction term between non-bank credit and distance to comparative advantage (NBCc,tz*DCAck,tz) confirms that when it comes to inefficient exporting, other financial institutions lack the disciplining impact of deposit-taking banks. This picture is further reinforced in Column (4) where the two proxies capturing the strength of banks and the strength of other financial institutions jointly enter the regression. While the interaction term between banks and distance to comparative advantage has the expected sign and a high level of statistical significance, the interaction of non-bank credit and distance to comparative advantage remains insignificant. In Columns (5) and (6) of Table IV, we distinguish between credit extended to private non-financial sector and credit extended to government. If our results are driven by the specific disciplining role of banks toward firms rather than by some general pattern of exporting during credit booms, credit to government (GCc,tz) should not have the same impact on terminating inefficient exports. And indeed, high levels of credit extended to government of the exporting country have actually a worsening impact on resource allocation, by disproportionately lowering the hazard rate of exports not corresponding to comparative advantage of exporting country. In particular, the interaction term between credit to government and distance to comparative advantage (GCc,tz*DCAck,tz) is negative and significant in Column (5) of Table IV. This result for credit to government remains the same in Column (6) where we allow our main interaction of banks and distance to comparative advantage to enter the regression. In contrast, the main interaction term itself is significant and of expected sign in Column (6). The last column of Table IV performs an additional robustness test due to the significantly lower number of observations in Columns (5) and (6). Data on government credit are available only for a smaller subset of countries, so that the results in Column (6) could be partially driven by multicollinearity between the bank credit to private non-financial sector and the credit to government in the small sample. In Column (7) of Table IV, we thus re-run the estimation on the smaller sample from previous two columns but allow only the interaction of banks and distance to comparative advantage to enter the regression. Even without the presence of government credit in the estimation, the disciplining impact of banks on export does survive, as can be seen from the positive and significant impact of the main interaction term (banks × distance to comparative advantage — BCc,tz*DCAck,tz) on the hazard rate of exports into USA. Overall, results in Table IV provide further evidence that a strong banking sector in the exporting country can enhance the disciplining impact of the US destination market when it comes to products not corresponding to the comparative advantage of the country exporting to the USA. While it does not matter whether the importance of banking sector in the economy is measured by credit extended to private non-financial sector or by the overall bank assets, other financial institutions do not seem to exert the same disciplining effect as banks. And the credit extended to government has, if anything, a negative impact on the resource allocation. These results are in line with conventional wisdom both about monitoring abilities of banks and about potential misallocative effects related to financial flows where either the lender or borrower has access to taxpayers’ money. We plan to look deeper into this issue in our future work. 5.1.1 Banks and Bond Markets When it comes to the disciplining role of external debt, existing literature often does not distinguish between private creditors (e.g., banks extending credit to firms) and public creditors (e.g., holders of publicly tradable corporate bonds). However, Denis and Mihov (2003) extend the arguments by Stulz (1990) and Berger, Ofek, and Yermack (1997) about the role of managerial discretion in choosing the debt/equity ratio to the case of public versus private debt and argue that private lenders like banks are better equipped to push firms’ management toward optimal level of investment than public debtholders are. This is due to their concentrated holdings of external debt and superior access to information. Lin et al. (2013) make a similar argument regarding the monitoring abilities of banks versus the holders of public debt.39 For completeness, Table A.II in the Appendix A reports additional regressions that include three proxies for the strength of bond markets in countries exporting to the USA: domestic private bonds (Columns 2 and 3), international private bonds (Columns 5 and 6), and international bonds issued by non-financial firms only (Columns 8 and 9) — all normalized by the GDP of the exporting country. Given the somewhat limited availability of internationally comparable data on bond markets, the number of observations is lower for all three bond markets proxies compared with our main bank proxy. For better comparison, Columns (1), (4), and (7) of Table A.II therefore report results in these reduced samples for the specification from Column (6) of Table III that focuses on our main banking channel. Columns (2), (5), and (8) report the regression results with the interactions of our three bond markets proxies with distance to comparative advantage measure. Finally, Columns (3), (6), and (9) allow for the possibility that both strong banks and well-developed bond markets in exporting countries increase the hazard rate in the US destination market disproportionately for those products that do not correspond to the comparative advantage of the exporting country. In Table A.II, the bank interaction term is always positive and significant when included, confirming the importance of the banking sector in the exporting country when it comes to enhancing the disciplining impact of the US destination market on products far away from the comparative advantage of the exporting country. When it comes to the bond markets, only the interaction term of the first bond proxy with distance to comparative advantage enters significantly the regressions in Columns (2) and (3). This results may be explained by the fact that both this proxy and the proxy entering Columns (5) and (6) of Table A.II include bonds issued by financial firms like banks themselves. These bonds can arguably contribute to financing bank operations that support exporting activities of firms and do not necessarily show up in data on bank loans (e.g., letters of credit, export insurance). In future research, we plan to look further into disentangling these channels. Importantly for the purposes of this paper, the interaction term of distance to comparative advantage with the one bond proxy that does capture solely the bonds issued by non-financial firms is not significant in Columns (8) and (9) of Table A.II. Overall, these results suggest that private debtholders like banks are indeed more effective than bondholders when it comes to disciplining firms’ export behavior. 5.2 The Role of Stock Markets In Table V, we repeat the estimations from Table III, but in the main interaction term we replace the ratio of private credit over GDP with the ratio of stockmarket capitalization over GDP. Specifications in Columns (1)–(3) allow for a different baseline hazard rate across 118 ISIC four-digit industries (Equation 13) while the regressions reported in Columns (4)–(6) allow the baseline hazard rate to vary across all 4,562 HS six-digit products (Equation 14). A positive coefficient for the resulting variable (stock markets × distance to comparative advantage — StMc,tz*DCAck,tz) would suggest that shareholders are also able to exert a disciplining influence on exports that do not correspond to the exporting country’s comparative advantage. The results in Table V do not support this hypothesis. The interaction term between stock market capitalization and distance to comparative advantage is never significant and once even enters the regression with the wrong sign. Table V. Stock markets and comparative advantage The dependent variable is the hazard rate on the US destination market for an export spell of product k from country c that started at time tz. All regressions are estimated using the Cox Proportional Hazard model (Cox, 1972) and control for country- and time-fixed effects. Estimations also allow for different baseline hazard rate across industries (Columns 1–3) and across products (Columns 4–6) by defining industry j and product k as strata variables. Stock markets development of country c is represented by the ratio of stock market capitalization over GDP (stock markets — StMc,tz). Other variables are defined in Table III. We report coefficients and not hazard ratios. Robust standard errors clustered at country*time (c*tz) level are in parentheses. *, **, and *** denote statistical significance at the 10%, 5%, and 1% level, respectively. Dep. Var.: hazard rate  (1)  (2)  (3)  (4)  (5)  (6)  Distance to comparative advantage  0.126***  0.131***  0.098***  0.145***  0.147***  0.108***  (0.007)  (0.011)  (0.010)  (0.008)  (0.012)  (0.011)  Stock markets×distance to comp. advantage    –0.000  0.004    0.005  0.006    (0.012)  (0.012)    (0.014)  (0.013)  Stock markets    0.031  0.025    0.026  0.024      (0.026)  (0.028)    (0.028)  (0.030)  GDPpc  0.141  0.111  0.210  0.207**  0.193*  0.271**    (0.099)  (0.115)  (0.132)  (0.096)  (0.112)  (0.132)  Physical capital      –0.110      –0.051        (0.121)      (0.119)  Human capital      0.057      0.076        (0.065)      (0.066)  Physical capital×phys. cap. intensity      0.847***      0.851***        (0.196)      (0.195)  Human capital×hum. cap. intensity      –0.110***      –0.118***        (0.010)      (0.011)  Initial export  –0.084***  –0.085***  –0.087***  –0.091***  –0.093***  –0.096***    (0.005)  (0.005)  (0.005)  (0.005)  (0.005)  (0.005)  Total export  –0.104***  –0.105***  –0.103***  –0.121***  –0.126***  –0.124***    (0.003)  (0.003)  (0.003)  (0.003)  (0.003)  (0.004)  Number of suppliers  –0.020***  –0.021***  –0.021***  0.007***  0.008***  0.006***    (0.001)  (0.001)  (0.001)  (0.001)  (0.001)  (0.001)  Multiple spell  1.024***  1.076***  1.116***  1.026***  1.076***  1.101***    (0.090)  (0.099)  (0.104)  (0.083)  (0.091)  (0.092)  Exporting country FE  Yes  Yes  Yes  Yes  Yes  Yes  Year FE  Yes  Yes  Yes  Yes  Yes  Yes  ISIC4 industry strata  Yes  Yes  Yes        HS6 product strata        Yes  Yes  Yes  Observations  220,041  203,649  182,592  220,041  203,649  182,592  Dep. Var.: hazard rate  (1)  (2)  (3)  (4)  (5)  (6)  Distance to comparative advantage  0.126***  0.131***  0.098***  0.145***  0.147***  0.108***  (0.007)  (0.011)  (0.010)  (0.008)  (0.012)  (0.011)  Stock markets×distance to comp. advantage    –0.000  0.004    0.005  0.006    (0.012)  (0.012)    (0.014)  (0.013)  Stock markets    0.031  0.025    0.026  0.024      (0.026)  (0.028)    (0.028)  (0.030)  GDPpc  0.141  0.111  0.210  0.207**  0.193*  0.271**    (0.099)  (0.115)  (0.132)  (0.096)  (0.112)  (0.132)  Physical capital      –0.110      –0.051        (0.121)      (0.119)  Human capital      0.057      0.076        (0.065)      (0.066)  Physical capital×phys. cap. intensity      0.847***      0.851***        (0.196)      (0.195)  Human capital×hum. cap. intensity      –0.110***      –0.118***        (0.010)      (0.011)  Initial export  –0.084***  –0.085***  –0.087***  –0.091***  –0.093***  –0.096***    (0.005)  (0.005)  (0.005)  (0.005)  (0.005)  (0.005)  Total export  –0.104***  –0.105***  –0.103***  –0.121***  –0.126***  –0.124***    (0.003)  (0.003)  (0.003)  (0.003)  (0.003)  (0.004)  Number of suppliers  –0.020***  –0.021***  –0.021***  0.007***  0.008***  0.006***    (0.001)  (0.001)  (0.001)  (0.001)  (0.001)  (0.001)  Multiple spell  1.024***  1.076***  1.116***  1.026***  1.076***  1.101***    (0.090)  (0.099)  (0.104)  (0.083)  (0.091)  (0.092)  Exporting country FE  Yes  Yes  Yes  Yes  Yes  Yes  Year FE  Yes  Yes  Yes  Yes  Yes  Yes  ISIC4 industry strata  Yes  Yes  Yes        HS6 product strata        Yes  Yes  Yes  Observations  220,041  203,649  182,592  220,041  203,649  182,592  Table V. Stock markets and comparative advantage The dependent variable is the hazard rate on the US destination market for an export spell of product k from country c that started at time tz. All regressions are estimated using the Cox Proportional Hazard model (Cox, 1972) and control for country- and time-fixed effects. Estimations also allow for different baseline hazard rate across industries (Columns 1–3) and across products (Columns 4–6) by defining industry j and product k as strata variables. Stock markets development of country c is represented by the ratio of stock market capitalization over GDP (stock markets — StMc,tz). Other variables are defined in Table III. We report coefficients and not hazard ratios. Robust standard errors clustered at country*time (c*tz) level are in parentheses. *, **, and *** denote statistical significance at the 10%, 5%, and 1% level, respectively. Dep. Var.: hazard rate  (1)  (2)  (3)  (4)  (5)  (6)  Distance to comparative advantage  0.126***  0.131***  0.098***  0.145***  0.147***  0.108***  (0.007)  (0.011)  (0.010)  (0.008)  (0.012)  (0.011)  Stock markets×distance to comp. advantage    –0.000  0.004    0.005  0.006    (0.012)  (0.012)    (0.014)  (0.013)  Stock markets    0.031  0.025    0.026  0.024      (0.026)  (0.028)    (0.028)  (0.030)  GDPpc  0.141  0.111  0.210  0.207**  0.193*  0.271**    (0.099)  (0.115)  (0.132)  (0.096)  (0.112)  (0.132)  Physical capital      –0.110      –0.051        (0.121)      (0.119)  Human capital      0.057      0.076        (0.065)      (0.066)  Physical capital×phys. cap. intensity      0.847***      0.851***        (0.196)      (0.195)  Human capital×hum. cap. intensity      –0.110***      –0.118***        (0.010)      (0.011)  Initial export  –0.084***  –0.085***  –0.087***  –0.091***  –0.093***  –0.096***    (0.005)  (0.005)  (0.005)  (0.005)  (0.005)  (0.005)  Total export  –0.104***  –0.105***  –0.103***  –0.121***  –0.126***  –0.124***    (0.003)  (0.003)  (0.003)  (0.003)  (0.003)  (0.004)  Number of suppliers  –0.020***  –0.021***  –0.021***  0.007***  0.008***  0.006***    (0.001)  (0.001)  (0.001)  (0.001)  (0.001)  (0.001)  Multiple spell  1.024***  1.076***  1.116***  1.026***  1.076***  1.101***    (0.090)  (0.099)  (0.104)  (0.083)  (0.091)  (0.092)  Exporting country FE  Yes  Yes  Yes  Yes  Yes  Yes  Year FE  Yes  Yes  Yes  Yes  Yes  Yes  ISIC4 industry strata  Yes  Yes  Yes        HS6 product strata        Yes  Yes  Yes  Observations  220,041  203,649  182,592  220,041  203,649  182,592  Dep. Var.: hazard rate  (1)  (2)  (3)  (4)  (5)  (6)  Distance to comparative advantage  0.126***  0.131***  0.098***  0.145***  0.147***  0.108***  (0.007)  (0.011)  (0.010)  (0.008)  (0.012)  (0.011)  Stock markets×distance to comp. advantage    –0.000  0.004    0.005  0.006    (0.012)  (0.012)    (0.014)  (0.013)  Stock markets    0.031  0.025    0.026  0.024      (0.026)  (0.028)    (0.028)  (0.030)  GDPpc  0.141  0.111  0.210  0.207**  0.193*  0.271**    (0.099)  (0.115)  (0.132)  (0.096)  (0.112)  (0.132)  Physical capital      –0.110      –0.051        (0.121)      (0.119)  Human capital      0.057      0.076        (0.065)      (0.066)  Physical capital×phys. cap. intensity      0.847***      0.851***        (0.196)      (0.195)  Human capital×hum. cap. intensity      –0.110***      –0.118***        (0.010)      (0.011)  Initial export  –0.084***  –0.085***  –0.087***  –0.091***  –0.093***  –0.096***    (0.005)  (0.005)  (0.005)  (0.005)  (0.005)  (0.005)  Total export  –0.104***  –0.105***  –0.103***  –0.121***  –0.126***  –0.124***    (0.003)  (0.003)  (0.003)  (0.003)  (0.003)  (0.004)  Number of suppliers  –0.020***  –0.021***  –0.021***  0.007***  0.008***  0.006***    (0.001)  (0.001)  (0.001)  (0.001)  (0.001)  (0.001)  Multiple spell  1.024***  1.076***  1.116***  1.026***  1.076***  1.101***    (0.090)  (0.099)  (0.104)  (0.083)  (0.091)  (0.092)  Exporting country FE  Yes  Yes  Yes  Yes  Yes  Yes  Year FE  Yes  Yes  Yes  Yes  Yes  Yes  ISIC4 industry strata  Yes  Yes  Yes        HS6 product strata        Yes  Yes  Yes  Observations  220,041  203,649  182,592  220,041  203,649  182,592  Table VI further explores whether stock markets in exporting countries play a role in enhancing the disciplining effects of the US product market. Columns (1) and (2) report the results of estimations analogous to Column (6) of Table V with alternative measures of the strength of stock markets in the exporting country. These measures are the value of stock market transactions relative to GDP—stock market value traded (StMVTc,tz) and the ratio of the value of total shares traded to market capitalization—stock market turnover (StMTc,tz). Both alternative measures are supposed to capture not only size but also liquidity of the stock markets in a given country (for a more detailed discussion of various stock market measures see, e.g., Strieborny and Kukenova, 2016). Like with all other financial proxies, the two alternative measures capturing the depth of stock markets in a given country exporting to the USA enter the regression both directly and interacted with the country-product measure of distance to comparative advantage. In Column (1) of Table VI, the interaction of stock market value traded and distance to comparative advantage is insignificant, confirming the results using stock market capitalization in Table V — deep stock markets do not seem to complement the disciplining effect of the US product market that is captured by the positive and significant direct impact of distance to comparative advantage. In Column (2), the interaction of stock market turnover and distance to comparative advantage is marginally significant but with the wrong sign. Table VI. Alternative measures of stock markets and comparative advantage The dependent variable is the hazard rate on the US destination market for an export spell of product k from country c that started at time tz. All regressions are estimated using the Cox Proportional Hazard model (Cox, 1972) and control for country- and time-fixed effects and allow for different baseline hazard rate across products by defining product k as strata variable. Stock markets is the ratio of stock market capitalization over GDP, stock market value traded is the value of stock market transactions relative to GDP, and stock market turnover is the ratio of the value of total shares traded to market capitalization. Other variables are defined in Table III. Full set of controls corresponds to Column (6) of Table III. We report coefficients and not hazard ratios. Robust standard errors clustered at country*time (c*tz) level are in parentheses. *, **, and *** denote statistical significance at the 10%, 5%, and 1% level, respectively. Dep. Var.: hazard rate  (1)  (2)  (3)  (4)  (5)  Distance to comparative advantage  0.117***  0.124***  0.086***  0.089***  0.094***    (0.010)  (0.010)  (0.015)  (0.015)  (0.016)  Banks×dist. to comp. advantage      0.064***  0.071***  0.054***        (0.023)  (0.022)  (0.020)  Stock markets×dist. to comp. advantage      –0.017            (0.015)      Stock market value traded×dist. to comp. advantage  –0.017      –0.047***    (0.014)      (0.017)    Stock market turnover×dist. to comp. advantage    –0.018*      –0.015    (0.009)      (0.009)  Banks      –0.060  –0.071  –0.050        (0.044)  (0.044)  (0.042)  Stock markets      0.053*            (0.030)      Stock market value traded  0.101***      0.132***      (0.028)      (0.030)    Stock market turnover    0.037*      0.020      (0.021)      (0.021)  Full set of controls  Yes  Yes  Yes  Yes  Yes  Exporting country FE  Yes  Yes  Yes  Yes  Yes  Year FE  Yes  Yes  Yes  Yes  Yes  HS6 product strata  Yes  Yes  Yes  Yes  Yes  Observations  180,479  179,311  175,481  173,368  172,200  Dep. Var.: hazard rate  (1)  (2)  (3)  (4)  (5)  Distance to comparative advantage  0.117***  0.124***  0.086***  0.089***  0.094***    (0.010)  (0.010)  (0.015)  (0.015)  (0.016)  Banks×dist. to comp. advantage      0.064***  0.071***  0.054***        (0.023)  (0.022)  (0.020)  Stock markets×dist. to comp. advantage      –0.017            (0.015)      Stock market value traded×dist. to comp. advantage  –0.017      –0.047***    (0.014)      (0.017)    Stock market turnover×dist. to comp. advantage    –0.018*      –0.015    (0.009)      (0.009)  Banks      –0.060  –0.071  –0.050        (0.044)  (0.044)  (0.042)  Stock markets      0.053*            (0.030)      Stock market value traded  0.101***      0.132***      (0.028)      (0.030)    Stock market turnover    0.037*      0.020      (0.021)      (0.021)  Full set of controls  Yes  Yes  Yes  Yes  Yes  Exporting country FE  Yes  Yes  Yes  Yes  Yes  Year FE  Yes  Yes  Yes  Yes  Yes  HS6 product strata  Yes  Yes  Yes  Yes  Yes  Observations  180,479  179,311  175,481  173,368  172,200  Table VI. Alternative measures of stock markets and comparative advantage The dependent variable is the hazard rate on the US destination market for an export spell of product k from country c that started at time tz. All regressions are estimated using the Cox Proportional Hazard model (Cox, 1972) and control for country- and time-fixed effects and allow for different baseline hazard rate across products by defining product k as strata variable. Stock markets is the ratio of stock market capitalization over GDP, stock market value traded is the value of stock market transactions relative to GDP, and stock market turnover is the ratio of the value of total shares traded to market capitalization. Other variables are defined in Table III. Full set of controls corresponds to Column (6) of Table III. We report coefficients and not hazard ratios. Robust standard errors clustered at country*time (c*tz) level are in parentheses. *, **, and *** denote statistical significance at the 10%, 5%, and 1% level, respectively. Dep. Var.: hazard rate  (1)  (2)  (3)  (4)  (5)  Distance to comparative advantage  0.117***  0.124***  0.086***  0.089***  0.094***    (0.010)  (0.010)  (0.015)  (0.015)  (0.016)  Banks×dist. to comp. advantage      0.064***  0.071***  0.054***        (0.023)  (0.022)  (0.020)  Stock markets×dist. to comp. advantage      –0.017            (0.015)      Stock market value traded×dist. to comp. advantage  –0.017      –0.047***    (0.014)      (0.017)    Stock market turnover×dist. to comp. advantage    –0.018*      –0.015    (0.009)      (0.009)  Banks      –0.060  –0.071  –0.050        (0.044)  (0.044)  (0.042)  Stock markets      0.053*            (0.030)      Stock market value traded  0.101***      0.132***      (0.028)      (0.030)    Stock market turnover    0.037*      0.020      (0.021)      (0.021)  Full set of controls  Yes  Yes  Yes  Yes  Yes  Exporting country FE  Yes  Yes  Yes  Yes  Yes  Year FE  Yes  Yes  Yes  Yes  Yes  HS6 product strata  Yes  Yes  Yes  Yes  Yes  Observations  180,479  179,311  175,481  173,368  172,200  Dep. Var.: hazard rate  (1)  (2)  (3)  (4)  (5)  Distance to comparative advantage  0.117***  0.124***  0.086***  0.089***  0.094***    (0.010)  (0.010)  (0.015)  (0.015)  (0.016)  Banks×dist. to comp. advantage      0.064***  0.071***  0.054***        (0.023)  (0.022)  (0.020)  Stock markets×dist. to comp. advantage      –0.017            (0.015)      Stock market value traded×dist. to comp. advantage  –0.017      –0.047***    (0.014)      (0.017)    Stock market turnover×dist. to comp. advantage    –0.018*      –0.015    (0.009)      (0.009)  Banks      –0.060  –0.071  –0.050        (0.044)  (0.044)  (0.042)  Stock markets      0.053*            (0.030)      Stock market value traded  0.101***      0.132***      (0.028)      (0.030)    Stock market turnover    0.037*      0.020      (0.021)      (0.021)  Full set of controls  Yes  Yes  Yes  Yes  Yes  Exporting country FE  Yes  Yes  Yes  Yes  Yes  Year FE  Yes  Yes  Yes  Yes  Yes  HS6 product strata  Yes  Yes  Yes  Yes  Yes  Observations  180,479  179,311  175,481  173,368  172,200  Columns (3)–(5) of Table VI report the results of direct horse-races between the effect of banks (measured by our main bank proxy BCc,tz) and the effect of stock markets (measured by the three alternative stock market proxies — StMc,tz, StMVTc,tz, StMTc,tz). In the third column of Table VI, we run an econometric horse-race between our main bank and stock market proxies and include the levels of banking and stock market development (banks — BCc,tz, stock markets — StMc,tz) and their interaction with distance to comparative advantage (BCc,tz*DCAck,tz, StMc,tz*DCAck,tz) simultaneously into our specification. The coefficients for distance to comparative advantage (DCAck,tz) and its interaction with banks (BCc,tz*DCAck,tz) are both positive and highly significant. In contrast, the interaction term between stock markets and distance to comparative advantage (StMc,tz*DCAck,tz) is non-significant in this specification, further confirming the results from Table V. In Columns (4) and (5) of Table VI, the econometric horse-race is between our main bank proxy and the alternative stock market measures from the first two columns of Table VI. The previously insignificant interaction term of stock market value traded from Column (1) becomes significant but with wrong sign in Column (4) while the previously significant interaction of stock market turnover from Column (2) gets insignificant when controlling for the bank channel in Column (5). In contrast, the coefficients for distance to comparative advantage and its interaction term with banks have once again the right sign and are highly significant in both Columns (4) and (5) of Table VI. Overall, the results in Tables V and VI suggest that stock markets do not exercise the same disciplining impact on the exports as banks and therefore do not complement the positive allocative impact of the competitive US markets in this regard, in accordance with the conjecture discussed at the end of Section 2. The occasional significance of the wrong sign for the interactions of distance to comparative advantage with some stock markets proxies in some specifications is somewhat puzzling and warrants further research. We will provide some preliminary evidence for a possible explanation in the next subsection where we look at alternative channels that could be correlated with the disciplining impact of product markets in the destination country and banks and stock markets in the country of origin. 5.3 Alternative Channels One concern in estimations reported in the previous two subsections is that the bank and stock market interaction terms may be picking up effects of other alternative channels that could be correlated with our mechanism. In Table VII, we therefore re-estimate our baseline specification that entails the bank interaction term (Column 6 in Table III), while adding the interaction term of our main stock market variable with distance to comparative advantage as well as additional controls capturing possible alternative channels. In Tables VIII and A.III in Appendix A, we further examine the interplay between our channel and the traditional channel from finance-trade literature that focuses on financial development alleviating financial frictions. Table VII. Banks, stock markets, and comparative advantage: alternative channels The dependent variable is the hazard rate on the US destination market for an export spell of product k from country c that started at time tz. All regressions are estimated using the Cox Proportional Hazard model (Cox, 1972) and control for country- and time-fixed effects and allow for different baseline hazard rate across products by defining product k as strata variable. Additional control variables include: the interaction between the bank credit over GDP in country c and the dependence of industry j on external finance (BCc,tz*ExFj), the interaction between GDP per capita of country c and the distance to comparative advantage (GDPpcc,tz*DCAck,tz), and the real exchange rate (RERc,tz) and its interaction with the distance to comparative advantage (RERc,tz*DCAck,tz). Other variables are defined in Tables III and V. Full set of controls corresponds to Column (6) of Table III. We report coefficients and not hazard ratios. Robust standard errors clustered at country*time (c*tz) level are in parentheses. *, **, and *** denote statistical significance at the 10%, 5%, and 1% level, respectively. Dep. Var.: hazard rate  (1)  (2)  (3)  (4)  Distance to comparative advantage  0.090***  0.163*  2.630***  2.650***    (0.014)  (0.090)  (0.812)  (0.858)  Banks×dist. to comp. advantage  0.051**  0.070***  0.042*  0.045*    (0.022)  (0.025)  (0.024)  (0.026)  Stock markets×dist. to comp. advantage  –0.016  –0.016  –0.037**  –0.033**    (0.015)  (0.015)  (0.015)  (0.015)  Banks×dep. on ext. finance  –0.222***      –0.245***    (0.037)      (0.040)  GDPpc×dist. to comp. advantage    –0.009    –0.037***      (0.011)    (0.012)  Real exchange rate×dist. to comp. advantage      1.126***  0.983**        (0.362)  (0.395)  Banks  0.008  –0.067  –0.075  –0.021    (0.043)  (0.047)  (0.047)  (0.051)  Stock markets  0.052*  0.052*  0.085***  0.081***    (0.030)  (0.030)  (0.029)  (0.029)  GDPpc  0.358***  0.362***  0.174  0.213    (0.132)  (0.133)  (0.166)  (0.167)  Real exchange rate      –0.693  –0.557        (0.862)  (0.875)  Full set of controls  Yes  Yes  Yes  Yes  Exporting country FE  Yes  Yes  Yes  Yes  Year FE  Yes  Yes  Yes  Yes  HS6 product strata  Yes  Yes  Yes  Yes  Observations  175,481  175,481  121,157  121,157  Dep. Var.: hazard rate  (1)  (2)  (3)  (4)  Distance to comparative advantage  0.090***  0.163*  2.630***  2.650***    (0.014)  (0.090)  (0.812)  (0.858)  Banks×dist. to comp. advantage  0.051**  0.070***  0.042*  0.045*    (0.022)  (0.025)  (0.024)  (0.026)  Stock markets×dist. to comp. advantage  –0.016  –0.016  –0.037**  –0.033**    (0.015)  (0.015)  (0.015)  (0.015)  Banks×dep. on ext. finance  –0.222***      –0.245***    (0.037)      (0.040)  GDPpc×dist. to comp. advantage    –0.009    –0.037***      (0.011)    (0.012)  Real exchange rate×dist. to comp. advantage      1.126***  0.983**        (0.362)  (0.395)  Banks  0.008  –0.067  –0.075  –0.021    (0.043)  (0.047)  (0.047)  (0.051)  Stock markets  0.052*  0.052*  0.085***  0.081***    (0.030)  (0.030)  (0.029)  (0.029)  GDPpc  0.358***  0.362***  0.174  0.213    (0.132)  (0.133)  (0.166)  (0.167)  Real exchange rate      –0.693  –0.557        (0.862)  (0.875)  Full set of controls  Yes  Yes  Yes  Yes  Exporting country FE  Yes  Yes  Yes  Yes  Year FE  Yes  Yes  Yes  Yes  HS6 product strata  Yes  Yes  Yes  Yes  Observations  175,481  175,481  121,157  121,157  Table VII. Banks, stock markets, and comparative advantage: alternative channels The dependent variable is the hazard rate on the US destination market for an export spell of product k from country c that started at time tz. All regressions are estimated using the Cox Proportional Hazard model (Cox, 1972) and control for country- and time-fixed effects and allow for different baseline hazard rate across products by defining product k as strata variable. Additional control variables include: the interaction between the bank credit over GDP in country c and the dependence of industry j on external finance (BCc,tz*ExFj), the interaction between GDP per capita of country c and the distance to comparative advantage (GDPpcc,tz*DCAck,tz), and the real exchange rate (RERc,tz) and its interaction with the distance to comparative advantage (RERc,tz*DCAck,tz). Other variables are defined in Tables III and V. Full set of controls corresponds to Column (6) of Table III. We report coefficients and not hazard ratios. Robust standard errors clustered at country*time (c*tz) level are in parentheses. *, **, and *** denote statistical significance at the 10%, 5%, and 1% level, respectively. Dep. Var.: hazard rate  (1)  (2)  (3)  (4)  Distance to comparative advantage  0.090***  0.163*  2.630***  2.650***    (0.014)  (0.090)  (0.812)  (0.858)  Banks×dist. to comp. advantage  0.051**  0.070***  0.042*  0.045*    (0.022)  (0.025)  (0.024)  (0.026)  Stock markets×dist. to comp. advantage  –0.016  –0.016  –0.037**  –0.033**    (0.015)  (0.015)  (0.015)  (0.015)  Banks×dep. on ext. finance  –0.222***      –0.245***    (0.037)      (0.040)  GDPpc×dist. to comp. advantage    –0.009    –0.037***      (0.011)    (0.012)  Real exchange rate×dist. to comp. advantage      1.126***  0.983**        (0.362)  (0.395)  Banks  0.008  –0.067  –0.075  –0.021    (0.043)  (0.047)  (0.047)  (0.051)  Stock markets  0.052*  0.052*  0.085***  0.081***    (0.030)  (0.030)  (0.029)  (0.029)  GDPpc  0.358***  0.362***  0.174  0.213    (0.132)  (0.133)  (0.166)  (0.167)  Real exchange rate      –0.693  –0.557        (0.862)  (0.875)  Full set of controls  Yes  Yes  Yes  Yes  Exporting country FE  Yes  Yes  Yes  Yes  Year FE  Yes  Yes  Yes  Yes  HS6 product strata  Yes  Yes  Yes  Yes  Observations  175,481  175,481  121,157  121,157  Dep. Var.: hazard rate  (1)  (2)  (3)  (4)  Distance to comparative advantage  0.090***  0.163*  2.630***  2.650***    (0.014)  (0.090)  (0.812)  (0.858)  Banks×dist. to comp. advantage  0.051**  0.070***  0.042*  0.045*    (0.022)  (0.025)  (0.024)  (0.026)  Stock markets×dist. to comp. advantage  –0.016  –0.016  –0.037**  –0.033**    (0.015)  (0.015)  (0.015)  (0.015)  Banks×dep. on ext. finance  –0.222***      –0.245***    (0.037)      (0.040)  GDPpc×dist. to comp. advantage    –0.009    –0.037***      (0.011)    (0.012)  Real exchange rate×dist. to comp. advantage      1.126***  0.983**        (0.362)  (0.395)  Banks  0.008  –0.067  –0.075  –0.021    (0.043)  (0.047)  (0.047)  (0.051)  Stock markets  0.052*  0.052*  0.085***  0.081***    (0.030)  (0.030)  (0.029)  (0.029)  GDPpc  0.358***  0.362***  0.174  0.213    (0.132)  (0.133)  (0.166)  (0.167)  Real exchange rate      –0.693  –0.557        (0.862)  (0.875)  Full set of controls  Yes  Yes  Yes  Yes  Exporting country FE  Yes  Yes  Yes  Yes  Year FE  Yes  Yes  Yes  Yes  HS6 product strata  Yes  Yes  Yes  Yes  Observations  175,481  175,481  121,157  121,157  In the first column of Table VII, we control for an alternative channel from finance to export survival. The seminal paper of Rajan and Zingales (1998) emphasizes the beneficial impact of a well-developed financial system for financially constrained industries that are highly dependent on external financing. Jaud, Kukenova, and Strieborny (2009) confirm the relevance of this mechanism in the context of export survival. They show that a strong financial sector facilitates export survival disproportionately for those sectors that do not generate enough cash-flow to cover their investment needs and therefore have to rely on external finance. In the first column, we therefore augment the specification by adding interaction of countries’ bank credit with industries’ dependence on external finance (BCc,tz*ExFj). The significant disciplining effects of the US destination market and banking sector in the country of origin are not affected by this additional variable. The stock markets in the country of origin lack a significant disciplining effect on products not congruent with the comparative advantage also in this specification, confirming the results from Section 5.2. The estimated coefficient for the control itself (banks × dependence on external finance — BCc,tz*ExFj) is negative and significant, confirming the findings of Jaud, Kukenova, and Strieborny (2009). The direct effect of banks (BCc,tz) on export survival remains insignificant while the direct effect of industry’s dependence on external finance (ExFj) is captured by the strata effects. Another bias might arise due to the high correlation between countries’ financial and overall economic development. Rather than the disciplining effects of banks, our main interaction term (banks × distance to comparative advantage — BCc,tz*DCAck,tz) may simply capture the effect of some unobservable features of rich countries that help them avoid using resources inefficiently for unpromising exports. In the second column of Table VII, in addition to exporting country’s GDP (GDPpcc,tz), we therefore also include its interaction with the distance to comparative advantage (GDPpcc,tz*DCAck,tz) to account for any direct and indirect effects of country’s economic development on export performance. Despite the high correlation between financial and economic development, the coefficients for the distance to comparative advantage (DCAck,tz) and its interaction with banks (BCc,tz*DCAck,tz) remain positive and significant.40 Both the interaction of stock markets and distance to comparative advantage (StMc,tz*DCAck,tz) and the interaction of the GDP per capita with the distance to comparative advantage (GDPpcc,tz*DCAck,tz) are insignificant. It seems to be indeed the disciplining impact of a well-developed banking system rather than some general feature of rich countries that prevents resource misallocation in the form of attempted exports not corresponding to the exporting country‘s factor endowment. In the third column of Table VII, we control for the level of real exchange rate of countries exporting their products to the USA. As countries develop both economically and financially, their currencies often appreciate due to the well-known Harrod–Balassa–Samuelson effect. The episodes of rapid financial development are also sometimes associated with inflows of foreign capital that can lead to currency appreciations. These developments could introduce bias with unclear sign into our regressions if changes in the currency levels affect exports of products that correspond to the comparative advantage of a given country differently from products that do not. The results in the third column of Table VII document the robustness of our main results to this alternative channel. Coefficients for both distance to comparative advantage and its interaction with banks remain significant with the expected positive sign. In contrast, the stock market interaction now becomes significant with the wrong sign. The result from Table VI where the other two stock market proxies were also sometimes significant with the wrong sign could thus be explained by some interplay between stock market development and real appreciations, for example during economic booms driven by inflows of foreign capital.41 In the fourth column of Table VII, we simultaneously control for all three alternative economic channels examined in Columns (1)–(3). In particular, we include as control variables the interaction term between countries’ banking sector development and industries’ dependence on external finance (BCc,tz*ExFj), the interaction term between countries’ economic development and the distance to comparative advantage (GDPpcc,tz*DCAck,tz), and the real exchange rate (RERc,tz) and its interaction with distance to comparative advantage (RERc,tz*DCAck,tz). Our main results are robust also in this extended specification. Both distance to comparative advantage (DCAck,tz) and its interaction term with banking development in the exporting country (banks × distance to comparative advantage — BCc,tz*DCAck,tz) remain positive and significant. In the rest of this subsection, we look closer at the interplay between the channel examined in this paper and the traditional role banks and stock markets play in alleviating financial constraints. The role of financial constraints and access to external finance is particularly important in the case of a costly and investment-intensive activity like exporting (see, e.g., Beck, 2002, 2003; Jaud, Kukenova, and Strieborny, 2009, 2015) and could therefore affect the ability of banks in the exporting countries to enhance the disciplining impact of the US destination market. At the same time, there seem to be two differences in how the traditional channel from finance-trade literature and the channel explored in our paper operate. First, the research on alleviating financial constraints mostly focuses on how financial development promotes entry into exporting while our channel looks into impact on long-term export survival. Second, the traditional channel looks into alleviation of financial constraints via the general financial development while it seems to be only the banking sector that enhances the long-term discipline of the product markets in our channel. In the remainder of this section, we explore these issues in more detail. The main focus of our paper is the role that banking sector plays in enhancing the disciplining impact of the product markets. In the first two columns of Table VIII, we explore whether the strength of this enhancing effect varies according to the severity of financial constraints firms face. Column (1) of Table VIII reports the results for the subset of exported products from industries that are in the bottom quartile of dependence on external finance. Column (2) of Table VIII reports results for the highly financially constrained industries from the top quartile of dependence on external finance. One could expect that external debtholders like banks in the exporting countries are better able to enhance the disciplining effect of the US destination market in industries where firms require a high level of external financing in the first place. And indeed, while the interaction term of banks and distance to comparative advantage has expected sign and is highly significant in Column (2), it is insignificant in Column (1). The coefficient capturing the direct effect of distance to comparative advantage is positive and significant in both Columns (1) and (2). The higher magnitude of this coefficient in Column (2) suggests that the disciplining effect of the US destination market might also be higher for the products from financially constrained industries. Again, deep stock markets in exporting countries do not seem to enhance the disciplining effect of the US destination market as the coefficient of the stock market interaction is insignificant for both financially constrained and unconstrained industries. Table VIII. Banks, stock markets, comparative advantage, and external finance dependence The dependent variable is the hazard rate on the US destination market for an export spell of product k from country c that started at time tz. In Columns (5) and (6), we replace the hazard rate of export exit with the hazard rate of entry as our dependent variable. All regressions are estimated using the Cox Proportional Hazard model (Cox, 1972) and control for country- and time-fixed effects and allow for different baseline hazard rate across products by defining product k as strata variable. Columns (1) and (2) report results for subsamples of products from the bottom and the top quartile of ISIC four-digit industries dependent on external finance. Columns (3) and (4) report results for subsamples of products from the bottom and the top quartile of the distance to comparative advantage. Additional control variables include: the interaction between the bank credit over GDP in country c and the dependence of industry j on external finance (BCc,tz*ExFj), the interaction between GDP per capita of country c and the distance to comparative advantage (GDPpcc,tz*DCAck,tz), and the real exchange rate (RERc,tz) and its interaction with the distance to comparative advantage (RERc,tz*DCAck,tz). Variables are defined in Tables III, V, and VII. Full set of controls corresponds to Column (6) of Table III. We report coefficients and not hazard ratios. Robust standard errors clustered at country*time (c*tz) level are in parentheses. *, **, and *** denote statistical significance at the 10%, 5%, and 1% level, respectively. Dep. Var.: hazard rate  (1)  (2)  (3)  (4)  (5)  (6)  External finance dependence   Distance to comp. advantage   Hazard rate of export entry   Low  High  Low  High      Distance to comparative advantage  0.046***  0.100***      –0.126***  –0.132***    (0.016)  (0.024)      (0.012)  (0.013)  Banks×dist. to comp. advantage  0.024  0.080**      –0.001  0.019    (0.021)  (0.034)      (0.016)  (0.018)  Stock markets×dist. to comp. advantage  0.006  –0.037        –0.013  (0.019)  (0.026)        (0.012)  Banks×dep. on external finance      –0.292***  –0.243***  0.226***  0.208***        (0.053)  (0.065)  (0.035)  (0.035)  Banks  0.021  –0.080  0.134**  0.071  –0.089**  –0.132***    (0.055)  (0.067)  (0.063)  (0.044)  (0.039)  (0.040)  Stock markets  –0.008  0.072**        0.016    (0.039)  (0.036)        (0.027)  Full set of controls  Yes  Yes  Yes  Yes  Yes  Yes  Exporting country FE  Yes  Yes  Yes  Yes  Yes  Yes  Year FE  Yes  Yes  Yes  Yes  Yes  Yes  HS6 product strata  Yes  Yes  Yes  Yes  Yes  Yes  Observations  48,158  50,504  48,495  47,876  165,190  144,912  Dep. Var.: hazard rate  (1)  (2)  (3)  (4)  (5)  (6)  External finance dependence   Distance to comp. advantage   Hazard rate of export entry   Low  High  Low  High      Distance to comparative advantage  0.046***  0.100***      –0.126***  –0.132***    (0.016)  (0.024)      (0.012)  (0.013)  Banks×dist. to comp. advantage  0.024  0.080**      –0.001  0.019    (0.021)  (0.034)      (0.016)  (0.018)  Stock markets×dist. to comp. advantage  0.006  –0.037        –0.013  (0.019)  (0.026)        (0.012)  Banks×dep. on external finance      –0.292***  –0.243***  0.226***  0.208***        (0.053)  (0.065)  (0.035)  (0.035)  Banks  0.021  –0.080  0.134**  0.071  –0.089**  –0.132***    (0.055)  (0.067)  (0.063)  (0.044)  (0.039)  (0.040)  Stock markets  –0.008  0.072**        0.016    (0.039)  (0.036)        (0.027)  Full set of controls  Yes  Yes  Yes  Yes  Yes  Yes  Exporting country FE  Yes  Yes  Yes  Yes  Yes  Yes  Year FE  Yes  Yes  Yes  Yes  Yes  Yes  HS6 product strata  Yes  Yes  Yes  Yes  Yes  Yes  Observations  48,158  50,504  48,495  47,876  165,190  144,912  Table VIII. Banks, stock markets, comparative advantage, and external finance dependence The dependent variable is the hazard rate on the US destination market for an export spell of product k from country c that started at time tz. In Columns (5) and (6), we replace the hazard rate of export exit with the hazard rate of entry as our dependent variable. All regressions are estimated using the Cox Proportional Hazard model (Cox, 1972) and control for country- and time-fixed effects and allow for different baseline hazard rate across products by defining product k as strata variable. Columns (1) and (2) report results for subsamples of products from the bottom and the top quartile of ISIC four-digit industries dependent on external finance. Columns (3) and (4) report results for subsamples of products from the bottom and the top quartile of the distance to comparative advantage. Additional control variables include: the interaction between the bank credit over GDP in country c and the dependence of industry j on external finance (BCc,tz*ExFj), the interaction between GDP per capita of country c and the distance to comparative advantage (GDPpcc,tz*DCAck,tz), and the real exchange rate (RERc,tz) and its interaction with the distance to comparative advantage (RERc,tz*DCAck,tz). Variables are defined in Tables III, V, and VII. Full set of controls corresponds to Column (6) of Table III. We report coefficients and not hazard ratios. Robust standard errors clustered at country*time (c*tz) level are in parentheses. *, **, and *** denote statistical significance at the 10%, 5%, and 1% level, respectively. Dep. Var.: hazard rate  (1)  (2)  (3)  (4)  (5)  (6)  External finance dependence   Distance to comp. advantage   Hazard rate of export entry   Low  High  Low  High      Distance to comparative advantage  0.046***  0.100***      –0.126***  –0.132***    (0.016)  (0.024)      (0.012)  (0.013)  Banks×dist. to comp. advantage  0.024  0.080**      –0.001  0.019    (0.021)  (0.034)      (0.016)  (0.018)  Stock markets×dist. to comp. advantage  0.006  –0.037        –0.013  (0.019)  (0.026)        (0.012)  Banks×dep. on external finance      –0.292***  –0.243***  0.226***  0.208***        (0.053)  (0.065)  (0.035)  (0.035)  Banks  0.021  –0.080  0.134**  0.071  –0.089**  –0.132***    (0.055)  (0.067)  (0.063)  (0.044)  (0.039)  (0.040)  Stock markets  –0.008  0.072**        0.016    (0.039)  (0.036)        (0.027)  Full set of controls  Yes  Yes  Yes  Yes  Yes  Yes  Exporting country FE  Yes  Yes  Yes  Yes  Yes  Yes  Year FE  Yes  Yes  Yes  Yes  Yes  Yes  HS6 product strata  Yes  Yes  Yes  Yes  Yes  Yes  Observations  48,158  50,504  48,495  47,876  165,190  144,912  Dep. Var.: hazard rate  (1)  (2)  (3)  (4)  (5)  (6)  External finance dependence   Distance to comp. advantage   Hazard rate of export entry   Low  High  Low  High      Distance to comparative advantage  0.046***  0.100***      –0.126***  –0.132***    (0.016)  (0.024)      (0.012)  (0.013)  Banks×dist. to comp. advantage  0.024  0.080**      –0.001  0.019    (0.021)  (0.034)      (0.016)  (0.018)  Stock markets×dist. to comp. advantage  0.006  –0.037        –0.013  (0.019)  (0.026)        (0.012)  Banks×dep. on external finance      –0.292***  –0.243***  0.226***  0.208***        (0.053)  (0.065)  (0.035)  (0.035)  Banks  0.021  –0.080  0.134**  0.071  –0.089**  –0.132***    (0.055)  (0.067)  (0.063)  (0.044)  (0.039)  (0.040)  Stock markets  –0.008  0.072**        0.016    (0.039)  (0.036)        (0.027)  Full set of controls  Yes  Yes  Yes  Yes  Yes  Yes  Exporting country FE  Yes  Yes  Yes  Yes  Yes  Yes  Year FE  Yes  Yes  Yes  Yes  Yes  Yes  HS6 product strata  Yes  Yes  Yes  Yes  Yes  Yes  Observations  48,158  50,504  48,495  47,876  165,190  144,912  While the presence of financial constraints should make the firms and their managers more responsive to the disciplining impact of external debtholders like banks, the alleviation of these constraints via financial development might facilitate exports of all kinds of products including the less efficient ones. We explore this conjecture in Columns (3) and (4) of Table VIII, which report the results for the subsamples of products from the bottom quartile and the top quartile of the distance to comparative advantage, respectively. The interaction term of countries’ banking sector development with industries’ dependence on external finance (BCc,tz*ExFj) has the expected negative sign and is significant in both columns. The results from the first four columns of Table VIII thus suggest that while being financially constrained matters for the disciplining channel explored in this paper, the standard channel from finance-trade literature works for all constrained industries independently on the alignment of their exports with the comparative advantage. Although there is some research on the importance of financial constraints for the long-term export survival (Jaud, Kukenova, and Strieborny, 2009, 2015), most of the literature on this traditional finance-trade channel focuses on how access to finance affects bilateral trade flows and/or how it helps financially constrained exporters to enter new destination markets (e.g., Minetti and Zhu, 2011; Becker, Chen, and Greenberg, 2013; Manova, 2013; Muûls, 2015).42 This is different from the channel explored in our paper where the issue of banking development enhancing the ultimate but slow-acting forces of competitive product markets naturally leads to empirical focus on the hazard rate of existing exports. In order to explore the issue of export entry and still stay in the framework of duration analysis, we thus look at the “hazard rate” of products’ entry rather than their exits in the last two columns of Table VIII.43 This means that all coefficients have now the opposite expected signs compared with the previous estimations. In Column (5), we focus on the impact of banks alone. In Column (6), we consider also the effect of the stock markets. In both specifications, the insignificant coefficient for our main interaction term (BCc,tz*DCAck,tz) suggests that the channel explored in our paper does not operate via the extensive margin of exports. This is consistent with our simple theoretical model where the agency costs influence termination of inefficient ongoing export projects but not the export entry. The result is also reassuring from the point of view of potential sample selection that we discuss in more detail in Section 6.1. The result for the disciplining channel explored in this paper does not mean that banking development does not play a role for export entry in general. The interaction term between countries’ banking sector development and industries’ dependence on external finance (BCc,tz*ExFj) has a positive and significant coefficient in both Columns (5) and (6), suggesting that banks disproportionately promote export entry in industries that do not generate enough internal funds relative to their investment needs. These results confirm the findings in existing literature about the important role banks play in promoting export entry by alleviating exporters’ financial constraints. In this section, we have provided ample evidence suggesting that disciplining channel examined in this paper operates only through strong banks and not through deep stock markets. This evidence is not likely to stem from a measurement problem. Similarly to the banking development, we have relied on standard and widely used empirical proxies also in the case of the stock markets development. Table A.III in the Appendix A reinforces this argument by confirming that the traditional channel of financial institutions and markets promoting exports via easing the liquidity constraints does work in the case of stock markets. The coefficient for the interaction term of stock markets and dependence on external finance (StMc,tz*ExFj) is negative and significant in all columns of Table A.III. This result is robust to controlling for our main channel as well as to controlling for the overall economic development in the exporting countries via the interaction term between GDP per capita and distance to comparative advantage. As predicted by theory and previous empirical work (e.g., Jaud, Kukenova, and Strieborny, 2009), deep stock markets indeed improve countries’ export performance (decrease the hazard rate of export spells) especially for those industries that heavily depend on external finance. Our results thus do not imply that stock markets are not important for export performance in general. It is only the case of the disciplining channel examined in this paper where banks play a pivotal role. A further analysis of possible interplay between this novel channel and the traditional channel of financial development alleviating financial constraints could represent a fruitful area for further research. 6. Robustness Tests In this section, we perform a series of additional regressions to confirm the robustness of our main results. As stock markets do not exert a significant disciplining effect on exports already in the previous specifications (as seen in Tables V, VI, VII, VIII, and A.III), we focus on the robustness of the bank channel. The point of departure is Column (6) of Table III. 6.1 Sample Selection In the first two sections of this paper, we argued both verbally and formally that the mechanism in our paper works in the long term, affecting the termination rates of existing export projects that are inefficient but might generate managerial perks. However, one could certainly come up with an alternative theoretical framework where banks also prevent export entry of products not corresponding to the comparative advantage of the exporting country. One empirical implication of such theoretical scenario would be less entry into exporting of inefficient products from countries with a well-developed banking system. The sample of existing export spells examined in our duration analysis framework would be then subject to a possible selection bias.44 We first test the possibility of the sample selection directly, by examining the hazard rate of export entry rather than export exits in Column (1) of Table IX. The expected coefficients have thus the opposite signs from our main estimations, similarly to the last two columns of Table VIII. While those two estimations examined the alternative channel from existing finance-trade literature, Column (1) of Table IX is based on our preferred specification from Column (6) of Table III. Confirming the results from Table VIII, there is no evidence for sample selection. The interaction term between banks and distance to comparative advantage is insignificant in Column (1) of Table IX, suggesting that export entry of inefficient products is not affected by the strength of the banking sector. Like in our model, banks enhance the disciplining impact of product markets only when it comes to the long-term survival of existing export projects. A possible explanation might be that government-run export agencies focus on export entry (Volpe Martincus and Carballo, 2008), possibly allowing some inefficient exports to evade the disciplining effect of non-governmental debtholders.45 Table IX. Robustness checks I: banks and comparative advantage The dependent variable is the hazard rate on the US destination market for an export spell of product k from country c that started at time tz. In Column (1), we replace the hazard rate of export exit with the hazard rate of entry as our dependent variable. All regressions are estimated using the Cox Proportional Hazard model (Cox, 1972) and control for country- and time-fixed effects and allow for different baseline hazard rate across products by defining product k as strata variable. Variables are defined in Table III. Full set of controls corresponds to Column (6) of Table III. In Columns (2) and (3), we drop the observations from exporting countries in the bottom 5th and the top 5th percentile of banking development. In Columns (4) and (5), we drop observations from exporting countries in the bottom 5th and the top 5th percentile of the export portfolio quality as measured by the distance to comparative advantage. We report coefficients and not hazard ratios. Robust standard errors clustered at country*time (c*tz) level are in parentheses. *, **, and *** denote statistical significance at the 10%, 5%, and 1% level, respectively. Dep. Var.:  (1)  (2)  (3)  (4)  (5)  Product introduction  Hazard rate   Financial development   DCA of export portfolio   Excluding  Excluding  Excluding  Excluding  bottom 5th  top 5th  bottom 5th  top 5th  Distance to comparative advantage  –0.122***  0.082***  0.089***  0.082***  0.081***  (0.012)  (0.014)  (0.015)  (0.014)  (0.014)  Banks×dist. to comp. advantage  –0.017  0.062***  0.054**  0.065***  0.063***  (0.016)  (0.019)  (0.025)  (0.020)  (0.019)  Banks  –0.021  –0.080*  –0.039  –0.087**  –0.080*    (0.039)  (0.043)  (0.061)  (0.043)  (0.043)  Full set of controls  Yes  Yes  Yes  Yes  Yes  Exporting country FE  Yes  Yes  Yes  Yes  Yes  Year FE  Yes  Yes  Yes  Yes  Yes  HS6 product strata  Yes  Yes  Yes  Yes  Yes  Observations  165,190  189,450  175,349  180,813  189,897  Dep. Var.:  (1)  (2)  (3)  (4)  (5)  Product introduction  Hazard rate   Financial development   DCA of export portfolio   Excluding  Excluding  Excluding  Excluding  bottom 5th  top 5th  bottom 5th  top 5th  Distance to comparative advantage  –0.122***  0.082***  0.089***  0.082***  0.081***  (0.012)  (0.014)  (0.015)  (0.014)  (0.014)  Banks×dist. to comp. advantage  –0.017  0.062***  0.054**  0.065***  0.063***  (0.016)  (0.019)  (0.025)  (0.020)  (0.019)  Banks  –0.021  –0.080*  –0.039  –0.087**  –0.080*    (0.039)  (0.043)  (0.061)  (0.043)  (0.043)  Full set of controls  Yes  Yes  Yes  Yes  Yes  Exporting country FE  Yes  Yes  Yes  Yes  Yes  Year FE  Yes  Yes  Yes  Yes  Yes  HS6 product strata  Yes  Yes  Yes  Yes  Yes  Observations  165,190  189,450  175,349  180,813  189,897  Table IX. Robustness checks I: banks and comparative advantage The dependent variable is the hazard rate on the US destination market for an export spell of product k from country c that started at time tz. In Column (1), we replace the hazard rate of export exit with the hazard rate of entry as our dependent variable. All regressions are estimated using the Cox Proportional Hazard model (Cox, 1972) and control for country- and time-fixed effects and allow for different baseline hazard rate across products by defining product k as strata variable. Variables are defined in Table III. Full set of controls corresponds to Column (6) of Table III. In Columns (2) and (3), we drop the observations from exporting countries in the bottom 5th and the top 5th percentile of banking development. In Columns (4) and (5), we drop observations from exporting countries in the bottom 5th and the top 5th percentile of the export portfolio quality as measured by the distance to comparative advantage. We report coefficients and not hazard ratios. Robust standard errors clustered at country*time (c*tz) level are in parentheses. *, **, and *** denote statistical significance at the 10%, 5%, and 1% level, respectively. Dep. Var.:  (1)  (2)  (3)  (4)  (5)  Product introduction  Hazard rate   Financial development   DCA of export portfolio   Excluding  Excluding  Excluding  Excluding  bottom 5th  top 5th  bottom 5th  top 5th  Distance to comparative advantage  –0.122***  0.082***  0.089***  0.082***  0.081***  (0.012)  (0.014)  (0.015)  (0.014)  (0.014)  Banks×dist. to comp. advantage  –0.017  0.062***  0.054**  0.065***  0.063***  (0.016)  (0.019)  (0.025)  (0.020)  (0.019)  Banks  –0.021  –0.080*  –0.039  –0.087**  –0.080*    (0.039)  (0.043)  (0.061)  (0.043)  (0.043)  Full set of controls  Yes  Yes  Yes  Yes  Yes  Exporting country FE  Yes  Yes  Yes  Yes  Yes  Year FE  Yes  Yes  Yes  Yes  Yes  HS6 product strata  Yes  Yes  Yes  Yes  Yes  Observations  165,190  189,450  175,349  180,813  189,897  Dep. Var.:  (1)  (2)  (3)  (4)  (5)  Product introduction  Hazard rate   Financial development   DCA of export portfolio   Excluding  Excluding  Excluding  Excluding  bottom 5th  top 5th  bottom 5th  top 5th  Distance to comparative advantage  –0.122***  0.082***  0.089***  0.082***  0.081***  (0.012)  (0.014)  (0.015)  (0.014)  (0.014)  Banks×dist. to comp. advantage  –0.017  0.062***  0.054**  0.065***  0.063***  (0.016)  (0.019)  (0.025)  (0.020)  (0.019)  Banks  –0.021  –0.080*  –0.039  –0.087**  –0.080*    (0.039)  (0.043)  (0.061)  (0.043)  (0.043)  Full set of controls  Yes  Yes  Yes  Yes  Yes  Exporting country FE  Yes  Yes  Yes  Yes  Yes  Year FE  Yes  Yes  Yes  Yes  Yes  HS6 product strata  Yes  Yes  Yes  Yes  Yes  Observations  165,190  189,450  175,349  180,813  189,897  While there does not seem to be any sample selection at the export entry, our results might still be driven by some influential outliers at the country level regarding banking development or the quality of export portfolio. In Columns (2) and (3) of Table IX, we thus drop the observations from exporting countries in the bottom 5th and the top 5th percentile of banking development. In Columns (4) and (5), we drop observations from countries with the worst and the best export portfolio, that is, countries whose export portfolios are in the bottom 5th and the top 5th percentile as measured by the average distance to comparative advantage. Otherwise the estimations correspond to Column (6) of Table III. The results remain qualitatively the same in all four columns and they do so also when we exclude the bottom and top 10th rather than 5th percentile (not reported). Finally, we provide some visual evidence. Figure 1 compares the export compositions of countries from the top and bottom quartiles of banking development. It shows that while the export composition of countries with a well-developed banking system does lean toward products close to their respective comparative advantage (which is to be expected given our results showing a higher hazard rate facing by inefficient exports from such countries), there is a sufficient overlap between the two distributions enabling estimation of our main interaction term (BCc,tz*DCAck,tz). Figure 1. View largeDownload slide Export composition of countries with different levels of financial development. Figure 1. View largeDownload slide Export composition of countries with different levels of financial development. 6.2 Alternative Empirical Proxies and an Alternative Destination Market Table X examines the robustness of our main results to alternative measures of physical and human capital intensity (first column), alternative definitions of the distance to comparative advantage (second–eighth column), and an alternative destination market (last column). Table X. Robustness checks II: banks and comparative advantage The dependent variable is the hazard rate on the US destination market for an export spell of product k from country c that started at time tz. All regressions are estimated using the Cox Proportional Hazard model (Cox, 1972) and control for country- and time-fixed effects and allow for different baseline hazard rate across products by defining product k as strata variable. Variables are defined in Table III. In Column (1), we use alternative measures of physical and human capital intensity. In Columns (2)–(8) we use alternative definitions of the distance to comparative advantage. In Column (9), we replace the USA with Germany as our destination market. Full set of controls corresponds to Column (6) of Table III. We report coefficients and not hazard ratios. Robust standard errors clustered at country*time (c*tz) level are in parentheses. *, **, and *** denote statistical significance at the 10%, 5%, and 1% level, respectively. Dep. Var.: hazard rate  (1)  (2)  (3)  (4)  (5)  (6)  (7)  (8)  (9)  Alt. RFI    Including land   Excluding origin country   Excluding USA   Germany    Abs.  Euclidean  Abs.  Euclidean  Abs.  Euclidean  Abs.    Distance to comparative advantage  0.061***  0.053***  0.067***  0.038***  0.074***  0.049***  0.080***  0.054***  0.048***  (0.017)  (0.009)  (0.014)  (0.008)  (0.014)  (0.010)  (0.013)  (0.009)  (0.013)  Banks×dist. to comparative advantage  0.083***  0.046***  0.080***  0.054***  0.063***  0.045***  0.063***  0.045***  0.045**  (0.019)  (0.013)  (0.019)  (0.012)  (0.019)  (0.013)  (0.019)  (0.013)  (0.019)  Banks  –0.104**  –0.080*  –0.134***  –0.141***  –0.080*  –0.078*  –0.081*  –0.078*  –0.109**    (0.042)  (0.042)  (0.048)  (0.049)  (0.043)  (0.042)  (0.043)  (0.042)  (0.053)  Full set of controls  Yes  Yes  Yes  Yes  Yes  Yes  Yes  Yes  Yes  Exporting country FE  Yes  Yes  Yes  Yes  Yes  Yes  Yes  Yes  Yes  Year FE  Yes  Yes  Yes  Yes  Yes  Yes  Yes  Yes  Yes  HS6 product strata  Yes  Yes  Yes  Yes  Yes  Yes  Yes  Yes  Yes  Observations  211,643  191,078  191,078  191,078  190,490  190,490  191,078  191,078  153,939  Dep. Var.: hazard rate  (1)  (2)  (3)  (4)  (5)  (6)  (7)  (8)  (9)  Alt. RFI    Including land   Excluding origin country   Excluding USA   Germany    Abs.  Euclidean  Abs.  Euclidean  Abs.  Euclidean  Abs.    Distance to comparative advantage  0.061***  0.053***  0.067***  0.038***  0.074***  0.049***  0.080***  0.054***  0.048***  (0.017)  (0.009)  (0.014)  (0.008)  (0.014)  (0.010)  (0.013)  (0.009)  (0.013)  Banks×dist. to comparative advantage  0.083***  0.046***  0.080***  0.054***  0.063***  0.045***  0.063***  0.045***  0.045**  (0.019)  (0.013)  (0.019)  (0.012)  (0.019)  (0.013)  (0.019)  (0.013)  (0.019)  Banks  –0.104**  –0.080*  –0.134***  –0.141***  –0.080*  –0.078*  –0.081*  –0.078*  –0.109**    (0.042)  (0.042)  (0.048)  (0.049)  (0.043)  (0.042)  (0.043)  (0.042)  (0.053)  Full set of controls  Yes  Yes  Yes  Yes  Yes  Yes  Yes  Yes  Yes  Exporting country FE  Yes  Yes  Yes  Yes  Yes  Yes  Yes  Yes  Yes  Year FE  Yes  Yes  Yes  Yes  Yes  Yes  Yes  Yes  Yes  HS6 product strata  Yes  Yes  Yes  Yes  Yes  Yes  Yes  Yes  Yes  Observations  211,643  191,078  191,078  191,078  190,490  190,490  191,078  191,078  153,939  Table X. Robustness checks II: banks and comparative advantage The dependent variable is the hazard rate on the US destination market for an export spell of product k from country c that started at time tz. All regressions are estimated using the Cox Proportional Hazard model (Cox, 1972) and control for country- and time-fixed effects and allow for different baseline hazard rate across products by defining product k as strata variable. Variables are defined in Table III. In Column (1), we use alternative measures of physical and human capital intensity. In Columns (2)–(8) we use alternative definitions of the distance to comparative advantage. In Column (9), we replace the USA with Germany as our destination market. Full set of controls corresponds to Column (6) of Table III. We report coefficients and not hazard ratios. Robust standard errors clustered at country*time (c*tz) level are in parentheses. *, **, and *** denote statistical significance at the 10%, 5%, and 1% level, respectively. Dep. Var.: hazard rate  (1)  (2)  (3)  (4)  (5)  (6)  (7)  (8)  (9)  Alt. RFI    Including land   Excluding origin country   Excluding USA   Germany    Abs.  Euclidean  Abs.  Euclidean  Abs.  Euclidean  Abs.    Distance to comparative advantage  0.061***  0.053***  0.067***  0.038***  0.074***  0.049***  0.080***  0.054***  0.048***  (0.017)  (0.009)  (0.014)  (0.008)  (0.014)  (0.010)  (0.013)  (0.009)  (0.013)  Banks×dist. to comparative advantage  0.083***  0.046***  0.080***  0.054***  0.063***  0.045***  0.063***  0.045***  0.045**  (0.019)  (0.013)  (0.019)  (0.012)  (0.019)  (0.013)  (0.019)  (0.013)  (0.019)  Banks  –0.104**  –0.080*  –0.134***  –0.141***  –0.080*  –0.078*  –0.081*  –0.078*  –0.109**    (0.042)  (0.042)  (0.048)  (0.049)  (0.043)  (0.042)  (0.043)  (0.042)  (0.053)  Full set of controls  Yes  Yes  Yes  Yes  Yes  Yes  Yes  Yes  Yes  Exporting country FE  Yes  Yes  Yes  Yes  Yes  Yes  Yes  Yes  Yes  Year FE  Yes  Yes  Yes  Yes  Yes  Yes  Yes  Yes  Yes  HS6 product strata  Yes  Yes  Yes  Yes  Yes  Yes  Yes  Yes  Yes  Observations  211,643  191,078  191,078  191,078  190,490  190,490  191,078  191,078  153,939  Dep. Var.: hazard rate  (1)  (2)  (3)  (4)  (5)  (6)  (7)  (8)  (9)  Alt. RFI    Including land   Excluding origin country   Excluding USA   Germany    Abs.  Euclidean  Abs.  Euclidean  Abs.  Euclidean  Abs.    Distance to comparative advantage  0.061***  0.053***  0.067***  0.038***  0.074***  0.049***  0.080***  0.054***  0.048***  (0.017)  (0.009)  (0.014)  (0.008)  (0.014)  (0.010)  (0.013)  (0.009)  (0.013)  Banks×dist. to comparative advantage  0.083***  0.046***  0.080***  0.054***  0.063***  0.045***  0.063***  0.045***  0.045**  (0.019)  (0.013)  (0.019)  (0.012)  (0.019)  (0.013)  (0.019)  (0.013)  (0.019)  Banks  –0.104**  –0.080*  –0.134***  –0.141***  –0.080*  –0.078*  –0.081*  –0.078*  –0.109**    (0.042)  (0.042)  (0.048)  (0.049)  (0.043)  (0.042)  (0.043)  (0.042)  (0.053)  Full set of controls  Yes  Yes  Yes  Yes  Yes  Yes  Yes  Yes  Yes  Exporting country FE  Yes  Yes  Yes  Yes  Yes  Yes  Yes  Yes  Yes  Year FE  Yes  Yes  Yes  Yes  Yes  Yes  Yes  Yes  Yes  HS6 product strata  Yes  Yes  Yes  Yes  Yes  Yes  Yes  Yes  Yes  Observations  211,643  191,078  191,078  191,078  190,490  190,490  191,078  191,078  153,939  When controlling for interaction terms between the exporting countries’ factor endowments and the corresponding factor intensities at the sectoral level (physical capital × physical capital intensity — κc,tz*CapIntj, human capital × human capital intensity — hc,tz*HumIntj), we generally rely on the measures of physical and human capital intensity from Braun (2003) as we prefer to stay as close as possible to existing literature in the case of variables that are merely controls in our context. As a robustness test, we use in Column (1) of Table X instead our own measures of RFI used in the computation of distance to comparative advantage. The results remain qualitatively the same. Columns (2)–(8) of Table X examine the robustness of our results to alternative ways of computing the distance to comparative advantage (DCAck,tz). In the second column, we replace the Euclidean distance with the absolute distance. The new formula for distance of product k from comparative advantage of exporting country c (Equation 8) now writes:   DCAck,tz=|std(κc,tz−κ^k,tz)|+|std(hc,tz−h^k,tz)|. (15) Results remain qualitatively the same. In the third and fourth column of Table X, we add arable land per worker as a third production factor when computing the distance to comparative advantage. We use the Euclidean distance in Column (3) and absolute distance in Column (4). Adding the third production factor does not qualitatively change our results for either the direct effect of distance to comparative advantage (DCAck,tz) or the main interaction term between banks and distance to comparative advantage (BCc,tz*DCAck,tz). In Columns (5)–(8), we address the potential bias that could emerge due to the fact that for all trade flows from any given exporting country to the USA, the factor endowments of both the country of origin and the destination country enter the computation of our distance to comparative advantage measure. In Columns (5) and (6), we drop the data from the country of origin c when computing the distance to comparative advantage (DCAck,tz). In Columns (7) and (8), we recompute DCAck,tz without data from the US destination market. Columns (5) and (7) report the results for recomputed Euclidean distance, while Columns (6) and (8) report the results for recomputed absolute distance. As one can see from Columns (5)–(8) of Table X, our main results are robust to exclusion of data from the country of origin and from the US destination market in computing the distance to comparative advantage (DCAck,tz). Most of our estimations look into exports from seventy-one countries at different stages of financial development into the US destination market, allowing us to better distinguish between the impact of well-developed banks and highly competitive product market on termination of exports not corresponding to the comparative advantage of exporting country. At the beginning of Section 4, we discussed in more detail our choice of the destination market. In Column (9) of Table X, we test the robustness of our results to the use of an alternative destination market. We have chosen Germany, which has also a well-developed, competitive, and large product market. Our results remain qualitatively the same, as they do in the case of the UK destination market (the latter not reported, results available upon request). 6.3 Alternative Samples Table XI examines whether our results are not driven by specific groups of countries. In the first column of Table XI, we drop all observations from small island countries as they often specialize in exports of only a few products.46 The reported results are similar to those in the sixth column of Table III. Products far away from the comparative advantage of the exporting country still face higher hazard rate on the competitive US market as documented by a positive and significant coefficient for distance to comparative advantage DCAck,tz. This pattern is stronger if the exporting country has a well-developed banking system (positive and significant coefficient for the main interaction term between banks and distance to comparative advantage BCc,tz*DCAck,tz). Our results are thus not driven by small countries in the sample. Table XI. Robustness checks III: banks and comparative advantage The dependent variable is the hazard rate on the US destination market for an export spell of product k from country c that started at time tz. All regressions are estimated using the Cox Proportional Hazard model (Cox, 1972) and control for country- and time-fixed effects and allow for different baseline hazard rate across products by defining product k as strata variable. We drop export observations from small islands in Column (1) and from the countries in the bottom 5th and the top 5th percentile according to GDP per capita in Columns (2) and (3), respectively. In Columns (4)–(7), we drop export observations from individual geographic regions. Variables are defined in Table III. Full set of controls corresponds to Column (6) of Table III. We report coefficients and not hazard ratios. Robust standard errors clustered at country*time (c*tz) level are in parentheses. *, **, and *** denote statistical significance at the 10%, 5%, and 1% level, respectively. Dep. Var.: hazard rate  (1)  (2)  (3)  (4)  (5)  (6)  (7)    GDPpc   Excl.   No small  Excluding  Excluding  Europe  America  Africa  Asia  islands  bottom 5th  top 5th          Distance to comparative advantage  0.082***  0.083***  0.074***  0.068***  0.094***  0.073***  0.077***  (0.014)  (0.014)  (0.014)  (0.016)  (0.018)  (0.014)  (0.015)  Banks×dist. to comp. advantage  0.061***  0.061***  0.081***  0.084***  0.041*  0.059***  0.085***  (0.019)  (0.019)  (0.021)  (0.029)  (0.022)  (0.019)  (0.021)  Banks  –0.081*  –0.079*  –0.106*  –0.219**  –0.014  –0.078*  –0.081*    (0.043)  (0.043)  (0.064)  (0.102)  (0.043)  (0.041)  (0.047)  Full set of controls  Yes  Yes  Yes  Yes  Yes  Yes  Yes  Exporting country FE  Yes  Yes  Yes  Yes  Yes  Yes  Yes  Year FE  Yes  Yes  Yes  Yes  Yes  Yes  Yes  HS6 product strata  Yes  Yes  Yes  Yes  Yes  Yes  Yes  Observations  181,612  189,915  178,595  122,616  128,479  178,236  143,903  Dep. Var.: hazard rate  (1)  (2)  (3)  (4)  (5)  (6)  (7)    GDPpc   Excl.   No small  Excluding  Excluding  Europe  America  Africa  Asia  islands  bottom 5th  top 5th          Distance to comparative advantage  0.082***  0.083***  0.074***  0.068***  0.094***  0.073***  0.077***  (0.014)  (0.014)  (0.014)  (0.016)  (0.018)  (0.014)  (0.015)  Banks×dist. to comp. advantage  0.061***  0.061***  0.081***  0.084***  0.041*  0.059***  0.085***  (0.019)  (0.019)  (0.021)  (0.029)  (0.022)  (0.019)  (0.021)  Banks  –0.081*  –0.079*  –0.106*  –0.219**  –0.014  –0.078*  –0.081*    (0.043)  (0.043)  (0.064)  (0.102)  (0.043)  (0.041)  (0.047)  Full set of controls  Yes  Yes  Yes  Yes  Yes  Yes  Yes  Exporting country FE  Yes  Yes  Yes  Yes  Yes  Yes  Yes  Year FE  Yes  Yes  Yes  Yes  Yes  Yes  Yes  HS6 product strata  Yes  Yes  Yes  Yes  Yes  Yes  Yes  Observations  181,612  189,915  178,595  122,616  128,479  178,236  143,903  Table XI. Robustness checks III: banks and comparative advantage The dependent variable is the hazard rate on the US destination market for an export spell of product k from country c that started at time tz. All regressions are estimated using the Cox Proportional Hazard model (Cox, 1972) and control for country- and time-fixed effects and allow for different baseline hazard rate across products by defining product k as strata variable. We drop export observations from small islands in Column (1) and from the countries in the bottom 5th and the top 5th percentile according to GDP per capita in Columns (2) and (3), respectively. In Columns (4)–(7), we drop export observations from individual geographic regions. Variables are defined in Table III. Full set of controls corresponds to Column (6) of Table III. We report coefficients and not hazard ratios. Robust standard errors clustered at country*time (c*tz) level are in parentheses. *, **, and *** denote statistical significance at the 10%, 5%, and 1% level, respectively. Dep. Var.: hazard rate  (1)  (2)  (3)  (4)  (5)  (6)  (7)    GDPpc   Excl.   No small  Excluding  Excluding  Europe  America  Africa  Asia  islands  bottom 5th  top 5th          Distance to comparative advantage  0.082***  0.083***  0.074***  0.068***  0.094***  0.073***  0.077***  (0.014)  (0.014)  (0.014)  (0.016)  (0.018)  (0.014)  (0.015)  Banks×dist. to comp. advantage  0.061***  0.061***  0.081***  0.084***  0.041*  0.059***  0.085***  (0.019)  (0.019)  (0.021)  (0.029)  (0.022)  (0.019)  (0.021)  Banks  –0.081*  –0.079*  –0.106*  –0.219**  –0.014  –0.078*  –0.081*    (0.043)  (0.043)  (0.064)  (0.102)  (0.043)  (0.041)  (0.047)  Full set of controls  Yes  Yes  Yes  Yes  Yes  Yes  Yes  Exporting country FE  Yes  Yes  Yes  Yes  Yes  Yes  Yes  Year FE  Yes  Yes  Yes  Yes  Yes  Yes  Yes  HS6 product strata  Yes  Yes  Yes  Yes  Yes  Yes  Yes  Observations  181,612  189,915  178,595  122,616  128,479  178,236  143,903  Dep. Var.: hazard rate  (1)  (2)  (3)  (4)  (5)  (6)  (7)    GDPpc   Excl.   No small  Excluding  Excluding  Europe  America  Africa  Asia  islands  bottom 5th  top 5th          Distance to comparative advantage  0.082***  0.083***  0.074***  0.068***  0.094***  0.073***  0.077***  (0.014)  (0.014)  (0.014)  (0.016)  (0.018)  (0.014)  (0.015)  Banks×dist. to comp. advantage  0.061***  0.061***  0.081***  0.084***  0.041*  0.059***  0.085***  (0.019)  (0.019)  (0.021)  (0.029)  (0.022)  (0.019)  (0.021)  Banks  –0.081*  –0.079*  –0.106*  –0.219**  –0.014  –0.078*  –0.081*    (0.043)  (0.043)  (0.064)  (0.102)  (0.043)  (0.041)  (0.047)  Full set of controls  Yes  Yes  Yes  Yes  Yes  Yes  Yes  Exporting country FE  Yes  Yes  Yes  Yes  Yes  Yes  Yes  Year FE  Yes  Yes  Yes  Yes  Yes  Yes  Yes  HS6 product strata  Yes  Yes  Yes  Yes  Yes  Yes  Yes  Observations  181,612  189,915  178,595  122,616  128,479  178,236  143,903  In Column (2) and (3) of Table XI, we drop the 5th percentile of the poorest and richest exporting countries, as measured by the GDP per capita. Our results remain the same in both columns and they do so also when we exclude the bottom and top 10th rather than 5th percentile (not reported). In the last four columns we drop exports from individual geographical regions to the USA — Europe in Column (4), America in Column (5), Africa in Column (6), and Asia in Column (7). Again, our main results are not affected by this set of robustness checks. 6.4 Econometric Robustness Tests We have also performed a battery of additional robustness tests that are important from the econometric point of view but might not considerably enhance the economic understanding of our channel. For space reasons, we report the results of these econometric tests in Online Appendix C. They include robustness tests regarding fixed effects, stratification and clustering of standard errors, robustness tests handling the issue of censoring, multiple spells, and tied-spells termination events that are characteristic for the duration analysis, as well as additional estimations relaxing the hazard proportionality assumption made under the Cox model.47 7. Conclusion This paper provides evidence for the allocative and disciplining role of finance. Banks do not promote exports in a sweeping non-discriminate way. They rather push exporters toward the optimal use of countries’ factor endowments, in compliance with the idea of comparative advantage. A well-developed banking system in the exporting country can thus reinforce the disciplining impact of a competitive destination market. In this way, finance helps to prevent inefficient export patterns with positive consequences for national and international allocation of available resources. To our knowledge, this is the first paper using disaggregated trade data to document that a well-functioning financial system can enhance the ultimate but often slow-acting discipline of product markets. Further research could investigate more in detail why banks seem to be more successful in this regard compared with stock and bond markets. Existing theories on disciplining role of debt and monitoring abilities of banks cited in this paper provide a possible economic rationale for this result. Compared with shareholders, creditors can also easier rely on the legal enforcement as their rights are usually more clearly defined and therefore better enforceable even by mediocre and poorly motivated courts (see, e.g., Harris and Raviv, 1990; Shleifer and Vishny, 1997). This is especially important in countries with insufficient judicial quality where the disciplining influence exercised by financial institutions and markets might be particularly needed. In these countries, it might also not be the general stock market development that makes the difference but rather the entry of foreign investors (Jaud, Kukenova, and Strieborny, 2016). Another venue for further research could involve exploiting the bilateral character of the highly disaggregated data on international trade to distinguish between the impact of financial and product markets also in other contexts than the one explored in this paper. Supplementary Material Supplementary data are available at Review of Finance online. Appendix A: Additional Tables Table A.I. Examples from the harmonized system (HS) product classification Harmonized system (HS) of tariff nomenclature is an internationally standardized system of names and numbers to classify traded products. Chapter  Heading  Sub-heading  Description  HS two-digit code  HS four-digit code  HS six-digit code  61      Articles of apparel and clothing accessories-knitted or crocheted    61.06    Women’s or girls’ blouses and shirts, knit or crocheted      6106.10  Women’s or girls’ blouses, shirts, of cotton, knitted or crocheted      6106.20  Women’s blouses, shirts, of man-made fibers, knitted or crocheted      6106.90  Women’s or girls, blouses, shirts, of other textile materials  84      Nuclear reactors, boilers, machinery and mechanical appliances, and computers    84.50    Washing machines, household, or laundry-type      8450.11  Fully-automatic washing machines, dry linen capacity not exceeding 10 kg      8450.12  Machines, with built-in centrifugal drier, capacity not exceeding 10 kg      8450.19  Other washing machines, each of a dry linen capacity not exceeding 10 kg      8450.20  Washing machines, each of a dry linen capacity exceeding 10 kg      8450.90  Parts of household or laundry-type washing machines  Chapter  Heading  Sub-heading  Description  HS two-digit code  HS four-digit code  HS six-digit code  61      Articles of apparel and clothing accessories-knitted or crocheted    61.06    Women’s or girls’ blouses and shirts, knit or crocheted      6106.10  Women’s or girls’ blouses, shirts, of cotton, knitted or crocheted      6106.20  Women’s blouses, shirts, of man-made fibers, knitted or crocheted      6106.90  Women’s or girls, blouses, shirts, of other textile materials  84      Nuclear reactors, boilers, machinery and mechanical appliances, and computers    84.50    Washing machines, household, or laundry-type      8450.11  Fully-automatic washing machines, dry linen capacity not exceeding 10 kg      8450.12  Machines, with built-in centrifugal drier, capacity not exceeding 10 kg      8450.19  Other washing machines, each of a dry linen capacity not exceeding 10 kg      8450.20  Washing machines, each of a dry linen capacity exceeding 10 kg      8450.90  Parts of household or laundry-type washing machines  Table A.I. Examples from the harmonized system (HS) product classification Harmonized system (HS) of tariff nomenclature is an internationally standardized system of names and numbers to classify traded products. Chapter  Heading  Sub-heading  Description  HS two-digit code  HS four-digit code  HS six-digit code  61      Articles of apparel and clothing accessories-knitted or crocheted    61.06    Women’s or girls’ blouses and shirts, knit or crocheted      6106.10  Women’s or girls’ blouses, shirts, of cotton, knitted or crocheted      6106.20  Women’s blouses, shirts, of man-made fibers, knitted or crocheted      6106.90  Women’s or girls, blouses, shirts, of other textile materials  84      Nuclear reactors, boilers, machinery and mechanical appliances, and computers    84.50    Washing machines, household, or laundry-type      8450.11  Fully-automatic washing machines, dry linen capacity not exceeding 10 kg      8450.12  Machines, with built-in centrifugal drier, capacity not exceeding 10 kg      8450.19  Other washing machines, each of a dry linen capacity not exceeding 10 kg      8450.20  Washing machines, each of a dry linen capacity exceeding 10 kg      8450.90  Parts of household or laundry-type washing machines  Chapter  Heading  Sub-heading  Description  HS two-digit code  HS four-digit code  HS six-digit code  61      Articles of apparel and clothing accessories-knitted or crocheted    61.06    Women’s or girls’ blouses and shirts, knit or crocheted      6106.10  Women’s or girls’ blouses, shirts, of cotton, knitted or crocheted      6106.20  Women’s blouses, shirts, of man-made fibers, knitted or crocheted      6106.90  Women’s or girls, blouses, shirts, of other textile materials  84      Nuclear reactors, boilers, machinery and mechanical appliances, and computers    84.50    Washing machines, household, or laundry-type      8450.11  Fully-automatic washing machines, dry linen capacity not exceeding 10 kg      8450.12  Machines, with built-in centrifugal drier, capacity not exceeding 10 kg      8450.19  Other washing machines, each of a dry linen capacity not exceeding 10 kg      8450.20  Washing machines, each of a dry linen capacity exceeding 10 kg      8450.90  Parts of household or laundry-type washing machines  Table A.II. Banks and bond markets The dependent variable is the hazard rate on the US destination market for an export spell of product k from country c that started at time tz. All regressions are estimated using the Cox Proportional Hazard model (Cox, 1972) and control for country- and time-fixed effects. Estimations also allow for different baseline hazard rate across HS6 products by defining product k as strata variable. Additional control variables include the interaction between three proxies of the strength of bond markets in the exporting country and the distance to comparative advantage. These three proxies are domestic private bonds (including bonds issued by banks) over GDP, international private bonds (including bonds issued by banks) over GDP, and nonfinancial international bonds (bonds issued by non-financial firms only) over GDP. Other variables are defined in Table III. Full set of controls corresponds to Column (6) of Table III. We report coefficients and not hazard ratios. Robust standard errors clustered at country * time (c*tz) level are in parentheses. *, **, and *** denote statistical significance at the 10%, 5%, and 1% level, respectively. Dep. Var.: hazard rate  (1)  (2)  (3)  (4)  (5)  (6)  (7)  (8)  (9)  Distance to comparative advantage  0.068***  0.077***  0.054***  0.080***  0.109***  0.081***  0.081***  0.116***  0.084***    (0.018)  (0.013)  (0.017)  (0.016)  (0.011)  (0.016)  (0.015)  (0.010)  (0.015)  Banks×dist. to comp. adv.  0.047**    0.037*  0.055***    0.054**  0.054***    0.056***    (0.020)    (0.021)  (0.020)    (0.023)  (0.019)    (0.020)  Banks  –0.052    –0.041  –0.068*    –0.038  –0.060    –0.075*    (0.040)    (0.039)  (0.041)    (0.049)  (0.041)    (0.043)  Dom. private bonds×dist. to comp. adv.    0.107***  0.092**                  (0.036)  (0.041)              Domestic private bonds to GDP    –0.075  –0.058                  (0.109)  (0.113)              Inter. private bonds×dist. to comp. adv.          0.075  –0.005                  (0.052)  (0.066)        International private bonds to GDP          0.261*  0.372***                  (0.140)  (0.139)        Nonfin. inter. bonds×dist. to comp. adv.                –0.005  –0.176                  (0.201)  (0.209)  Nonfinancial international bonds to GDP                –0.695  –0.518                  (0.518)  (0.530)  Full set of controls  Yes  Yes  Yes  Yes  Yes  Yes  Yes  Yes  Yes  Exporting country FE  Yes  Yes  Yes  Yes  Yes  Yes  Yes  Yes  Yes  Year FE  Yes  Yes  Yes  Yes  Yes  Yes  Yes  Yes  Yes  HS6 product strata  Yes  Yes  Yes  Yes  Yes  Yes  Yes  Yes  Yes  Observations  127, 260  127, 260  127, 260  158, 291  158, 291  158, 291  168, 113  168, 113  168, 113  Dep. Var.: hazard rate  (1)  (2)  (3)  (4)  (5)  (6)  (7)  (8)  (9)  Distance to comparative advantage  0.068***  0.077***  0.054***  0.080***  0.109***  0.081***  0.081***  0.116***  0.084***    (0.018)  (0.013)  (0.017)  (0.016)  (0.011)  (0.016)  (0.015)  (0.010)  (0.015)  Banks×dist. to comp. adv.  0.047**    0.037*  0.055***    0.054**  0.054***    0.056***    (0.020)    (0.021)  (0.020)    (0.023)  (0.019)    (0.020)  Banks  –0.052    –0.041  –0.068*    –0.038  –0.060    –0.075*    (0.040)    (0.039)  (0.041)    (0.049)  (0.041)    (0.043)  Dom. private bonds×dist. to comp. adv.    0.107***  0.092**                  (0.036)  (0.041)              Domestic private bonds to GDP    –0.075  –0.058                  (0.109)  (0.113)              Inter. private bonds×dist. to comp. adv.          0.075  –0.005                  (0.052)  (0.066)        International private bonds to GDP          0.261*  0.372***                  (0.140)  (0.139)        Nonfin. inter. bonds×dist. to comp. adv.                –0.005  –0.176                  (0.201)  (0.209)  Nonfinancial international bonds to GDP                –0.695  –0.518                  (0.518)  (0.530)  Full set of controls  Yes  Yes  Yes  Yes  Yes  Yes  Yes  Yes  Yes  Exporting country FE  Yes  Yes  Yes  Yes  Yes  Yes  Yes  Yes  Yes  Year FE  Yes  Yes  Yes  Yes  Yes  Yes  Yes  Yes  Yes  HS6 product strata  Yes  Yes  Yes  Yes  Yes  Yes  Yes  Yes  Yes  Observations  127, 260  127, 260  127, 260  158, 291  158, 291  158, 291  168, 113  168, 113  168, 113  Table A.II. Banks and bond markets The dependent variable is the hazard rate on the US destination market for an export spell of product k from country c that started at time tz. All regressions are estimated using the Cox Proportional Hazard model (Cox, 1972) and control for country- and time-fixed effects. Estimations also allow for different baseline hazard rate across HS6 products by defining product k as strata variable. Additional control variables include the interaction between three proxies of the strength of bond markets in the exporting country and the distance to comparative advantage. These three proxies are domestic private bonds (including bonds issued by banks) over GDP, international private bonds (including bonds issued by banks) over GDP, and nonfinancial international bonds (bonds issued by non-financial firms only) over GDP. Other variables are defined in Table III. Full set of controls corresponds to Column (6) of Table III. We report coefficients and not hazard ratios. Robust standard errors clustered at country * time (c*tz) level are in parentheses. *, **, and *** denote statistical significance at the 10%, 5%, and 1% level, respectively. Dep. Var.: hazard rate  (1)  (2)  (3)  (4)  (5)  (6)  (7)  (8)  (9)  Distance to comparative advantage  0.068***  0.077***  0.054***  0.080***  0.109***  0.081***  0.081***  0.116***  0.084***    (0.018)  (0.013)  (0.017)  (0.016)  (0.011)  (0.016)  (0.015)  (0.010)  (0.015)  Banks×dist. to comp. adv.  0.047**    0.037*  0.055***    0.054**  0.054***    0.056***    (0.020)    (0.021)  (0.020)    (0.023)  (0.019)    (0.020)  Banks  –0.052    –0.041  –0.068*    –0.038  –0.060    –0.075*    (0.040)    (0.039)  (0.041)    (0.049)  (0.041)    (0.043)  Dom. private bonds×dist. to comp. adv.    0.107***  0.092**                  (0.036)  (0.041)              Domestic private bonds to GDP    –0.075  –0.058                  (0.109)  (0.113)              Inter. private bonds×dist. to comp. adv.          0.075  –0.005                  (0.052)  (0.066)        International private bonds to GDP          0.261*  0.372***                  (0.140)  (0.139)        Nonfin. inter. bonds×dist. to comp. adv.                –0.005  –0.176                  (0.201)  (0.209)  Nonfinancial international bonds to GDP                –0.695  –0.518                  (0.518)  (0.530)  Full set of controls  Yes  Yes  Yes  Yes  Yes  Yes  Yes  Yes  Yes  Exporting country FE  Yes  Yes  Yes  Yes  Yes  Yes  Yes  Yes  Yes  Year FE  Yes  Yes  Yes  Yes  Yes  Yes  Yes  Yes  Yes  HS6 product strata  Yes  Yes  Yes  Yes  Yes  Yes  Yes  Yes  Yes  Observations  127, 260  127, 260  127, 260  158, 291  158, 291  158, 291  168, 113  168, 113  168, 113  Dep. Var.: hazard rate  (1)  (2)  (3)  (4)  (5)  (6)  (7)  (8)  (9)  Distance to comparative advantage  0.068***  0.077***  0.054***  0.080***  0.109***  0.081***  0.081***  0.116***  0.084***    (0.018)  (0.013)  (0.017)  (0.016)  (0.011)  (0.016)  (0.015)  (0.010)  (0.015)  Banks×dist. to comp. adv.  0.047**    0.037*  0.055***    0.054**  0.054***    0.056***    (0.020)    (0.021)  (0.020)    (0.023)  (0.019)    (0.020)  Banks  –0.052    –0.041  –0.068*    –0.038  –0.060    –0.075*    (0.040)    (0.039)  (0.041)    (0.049)  (0.041)    (0.043)  Dom. private bonds×dist. to comp. adv.    0.107***  0.092**                  (0.036)  (0.041)              Domestic private bonds to GDP    –0.075  –0.058                  (0.109)  (0.113)              Inter. private bonds×dist. to comp. adv.          0.075  –0.005                  (0.052)  (0.066)        International private bonds to GDP          0.261*  0.372***                  (0.140)  (0.139)        Nonfin. inter. bonds×dist. to comp. adv.                –0.005  –0.176                  (0.201)  (0.209)  Nonfinancial international bonds to GDP                –0.695  –0.518                  (0.518)  (0.530)  Full set of controls  Yes  Yes  Yes  Yes  Yes  Yes  Yes  Yes  Yes  Exporting country FE  Yes  Yes  Yes  Yes  Yes  Yes  Yes  Yes  Yes  Year FE  Yes  Yes  Yes  Yes  Yes  Yes  Yes  Yes  Yes  HS6 product strata  Yes  Yes  Yes  Yes  Yes  Yes  Yes  Yes  Yes  Observations  127, 260  127, 260  127, 260  158, 291  158, 291  158, 291  168, 113  168, 113  168, 113  Table A.III. Stock markets, comparative advantage, and external finance dependence The dependent variable is the hazard rate on the US destination market for an export spell of product k from country c that started at time tz. All regressions are estimated using the Cox Proportional Hazard model (Cox, 1972) and control for country- and time-fixed effects. Estimations also allow for different baseline hazard rate across industries (Columns 1–2) and across products (Columns 3–4) by defining industry j and product k as strata variables. Additional control variables include the interaction between the development of stock markets in country c and the dependence of industry j on external finance (stock markets×dependence on external finance − StMc,tz*ExFj), the interaction between the GDP per capita of country c and the distance to comparative advantage (GDPpcc,tz*DCAck,tz). Other variables are defined in Table III and Table V. Full set of controls corresponds to Column (6) of Table III. We report coefficients and not hazard ratios. Robust standard errors clustered at country * time (c*tz) level are in parentheses. *, **, and *** denote statistical significance at the 10%, 5%, and 1% level, respectively. Dep. Var.: hazard rate  (1)  (2)  (3)  (4)  Distance to comparative advantage  0.102***  –0.022  0.113***  0.035    (0.010)  (0.071)  (0.011)  (0.081)  Stock markets×distance to comparative advantage  –0.006  –0.014  –0.008  –0.012    (0.011)  (0.012)  (0.013)  (0.013)  Stock markets×dependence on external finance  –0.172***  –0.176***  –0.209***  –0.212***    (0.028)  (0.028)  (0.031)  (0.031)  GDPpc×distance to comparative advantage    0.014*    0.009      (0.008)    (0.009)  Stock markets  0.076***  0.087***  0.088***  0.093***    (0.029)  (0.030)  (0.032)  (0.032)  GDPpc  0.210  0.195  0.271**  0.262**    (0.132)  (0.133)  (0.133)  (0.133)  Full set of controls  Yes  Yes  Yes  Yes  Exporting country FE  Yes  Yes  Yes  Yes  Year FE  Yes  Yes  Yes  Yes  ISIC4 industry strata  Yes  Yes      HS6 product strata      Yes  Yes  Observations  182,592  182,592  182,592  182,592  Dep. Var.: hazard rate  (1)  (2)  (3)  (4)  Distance to comparative advantage  0.102***  –0.022  0.113***  0.035    (0.010)  (0.071)  (0.011)  (0.081)  Stock markets×distance to comparative advantage  –0.006  –0.014  –0.008  –0.012    (0.011)  (0.012)  (0.013)  (0.013)  Stock markets×dependence on external finance  –0.172***  –0.176***  –0.209***  –0.212***    (0.028)  (0.028)  (0.031)  (0.031)  GDPpc×distance to comparative advantage    0.014*    0.009      (0.008)    (0.009)  Stock markets  0.076***  0.087***  0.088***  0.093***    (0.029)  (0.030)  (0.032)  (0.032)  GDPpc  0.210  0.195  0.271**  0.262**    (0.132)  (0.133)  (0.133)  (0.133)  Full set of controls  Yes  Yes  Yes  Yes  Exporting country FE  Yes  Yes  Yes  Yes  Year FE  Yes  Yes  Yes  Yes  ISIC4 industry strata  Yes  Yes      HS6 product strata      Yes  Yes  Observations  182,592  182,592  182,592  182,592  Table A.III. Stock markets, comparative advantage, and external finance dependence The dependent variable is the hazard rate on the US destination market for an export spell of product k from country c that started at time tz. All regressions are estimated using the Cox Proportional Hazard model (Cox, 1972) and control for country- and time-fixed effects. Estimations also allow for different baseline hazard rate across industries (Columns 1–2) and across products (Columns 3–4) by defining industry j and product k as strata variables. Additional control variables include the interaction between the development of stock markets in country c and the dependence of industry j on external finance (stock markets×dependence on external finance − StMc,tz*ExFj), the interaction between the GDP per capita of country c and the distance to comparative advantage (GDPpcc,tz*DCAck,tz). Other variables are defined in Table III and Table V. Full set of controls corresponds to Column (6) of Table III. We report coefficients and not hazard ratios. Robust standard errors clustered at country * time (c*tz) level are in parentheses. *, **, and *** denote statistical significance at the 10%, 5%, and 1% level, respectively. Dep. Var.: hazard rate  (1)  (2)  (3)  (4)  Distance to comparative advantage  0.102***  –0.022  0.113***  0.035    (0.010)  (0.071)  (0.011)  (0.081)  Stock markets×distance to comparative advantage  –0.006  –0.014  –0.008  –0.012    (0.011)  (0.012)  (0.013)  (0.013)  Stock markets×dependence on external finance  –0.172***  –0.176***  –0.209***  –0.212***    (0.028)  (0.028)  (0.031)  (0.031)  GDPpc×distance to comparative advantage    0.014*    0.009      (0.008)    (0.009)  Stock markets  0.076***  0.087***  0.088***  0.093***    (0.029)  (0.030)  (0.032)  (0.032)  GDPpc  0.210  0.195  0.271**  0.262**    (0.132)  (0.133)  (0.133)  (0.133)  Full set of controls  Yes  Yes  Yes  Yes  Exporting country FE  Yes  Yes  Yes  Yes  Year FE  Yes  Yes  Yes  Yes  ISIC4 industry strata  Yes  Yes      HS6 product strata      Yes  Yes  Observations  182,592  182,592  182,592  182,592  Dep. Var.: hazard rate  (1)  (2)  (3)  (4)  Distance to comparative advantage  0.102***  –0.022  0.113***  0.035    (0.010)  (0.071)  (0.011)  (0.081)  Stock markets×distance to comparative advantage  –0.006  –0.014  –0.008  –0.012    (0.011)  (0.012)  (0.013)  (0.013)  Stock markets×dependence on external finance  –0.172***  –0.176***  –0.209***  –0.212***    (0.028)  (0.028)  (0.031)  (0.031)  GDPpc×distance to comparative advantage    0.014*    0.009      (0.008)    (0.009)  Stock markets  0.076***  0.087***  0.088***  0.093***    (0.029)  (0.030)  (0.032)  (0.032)  GDPpc  0.210  0.195  0.271**  0.262**    (0.132)  (0.133)  (0.133)  (0.133)  Full set of controls  Yes  Yes  Yes  Yes  Exporting country FE  Yes  Yes  Yes  Yes  Year FE  Yes  Yes  Yes  Yes  ISIC4 industry strata  Yes  Yes      HS6 product strata      Yes  Yes  Observations  182,592  182,592  182,592  182,592  Appendix B: Full Sample of Countries Exporting to the USA Argentina, Australia, Austria, Bangladesh, Benin, Bolivia, Brazil, Cameroon, Canada, Chile, China, Colombia, Congo, Costa Rica, Denmark, Dominican Republic, Ecuador, Egypt, El Salvador, Finland, France, Gambia, Germany, Ghana, Greece, Guatemala, Haiti, Honduras, India, Indonesia, Ireland, Italy, Jamaica, Japan, Jordan, Kenya, Korea, Malawi, Malaysia, Mali, Mauritius, Mexico, Mozambique, Nepal, Netherlands, New Zealand, Nicaragua, Niger, Norway, Pakistan, Panama, Paraguay, Peru, Philippines, Portugal, Rwanda, Senegal, Spain, Sri Lanka, Sweden, Switzerland, Thailand, Togo, Trinidad and Tobago, Tunisia, Turkey, United Kingdom, Uruguay, Venezuela, Zambia, Zimbabwe. Footnotes 1 Besides complementing the discipline imposed by the product markets, financial markets and institutions can improve resource allocation also by monitoring the firms and providing valuable information about them to other agents, thereby decreasing information asymmetries in the economy. This task can be executed both by banks (Diamond, 1984; Ramakrishnan and Thakor, 1984; Bhattacharya and Pfleiderer, 1985; Boyd and Prescott, 1986; Allen, 1990) and by stock markets (Grossman and Stiglitz, 1980; Merton, 1987; Holmstrom and Tirole, 1993). 2 Berman and Héricourt (2010) address this issue when examining the selection role of finance with respect to exporting. They look at firm’s productivity rather than just its financial health and show that productivity is an important determinant of export decision only after some threshold of financial development is reached. However, subsidies or political connections could still affect both productivity and export performance at the firm level. 3 Most of our estimations focus on the physical and human capital. Table X confirms that our results extend to natural resources as well. 4 Economists since Smith (1776) and Hicks (1935) have argued that competition in product markets is an important source of discipline, ensuring that firms and individuals make the best possible use of the resources at their disposal. Other contributions stressing the disciplining impact of product markets include Leibenstein (1966), Machlup (1967), Winter (1971), Hart (1983), Nickell (1996), Schmidt (1997), Raith (2003), or Giroud and Mueller (2010, 2011). Various papers examine external debt as additional incentive mechanism for self-interested managers (Stiglitz, 1974; Jensen and Meckling, 1976; Grossman and Hart, 1982; Jensen, 1986; Stulz, 1990; Hart and Moore, 1995; Zwiebel, 1996; Lang, Ofek, and Stulz, 1996). 5 The utility maximization of a risk-neutral manager amounts to the maximizing of her expected perks in the two periods. Assuming a risk-averse manager with a concave rather than linear utility function would not materially affect the implications of the model. 6 We interpret managerial export perks broadly, including the well-documented wage premium and better working conditions offered by the exporting firms (e.g., Bernard and Jensen, 1995; Verhoogen, 2008; Frias, Kaplan, and Verhoogen, 2009; Brambilla, Lederman, and Porto, 2010), which can partially persist even after a switch to another employer. In particular, Mion and Opromolla (2011) find a 15% wage premium for managers who have previously worked for an exporting firm. Managing export projects also involves non-monetary rewards like traveling abroad or spending time in luxury hotels. All these perks and benefits provide incentives for managers to keep alive even exports that they know are not efficient. 7 For example, the manager needs to request a renewal of an existing line of credit or attempt to secure another form of trade finance provided by the banking sector like letter of credit, factoring services, or bank guarantee. 8 Diamond (1984) shows how banks can overcome the problem of weak incentives of dispersed securities holders to monitor the firms. Fama (1985) argues that banks are well placed to provide monitoring due to their privileged access to firms’ internal information. And Stiglitz and Weiss (1983) document that bank loans can provide the right incentives to managers even if banks cannot directly observe the chosen project. 9 In a frictionless Heckscher–Ohlin world (with complete frictionless markets, no tariffs, zero transportation costs, etc.), the projects perfectly aligned with the comparative advantage of the exporting country would always survive in the destination market (μE=1). 10 The manager will never attempt to terminate an efficient project. She would namely need to pay the cost kX just to be reassigned to another ongoing project that is associated with the same level of perks X and faces the same (with probability λ) or even lower (with probability 1−λ) chances of generating those perks over the two periods. Similarly, the manager will never find it optimal to initiate termination of an inefficient project and pay the costs kX after obtaining bank funding at time t = 1 if she did not find it optimal already at time t = 0 before securing the funding. 11 As mentioned before, the model deliberately does not deal with the issue of export entry. In particular, the “switching” costs k (effort needed to terminate an already existing project, uncertain reassignment to another ongoing project, etc.) do not apply when it comes to new export projects while effort related to export entry arguably does not differ across efficient and inefficient projects. The manager would therefore always prefer to enter the export market with an efficient project that is more likely to sustain the competitive pressures and thus generate export perks in the long term. Following the same logic, the export entry would be also efficient if we interpreted our model in a broader sense where the manager could initiate a new export project instead of pursuing reassignment to another ongoing project. 12 We assume that if indifferent (UT = UC), the manager will do the “right thing” and initiate the termination of an inefficient project. 13 As for the direct impact of financial development on the managerial benefit of initiating termination of an inefficient project, the derivative of the LHS of Equation (5) with respect to f has an ambiguous sign: ∂φ(0,f)∂f{1+μE+[1−μE]l}−∂φ(d,f)∂f{1+μI(d)+[1−μI(d)]l}⋛0  This is due to the fact that better access to external funding improves the probability of obtaining bank funding for both efficient and inefficient projects (∂φ(0,f)∂f>0, ∂φ(d,f)∂f>0). 14 In practice, banks are the principal suppliers of export financing (Amiti and Weinstein, 2011). Bank lending also clearly dominates debt financing around the world (Shleifer and Vishny, 1997). 15 For instance, if a country exports a given product to the USA during 10 years starting in 1992, we take this as one spell with a length of 10 years and initiation year 1992. If a country exports a given product to the USA during 1995–1998 and then again during 2000–2005, we treat these as two independent spells of 4 and 6 years with initiation years of 1995 and 2000, respectively. 16 Like in the two-country case, this reasoning is valid only if factor price equalization does not hold and the world is thus divided into multiple diversification cones. In Heckscher–Ohlin framework with more products than production factors, equalization of factor prices would namely lead to indeterminacy of both production and trade (see also Dornbusch, Fischer, and Samuelson, 1980). More recent papers on diversification cones include Schott (2003, 2004) and Xiang (2007). 17 Some scholars suggest the different dimensions of institutional quality as another possible source of comparative advantage. Nunn and Trefler (2014) provide a thorough overview of this literature. 18 Costinot, Donaldson, and Komunjer (2012) provide a modern interpretation of the revealed comparative advantage in the context of the Ricardian theory, testing it in a sample of twenty-one rich and middle-income countries and thirteen broadly defined industries. 19 To be fair, Morrow (2010) himself calls the evidence about potential bias in empirical work focusing on Ricardian comparative advantage “weak and mixed.” Costinot, Donaldson, and Komunjer (2012) show evidence for the orthogonality of Ricardian and Heckscher–Ohlin motives in their data, although this is not the main focus of their paper. 20 Note that we differ in this respect from the original measure in Cadot, Carrère, and Strauss-Kahn (2011). 21 The formulation of ωck,tz used by Hausmann, Hwang, and Rodrik (2007) and Cadot, Tumurchudur, and Shirotori (2009) slightly differs from the original index of revealed comparative advantage by Balassa (1965). This modified formulation ensures that the weights add up to one: ∑cωck,tz=∑cXck,tz/Xc,tz∑cXck,tz/Xc,tz=∑cXck,tz/Xc,tz∑cXck,tz/Xc,tz=1 22 We do not need export values for measuring the export survival rate but we do use them for constructing some trade-related control variables and the dependent variable in one of the robustness tests. 23 In particular, we are using BACI in HS classification from 1992 that covers the period 1994–2005. 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 1995–2005. 24 Dependence on external finance as well as strata effects in our econometric analysis are defined at the finer four-digit level of the ISIC classification. In the text, we therefore generally refer to j as denoting ISIC four-digit industry. 25 Romalis (2004) uses US SIC classification instead. 26 Our product-level variables are at the HS six-digit level and there is no direct concordance between the industry-level ISIC four-digit Revision 2 classification and the product-level HS classification. In order to match the data, we therefore first apply the concordance between ISIC Revision 2 and ISIC Revision 3 classification and then the concordance between the latter and the HS six-digit classification. 27 This approach is also in accordance with our model where the intensity of product market competition in the destination market (represented by μE,μI,∂μI(.)∂d<0) is held constant while the variation in financial development (f) determines the probability of obtaining export financing in a given exporting country and thus affects the incentives of self-interested managers to initiate termination of inefficient exports. 28 Due to its political and military power, the USA is also well-placed to withstand political pressures from other countries trying to place their products on international markets. For example, Berger et al. (2013) show that successful CIA interventions in foreign countries lead to dramatic export increases from the USA to these countries but have no impact on the exports from the affected countries to the US destination market. 29 Besedes and Prusa (2006a) were the first to apply the analytical tools of survival analysis in the context of international trade, triggering an extensive follow-up research on trade survival (see, e.g., Besedes and Prusa, 2006b, 2011; Carballo and Volpe Martincus, 2008; Jaud, Kukenova, and Strieborny, 2009, 2015; Nitsch, 2009; Brenton, Saborowski, and von Uexkull, 2010; Obashi, 2010; Fugazza and Molina, 2011; Brenton, Cadot, and Pierola, 2012; Carrere and Strauss-Khan, 2012; Chen, 2012; Besedes, 2013; Cadot et al., 2013). 30 In one of the robustness checks in the Online Appendix, we also run a linear probability model of exiting the US destination market as an alternative to the formal survival framework. 31 Ai and Norton (2003) discuss the issues associated with estimating a non-linear model with interaction terms. As a robustness check, we define a time-varying dummy variable marking countries with high level of banking development. We set λtz at the 75th percentile of the cross-sectional distribution of countries’ level of banking development. We split our measure of distance to comparative advantage in two, one for countries in the bottom 75th percentile (B75*DCAck,tz) and one for countries in the top 25th percentile (T25*DCAck,tz) regarding the level of banking development. We then re-run our estimations, replacing the interaction term between the level of banking sector development and the measure of distance to comparative advantage (BCc,tz*DCAck,tz) with these two new distance measures (B75*DCAck,tz,T25*DCAck,tz). Results confirm the additional effect of banking development. The coefficients are increasing in magnitude from the bottom 75th percentile to the top 25th percentile and are statistically significantly different from each other. Results are available upon request. 32 Online Appendix C.2 reports results for alternative methods of handling multiple spells. 33 See Hess and Persson (2012) for a full discussion of the drawbacks of Cox regressions in the context of trade data and alternative estimators. 34 Stratification in the continuous Cox model is equivalent to modeling frailty with a gamma distribution in a discrete-time duration model (Hosmer, Lemeshow, and May, 2008). 35 Countries’ factor endowments are time-varying variables measured in the initial year of a trade relationship. Direct impact of the physical and human capital (κc,tz, hc,tz) is therefore not absorbed by the exporting country fixed effects. The same logic also applies to other time-varying characteristics of the exporting countries like the bank credit over GDP (BCc,tz) or the GDP per capita (GDPpcc,tz). 36 To be precise, this capturing is not complete due to application of concordances in order to match data measured in different classifications as described in the data section. We do not report the coefficients for direct effect of factor intensities as these do not have a meaningful economic interpretation and are a mere artifact of the circumstance that official concordances sometimes match an ISIC three-digit industry according to Revision 2 to more than one ISIC four-digit industry according to Revision 3. 37 The estimated coefficient for DCAck,tz in Column (4) of Table III that abstracts from the impact of banks is 0.145, implying a following computation: 0.145*(1.90−0.71)≃0.173. 38 To put our result in the broader context of existing research on financial development and the real economy, let us mention the economic magnitude found in one of the seminal papers in this area. Rajan and Zingales (1998, p. 574) show that an industry at the 75th percentile of financial dependence would grow 1.3% faster than an industry at the 25th percentile of financial dependence if it were located in Italy as opposed to less financially developed Philippines. The 1.3% should in this case be compared with an average growth rate of 3.4%. 39 Recent papers that examine bank loans versus public bonds from different perspectives than disciplining role of debt include Gomes and Phillips (2012); Colla, Ippolito, and Li (2013); and Morellec, Valta, and Zhdanov (2015). 40 The strength of banking sector (BCc,tz) and the GDP per capita (GDPpcc,tz) are correlated at 61% in our sample. 41 In terms of our simple model, the domestic stock markets could so heavily misallocate the excessive inflows of foreign capital that it became disproportionately easier for inefficient export projects to receive external funding (∂2φ(d,f)∂d∂f>0) and the whole expression in Equation (7) could become negative. We leave the exploration of this possibility for further research. 42 We are grateful to the editor for drawing our attention to this fact. 43 We are grateful to the referee for suggesting this empirical specification that allows us to explore the issue of export entry within the empirical framework of duration analysis. 44 We are grateful to the referee for making us aware of this possibility. 45 Görg, Henry, and Strobl (2008) provide some evidence that general government subsidies like R&D grants promote also the intensive margin of exports. In either case, the product market competition arguably prevails in the long run as most countries lack resources to subsidize exports of non-competitive products indefinitely. Distortionary government interventions in exporting countries can therefore be expected to affect the short-term export dynamics more than the long-term export survival. We leave this issue for further research. 46 The excluded microstate islands are The Dominican Republic, Haiti, Jamaica, Mauritius, and Trinidad and Tobago. 47 A reader interested in the background of the performed robustness tests can consult various papers cited in the Online Appendix (Breslow, 1974; Efron, 1988; Moffitt and Rendall, 1995; Bertrand, Duflo, and Mullainathan, 2004; Cameron, Gelbach and Miller, 2006; Araujo, Mion, and Ornelas, 2014). References Aghion P., Dewatripont M., Rey P. ( 1999) Competition, financial discipline and growth, Review of Economic Studies  66, 825– 852. Google Scholar CrossRef Search ADS   Ai C., Norton E. C. ( 2003) Interaction terms in logit and probit models, Economics Letters  80, 123– 129. Google Scholar CrossRef Search ADS   Allen F. ( 1990) The market for information and the origin of financial intermediaries, Journal of Financial Intermediation  1, 3– 30. Google Scholar CrossRef Search ADS   Allen F. ( 1993) Stock markets and resource allocation, in: Mayer C., Vives X. (eds.), Capital Markets and Financial Intermediation , Cambridge University Press, Cambridge, UK, pp. 81– 108. Google Scholar CrossRef Search ADS   Amiti M., Weinstein D. E. ( 2011) Exports and financial shocks, Quarterly Journal of Economics  126, 1841– 1877. Google Scholar CrossRef Search ADS   Araujo L., Mion G., Ornelas E. ( 2014) Institutions and export dynamics. Mimeo, London School of Economics. Balassa B. ( 1965) Trade liberalisation and “revealed” comparative advantage, The Manchester School of Economic and Social Studies  33, 99– 123. Google Scholar CrossRef Search ADS   Beck T. ( 2002) Financial development and international trade: is there a link?, Journal of International Economics  57, 107– 131. Google Scholar CrossRef Search ADS   Beck T. ( 2003) Financial dependence and international trade, Review of International Economics  11, 296– 316. Google Scholar CrossRef Search ADS   Beck T., Demirgüç-Kunt A., Levine R. ( 2000) A new database on financial development and structure, World Bank Economic Review  14, 597– 605. Google Scholar CrossRef Search ADS   Beck T., Levine R. ( 2002) Industry growth and capital allocation: does having a market- or bank-based system matter?, Journal of Financial Economics  64, 147– 180. Google Scholar CrossRef Search ADS   Becker B., Chen J., Greenberg D. ( 2013) Financial development, fixed costs, and international trade, Review of Corporate Finance Studies  2, 1– 28. Google Scholar CrossRef Search ADS   Bekaert G., Harvey C. R., Lundblad C., Siegel S. ( 2007) Global growth opportunities and market integration, Journal of Finance  62, 1081– 1137. Google Scholar CrossRef Search ADS   Berger D., Easterly W., Nunn N., Satyanath S. ( 2013) Commercial imperialism? Political influence and trade during the Cold War, American Economic Review  103, 863– 896. Google Scholar CrossRef Search ADS   Berger P. G., Ofek E., Yermack D. L. ( 1997) Managerial entrenchment and capital structure decisions, Journal of Finance  52, 1411– 1438. Google Scholar CrossRef Search ADS   Berman N., Héricourt J. ( 2010) Financial factors and the margins of trade: evidence from cross-country firm-level data, Journal of Development Economics  93, 206– 217. Google Scholar CrossRef Search ADS   Bernard A. B., Jensen J. B. ( 1995) Exporters, jobs and wages in U.S. manufacturing, 1976–87, in: Baily M. N., Reiss P. C., Winston C. (eds.), Brookings Papers on Economic Activity: Microeconomics, Volume 1995 , Brooking Institutions Press, Washington, DC, pp. 67– 119. Bertrand M., Duflo E., Mullainathan S. ( 2004) How much should we trust differences-in-differences estimates?, Quarterly Journal of Economics  119, 249– 275. Google Scholar CrossRef Search ADS   Bertrand M., Mullainathan S. ( 2003) Enjoying the quiet life? Managerial behavior following anti-takeover legislation, Journal of Political Economy  111, 1043– 1075. Google Scholar CrossRef Search ADS   Besedes T. ( 2013) The role of NAFTA and returns to scale in export duration, CESifo Economic Studies  59, 306– 336. Google Scholar CrossRef Search ADS   Besedes T., Prusa T. J. ( 2006a) Ins, outs, and the duration of trade, Canadian Journal of Economics  39, 266– 295. Google Scholar CrossRef Search ADS   Besedes T., Prusa T. J. ( 2006b) Product differentiation and duration of US import trade, Journal of International Economics  70, 339– 358. Google Scholar CrossRef Search ADS   Besedes T., Prusa T. J. ( 2011) The role of extensive and intensive margins and export growth, Journal of Development Economics  96, 371– 379. Google Scholar CrossRef Search ADS   Besedes T., Kim B. C., Lugovskyy V. ( 2014) Export growth and credit constraints, European Economic Review  70, 350– 370. Google Scholar CrossRef Search ADS   Bhagwati J. N. ( 1972) The Heckscher–Ohlin theorem in the multi-commodity case, Journal of Political Economy  80, 1052– 1055. Google Scholar CrossRef Search ADS   Bhattacharya S., Pfleiderer P. ( 1985) Delegated portfolio management, Journal of Economic Theory  36, 1– 25. Google Scholar CrossRef Search ADS   Boyd J. H., Prescott E. C. ( 1986) Financial intermediary-coalitions, Journal of Economic Theory  38, 211– 232. Google Scholar CrossRef Search ADS   Brambilla I., Lederman D., Porto G. ( 2010) Exports, export destinations, and skills, American Economic Review  102, 3406– 3488. Google Scholar CrossRef Search ADS   Braun M. ( 2003): Financial contractibility and asset hardness. Mimeo, University of California—Los Angeles. Brenton P., Cadot O., Pierola M. D. ( 2012) Pathways to African Export Sustainability , World Bank, Washington, DC. Google Scholar CrossRef Search ADS   Brenton P., Saborowski C., von Uexkull E. ( 2010) What explains the low survival rate of developing country export flows?, World Bank Economic Review  24, 474– 499. Google Scholar CrossRef Search ADS   Breslow N. ( 1974) Covariance analysis of censored survival data, Biometrics  30, 89– 99. Google Scholar CrossRef Search ADS   Bricongne J. C., Fontagné L., Gaulier G., Taglioni D., Vicard V. ( 2010) Firms and the global crisis: French exports in the turmoil. ECB Working Paper No. 1245. Cadot O., Tumurchudur B., Shirotori M. ( 2009) Revealed factor intensity indices at the product level. Mimeo, United Nations Conference on Trade and Development. Cadot O., Carrère C., Strauss-Kahn V. ( 2011) Export diversification: What’s behind the hump?, Review of Economics and Statistics  93, 590– 605. Google Scholar CrossRef Search ADS   Cadot O., Iacovone L., Pierola M. D., Rauch F. ( 2013) Success and failure of African exporters, Journal of Development Economics  101, 284– 296. Google Scholar CrossRef Search ADS   Carballo J., Volpe Martincus C. ( 2008) Survival of new exporters in developing countries: does it matter how they diversify?, Inter-American Development Bank Working Paper No. 9291. Carrere C., Strauss-Khan V. ( 2012) Exports dynamics: raising developing countries exports survival through experience. FERDI Working Paper No. P35-A. Cameron A. C., Gelbach J. B., Miller D. L. ( 2006) Robust inference with multi-way clustering. NBER Technical Working Paper No. 327. Cameron A. C., Trivedi P. K. ( 2005) Microeconometrics: Methods and Applications , Cambridge University Press, New York. Google Scholar CrossRef Search ADS   Cassey A., Schmeiser K. ( 2010) The agglomeration of exporters by destination. FREIT Working Paper No. 233. Cihak M., Demirgüç-Kunt A., Feyen E., Levine R. ( 2012) Benchmarking financial systems around the world. World Bank Policy Research Working Paper No. 6175. Chor D., Manova K. ( 2012) Off the cliff and back? Credit conditions and international trade during the global financial crisis, Journal of International Economics  87, 117– 133. Google Scholar CrossRef Search ADS   Colla P., Ippolito F., Li K. ( 2013) Debt specialization, Journal of Finance  68, 2117– 2141. Google Scholar CrossRef Search ADS   Costinot A. ( 2009) An elementary theory of comparative advantage, Econometrica  77, 1165– 1192. Google Scholar CrossRef Search ADS   Costinot A., Donaldson D., Komunjer I. ( 2012) What goods do countries trade? A quantitative exploration of Ricardo’s ideas, Review of Economic Studies  79, 581– 608. Google Scholar CrossRef Search ADS   Cox D. R. ( 1972) Regression models and life tables, Journal of the Royal Statistical Society, Series B  34, 187– 220. Deardorff A. V. ( 1979) Weak links in the chain of comparative advantage, Journal of International Economics  9, 197– 209. Google Scholar CrossRef Search ADS   Denis D. J., Mihov V. T. ( 2003) The choice among bank debt, non-bank private debt, and public debt: evidence from new corporate borrowings, Journal of Financial Economics  70, 3– 28. Google Scholar CrossRef Search ADS   Diamond D. ( 1984) Financial intermediation and delegated monitoring, Review of Economic Studies  51, 393– 414. Google Scholar CrossRef Search ADS   Do Q. T., Levchenko A. A. ( 2007) Comparative advantage, demand for external finance, and financial development, Journal of Financial Economics  86, 796– 834. Google Scholar CrossRef Search ADS   Dornbusch R., Fischer S., Samuelson P. A. ( 1977) Comparative advantage, trade, and payments in a Ricardian model with a continuum of goods, American Economic Review  67, 823– 839. Dornbusch R., Fischer S., Samuelson P. A. ( 1980) Heckscher–Ohlin trade theory with a continuum of goods, Quarterly Journal of Economics  95, 203– 224. Google Scholar CrossRef Search ADS   Eaton J., Kortum S. ( 2002) Technology, geography, and trade, Econometrica  70, 1741– 1779. Google Scholar CrossRef Search ADS   Efron B. ( 1988) Logistic regression, survival analysis, and the Kaplan–Meier curve, Journal of the American Statistical Association  83, 414– 425. Google Scholar CrossRef Search ADS   Fama E. F. ( 1985) What’s different about banks?, Journal of Monetary Economics  15, 29– 39. Google Scholar CrossRef Search ADS   Fisman R., Love I. ( 2004) Financial development and intersectoral allocation: a new approach, Journal of Finance  59, 2785– 2807. Google Scholar CrossRef Search ADS   Frias J., Kaplan D., Verhoogen E. ( 2009) Exports and wage premia: evidence from Mexican employer–employee data. Mimeo, Columbia University. Fugazza M., Molina A. C. ( 2011) On the determinants of exports survival. UNCTAD Blue Series Papers No. 47, United Nations Conference on Trade and Development. Gaulier G., Zignago S. ( 2010) BACI: International trade database at the product level. The 1994–2007 version. CEPII Working Paper No. 2010-23. Giroud X., Mueller H. M. ( 2010) Does corporate governance matter in competitive industries?, Journal of Financial Economics  95, 312– 331. Google Scholar CrossRef Search ADS   Giroud X., Mueller H. M. ( 2011) Corporate governance, product market competition, and equity prices, Journal of Finance  66, 563– 600. Google Scholar CrossRef Search ADS   Gomes A., Phillips G. ( 2012) Why do public firms issue private and public securities?, Journal of Financial Intermediation  21, 619– 658. Google Scholar CrossRef Search ADS   Görg H., Henry M., Strobl E. ( 2008) Grant support and exporting activity, Review of Economics and Statistics  90, 168– 174. Google Scholar CrossRef Search ADS   Greenaway D., Guariglia A., Kneller R. ( 2007) Financial actors and exporting decisions, Journal of International Economics  73, 377– 395. Google Scholar CrossRef Search ADS   Grossman S., Stiglitz J. ( 1980) On the impossibility of informationally efficient markets, American Economic Review  70, 393– 408. Grossman S., Hart O. ( 1982) Corporate financial structure and managerial incentives, in: McCall J. (ed.), The Economics of Information and Uncertainty , University of Chicago Press, Chicago. Harris M., Raviv A. ( 1990) Capital structure and the informational role of debt, Journal of Finance  45, 321– 349. Google Scholar CrossRef Search ADS   Hart O. ( 1983) The market mechanism as an incentive scheme, Bell Journal of Economics  14, 366– 382. Google Scholar CrossRef Search ADS   Hart O., Moore J. ( 1995) Debt and seniority: an analysis of the role of hard claims in constraining management, American Economic Review  85, 567– 585. Hausmann R., Hwang J., Rodrik D. ( 2007) What you export matters, Journal of Economic Growth  12, 1– 25. Google Scholar CrossRef Search ADS   Hess W., Persson M. ( 2012) The duration of trade revisited: Continuous-time vs. discrete-time hazards, Empirical Economics  43, 1083– 1107. Google Scholar CrossRef Search ADS   Hicks J. ( 1935) Annual survey of economic theory: the theory of monopoly, Econometrica  3, 1– 20. Google Scholar CrossRef Search ADS   Holmstrom B., Tirole J. ( 1993) Market liquidity and performance monitoring, Journal of Political Economy  101, 678– 709. Google Scholar CrossRef Search ADS   Hosmer D. W.Jr, Lemeshow S., May S. ( 2008) Applied Survival Analysis: Regression Modeling of Time to Event Data , 2th edition, John Wiley and Sons Inc, Hoboken. Google Scholar CrossRef Search ADS   Jaud M., Kukenova M., Strieborny M. ( 2009) Financial dependence and intensive margin of trade. PSE Working Paper No. 2009-35. Jaud M., Kukenova M., Strieborny M. ( 2015) Financial development and sustainable exports: evidence from firm–product data, World Economy  38, 1090– 1114. Google Scholar CrossRef Search ADS   Jaud M., Kukenova M., Strieborny M. ( 2016) Stock market liberalizations and export dynamics. Mimeo, Lund University. Jensen M. C. ( 1986) Agency costs of free cash flow, corporate finance, and takeovers, American Economic Review  76, 323– 329. Jensen M. C. ( 1993) The modern industrial revolution, exit, and the failure of internal control systems, Journal of Finance  48, 831– 880. Google Scholar CrossRef Search ADS   Jensen M. C., Meckling W. H. ( 1976) Theory of the firm: managerial behavior, agency costs and ownership structure, Journal of Financial Economics  3, 305– 360. Google Scholar CrossRef Search ADS   Jones R. W. ( 1956) Factor proportions and the Heckscher–Ohlin theorem, Review of Economic Studies  24, 1– 10. Google Scholar CrossRef Search ADS   Ju J., Wei S. J. ( 2005) Endowment versus finance: a wooden barrel theory of international trade. CEPR Discussion Paper No. 5109. Kiefer N. ( 1988) Economic duration data and hazard functions, Journal of Economic Literature  2, 646– 679. Koenig P. ( 2009) Agglomeration and the export decisions of French firms, Journal of Urban Economics  66, 186– 195. Google Scholar CrossRef Search ADS   Lang L., Ofek E., Stulz R. M. ( 1996) Leverage, investment, and firm growth, Journal of Financial Economics  40, 3– 29. Google Scholar CrossRef Search ADS   Leibenstein H. ( 1966) Allocative efficiency vs. “X–efficiency”, American Economic Review  56, 392– 415. Levine R. ( 2005) Finance and growth: theory and evidence, in: Aghion P., Durlauf S. (eds.), Handbook of Economic Growth , North Holland, Amsterdam, pp. 865– 934. Levchenko A. A., Lewis L. T., Tesar L. L. ( 2011) The role of trade finance in the U.S. trade collapse: a skeptic’s view, in: Chauffour J.-P., Malouche M. (eds.), Trade Finance During the Great Trade Collapse , World Bank, Washington, DC, pp. 133– 147. Lin C., Ma Y., Malatesta P., Xuan Y. ( 2013) Corporate ownership structure and the choice between bank debt and public debt, Journal of Financial Economics  109, 517– 534. Google Scholar CrossRef Search ADS   Love I. ( 2003) Financial development and financing constraints: international evidence from the structural investment model, Review of Financial Studies  16, 765– 791. Google Scholar CrossRef Search ADS   Machlup F. ( 1967) Theories of the firm: marginalist, behavioral, managerial, American Economic Review  57, 1– 33. Manova K. ( 2008) Credit constraints, equity market liberalizations and international trade, Journal of International Economics  76, 33– 47. Google Scholar CrossRef Search ADS   Manova K. ( 2013) Credit constraints, heterogeneous firms, and international trade, Review of Economic Studies  80, 711– 744. Google Scholar CrossRef Search ADS   Merton R. ( 1987) A simple model of capital market equilibrium with incomplete information, Journal of Finance  42, 483– 510. Google Scholar CrossRef Search ADS   Minetti R., Zhu S. C. ( 2011) Credit constraints and firm export: microeconomic evidence from Italy, Journal of International Economics  83, 109– 125. Google Scholar CrossRef Search ADS   Mion G., Opromolla L. D. ( 2011) Managers’ mobility, trade status, and wages. CEPR Discussion Paper No. 8230. Moffitt R. A., Rendall M. S. ( 1995) Cohort trends in the lifetime distribution of female family headship in the United States, 1968–1985, Demography  32, 407– 424. Google Scholar CrossRef Search ADS   Morellec E., Valta P., Zhdanov A. ( 2015) Financing investment: the choice between bonds and bank loans, Management Science  61, 2580– 2602. Google Scholar CrossRef Search ADS   Morrow P. M. ( 2010) Ricardian–Heckscher–Ohlin comparative advantage: theory and evidence, Journal of International Economics  82, 137– 151. Google Scholar CrossRef Search ADS   Moulton B. R. ( 1986) Random group effects and the precision of regression estimates, Journal of Econometrics  32, 385– 397. Google Scholar CrossRef Search ADS   Moulton B. R. ( 1990) An illustration of a pitfall in estimating the effects of aggregate variables on micro unit, Review of Economics and Statistics  72, 334– 338. Google Scholar CrossRef Search ADS   Muûls M. ( 2015) Exporters, importers and credit constraints, Journal of International Economics  95, 333– 343. Google Scholar CrossRef Search ADS   Nickell S. J. ( 1996) Competition and corporate performance, Journal of Political Economy  104, 724– 746. Google Scholar CrossRef Search ADS   Nitsch V. ( 2009) Die another day: duration in German import trade, Review of World Economics  145, 133– 154. Google Scholar CrossRef Search ADS   Nunn N., Trefler D. ( 2014) Domestic institutions as a source of comparative advantage, in: Gopinath G., Helpman E., Rogoff K. (eds.), Handbook of International Economics, Volume 4 , North Holland, Amsterdam, pp. 263– 315. Obashi A. ( 2010) Stability of production networks in East Asia: duration and survival of trade, Japan and the World Economy  22, 21– 30. Google Scholar CrossRef Search ADS   Paravisini D., Rappoport V., Schnabl P., Wolfenzon D. ( 2015) Dissecting the effect of credit supply on trade: evidence from matched credit–export data, Review of Economic Studies  82, 333– 359. Google Scholar CrossRef Search ADS   Raddatz C. ( 2006) Liquidity needs and vulnerability to financial underdevelopment, Journal of Financial Economics  80, 677– 722. Google Scholar CrossRef Search ADS   Raith M. ( 2003) Competition, risk, and managerial incentives, American Economic Review  93, 1425– 1436. Google Scholar CrossRef Search ADS   Rajan R. G., Zingales L. ( 1998) Financial dependence and growth, American Economic Review  88, 559– 586. Ramakrishnan R. T. S., Thakor A. ( 1984) Information reliability and a theory of financial intermediation, Review of Economic Studies  51, 415– 432. Google Scholar CrossRef Search ADS   Romalis J. ( 2004) Factor proportions and the structure of commodity trade, American Economic Review  94, 67– 97. Google Scholar CrossRef Search ADS   Scharfstein D. ( 1988) The disciplinary role of takeovers, Review of Economic Studies  55, 185– 199. Google Scholar CrossRef Search ADS   Schmidt K. M. ( 1997) Managerial incentives and product market competition, Review of Economic Studies  64, 191– 213. Google Scholar CrossRef Search ADS   Schott P. K. ( 2003) One size fits all? Heckscher–Ohlin specialization in global production, American Economic Review  93, 686– 708. Google Scholar CrossRef Search ADS   Schott P. K. ( 2004) Across-product versus within-product specialization in international trade, Quarterly Journal of Economics  119, 647– 678. Google Scholar CrossRef Search ADS   Shleifer A., Vishny R. W. ( 1997) A survey of corporate governance, Journal of Finance  52, 737– 783. Google Scholar CrossRef Search ADS   Smith A. ( 1776): An Inquiry into the Nature and Causes of the Wealth of Nations , Strahan W., Cadell T., London. Google Scholar CrossRef Search ADS   Stein J. ( 1988) Takeover threats and managerial myopia, Journal of Political Economy  96, 61– 80. Google Scholar CrossRef Search ADS   Stiglitz J. E. ( 1974) Incentives and risk sharing in sharecropping, Review of Economic Studies  41, 219– 255. Google Scholar CrossRef Search ADS   Stiglitz J. E., Weiss A. ( 1983) Incentive effects of terminations: applications to the credit and labor markets, American Economic Review  73, 912– 927. Strieborny M., Kukenova M. ( 2016) Investment in relationship-specific assets: does finance matter?, Review of Finance  20, 1487– 1515. Google Scholar CrossRef Search ADS   Stulz R. M. ( 1990) Managerial discretion and optimal financing policies, Journal of Financial Economics  26, 3– 27. Google Scholar CrossRef Search ADS   Verhoogen E. ( 2008) Trade, quality upgrading, and wage inequality in the Mexican manufacturing sector, Quarterly Journal of Economics  123, 489– 530. Google Scholar CrossRef Search ADS   Volpe Martincus C., Carballo J. ( 2008) Is export promotion effective in developing countries? Firm-level evidence on the intensive and the extensive margins of exports, Journal of International Economics  76, 89– 106. Google Scholar CrossRef Search ADS   Wurgler J. ( 2000) Financial markets and the allocation of capital, Journal of Financial Economics  58, 187– 214. Google Scholar CrossRef Search ADS   Xiang C. ( 2007) Diversification cones, trade costs and factor market linkages, Journal of International Economics  71, 448– 466. Google Scholar CrossRef Search ADS   Chen W. C. ( 2012) Innovation and duration of exports, Economics Letters  115, 305– 308. Google Scholar CrossRef Search ADS   Winter S. G. ( 1971) Satisficing, selection, and the innovating remnant, Quarterly Journal of Economics  85, 237– 261. Google Scholar CrossRef Search ADS   Zwiebel J. ( 1996) Dynamic capital structure under managerial entrenchment, American Economic Review  86, 1197– 1215. © The Authors 2017. Published by Oxford University Press on behalf of the European Finance Association. All rights reserved. For Permissions, please email: journals.permissions@oup.com This article is published and distributed under the terms of the Oxford University Press, Standard Journals Publication Model (https://academic.oup.com/journals/pages/about_us/legal/notices)
TI - Finance, Comparative Advantage, and Resource Allocation
JF - Review of Finance
DO - 10.1093/rof/rfx047
DA - 2018-05-01
UR - https://www.deepdyve.com/lp/oxford-university-press/finance-comparative-advantage-and-resource-allocation-raQJDpXzyT
SP - 1011
EP - 1061
VL - 22
IS - 3
DP - DeepDyve
ER -