Financial Factors and the Margins of Trade: Evidence ...

Viewer
Transcript

Financial Factors and the Margins of Trade: Evidence from Cross-Country Firm-Level Data∗ Nicolas Berman†

J´erˆome H´ ericourt‡

European University Institute

University of Lille I

Abstract

Using a large cross-country, ﬁrm level database containing 5,000 ﬁrms in 9 developing and emerging economies, we study how ﬁnancial factors aﬀect both ﬁrms’ export decisions and the amount exported by ﬁrms. First, our results highlight the importance of the impact of ﬁrms’ access to ﬁnance on their entry decision into the export market. However, better ﬁnancial health neither increases the probability of remaining an exporter once the ﬁrm has entered, nor the size of exports. Second, we ﬁnd that ﬁnancial constraints create a disconnection between ﬁrms’ productivity and their export status: productivity is only a signiﬁcant determinant of the export decision if the ﬁrm has a suﬃcient access to external ﬁnance. Finally, an increase in a country’s ﬁnancial development dampens this disconnection, thus acting positively both on the number of exporters and on the exporters’ selection process. These results contribute to the literature documenting the role of ﬁxed costs and of the extensive margin of trade in total trade adjustment, and provide microlevel evidence of the positive impact of ﬁnancial development on trade found by previous literature.

JEL classiﬁcation: D24, F14, F10 Keywords: Export decisions, margins of trade, ﬁnancial constraints

∗ For very useful comments and discussions, we thank Kenza Benhima, Anne-C´elia Disdier, Jonathan Eaton, Ann Harrison, Jean Imbs, Thierry Mayer, Farid Toubal, Thierry Verdier, two anonymous referees, as well as participants at the 10th ETSG conference, RIEF VIIIth meetings, and SMYE 2008 conference. All remaining errors are ours. † Corresponding author, Graduate Institute of International and Development Studies (IHEID). Address: Case Postale 136, CH - 1211, Geneva 21 - Switzerland. Tel: (0041) 22 908 5935. E-mail: [email protected]. ‡ ´ EQUIPPE-University of Lille. Universit´e de Lille 1, Facult´e des Sciences Economiques et Sociales, USTL - Cit´e Scientiﬁque - Bˆ at SH2, 59655 Villeneuve d’Ascq Cedex, France. Tel/Fax: (33) 1 44 07 82 71/47, Email: [email protected]

I

Introduction

Three important facts have been pointed out by recent literature in international trade. First, ﬁrms may face signiﬁcant ﬁxed, start-up costs when entering the export market, to gather information on foreign markets, establish a distribution system, or more generally to adapt their product to foreign tastes and environments. Second, ﬁrms are greatly heterogenous in terms of productivity. Together with the existence of ﬁxed costs, this heterogeneity helps to explain why not all ﬁrms engage in international trade, why exporters are more productive than domestic producers, and why an important share of variations in total exports comes from the adjustment of the extensive margin of trade - i.e. of the number of exporters (Melitz, 2003, Eaton et al., 2004, Bernard and Jensen, 2004). Third, at the macro level, ﬁnancial development exerts a signiﬁcant and positive impact on bilateral trade ﬂows (Beck, 2002, Berthou, 2006, Manova, 2007 among others), on both the number of bilateral ﬂows and the mean value of shipments, suggesting that heterogeneity in terms of access to ﬁnance (within and between countries) may be an important determinant of exporting behavior at the micro level. As they stand, ﬁrm-level studies have generally focused on the role of ﬁxed costs and productivity on the decision to export. The eﬀect of ﬁnancial constraints on international trade has been mainly studied at the country or sectoral level. There is a striking lack of evidence of both the impact of ﬁrms’ access to ﬁnance on their exporting behavior and the elements underlying the positive inﬂuence of ﬁnancial development on international trade. A comprehensive study of the relationship between ﬁnance and trade at the ﬁrm-level may thus improve our understanding of the impact of ﬁnancial development on trade, as well as of the role of productivity and ﬁxed costs on exporting behavior.

The study of the impact of ﬁnancial constraints, productivity and ﬁnancial development on both the extensive (i.e. the exporting decision) and the intensive (i.e. the amount exported by ﬁrm) margins of trade presented in this paper aims to ﬁll these gaps. We study both the determinants of ﬁrm-level exporting behavior and the impact of ﬁnancial development on trade at the ﬁrm level using a cross-country ﬁrm-level database containing 5,000 ﬁrms in 9 developing and emerging countries. Our ﬁndings can be summarized as follows: ﬁrst, larger productivity and lower ﬁnancial constraints both have a positive impact on export market participation (the extensive margin of trade). Interestingly, these two elements interact positively: productivity becomes increasingly important for exporting decisions as ﬁnancial constraints decrease. Productivity does not matter for ﬁrms holding a very low level of liquidity, or which are prevented from entering the export market because of their lack of access to external ﬁnancing. Put diﬀerently, credit constraints generate a disconnection

between ﬁrms’ productivity and the export status. This explains the imperfect selection of ﬁrms in the export market in terms of their productivity: while recent international trade theories, following Melitz (2003), predict that, for a given foreign market, all ﬁrms with productivity higher than a given threshold will export, evidence shows that, whatever the measure used to compute productivity, this is far from being true (see for example Eaton et al., 2004). Even if productivity and exporting status are overall positively correlated, an important number of low productivity ﬁrms do export, and some high productivity ﬁrms do not. Therefore, we propose in this paper an explanation of this fact based on the imperfect correlation between ﬁrms’ productivity and access to ﬁnance. Second, a direct implication of this result is that ﬁnancial development exerts a positive impact on trade ﬂows not only because it allows more ﬁrms to export, but also because it reduces this artiﬁcial disconnection between productivity and exports generated by liquidity constraints. Interacting ﬁrms’ productivity with the country’s level of ﬁnancial development, we ﬁnd that productivity is a stronger determinant of the decision to export in more ﬁnancially developed countries. More strikingly, productivity is actually insigniﬁcant for countries that are below the median of the sample in terms of ﬁnancial development. By reducing the strength of credit constraints in the economy, ﬁnancial development allows ﬁrms with a suﬃcient level of productivity to enter foreign markets. It may thus act positively both on the number and on the selection of ﬁrms: ﬁrms that enter the export market when the ﬁnancial market becomes deeper are the most productive of the ﬁnancially constrained ﬁrms. As a result, both the extensive and the intensive margins increase. Third, the role of ﬁnancial constraints on margins of trade is concentrated at the time of entry. Only productivity signiﬁcantly increases intensive margin. Neither the quantity exported (or the share of exports over total sales), nor the probability of remaining an exporter are aﬀected by ﬁnancial constraints. This is in line with the existence of large sunk costs which have to be paid to access the export market for the ﬁrst time. Our results suggest that once the ﬁrms have become exporters, the potential ﬁxed cost that they have to pay at each subsequent period is dramatically lower, or that ﬁnancial constraints matter less for continuing decisions - possibly because the exporting activity is viewed as less risky by ﬁnancial institutions in this case. The existence of sunk costs have be shown to importantly aﬀect the dynamics of aggregate trade ﬂows, in particular by generating export hysteresis (Baldwin, 1988, Baldwin and Krugman, 1989, Dixit, 1989). In combination with ﬁnancial constraints, their impact on the level of extensive margin of trade may be exacerbated.

Our paper is related to the literature that studies the eﬀect of ﬁnancial constraints on investment and the studies dealing with the determinants of international trade patterns. Sunk costs of exports can indeed be seen as a form of investment in intangible capital. The discussion on the impact of 2

ﬁnancial constraints on trade can therefore take place in the more general frame of the impact of ﬁnancial markets’ imperfections on investment. The seminal work of Modigliani and Miller (1958) indeed suggests that in the presence of perfect capital and credit markets, the ﬁnancing decision of a ﬁrm is irrelevant for its investment behavior. On the contrary, in the presence of market imperfections, ﬁnancing constraints will be reﬂected in ﬁrms’ investment decisions. The literature typically estimates models where investment (physical or intangible, such as R&D investment) is a function of the ﬁrm’s cash ﬂow. A signiﬁcant eﬀect of cash ﬂow is interpreted as an indicator of ﬁnancial constraints, since ﬁnancially constrained ﬁrms will only be able to invest if they own enough liquidity. Many ﬁrm-level studies have supported the existence of ﬁnancing constraints, but mainly in developed countries (see Greenaway et al., 2007, for a list of references concerning the US, the UK and some European countries). The papers dealing with developing countries are certainly less numerous (see Jaramillo and Weiss, 1996 on Ecuador, Harrison and McMillan, 2003 on the Ivory Coast or H´ericourt and Poncet, 2009 on China). Regarding the relationship between ﬁnancial constraints and trade, studies at the ﬁrm-level are much more scarce. To our knowledge, only two papers looks at the impact of ﬁnancial constraints on trade at the ﬁrm-level. The ﬁrst is Greenaway et al. (2007), which studies the impact of ﬁrms’ ﬁnancial health on their exporting decisions in a panel of UK manufacturing ﬁrms. Greenaway et al. (2007) ﬁnd that exporters signiﬁcantly display a better ﬁnancial health than non-exporters. However, this result seems to be mainly driven by the fact that ﬁnancial health is improved by participation in exporting activities. At the time of entry, exporters do not seem to be ﬁnancially healthier than domestic producers. On the contrary, exporting probability was found to be negatively and signiﬁcantly aﬀected by ﬁnancial constraints on a panel of Italian ﬁrms by Forlani (2008), and on Belgian ﬁrms by Muˆ uls (2008). We focus here on developing countries, in which ﬁnancial constraints may be much stronger and more binding; more importantly, we depart from these previous works by studying both margins of trade, by considering the possible interaction between productivity and access to ﬁnance, and the impact of ﬁnancial development at the micro level. Our results are consistent with recent theoretical works studying the eﬀect of ﬁnancial constraints and productivity on international trade (Chaney, 2005, Suwantaradon, 2008).

This paper contributes to the existing literature at various levels. First, it improves the understanding of the relationship between ﬁrms’ ﬁnancial constraints and trade at the ﬁrm level. Second, it provides micro evidence of the role of ﬁnancial development on trade. Third, our results support the view of sunk costs of entry as an important feature of international trade, especially in the presence of ﬁnancial market imperfections. The joint consideration of the extensive and intensive margins of trade is 3

thus especially important for the understanding of trade ﬂows and of their impact on macroeconomic adjustment. In the next section, we present our database, before discussing our general empirical methodology to test the impact of ﬁnancial constraints on both margins of trade in section III. Results are presented in sections IV to VI. We consecutively study the impact of ﬁnancial constraints on (i) export market participation (ii) entry-exit into the export market (iii) the intensive margin of trade. Finally, section VII contains robustness checks, and section VIII concludes.

II

Data

1

Database

We want to assess the role of ﬁnancial constraints on exporting behavior at the ﬁrm level. We therefore need information on both exports and ﬁnancial health at the ﬁrm level, with at least some time variation within ﬁrms, since we are interested in both participation and entry-exit behavior in the export market. We have constructed a large ﬁrm-level database from diﬀerent Investment Climate Surveys made by the WorldBank.1 Keeping only the countries for which the data is available for more than one year, we are left with nine countries and around 5,000 ﬁrms, each of which is present in the database for three consecutive years.2 In each country, industries were selected non-randomly in order to focus on the main producing sectors. Within each industry, ﬁrms were chosen randomly and their composition is therefore representative of the population. The time period diﬀers across the countries, but always lies between 1998 and 2004. The data includes accounting information on sales, inputs, labor, capital stock, investment and several other expenditures; broader information is also included, such as ownership structures, labor force characteristics, relations with competitors, clients and suppliers, innovation, market environment and investment climate. As the data is denominated in home currency, we have converted it into US dollars using yearly exchange rates from the International Financial Statistics. This is mainly done in order to provide some descriptive statistics (see below), however, our empirical estimations will systematically include country×year dummies controlling for exchange rate changes. Finally, we have restricted our sample to ﬁrms that present strictly positive sales and assets, and positive or null debt and interest payments. 1

Available at http://www.enterprisesurveys.org. The countries and time-periods are: Bangladesh (2000-2002), China (1998-2000, 2000-2002), India (1999-2001), Indonesia (2000-2002), Morocco (2000-2002), Philippines (2000-2002), South Africa (2000-2002), Thailand (2000-2002), Vietnam (2002-2004). 2

4

2

Variables of interest

Financial constraints are proxied in two diﬀerent ways. Previous works typically compute the correlation between investment and measures of internal (cash ﬂow) or external (debt) funds, after controlling for other factors to identify credit constraints. Signiﬁcant correlations are usually attributed to capital market imperfections and therefore suggest the presence of ﬁnancing constraints.3 In this paper, we ﬁrst use the ratio of total debt over total assets which can be interpreted as both a measure of the ﬁrm’s lack of collateral and of the ﬁrm’s current demand for borrowing relative to its capacity to borrow. Secondly, the ratio of cash flow over total assets gives an indication of the volume of funds that can be mobilized on a very short-term basis by the ﬁrm. These two indicators have been widely used in the literature dealing with ﬁnancial constraints (see e.g. Harrison and McMillan, 2003). We take the inverse of the ﬁrst ratio, so that an increase of each measure represents a decrease in ﬁnancial constraints. Productivity is measured by the ratio of value added over the number of workers.4 Finally, based on previous ﬁrm-level studies in international trade5 , we also include the ﬁrm’s size (measured by the logarithm of assets) and nationality (binary variable, which takes the value of one if the company is foreign-owned6 , 0 otherwise) as control variables. Note also that the ﬁrm size can also be interpreted as an additional proxy for ﬁnancial constraints, but is also correlated with productivity, as we discussed more extensively in the following sections. Descriptive statistics are given in Table 1. Around 40% of the ﬁrms in our sample export during at least one year over the studied period. This share is relatively homogeneous across countries (generally between 30 and 50%). Note that this share is actually larger than the share of exporters generally observed by ﬁrm-level studies (see for instance Mayer and Ottaviano, 2008 for European countries). Thus, if there is a bias in our sample, it is toward larger ﬁrms, which should face less ﬁnancial constraints; signiﬁcant results on this sample should therefore strengthen the view of ﬁnancial constraints as an important determinant of ﬁrm-level exporting behavior. We have separated the sample into four categories: continuous exporters, starters, continuous nonexporters and stoppers. Consistent with the literature, exporter are twice as large as non-exporters, and display a much higher productivity. They also display a higher average ratio of cash ﬂow over total assets. Entrants are found to display an even higher ratio (0.44), while stoppers have the lowest ratio (0.3). These results contrast with those of Greenaway et al. (2007) who ﬁnd that starters actually 3

See for instance the survey by Hubbard (1998). The time dimension of our data prevents us from computing proper total factor productivity. However, as robustness checks, we have included the labor to capital ratio in our estimations. The results are unchanged. 5 See among others Bernard and Jensen (2004), Bernard and Wagner (1998), Greenaway et al. (2007). 6 We deﬁne a ﬁrm as foreign when foreign participation in its capital is at least 49 percent; it is otherwise deﬁned as domestic. 4

5

have a lower cash ﬂow than the other types of ﬁrms. The main reason for this diﬀerence may lie with the country coverage of our data: while Greenaway et al. (2007) use UK data, we focus on developing countries in which the ﬁnancial constraints are more stringent. It may be the case that the role of ﬁnancial health on exporting decisions is larger in these countries, which is indeed what our econometric analysis will suggest. [Table 1 about here] The cross-country dimension also allows us to study the relationship between ﬁnancial development (proxied by the ratio of private credit over GDP) and export propensity, i.e. the percentage of exporters in our database. As shown in Figure 1 (left), the correlation is slightly positive. This is also the case for the relationship between export propensity and the sector-year speciﬁc mean of the ratio of cash ﬂow over total assets (Figure 1, left): sectors in which ﬁrms are ﬁnancially healthier seem to have a larger proportion of exporters.7 [Figure 1 about here]

III

Empirical Methodology

We study the eﬀect of ﬁnancial constraints on both margins of trade, i.e. on the probability of being an exporter (the extensive margin) and on the size of export by ﬁrm (the intensive margin) separately.

1

Extensive Margin

As the exporting decision is a discrete variable by deﬁnition equal to 0 or 1, the probit model is appropriate. Hence, the exporting probability of a ﬁrm i operating in country c during year t is:

P rob(Xict > 0) =

1

if

αi Uict + βϕict + γΩict + ηckt + ict > 0

0

otherwise

(1)

where ϕict is the logarithm of the ﬁrm’s productivity, Ωict is our proxy for ﬁrm i’s access to external ﬁnance, and Uict is a vector of control variables including ﬁrms’ size and nationality. The coeﬃcients are estimated by maximum likelihood procedures. 7 In ﬁgure 1 (right) only the sectors for which information is available for more than 100 ﬁrms are shown. The inclusion of other sectors modiﬁes the representation since in this case the export propensity is equal to zero or one for some sectors.

6

We estimate speciﬁcation (1) with a full set of country×industry×year dummies to account for unobservable characteristics (ηckt ). These dummies capture in particular sector-country speciﬁc variations in deﬂators, as well as exchange rate changes, and therefore make a comparison between ﬁrms producing in diﬀerent countries possible. Due to the limited degrees of freedom, we cannot account for ﬁrm ﬁxed eﬀects, even in a logit speciﬁcation.8 A ﬁrm’s productivity, size and nationality are expected to have a positive impact on the exporting decision (β, αi > 0). We expect the signs of estimated coeﬃcients on our proxy for access to ﬁnance (either the ratio of total assets over total debt or cash ﬂow over total assets) to be positive, i.e. γ > 0. We expect a positive sign on the ﬁrst proxy, i.e. on the inverse of the ratio of total debt over total assets, since the more a ﬁrm is indebted, the less it is likely to be able to face the ﬁxed costs of entering the export market. The rationale for the expected positiveness of the second proxy - the ratio of cash ﬂow over total assets - is symmetric: the more short-term liquidity the ﬁrm has, the more likely it will be to be able to enter the export market.

Econometric Issues. Since our sample generally contains three years per ﬁrm (cf. Data section supra), we are left with an insuﬃcient time variance to include ﬁrms ﬁxed eﬀects in a logit model estimation. We decided therefore to estimate equation (1) with a probit model in levels with additional controls (country×industry×year dummies, ﬁrm-level control variables such as size and foreign status). Still, due to the limited degrees of freedom, we cannot perform any non-linear instrumental variables estimation for the exporting probability. We then test the endogeneity of the right-hand side variables. We regress each instrumented variable on its own lags and on other exogenous variables. Residuals of all these ﬁrst stage regressions are introduced afterwards in equations (1), and a Fisher test for joint signiﬁcance of all residuals is implemented to provide a diagnosis on endogeneity. Since the null hypothesis cannot be rejected in a number of cases, we decide to report a double set of estimates for each speciﬁcation. A pooled probit based on contemporaneous values of regressors is ﬁrst reported, completed by another one relying on ﬁrst lagged values of right-hand side variables. A second way to account for the endogeneity of our ﬁnancial proxies is to use a methodology similar to Rajan and Zingales (1998). We use the sectoral ﬁnancial dependence indicator of Rajan and Zingales (1998), and run the last set of estimations on two diﬀerent sub-samples, respectively consisting of sectors with analogs in the US9 that are more (above median) and less (below median) 8

See the next sub-section for more details. The sectoral data has been constructed by Rajan and Zinagales (1998) from US data. External ﬁnancial dependence is deﬁned by these authors as the fraction of capital expenditures not financed with cash flow from operations. The external dependence of US industries is considered as being an optimal one, given the high level of ﬁnancial development in the US and the low probability of ﬁrms to be ﬁnancially constrained. The level of each industry’s external ﬁnancial dependence in the US should therefore represent the actual demand of external ﬁnance by those industries, in each country. Consequently, we apply this measure of external dependence to all countries in our database. 9

7

ﬁnancially dependent. The intuition behind this is twofold. First, there is a priori no reason for the endogeneity bias between export participation and our ﬁnancial proxies to be diﬀerently distributed across sectors with diﬀerent levels of external dependence. Second, the previous results should be exacerbated in more ﬁnancially dependent sectors. Hence, achieving more signiﬁcant results on the ﬁnancially dependent sub-sample would suggest both that the endogeneity bias is weak and that the eﬀect of our ﬁnancial proxies is indeed due to ﬁnancial constraints. Importantly, our ﬁnancial proxies are not correlated with the sectoral degree of ﬁnancial external dependence. Finally, standard errors are clustered at the ﬁrm-level and robust to heteroskedasticity.

2

Intensive Margin

The impact of ﬁnancial factors on the intensive margin of trade is estimated by replacing the dependent variable in equation (1) by the value of export. The estimated relationship is a standard linear equation which can be written as follows:

log(Xict ) = θi Uict + ρϕict + δΩict + ηckt + μict

if

Xict > 0

(2)

This equation is estimated only on the sub-sample of exporting ﬁrms (i.e. if Xict > 0). Here again, we expect the signs of the coeﬃcients on a ﬁrm’s productivity, size and nationality to be positive (ρ, θi > 0). If the impact of ﬁnancial variables is more important on the extensive margin than on the intensive one because of the existence of ﬁxed costs of exports, then we expect our ﬁnancial proxies to have either no impact (δ = 0) or a lower impact (δ < γ) on the intensive margin. The diﬀerence in terms of signiﬁcance between γ and δ should give some information on the respective roles of ﬁxed and marginal costs in the relationship between ﬁnance and trade. More precisely, a larger γ will suggest the role of ﬁnance on export is mainly due to the existence of ﬁxed costs of exports. Alternatively, we use another deﬁnition of the intensive margin as a dependent variable, namely the share of exports over total sales. The impact of productivity is less clear in this case: in a two-country Melitz-type model, productivity has the same impact on domestic sales and exports, so that the eﬀect of this variable on the share of exports over total sales is expected to be insigniﬁcant. On the other hand, in a N-country Melitz-type model (such as Chaney, 2008), the most productive ﬁrms reach more destinations, so that outward orientation increases with productivity.

Econometric issues. The linear form of equation 2 allows the inclusion of ﬁxed eﬀects in spite 8

of the short time dimension, at least for standard OLS estimation. Indeed, the control for a possible endogeneity bias requires a two-stage least squares (2SLS) estimation to be performed. Two alternative sets of instruments are used. The ﬁrst one contains the ﬁrst and the second lags of current period regressors. The second one relies on the second lags of current period regressors, the age of the ﬁrm and the dependence to external ﬁnance at the sectoral level interacted with ﬁnancial development at the country level. The ﬁrst is correlated with ﬁnancial factors, size and productivity (see for example Cooley and Quadrini, 2001, for a theoretical approach to this correlation). The second is correlated with ﬁnancial constraints (Rajan and Zingales, 1998). As our sample contains three years per ﬁrm, the 2SLS are therefore performed over a single year. Once again, we correct for clustering at the ﬁrm-level. We check the validity of our instruments using two diﬀerent tests. Robust to heteroskedasticity and clustering, Hansen’s J statistics of overidentifying restrictions are unable to reject our set of instruments. We also report the F-stat form of the Kleibergen-Paap statistic, the heteroskedastic and clustering robust version of the Cragg-Donald statistic suggested by Stock and Yogo (2005) as a test for weak instruments. All statistics are well above the critical values, conﬁrming that our choice of instruments is appropriate. The next step is to perform the Durbin-Wu-Hausman test for exogeneity of regressors. The null hypothesis of exogeneity cannot be rejected in most speciﬁcations. We will then report both OLS and 2SLS estimates.

IV 1

Finance, Productivity and Participation in the Export Market Direct impact of financial constraints

As mentioned in the introduction, the main reason why ﬁnancial constraints may have an important eﬀect on exports is related to the existence of ﬁxed costs of exports. The so-called “New New Trade Theory”, pioneered by Melitz (2003), and followed by many others, assumes that ex-ante diﬀerences across ﬁrms in terms of productivity determines participation in the export market. Beyond a certain cutoﬀ of productivity, ﬁrms can proﬁtably export, since the value of their sales on the foreign markets exceeds the level of ﬁxed costs. Therefore, all ﬁrms with a productivity level above the cutoﬀ do export. This prediction is however contradicted by recent empirical research, which has shown that the selection of ﬁrms into the export market is imperfectly correlated to ﬁrms’ productivity: even when considering a very disaggregated sectoral decomposition, there are a lot a high productivity ﬁrms which do not export, and low productivity ﬁrms which do so. An explanation was already provided by the sunk costs / hysteresis literature (Baldwin, 1988, Baldwin and Krugman, 1989, Dixit, 1989): in a dynamic model with export hysteresis, export status and productivity are less correlated, 9

as negative shocks on productivity do not necessarily translates into ﬁrms’ exit. Another explanation of this imperfect correlation is the existence of ﬁnancial constraints. Standard international trade models - including model with export hysteresis - assume perfect ﬁnancial markets: if they do not have enough liquidity ex-ante, ﬁrms can always borrow enough funds to pay the ﬁxed costs and enter the export market if it is proﬁtable for them to do so. Under imperfect ﬁnancial markets, on the other hand, access to ﬁnance matters: only ﬁrms that own suﬃcient liquidity or access to external ﬁnance will be able to enter. In this case, both productivity and access to ﬁnance may play a role. As soon as these two elements are imperfectly correlated, productivity will be an imperfect predictor of the participation in export markets. Put diﬀerently, ﬁnancial constraints generate a disconnection between productivity and export market participation. This is the ﬁrst proposition that we will take to the data. First note that although positively correlated, the relationship between these two dimensions of ﬁrm heterogeneity is indeed only imperfect: as shown in Table 2, an important number of high productivity ﬁrms for instance, display a low ratio of cash ﬂow over total assets.10 Table 3 presents the results of the basic estimations, based on equation (1), i.e. the impact of ﬁrms’ ﬁnancial constraints and physical characteristics (size, productivity, and nationality) on their exporting probability. Columns (a) to (d) present the results without using our ﬁnancial proxies; columns (e) to (i) use the ﬁrst ﬁnancial proxy, i.e. the ratio of total assets over total debt, columns (j) to (o) contain the estimations using the second proxy, i.e. the ratio of cash ﬂow over total assets. Columns (h), (i), (n) and (o) divide the sample according to the sectoral level of external ﬁnancial dependence. The reported coeﬃcients are marginal eﬀects computed at means for continuous regressors. If signiﬁcant, the coeﬃcients on all ﬁnancial proxies are expected to be positive, reﬂecting the impact of better access to ﬁnance on exporting probability. [Table 3 about here] The estimated coeﬃcients of the traditional determinants of export decisions show the expected signs. Exporters are found to be larger and more productive than domestic producers. Foreign-owned ﬁrms export more than domestic ones. Coeﬃcients on our ﬁnancial proxies are positive and signiﬁcant. Finally, the use of the Rajan and Zingales (1998) sectoral decomposition strengthens our results: the coeﬃcient on our ﬁnancial proxies are signiﬁcantly higher in sectors characterized by a high degree of ﬁnancial dependence. A 10% increase in our ﬁnancial proxies is found to increase the exporting probability between 0.5 and 1%. 10

”High” (resp. ”Low”) means above the third quartile (resp. below the ﬁrst quartile) in terms of productivity, computed by sector-country-year.

10

The pseudo R-squared in these estimations does not increase dramatically when we include the ﬁnancial variables. Note however that the same observation can be made for productivity: neither productivity nor ﬁnancial variables increase the R-squared by more than one percent. This is not the case for the variable size (the logarithm of total assets) which increases it by more than 4 percent. One could actually interpret this variable as another proxy for ﬁnancial constraints. Interestingly, the results on this variable in the rest of the paper are generally similar to those found on our ﬁnancial proxies. One should however be cautious with this interpretation, as the size variable may also be correlated with some unobserved components of ﬁrms’ productivity - even if we deﬁne size as the logarithm of total assets, not sales. In the same way, we can see in columns (h), (i), (n) and (o) that the ﬁrm size coeﬃcient is signiﬁcantly high in more ﬁnancially dependent sectors.

2

Indirect impacts of financial constraints

As emphasized recently by Suwantaradon (2008), another consequence of the consideration of access to ﬁnance as a second source of heterogeneity is that productivity and access to ﬁnance interact with each other. Productivity will only matter for exporting decisions if access to ﬁnance is high enough, as access to ﬁnance will only matter for ﬁrms with high productivity. The impact of productivity on the probability to export should therefore increase with access to ﬁnance. When ﬁrms are more ﬁnancially constrained, variations in productivity will have a lower eﬀect on exporting decisions. Table 4 includes, in addition to the previous regressors, interacted terms between each of our ﬁnancial proxies and productivity. The coeﬃcient on the interaction term is thus expected to be positive, i.e. an increase in productivity should have a stronger impact on exporting probability when ﬁnancial constraints are low (for high values of our ﬁnancial proxies). [Table 4 about here] The results are in line with our intuition: coeﬃcients on the interaction term between ﬁnancial health and productivity are positive and signiﬁcant in most of our speciﬁcations. Endogeneity concerns lead us to prefer the lagged speciﬁcations in which the interaction term is always signiﬁcant (columns (d) and (j)). The impact of productivity on exporting decision is thus stronger in ﬁnancially healthy ﬁrms. Importantly, these results only hold in ﬁnancially dependent sectors (columns (e) and (k)). The interaction terms are insigniﬁcant in other sectors (columns (f) and (l)). [Table 5 about here] Another way to look at this disconnection between productivity and export status is to run our basic estimation on diﬀerent sub-samples, deﬁned according to the level our diﬀerent ﬁnancial proxies. 11

In table 5 we run separate estimations on ﬁrms that have a low (respectively high) access to ﬁnance, as deﬁned by the median of each of our ﬁnancial proxies.11 The eﬀect of productivity strongly diﬀers on both sub-samples: while ﬁrms’ size and foreign status positively and signiﬁcantly aﬀect exporting probability in all estimations, productivity is only found to improve the exporting probability in ﬁrms characterized by a suﬃcient access to ﬁnance in our preferred speciﬁcations (columns (c), (d), (g) and (h) using the lagged regressors). The impact of our ﬁnancial proxies is also interesting, as they are found to be positive and signiﬁcant only when they are above the median, whatever the speciﬁcation. This provides evidence of a non-linear eﬀect of ﬁnancial constraints on exporting behavior. As we will show later, this non-linearity may be especially important to understand the impact of ﬁnance on export dynamics.

3

On the impact of financial development

Hence, ﬁnancial constraints have a twofold eﬀect: they have an overall negative eﬀect on the extensive margin of trade (i.e. the number of exporters) and they aﬀect the selection of ﬁrms in terms of their productivity. In this context, a reduction in ﬁnancial constraints at the country level, e.g. through an increase in the level of ﬁnancial development, will also have a twofold eﬀect. First, it will allow more ﬁrms to export, increasing total trade (Chaney, 2005). Second and more importantly, countries characterized by a more developed ﬁnancial market should display patterns of international trade closer to the Melitz (2003) model: productivity should have a larger role on export market participation. By decreasing the average level of ﬁnancial constraints, a deepening of ﬁnancial markets ﬁrst beneﬁts those ﬁrms whose productivity is high. Table 6 includes the interacted terms between productivity and ﬁnancial development, proxied by the ratio of total private credit to GDP (columns (e) and (f))12 , and further separates the sample according to the level of ﬁnancial development (columns (g) and (h)). The ﬁnancial development variable alone does not appear, since it is captured by the country×year dummies. Following the literature on ﬁnance, growth and trade13 , we use a proxy for ﬁnancial development, showing the ratio of private credit to GDP.14 We expect the coeﬃcient on the interaction term to be positive, reﬂecting the fact that ﬁnancial development disproportionately increases the exporting probability of productive ﬁrms. [Table 6 about here] 11

Median are country-year speciﬁc. Other decompositions (e.g. sector-speciﬁc) do not modify the results. The data comes from Beck (2002). 13 See among others King and Levine (1993), Beck (2002). 14 To ensure that our results are not biased due to omitted variables correlated with a country’s ﬁnancial development, we have included interacted terms between our ﬁrm-level variables and other country-speciﬁc variables, including GDP growth, inﬂation, and proxies for institutions. The results were unchanged. 12

12

As expected, ﬁnancial development magniﬁes the impact of productivity on exporting probability. The coeﬃcient on the interaction term in column (e) is strongly signiﬁcant. More strikingly, the coeﬃcient on the productivity variable is only signiﬁcant for more ﬁnancially developed countries (columns (g) and (h)). This suggests that the positive impact of ﬁnancial development found by previous literature comes both from the increase in the number of exporters and the selection of ﬁrms going into the export market. Indeed, according to our results, the average productivity of exporters should increase following an increase in ﬁnancial development, thus positively inﬂuencing the intensive margin of trade at the aggregate level. Note that because our data is not exhaustive, we cannot test for the sign and size of this selection eﬀect on aggregate exports, but only for the interaction between ﬁnancial development and productivity at the ﬁrm level. There are however many reasons to believe that it is positive: for instance, by preventing high productivity ﬁrms from exporting, a low ﬁnancial development may decrease the level of competition and allow low productivity ﬁrms (with enough access to ﬁnance) to export - lowering the cutoﬀ of productivity from which ﬁrms export. An increase in ﬁnancial development may in turn increase this cutoﬀ, hence increasing the average productivity of exporters. We have also introduced interacted terms between our ﬁnancial proxies and ﬁnancial development (columns (a) to (f)). When signiﬁcant, these proxies are positive, meaning that a higher ﬁnancial development magniﬁes the eﬀect of ﬁrm-speciﬁc ﬁnancial health on its exporting decision. For a very low level of ﬁnancial development, our ﬁnancial proxies have very little impact on the export decision, since most ﬁrms are credit constrained whatever their level of collateral. The magniﬁcation of the impact of ﬁnancial health through ﬁnancial development may come from the fact that ﬁrms in poorly ﬁnancially developed countries may face important credit constraints uncorrelated with their ﬁnancial health. These results suggest that ﬁnancial development not only allows more ﬁrms to enter the export market but also has an impact on the exporters’ selection process: in ﬁnancially developed countries, ﬁrms are both ﬁnancially healthier and more productive. More ﬁrms export, and in larger quantities.

V

Finance and entry-exit into the export market

So far, we have implicitly only considered a static framework, and have not distinguished between ﬁxed and sunk costs of exports. There are however reasons to believe that sunk costs, i.e. the costs of entering the export market for the ﬁrst time, and ﬁxed continuing costs, i.e. the costs of maintaining an activity on the export market, are fundamentally diﬀerent, and that the ﬁrst set of costs is higher. Recent evidence in line with this view is provided by Das et al. (2007), who ﬁnd, using Colombian ﬁrm13

level data, that sunk costs of entry into the export market represent between 18.4 and 41.2 percent of the annual value of a ﬁrm’s exports. The ﬁxed costs associated with the continuation of the exporting activity are estimated to be considerably lower, i.e. around 1% of the value of exports. Finance should then matter more for ﬁrms entering the market for the ﬁrst time. This is especially likely to be the case, as evidence presented in Table 5 already suggested that the impact of ﬁnancial constraints may be non-linear: our ﬁnancial proxies only have an impact on exporting probability if they exceed a given threshold. If continuing costs are signiﬁcantly lower than entry costs, ﬁnancial constraints are therefore more likely to aﬀect only entry decisions. To account for this point, we further estimate speciﬁcation (1) on two diﬀerent sub-samples, characterized by the conditions Xi,t−1 > 0 and Xi,t−1 = 0 respectively. The results on the ﬁrst sub-sample give the role of our regressors on the probability of remaining an exporter (which is the opposite of the probability of exiting the export market), while the second sub-sample gives their impact on the ﬁrm’s entry decision into the export market. Results are presented in tables 7 and 8, in which we decompose the eﬀect on exporting probability between the entry probability (if Xt−1 = 0, columns (a) to (f)) and the probability of remaining an exporter (Xt−1 > 0, columns (g) to (l)). Tables 7 and 8 respectively contain the results using the ﬁrst and the second proxies for ﬁnancial constraints. Here again, as exogeneity is occasionally rejected, we present both the current values and the lagged speciﬁcations, the latter being thus preferred. Our ﬁnancial proxies are mainly signiﬁcant in explaining the entry probability, but do not greatly inﬂuence the probability of remaining an exporter once the ﬁrm has entered (columns (d) and (h) of each table). Interestingly, size is also more signiﬁcant in determining the entry decision than the probability of remaining an exporter. Finally, the decomposition of sectors according to the degree of ﬁnancial dependence leads to the same results: only sectors which are more dependent upon external ﬁnance are aﬀected by our ﬁnancial proxies, and only at the time of entry. [Tables 7 and 8 about here] Hence, these results support the existence of sunk costs of exports. Consistent with the early models of export hysteresis, our results suggest that once the ﬁrms become exporters, the potential maintenance costs that they have to pay in each subsequent period are dramatically lower, thus dampening the impact of ﬁnancial constraints on exporting probability. Another potential explanation of our ﬁndings is simply that ﬁnancial constraints matter less for continuing decisions, for instance because the ﬁrms is viewed as a successful exporter, and its activity less risky by ﬁnancial institutions. This evidence is also in line with Suwantaradon (2008), who ﬁnds that ﬁnancial frictions can have persistent eﬀects on ﬁrms’ dynamics. In her model, productive ﬁrms with very low starting net worth 14

will never accumulate enough to overcome credit constraints and, therefore, never export even if they are very productive. The eﬀect of ﬁnancial constraints is therefore concentrated at the time of entry.

VI

Finance and the Intensive Margin

In the last section we have assessed the eﬀect of ﬁnancial constraints on the extensive margin of trade, i.e. how they aﬀect the ﬁnancing of the ﬁxed cost of exports and entry decisions. However, access to ﬁnance may also inﬂuence the intensive margin, i.e. the size of exports, if prevents ﬁrms from paying their marginal costs of trade and production. The comparison between the eﬀect of ﬁnancial constraints on the extensive and intensive margins is of interest for several reasons. First because it should provide insights about the importance of ﬁxed costs of export relative to other (marginal) trade costs. Second, from a policy point of view, if ﬁnancial constraints mainly aﬀect the extensive margin, governments actively providing trade subsidies should do it in order to facilitate entry rather than ﬁnancing existing exports. Tables 9 and 1015 present the results on the intensive margin, deﬁned alternatively as the log of total exports by ﬁrm and the ratio of exports over total sales; columns (a) to (e) present the results using the ﬁrst ﬁnancial proxy, while columns (f) to (j) use the second one; in each table, columns (d) and (e) and (i) and (j) divide the sample between ﬁrms that have a low (respectively high) access to ﬁnance, as deﬁned by the median of each of our ﬁnancial proxies. We only consider here exporting ﬁrms, thus estimating equation (2) above under the condition Xt > 0.

[Table 9 and 10 about here]

The comparison of Tables 9 and 10 enlightens some diﬀerences according to the deﬁnition of the intensive margin. When the latter is deﬁned as the log of total exports (Table 9), ﬁrms’ size, productivity and foreign status are strongly signiﬁcant in explaining the size of exports. The evidence emerging from the coeﬃcients on our ﬁnancial proxies, however, is rather weak. Indeed, the ratio of total asset over total debt is never signiﬁcant. If the ratio of cash ﬂows over total assets emerges as signiﬁcant at the 5% level over the whole sample (columns (f) and (g)), but it sign is not stable, and it falls to 10% when the sample is split between higher and lower ﬁnancially dependent sectors. When the intensive margin is deﬁned as the ratio of exports over total sales (Table 10), ﬁrms’ foreign status still matters strongly, and productivity is generally signiﬁcant. This tends to support the N-country version of the Melitz model ((Chaney, 2008)) in which the ratio of exports over total 15

For clarity purposes, the results from the 2SLS estimates relying on the second set of instruments are not reported in table 10. Available upon request to the authors, they are almost identical to the ones built on the ﬁrst set of instruments.

15

sales increases with productivity. On the contrary, both the coeﬃcients on our ﬁnancial proxies and the one on ﬁrm’s size become clearly insigniﬁcant. These estimates clearly show that the impact of ﬁnancial factors on the intensive margin of trade is either very low or insigniﬁcant. More generally, results from tables 9 and 10 support the fact that ﬁnance has a small or null impact on the intensive margin. These outcomes are therefore more in line with the sunk costs hypothesis, according to which the role of ﬁnance on trade is mainly concentrated at the time of entry.

VII

General discussion and robustness checks

Put together, Tables 3 to 10 shed new light on the role of ﬁnance on international trade patterns. Our results suggest that better ﬁnancial health exerts a positive role on the probability of becoming an exporter - or in other words that ﬁnancially healthy ﬁrms are more able to meet the sunk entry costs of export. This result ﬁts well with the ones coming from the literature connecting ﬁnancial constraints to ﬁrm investment in developing countries (see e.g. Jaramillo and Weiss, 1996, Harrison and McMillan, 2003 or H´ericourt and Poncet, 2009). From that perspective, the sunk cost, which can be viewed as an investment in export capacity, is no exception. Once ﬁrms become exporters, ﬁnancial health does not help them to remain on the foreign market. Financial health has a negligible impact on the value of exports, meaning that almost all the eﬀect on total export comes from the increasing probability of entering the export market. Finally, both heterogeneities in terms of productivity and access to ﬁnance matter: when a ﬁrm faces important ﬁnancial constraints, productivity will not matter for its exporting decision. Finally, we have checked the robustness of our results to diﬀerent alternative speciﬁcations. Our results are unaﬀected by: (i) controlling for ﬁrms’ innovating behavior (number of new products and R&D investment) and technology (capital intensity); (ii) dropping each country separately (meaning that our results are not driven by one particular country in our sample); (iii) controlling for other country-speciﬁc characteristics that may be correlated with ﬁnancial development, including GDP, GDP per capita, inﬂation and quality of institutions (for the last set of results, Table 6); (iv) using Random Eﬀects probit estimations instead of pooled probit; (v) using the methodology of Ai and Norton (2003) to compute interacted eﬀects; (vi) controlling for selection on the intensive margin through the use of a Heckman model16 ; (vii) dropping the ﬁrms which are 100 percent exporters (that represent an important part of our sample); (viii) controlling for single versus multi-destination exporters; (ix) using another proxy for access to ﬁnance, i.e. the ratio of tangible over total assets. 16

This check (and some of the others) is available in the working paper version of the paper (Berman and H´ericourt, 2008). Evidence in favor of selection is rather weak.

16

VIII

Conclusions

Using a large cross-country ﬁrm-level database on nine developing and emerging economies, we have studied the impact of ﬁnancial factors on both the intensive and extensive margins of trade. Our results stress the important role of ﬁrms’ access to ﬁnance on their entry decision into the export market. However, better ﬁnancial health neither increases the probability of remaining an exporter once the ﬁrm has entered, nor the size of its exports. We also ﬁnd that productivity and access to ﬁnance interact positively, productivity only becoming a signiﬁcant determinant of the exporting decision above a given threshold of access to ﬁnance; in the presence of important credit constraints, productivity and exporting status are disconnected. Financial development reduces this disconnection, thus positively acting both on the number of exporters and on the selection process into the export market. In ﬁnancially developed countries, exporting ﬁrms are more productive, and thus export in larger quantities. Financial development may therefore positively aﬀect the extensive margin, and (indirectly, through better selection of ﬁrms) the intensive margin. This paper contributes to the recent literature which documents the signiﬁcant role played by sunk costs in aﬀecting trade level and dynamics. The role of ﬁnance on trade is mainly concentrated at the time of entry. This is in line with the existence of sunk costs which have to be paid to access the export market. Finally, while recent international models only emphasize ﬁrms’ productivity heterogeneity as a determinant of exporting status, the present work points out the importance of also considering heterogeneity in terms of access to external ﬁnance. Our results have other important implications. In Melitz (2003), trade liberalization leads to intra-sectoral ﬁrm reallocations in favor of the most productive ones, and in turn generate aggregate productivity gains. Here, if ﬁnancial constraints and productivity are not perfectly correlated, this positive eﬀect of trade liberalization may be altered, and may depend on the country’s level of credit constraints. However, the general equilibrium eﬀects of the introduction of ﬁnancial constraints within a Melitz-type model of international trade remain to be more carefully studied, and constitute surely an interesting area for future research. It is also the case for the eﬀect of ﬁnancial development: as discussed before, a low ﬁnancial development may decrease the level of competition and allow low productivity ﬁrms (with enough access to ﬁnance) to export - lowering the cutoﬀ of productivity from which ﬁrms export. In turn, an increase in ﬁnancial development may increase this cutoﬀ, hence increasing the average productivity of exporters.

17

References Ai, C. and Norton, E. C. (2003), “ Interaction terms in logit and probit models ”, Economic Letters, vol. 80 no 1: pp. 123–129. Baldwin, R. (1988), “ Hyteresis in Import Prices: The Beachhead Eﬀect ”, American Economic Review, vol. 78 no 4: pp. 773–85. Baldwin, R. and Krugman, P. (1989), “ Persistent Trade Eﬀects of Large Exchange Rate Shocks ”, Quarterly Journal of Economics, vol. 419: pp. 635–654. Beck, T. (2002), “ Financial Development and International trade: is there a link? ”, Journal of International Economics, vol. 57: pp. 107–131. Berman, N. and H´ ericourt, J. (2008), “ Financial Constraints and the Margins of Trade: Evidence from Cross-Country Firm-Level Data ”, Ces working paper 2008-50. Bernard, A. and Wagner, J. (1998), “ Export Entry and Exit by German Firms ”, NBER Working Paper W6538. Bernard, A. B. and Jensen, J. B. (2004), “ Why Some Firms Export? ”, The Review of Economics and Statistics, vol. 86 no 2: pp. 561–569. Berthou, A. (2006), “ Credit Constraints and Zero Trade Flows: the Role of Financial Development ”, Mimeo University of Paris I. Chaney, T. (2005), “ Liquidity Constrained Exporters ”, Mimeo, University of Chicago. Chaney, T. (2008), “ Distorted Gravity: The Intensive and Extensive Margins of International Trade ”, American Economic Review, vol. Forthcoming. Cooley, T. F. and Quadrini, V. (2001), “ Financial Markets and Firm Dynamics ”, American Economic Review, vol. 91 no 5: pp. 1286–1310. Das, S., Roberts, M. J. and Tybout, J. R. (2007), “ Market Entry Costs, Producer Heterogeneity, and Export Dynamics ”, Econometrica, vol. 75 no 3: pp. 837–873. Dixit, A. K. (1989), “ Entry and Exit Decisions under Uncertainty ”, Journal of Political Economy, vol. 97 no 3: pp. 620–38. Eaton, J., Kortum, S. and Kramarz, F. (2004), “ Dissecting Trade: Firms, Industries, and Export Destinations ”, American Economic Review Papers and Proceedings, vol. 94 no 2: pp. 150–154. 18

Forlani, E. (2008), “ Firms Credit Constraints and Export Propensity ”, Mimeo CORE/Universit´e Catholique de Louvain. Greenaway, D., Guariglia, A. and Kneller, R. (2007), “ Financial Factors and Exporting Decisions ”, Journal of International Economics, vol. 73 no 2: pp. 377–395. Harrison, A. and McMillan, M. (2003), “ Does direct foreign investment aﬀect domestic ﬁrm credit constraints? ”, Journal of International Economics, vol. 61 no 1: pp. 73–100. H´ ericourt, J. and Poncet, S. (2009), “ FDI and credit constraints: ﬁrm level evidence in China ”, Economic Systems, vol. 33 no 1: pp. 1–21. Hubbard, G. (1998), “ Capital Market Imperfections and Investment ”, Journal of Economic Literature, vol. 36 no 3: pp. 193–225. Jaramillo, S. F., F. and Weiss, A. (1996), “ Capital market imperfections before and after ﬁnancial liberalization: An Euler equation approach to panel data for Ecuadorian ﬁrms ”, Journal of Development Economics, vol. 51 no 2: pp. 367–386. King, R. G. and Levine, R. (1993), “ Finance and Growth: Schumpeter Might Be Right ”, The Quarterly Journal of Economics, vol. 108 no 3: pp. 717–37. Manova, K. (2007), “ Credit Constraints, Heterogeneous Firms and International Trade ”, Mimeo Harvard University. Mayer, T. and Ottaviano, G. (2008), “ The Happy Few: The Internationalisation of European Firms ”, Intereconomics: Review of European Economic Policy, vol. 43 no 3: pp. 135–148. Melitz, M. (2003), “ The Impact of Trade on Intra-Industry Reallocations and Aggregate Industry Productivity ”, Econometrica, vol. 71 no 6: pp. 1695–1725. Modigliani, F. and Miller, M. H. (1958), “ The cost of capital, corporation ﬁnance and the theory of investment ”, American Economic Review, vol. 48 no 3: pp. 261–97. ˆ ls, M. (2008), “ Exporters and credit constraints. A ﬁrm-level approach ”, Technical report Muu 200809-22, National Bank of Belgium. Rajan, R. G. and Zingales, L. (1998), “ Financial Dependence and Growth ”, American Economic Review, vol. 88 no 3: pp. 559–86.

19

Stock, J. H. and Yogo, M. (2005), “ Testing for Weak Instruments in Linear IV Regression ”, in Andrews, D. W. and Stock, J. H. (editors), Identification and Inference for Econometric Models: Essays in Honor of Thomas Rothenberg, Cambridge: Cambridge University Press, p. 80108. Suwantaradon, R. (2008), “ Financial Frictions and International Trade ”, Mimeo University of Minnesota.

Appendix

0

.1

Export Propensity (% of exporters) .2 .4 .6

Export Propensity (% of exporters) .2 .3 .4 .5

.8

IX

.2

.4 .6 .8 1 Financial Development (Private credit / GDP)

1.2

−.5

Fitted relationship

0 .5 Cash Flow / Total Assets (Country−sector average)

1

Fitted relationship

Figure 1: Financial Development (left), Financial Constraints (right) and Export Propensity

20

Table 1: Descriptive Statistics Variable

Obs.

Mean

S.D

Q1

Median Q3

15351 14087 8662 10853

392.6 3.99 0.38 0.04

1658.7 3 0.65 0.07

26 1.86 0.05 0

88 2.95 0.23 0.01

309 4.86 0.51 0.04

6098 5821 3595 4016 6672

579.7 4.65 0.41 0.04 60.19

1223.2 3.22 0.69 0.08 37.72

68 2.2 0.08 0 23

208 3.41 0.25 0.01 60

579 7.45 0.52 0.04 100

268 250 185 213 299

562.3 4.03 0.44 0.06 20.5

1509.1 2.42 0.62 0.09 28.36

55.5 2.44 0.09 0.01 2

150 3.36 0.27 0.02 10

471 4.82 0.55 0.06 26.2

8888 7922 4810 6538

260.7 3.52 0.36 0.04

1900.5 2.76 0.61 0.06

17 1.62 0.04 0

45 2.63 0.21 0.01

152 4.24 0.51 0.04

97 94 72 86

272.8 3.1 0.3 0.03

543.9 1.8 0.74 0.05

73 2.45 0.09 0

189 2.7 0.1 0.01

203 3.5 0.27 0.03

All observations Employees Labor Productivity Cash Flow / Total Asset Tangible Assets / Total Assets Continuous Exporters Employees Labor Productivity Cash Flow / Total Asset Tangible Assets / Total Assets Exports / Total Sales Starters Employees Labor Productivity Cash Flow / Total Asset Tangible Assets / Total Assets Exports / Total Sales Continuous Non-Exporters Employees Labor Productivity Cash Flow / Total Asset Tangible Assets / Total Assets Stoppers Employees Labor Productivity Cash Flow / Total Asset Tangible Assets / Total Assets

Source: Authors’ computations from the World Bank Enterprize Surveys.

21

Table 2: Productivity and Financial Constraints

Cash Flow / Total Assets

Productivity

High

Low

High Low

32% 17%

14% 37%

Percentages of observations of the sample when we keep only ”low” and ”high” types ﬁrms. ”Low” (resp. ”high”) means below (resp. above) the ﬁrst (resp. last) quartile of the variable, computed by sector-country-year. .

22

Dep. Var. Financial Dep.

Foreign Size Productivity Size(t-1) Productivity(t-1) Total Assets / Total Debt (Total Assets / Total Debt)(t-1) Cash Flow / Total Assets (Cash Flow / Total Assets)(t-1)

No. Obs. Estimation Fisher statistic p-value Pseudo R2

(b)

(c)

(d)

0.661a (0.067) 0.235a (0.013) 0.081a (0.019)

(e)

13354

0.80 0.67 0.349

9746

13.3 0.004 0.319

9746

P (X > 0) High

(i)

Low (j)

(k)

(g)

0.289a (0.034)

0.112a (0.008) 0.015 (0.011)

(h)

(f)

0.258a (0.026)

0.095a (0.008) 0.019 (0.012)

0.215a (0.037)

0.264a (0.026)

0.101a (0.005) 0.022a (0.008)

0.235a (0.025) 0.105a (0.006) 0.024a (0.009) 0.098a (0.005) 0.024a (0.008)

2897

0.316

4063

3.63 0.16 0.32

8567

9.35 0.025 0.304

8567

0.030c (0.018)

7335

0.288

0.070a (0.015)

7335 Probit

0.318

0.047a (0.011)

0.321

(m)

0.287a (0.032)

(n)

High

0.208a (0.036)

(o)

Low

P (X > 0)

(l)

0.245a (0.025)

0.041 (0.028)

0.247a (0.025)

0.108a (0.025)

3937

0.117a (0.009) 0.021c (0.012)

0.073a (0.019)

2898

0.105a (0.009) 0.006 (0.013)

6939

0.296

0.110a (0.006) 0.018b (0.009)

6939 Probit

0.309

0.105a (0.006) 0.026a (0.008)

0.306

0.087a (0.016)

0.240a (0.025) 0.099a (0.005) 0.035a (0.008)

Table 3: Financial constraints and exporting probability

(a)

0.267a (0.025) 0.089a (0.005) 0.034a (0.007)

P (X > 0)

0.321a (0.020)

0.276a (0.022) 0.087a (0.004)

0.263a (0.023) 0.079a (0.004) 0.034a (0.006)

13354 Probit

0.347

0.085a (0.006)

13354

0.344

0.085a (0.024)

0.306

High and low mean respectively above and below the median of the sample.

Note: Marginal eﬀects computed at means. All estimations include year×country×sector dummies. Robust standard errors into parentheses. Signiﬁcance levels: c 10%, 5%, a 1%. Intercept not reported. Froot (1989) correction for ﬁrm-level cluster correlation. Rajan and Zingales (1998) data for sectoral external ﬁnancial dependence. b

23

Dep. Var. Financial Dep.

Foreign Size Productivity Size(t-1) Productivity(t-1) Total Assets / Total Debt (TA/TD)*Prod. (Total Assets / Total Debt)(t-1) (TA/TD*Prod.)(t-1) Cash Flow / Total Assets (CF/TA)*Prod. (Cash Flow / Total Assets)(t-1) CF/TA*Prod.(t-1)

No. Obs. Estimation Fisher statistic p-value Pseudo R2

(a)

0.268a (0.027) 0.087a (0.005) 0.033a (0.007)

(b)

P (X > 0) High (f)

Low (g)

(h)

(d)

0.302a (0.035)

(e)

(c)

0.269a (0.029)

0.110a (0.008) 0.019c (0.011)

0.216a (0.041)

0.254a (0.027)

0.087a (0.008) 0.021c (0.012)

0.235a (0.025) 0.105a (0.006) 0.024a (0.009)

0.096a (0.006) 0.027a (0.008)

0.235a (0.025) 0.105a (0.006) 0.024a (0.009)

Table 4: Liquidity, Productivity, and Exporting Decisions

0.259a (0.025) 0.091a (0.005) 0.032a (0.007) 0.102a (0.006) 0.024a (0.008)

0.236a (0.026)

(i)

0.116a (0.006) 0.008 (0.010)

0.233a (0.025)

(j)

0.116a (0.009) -0.011 (0.015)

0.278a (0.035)

(k)

High

0.119a (0.009) 0.016 (0.013)

0.200a (0.035)

(l)

Low

P (X > 0)

0.110a (0.006) 0.021b (0.009)

7366 Probit

0.309

3109

0.029 (0.048) 0.029b (0.012)

0.29

4151

0.024 (0.056) 0.008 (0.012)

0.033a (0.009)

6580

0.301

0.013 (0.019) 0.005 (0.004)

8567

0.307

0.048a (0.012)

8567

2.15 0.71 0.32

-0.001 (0.034) 0.006 (0.005)

3653

9.35 0.025 0.32

0.004 (0.032) 0.014c (0.007)

2600

0.288

0.005 (0.022) 0.009b (0.004)

6392 Probit

0.326

0.087a (0.016)

7059

0.328

0.089a (0.032) -0.001 (0.007)

9025

0.322

0.077a (0.020)

9746

11.03 0.03 0.355

0.033 (0.038) 0.015c (0.009)

13.3 0.04 0.349

Zingales (1998) data for sectoral external ﬁnancial dependence. High and low mean respectively above and below the median of the sample.

Robust standard errors into parentheses. Signiﬁcance levels: c 10%, b 5%, a 1%. Intercept not reported. Froot (1989) correction for ﬁrm-level cluster correlation. Rajan and

Note: Prod.: Productivity; TA: Total Assets. TD: Total Debt; CF: Cash Flow. Marginal eﬀects computed at means. All estimations include year×country×sector dummies.

24

Table 5: Productivity and the impact of ﬁnance on exporting probability P (X > 0)

P (X > 0)

P (X > 0)

P (X > 0)

Total Assets/ TD

Total Assets/ TD

CF/Total Assets

CF/Total Assets

Dep. Var

Foreign Size Productivity TA / TD

Low

High

Low

High

Low

High

Low

High

(a)

(b)

(c)

(d)

(e)

(f)

(g)

(h)

0.229a (0.037) 0.088a (0.007) 0.026a (0.010) 0.032 (0.035)

0.304a (0.034) 0.091a (0.007) 0.036a (0.011) 0.047a (0.013)

0.239a (0.037)

0.300a (0.036)

0.214a (0.037) 0.113a (0.008) 0.019 (0.012)

0.246a (0.033) 0.111a (0.008) 0.019 (0.013)

0.259a (0.034)

0.225a (0.034)

-0.119 (0.077)

0.063a (0.021)

0.124a (0.009) -0.001 (0.012)

0.113a (0.008) 0.027b (0.012)

-0.254a (0.082)

0.062a (0.024)

CF / TA 0.095a (0.008) 0.015 (0.011) 0.077b (0.037)

Size(t-1) Productivity(t-1) (TA / TD)(t-1)

0.106a (0.008) 0.028b (0.011) 0.047a (0.015)

(CF/ TA)(t-1)

No. Obs. Pseudo R2 Estimation

4801 0.32

4694 0.377 Probit

3616 0.287

3558 0.357 Probit

4174 0.317

4199 0.324 Probit

3413 0.305

3397 0.318 Probit

Note: Note: Marginal eﬀects computed at means. Prod.: Productivity; TD: Total Debt; TA: Tangible Assets; CF: Cash Flow. low and high means respectively below and above the median of each ﬁnancial proxy. Marginal eﬀects computed at means. All estimations include year×country×sector dummies. Robust standard errors into parentheses. Signiﬁcance levels: c 10%, b 5%, a 1%. Intercept not reported. Froot (1989) correction for ﬁrm-level cluster correlation.

25

Table 6: Financial Development, ﬁnance and exports P (X > 0)

P (X > 0)

P (X > 0)

Fin. Dev.

Foreign Size

(TA/TD)*Fin.Dev.

(c)

(d)

(e)

(f)

(g)

(h)

0.259a (0.025) 0.090a (0.005)

0.258a (0.026)

0.211a (0.033) 0.121a (0.008)

0.230a (0.034)

0.260a (0.023) 0.080a (0.004)

0.264a (0.022)

0.320a (0.033)

0.221a (0.025)

0.085a (0.004)

0.105a (0.008)

0.079a (0.005)

0.013 (0.015) 0.015 (0.016)

-0.006 (0.012)

0.033a (0.008)

-0.177a (0.054) 0.195a (0.049)

(TA/TD)(t-1)*Fin.Dev.

0.101a (0.005)

0.122a (0.008)

-0.118b (0.058) 0.154a (0.053)

CF / TA

-0.023 (0.042) 0.060 (0.058)

(CF/TA)*Fin.Dev. (CF / TA)(t-1) (CF/TA)(t-1)*Fin.Dev. 0.031a (0.007)

-0.130b (0.054) 0.158b (0.065)

-0.002 (0.012)

0.001 (0.013) 0.038a (0.014)

Prod.*Fin.Dev. 0.022a (0.008)

Productivity(t-1)

0.003 (0.012)

Prod.(t-1)*Fin.Dev.

No. Obs. Estimation Pseudo R2

Low

(b)

(TA / TD)(t-1)

Productivity

High (a)

Size(t-1) TA / TD

P (X > 0)

9746

7335 Probit 0.351 0.322

5040 0.26

4441 Probit 0.26

13354

10037 Probit 0.348 0.325

3968

6896 Probit 0.358 0.295

Note: Prod.: Productivity; TA: Total Assets. TD: Total Debt; CF: Cash Flow. Fin. Dev.: Country’s ﬁnancial development (Private Credit / GDP). Marginal eﬀects computed at means. All estimations include year×country×sector dummies. Robust standard errors into parentheses. Signiﬁcance levels: c 10%, b 5%, a 1%. Intercept not reported. Froot (1989) correction for ﬁrm-level cluster correlation. High and low mean respectively above and below the median of the sample in terms of ﬁnancial development.

26

Dep. Var.

Financial Dep.

Foreign Size Productivity Size(t-1) Productivity(t-1) Total Assets / Total Debt (Total Assets / Total Debt)(t-1)

No. Obs. Estimation Fisher statistic p-value Pseudo R2

Table 7: Financial constraints and entry probability (1/2)

(e)

High

0.002 (0.012)

(f)

Low (g)

(h)

0.020a (0.007)

(i)

0.019a (0.007)

(j)

0.014 (0.009)

(k)

High

0.024b (0.010)

(l)

Low

P (X > 0)

(d)

0.062 (0.039)

P (X > 0)

(c)

0.021 (0.015)

0.022a (0.006) 0.002 (0.001) 0.007a (0.002)

Xt−1 > 0

(b)

0.023 (0.015)

0.022a (0.006) 0.002 (0.001) 0.007a (0.002)

0.003 (0.003) 0.009 (0.006)

Xt−1 = 0

(a)

0.018 (0.011) 0.007a (0.001) 0.003b (0.001)

0.005b (0.002) -0.000 (0.004)

0.006a (0.002) -0.005c (0.002)

0.005b (0.002) 0.003 (0.003)

0.017a (0.004) -0.000 (0.005)

0.004b (0.002) 0.004 (0.004)

0.008a (0.001) -0.001 (0.002)

0.000 (0.003)

0.007a (0.001) -0.001 (0.002)

0.020c (0.012) 0.006a (0.001) 0.003b (0.001)

0.004b (0.002)

0.004 (0.007)

-0.003 (0.004)

0.003 (0.004)

0.018a (0.006)

0.005 (0.004)

0.006b (0.002)

1212 Probit

512

1212

689

1787

2907 Probit

1787

2907

1659

4273

972

4273

0.133

0.146

0.13

0.087

2.12 0.55 0.14 0.222

1.69 0.43 0.14 0.216

0.147

8.07 0.04 0.25

0.246

4.61 0.10 0.247

5%, a 1%. Intercept not reported. Froot (1989) correction for ﬁrm-level cluster correlation. Rajan and Zingales (1998) data for sectoral external ﬁnancial dependence.

High and low mean respectively above and below the median of the sample.

b

Note: Marginal eﬀects computed at means. All estimations include year×country×sector dummies. Robust standard errors into parentheses. Signiﬁcance levels: c 10%,

27

Dep. Var.

Financial Dep.

Foreign Size Productivity Size(t-1) Productivity(t-1) Cash Flow / Total Assets (Cash Flow / Total Assets)(t-1)

No. Obs. Estimation Fisher statistic p-value Pseudo R2

(a)

0.010 (0.011) 0.008a (0.001) 0.003 (0.002)

(b)

Low (i)

0.012c (0.007)

(j)

0.005b (0.002) 0.002 (0.003)

0.016b (0.008)

(k)

High

-0.004 (0.007)

0.005 (0.004) 0.011c (0.007)

0.009 (0.012)

(l)

Low

P (X > 0)

High (f)

0.013c (0.007)

0.005b (0.002) 0.007c (0.003)

0.013b (0.005)

682

P (X > 0)

(e)

0.005 (0.013)

0.005b (0.002) 0.007b (0.003)

0.001 (0.005)

794

Xt−1 > 0

(d)

0.027 (0.032)

1476 Probit

0.014c (0.008)

(h)

(c)

0.011 (0.013)

0.005a (0.002) 0.000 (0.002)

1476

(g)

0.009 (0.012)

0.018a (0.004) -0.004 (0.006)

1654

0.017b (0.007) 0.005a (0.002) 0.003 (0.003) 0.007a (0.002) 0.000 (0.002)

-0.003 (0.009)

1654

0.019a (0.007) 0.004b (0.002) 0.006c (0.003)

0.006a (0.002) 0.003 (0.002)

0.039a (0.011)

1608

Xt−1 = 0

Table 8: Financial constraints and entry probability (2/2)

0.010 (0.011) 0.008a (0.001) 0.004b (0.002)

0.004 (0.004)

0.015a (0.005)

885

0.192

2585 Probit

0.15

2585

0.183

3558

0.183

3558

1.57 0.67 0.198 0.252

0.04 0.98 0.188 0.243

0.182

0.76 0.86 0.234

0.252

5.93 0.05 0.234

5%, a 1%. Intercept not reported. Froot (1989) correction for ﬁrm-level cluster correlation. Rajan and Zingales (1998) data for sectoral external ﬁnancial dependence.

High and low mean respectively above and below the median of the sample.

b

Note: Marginal eﬀects computed at means. All estimations include year×country×sector dummies. Robust standard errors into parentheses. Signiﬁcance levels: c 10%,

28

Table 9: Financial variables and the intensive margin of trade: Export Value

Log(X)

Dep. Var. Financial Dep. (a)

Foreign Size Productivity TA / TD

(b)

0.557a (0.102) 0.139a 0.767a (0.037) (0.031) 0.746a 0.531a (0.081) (0.069) 0.039 0.121 (0.028) (0.095)

High

Low

(c)

(d)

(e)

0.388b (0.17) 0.654a (0.068) 0.616a (0.139) -0.082 (0.246)

359 2SLS No 2.04 0.36 35.7 9.53 3.6 0.31 0.68

CF / TA

No. Obs. 4154 Estimation OLS Firm F.E. Yes Hansen stat. p-value Kleibergen-Paap Stat. Critical value (5%) (1) Durbin-Wu-Hausman p-value 0.49 R2

1268 2SLS No 6.397 0.0938 101.3 12.2 5.2 0.157 0.91

Log(X) High

Low

(h)

(i)

(j)

0.714a (0.091) 0.086 0.137a 0.111b 0.738a (0.07) (0.039) (0.051) (0.03) 0.538a 0.931a 0.844a 0.397a (0.15) (0.052) (0.051) (0.058) 0.015 0.029 (0.065) (0.027) 0.142b 0.31b (0.059) (0.121)

0.373c (0.196) 0.648a (0.095) 0.671a (0.206)

0.114 (0.075) 0.788a (0.081)

0.113 (0.069) 0.903a (0.067)

1907 OLS Yes

2197 OLS Yes

3834 OLS Yes

0.30

0.67

0.58

269 2SLS No 0.75 0.69 41.47 9.53 1.78 0.62 0.70

(f)

(g)

1509 2SLS No 3.5 0.32 41.1 12.2 5.37 0.14 0.90

-0.496c 0.172c 0.107c (0.282) (0.098) (0.057)

1736 OLS Yes

2073 OLS Yes

0.53

0.64

Note: TA: Total Assets. TD: Total Debt; CF: Cash Flow. Robust standard errors into parentheses. All estimations include year×country×sector dummies. Signiﬁcance levels:

c

10%, b 5%,

a

1%. Intercept not reported. Froot (1989)

correction for ﬁrm-level cluster correlation. First and second order lagged values of regressors used as instruments in 2SLS speciﬁcations is columns (b) and (g); second order lags, age and ﬁnancial dependence×Financial Development in columns (c) and (h). Therefore, 2SLS are estimated over a single year and ﬁrms individual eﬀects cannot enter the estimation. (1) Critical values for the weak instruments test based on a 5% 2SLS bias at the 5% signiﬁcance level (see Stock and Yogo (2005)). Rajan and Zingales (1998) data for sectoral external ﬁnancial dependence. High and low mean respectively above and below the median of the sample.

29

Table 10: Financial variables and the intensive margin of trade: Export Share

Dep. Var.

Exports / Total Sales

Financial Dep.

Foreign Size Productivity TA / TD

High

Low

(d)

(e)

(a)

(b)

0.103a (0.014) -0.005 (0.003) 0.015a (0.005) 0.006 (0.006)

0.101a 0.138a 0.111a (0.014) (0.023) (0.02)

CF / TA

No. Obs. 4363 Estimation OLS Hansen stat. p-value Kleibergen-Paap Stat. Critical value (5%) (1) Durbin-Wu-Hausman p-value 0.54 R2

(c)

Exports / Total Sales

(f)

(g)

(h)

High

Low

(i)

(j)

0.092a 0.11a (0.019) (0.014) -0.005 (0.004) 0.012a 0.022c 0.006 0.017b 0.016a (0.004) (0.012) (0.006) (0.007) (0.005) 0.008 0.015 0.013c -0.003 (0.005) (0.024) (0.008) (0.01) -0.014c (0.008)

0.013a 0.001 0.007 0.019b (0.005) (0.009) (0.007) (0.008)

4363 OLS

0.54

1268 2SLS 2.52 0.28 194.3 11.04 3.6 0.17 0.38

0.11a 0.144a 0.124a 0.091a (0.014) (0.018) (0.021) (0.021)

-0.011 -0.016 -0.01 -0.021 (0.008) (0.015) (0.009) (0.013)

1907 OLS

2197 OLS

4020 OLS

4020 OLS

0.57

0.48

0.51

0.14 0.51

1509 2SLS 0.44 0.80 71.2 11.04 0.06 0.97 0.42

1736 OLS

2073 OLS

0.54

0.45

Note: TA: Total Assets. TD: Total Debt; CF: Cash Flow. Robust standard errors into parentheses. All estimations include year×country×sector dummies. Signiﬁcance levels:

c

10%, b 5%,

a

1%. Intercept not reported. Froot (1989)

correction for ﬁrm-level cluster correlation. First and second order lagged values of regressors used as instruments in 2SLS speciﬁcations is columns (b) and (g). (1) Critical values for the weak instruments test based on a 5% 2SLS bias at the 5% signiﬁcance level (see Stock and Yogo (2005)).

30

Price-cost Margins and Shares of Fixed Factors