Export Activity and Firms’ Financial Constraints

Abstract

This chapter assesses the role of firms’ financial resources used to begin export activity. We propose a new methodology to identify a priori constrained firms, exploiting a rich dataset on Italian firms’ assets and liabilities. We provide evidence that the entry probability in the export market is affected by the level of cash stock for the constrained firms: an increase of 10% in the cash stock of constrained firms raises by an additional 0.17% the entry probability of rationed firms, compared to unconstrained ones. We also find that internal liquidity is positively correlated with the extensive margin of trade: an expansion in new destination market is associated to higher liquidity levels. In particular, liquidity is mainly used for investments in the development of new products for foreign markets.

1 Introduction

The sunk costs associated with the export activity are a fundamental characteristic of the current literature in international trade and industrial organization. Both empirical and theoretical evidences underline the role of fixed cost. Firms that overcome these costs become exporter. Therefore, it becomes crucial to understand if and how firms are able to face fixed costs associated to exports.

Investments’ structure contemplates a temporal discrepancy between present cost and expected future profits. In the case of exporting (sunk) costs are certain and immediately paid, while revenues are uncertain and postponed in the future. Imperfect capital markets (e.g., information asymmetries) may decrease the probability to start the export activity. Lenders and borrowers may not own the same information set. Thus, potential lenders are not able to evaluate the investments’ value, given the uncertainty about future profits, and firms cannot gather enough resources to overcome fixed costs. For example, Das et al. (2007) estimate, for a sample of Mexican firms, an average fixed investment of $400,000 for potential exporters.

This chapter aims at analyzing the role of internal financing on export activity for credit-constrained firms. In the first instance, financially constrained firms are those firms for which internal source of financing cost less than external sources. If it is the case, investments (as exports) are sensitive to the availability of internal resources. This does not imply that the “non-constrained” firms do not use internal funds to implement/increase investments: also “healthy” firms show a positive correlation between investments and internal financial resources (Kaplan and Zingales 1997).

The key point is to understand how much the internal (financial) resources are relevant for the export activity of credit-constrained firms (compared to unconstrained ones). Therefore, the chapter addresses also the question of which type of firms are more likely to face financial constraints.¹

Using a representative sample of Italian firms, we analyze if financially constrained firms increase their entry probability in the export market, once they own a larger amount of internal financial resources. Since that a credit-constrained firm finds less costly internal resources, we expect a positive effect of the firm’s cash flows on the process of internationalization for these firms.²

The chapter covers two important issues. First, it is necessary to define a methodology to identify a priori the extent of a firm’s credit constraints. Employing a detailed information on asset and liabilities, a firm’s credit status is defined as financial reliability in the long and in the short run. This approach consists in evaluating the riskiness of firm from the point of view of a potential lender (bank) using ratio indices. The methodology puts light on the mechanism behind credit constraints, and it allows to understand how the relationship between firms’ and banks may affect the investments’ choice for the former.³

In such a framework, it is possible to offer additional insights for the economic policy analysis. It would be feasible to evaluate the implementation of a more stringent credit requirements, if these requirements rely, among other things, on balance sheet indices.⁴

The present chapter can be ideally placed in the between of two streams of literature: the first one concerns the investments’ sensitivity to cash flows as measure of credit constrains, and the second one regards the relationship between exporting and credit constraints. In the former group, since Fazzari et al. (1988), there existed a large body of literature that analyzes the sensitivity of investments to internal resources.⁵ Similarly, the entry in the export market is considered as an investment, and consequently entry decision can be sensitive to the level of internal financing.

The second stream of research focuses on the relationship between export and financial health. Such stream may be classified into three subgroups of analysis. The first one analyzes how credit availability affects the export’s decisions (Campa and Shaver 2002; Chaney 2016; Manova 2013; Muùls 2015); the second describes whether the export activity eases credit constraints (Manole and Spatareanu 2009); the third observes how financial health changes before and after entry into the export market (Greenaway et al. 2007; Bellone et al. 2010; Wagner 2014 for literature review).

From a theoretical point of view, Chaney (2016) introduces liquidity constraints into a model of international trade with heterogeneous firms (Melitz 2003), so that liquidity becomes a second source of heterogeneity across firm.⁶ In an empirical framework, the role of credit constraints has been demonstrated crucial to explain some features of international markets. Manova (2013) shows that credit constraints determine both the zeros in bilateral trade flows, and the variations in the number of exported products as well as the number of destination markets. Berman and Héricourt (2010) find evidence that credit access is an important factor in determining the entry into the export market for firms in developing countries; however, they also show that exporting does not improve firms’ financial health ex post.

Despite the increasing literature, the main conclusion remains contrasting. Greenaway et al. (2007), using a dataset for British firms, find that new exporters do not show a larger pool of financial resources than domestic firms before the entry, but long-term exporters own more liquidity than domestic firms.⁷ Differently, Bellone et al. (2010), using French data, empirically show that new exporters have an ex ante financial advantage compared to domestic firms, but not an ex post effect.

Similarly to Bellone et al. (2010), in the present chapter we define an index of credit constraints using information on asset and liabilities; however, we use thresholds for balance sheet indices to define a clear-cut rule for a firm’s financial reliability. These thresholds are commonly defined as rule of thumb in business economics. As we illustrate in the next sections, we assess credit constraints analyzing the firms from the point of view of a potential lender (bank).

Two papers are close to the present chapter, in terms of both data and research questions. Firstly, Minetti and Zhou (2011) show that the probability of exporting and the level of foreign sales are lower for credit-constrained firms. They evaluate credit rationing using firms’ responses to survey questions about their credit status. Differently from them, we assess credit status exploiting the information in the balance sheet data rather than using survey question.

The second one is by Caggese and Cunat (2013), where they develop a dynamic industry model where financing frictions affect the entry decision in the home market as well as the riskiness of firms’ activity. Calibrating the model, they predict that financing friction reduce the likelihood of a given firm to become an exporter, but overall they have an ambiguous effect on the number of firms starting to export. In addition, they find that financing constraints distort selection in the export reducing the aggregate gains due to trade liberalization. Using a similar dataset to Minetti Zhou (2011), their empirical analysis confirms the calibration findings.

The analysis is composed of two parts. In the first one, we develop the methodology to construct an index that allows to identify a priori the firm’s financial status. We consider a firm’s financial reliability both in a long-term and in a short-term perspective. In the second part, we empirically show that the amount of internal resources affects the entry probability into the export market for those firms identified as highly credit constrained (or without long-term reliability).

From a methodological point of view, we suggest a different strategy for testing the hypothesis of liquidity constraints and export. We classify firms in four groups. The firm clustering can be viewed also as a credit score: depending on firm classification a firm’s financial score changes and consequently also its financial reliability. We directly estimate the impact of liquidity across group of firms. Indirectly, we are also able to understand the effect of more stringent criteria, if changes in criteria changes firms’ classification.⁸

Finally, we control for potential endogeneity in the clustering process (exogenous to the entry in the export market). As Minetti and Zhou (2011), we use the same instrument set, but we proceed in a more rigorous way; since that we estimate a nonlinear model (probit) we prefer to follow a two-stage residual inclusion approach (2SRI, Terza et al. (2008)) rather than a more standard two-stage predictor inclusion.

The chapter provides two main results. First, we find that the entry in the export market is affected by the level of internal liquidity: for the more constrained firms, or firms which are not reliable in the long run (from lenders perspectives), exporting is sensitive to cash flows availability. The entry probability for constrained firms raise, compared to unconstrained firms, as the level of liquidity increases. The value of marginal effects remains constant across the different specifications; when we correct for the endogeneity bias in the clustering process, the magnitude of marginal effect increases.

Second, we find that an expansion in additional markets is affected by internal liquidity. However, the effect is not sensitive to firm’s financial status. Using a different subsample of firms (only continuous exporters), we find that the entry in new markets is positively correlated with the internal level of liquidity, for every group of firms. Finally, the export activity in close market (EU15) does not depend on internal cash, while exporting in more distant market depend on it.

The results are robust to different thresholds used to identify credit-constrained firms, as well as to financial indices employed to evaluate the level of financial reliability. Independently from the definition of credit constraints we use, the main massage does not change.

The rest of chapter is structured as follows. In Sect. 7.2, we present the data, describing the relevant characteristics and descriptive statistics. In Sect. 7.3, we introduce the motivations for the methodology proposed, and the strategy for identifying the credit-constrained firms. In Sect. 7.4, we present the empirical specifications and we discuss the results. Finally, Sect. 7.5 deals with the endogeneity of clustering process, and Sect. 7.6 concludes.

2 Data

The main data source is the “Indagine sulle Imprese Manifatturiere,” a survey conducted by the Italian bank Capitalia.⁹ Each survey was collected every three years. In the present chapter, we are going to consider the eighth and the ninth wave of the survey, which cover respectively the period 1998–2000 and 2001–2003. Each wave collects data for manufacturing firms with more than 10 employees. A survey includes the universe of large firms, and a stratified sample of firms with less than 500 employees.¹⁰ Each survey includes of 4680 firms, and the surveys can be matched among them every two waves (as in our case eighth and ninth).

An important feature of the survey is that it represents quite well the heterogeneity in the Italian manufacturing sector. Moreover, it allows to focus our analysis on medium- and small-sized firms: the median firm in the sample has 25 employees. The survey investigates different firms’ activities such as trade, R&D, and financial activities. The data are relative to year 2000 (eighth wave) or 2003 (ninth wave). It means that it is possible to observe only two time periods, even if the survey covers a three-year period.¹¹

The second main data source is the balance sheet dataset associated to surveys. The balance sheet dataset is collected on yearly basis, and it provides information on firms’ item as fixed assets or revenues.¹² Most importantly, it collects detailed data on firms’ financial activities such as short- and long-term debts, assets, and equity.

Given that, survey data are collected every three years, there exists a problem of matching survey information with the balance sheet data (defined on yearly basis). A researcher cannot associate a survey data (export status) with the balance sheet data for a specific year. To deal with it, we calculate the average value of balance sheet items on a three-year basis (i.e., average for periods 1998–2000, and 2001–2003). Then, averaged data (from balance sheet) are merged with the corresponding survey.

Finally, the match between the eighth and the ninth wave allows us to follow 2263 firms. Table 7.1 reports the descriptive statistics for the matched observations (firms are classified according with a two-digit ATECO 2002 industrial classification), while Table 7.8 (Appendix) presents the description of data used in the analysis.¹³ Finally, we integrate our dataset with “Struttura funzionale e territoriale del sistema bancario italiano, 1936–1974” (SFT) from Bank of Italy, that includes our instrumental variables (Sect. 7.5).

Table 7.1 Averages by sectors

3 Methodology

Our main hypothesis is that the availability of financial resources affects the entry in the export market, through sunk costs.¹⁴ Fixed investment is paid at the begin of export activity, while profits are uncertain and realized in the future. In this framework, asymmetric information and capital market friction may create a wedge in the cost of financing between internal and external sources. Therefore, the entry probability (in the export market) can be sensitive to the level of internal liquidity for credit rationed firms, for whom external funds are relatively more expensive.

In order to analyze export sensitivity, we proceed similarly to Euler equation’s models testing the effect of credit constraints on investments’ level (Bond and Van Reenen 2005).¹⁵ In these class models, financially constrained firms pay higher prices for external source of financing (issue new equity, or debt).¹⁶ Therefore, internal liquidity affects the rate of inter-temporal substitution between investment today and investment tomorrow; the more constrained the firm is, the larger (and positive) is the impact of cash availability on the investment level.

For the empirical estimation, it is crucial to identify a priori firms’ credit status, because the relationship between liquidity and investment varies in function of firms’ characteristics. Therefore, we analyze the role of liquidity for exporting, by clustering firms according to their level of financial reliability.

The direct estimate of liquidity for the entry choice is biased. For example, if we estimate the impact of cash stock (CS) on the entry probability (Enter) for firm i as follows,

$$ \Pr {\left(\mathrm{Enter}|X, CS\right)}_i=\alpha {X}_i+\beta C{S}_i+{\epsilon}_i $$

(7.1)

where X _i is a set of control variables. We have no a priori on β coefficient. If constrained and unconstrained firms are not differentiated in the empirical model, the effect of internal liquidity can be biased. We may identify three different potential situations. First, a not-constrained firm enters into the export market even with a low level of liquidity, because the sources of external financing are not too costly. Second, a healthy firm can also self-finance its own export activity (Kaplan and Zingales 1997): in this case, we observe a positive correlation between liquidity and the entry probability. Finally, a credit-constrained firm must rely on internally generated resources: also, in this case, we expect that entry is sensitive (positively) to internal liquidity.

Therefore, it is crucial to identify a priori firms’ financial status to estimate β in Eq. 7.1 across different types of firms (class of financial status). For this reason, we cluster firms in four groups according to their level of financial status, and for each group we assess the role of internal liquidity in the internationalization’s process.¹⁷

In the existing literature, many indices have been used to assess the financial health of a firm, as liquidity ratio or leverage ratio (Greenaway et al. 2007). However, as Bellone et al. (2010) underline, these indices do not capture the differences between short-term and long-term financial stability. Conversely, we define credit status from long- and short-term perspectives. To do that, we exploit information in the balance sheet to assess the degree of credit constraints.

Similarly to external investors, using balance sheet data, we can assess a firm’ financial reliability calculating financial ratios. In business economics, such ratios are often employed to determine the “goodness” of an investment.¹⁸ More recently, financial ratios are used by banks (among other procedures) to assess the riskiness of granted loans; according to the principles imposed by Basel III agreement (Bank for International Settlements 2006), banks have to manage the risk of credit by using objective criteria.

This approach allows to define an exogenous clustering process (exogenous to investment choice); the financial reliability is assessed by criteria external to firm’s decision process.¹⁹ To simplify the clustering process, we consider two indices, for which conventional thresholds exist. The two ratios consider respectively a firm’s financial reliability in the long run and in the short run.²⁰

• The Equity Ratio (ER hereafter) is used to assess long-term financial reliability. It is defined as the ratio between the total amount of internal resources (equity plus profits and reserves) and the total amount of capital invested (total assets). ER measures the proportion of the total assets that are financed by internal funds: it evaluates to what extent a firm is self-financing its economic activities. A ratio lower than 0.33 suggests a situation of sub-optimality, because a firm has a low capacity to self-financing; at least one-third of firm’s assets have be covered by internal resources in order to reach a financial stable situation in the long run (Brealey and Myers 1999).

• The Quick Ratio (QR hereafter) assesses short-term financial reliability, and it is a rough indicator of cash’s availability; QR measures a company’s ability to meet its short-term obligations with its most liquid assets. It is defined as the ratio of instantaneous liquidity or cash assets (cash, bank, and current account) to short-term debts (interests, furniture, wages etc.). The optimal value is fixed as greater than 1: if QR meets this criterion, a firm owns enough resources to face the daily cost of production process. The ratio indicates a firm’s chances of paying off short-term debts without the need for additional external funds.

A firm’s financial health improves when the ratios increase. Nonetheless, we test if the indices are reliable indicators for a firm’s financial health. Therefore, we exploit information on credit rationing, provided by the survey data. Each survey (the eighth and the ninth survey) report firms’ response to the following questions.

• (a) “In 2000 (or 2003), would the firm have liked to obtain more credit at the market interest rate ?” In case of a positive answer, the following question is asked:

• (b) “In 2000 (or 2003), did the firm demand more credit than it actually obtained?”

According to question (a) and (b), we create two dummy variables, Des and Ask, respectively. Des is equal to 1 if a firm replies yes to question (a), otherwise 0; similarly Ask is equal to 1 if a firm replies yes to question (b), otherwise 0. We use such information to understand if ER and QR can approximate a firm’s credit constraints.²¹

We expect that for high values of ER and QR correspond a lower probability to answer yes to questions (a) and (b). We estimate

$$ {Y}_i={\alpha}_0+{\alpha}_1\delta \mathrm{Inde}{\mathrm{x}}_i+\gamma \overline{X_i}+{\epsilon}_i, $$

(7.2)

where Y represents the binary information Des and Ask. δIndex takes value of 1 if ER or QR criteria are meet, and $\overline{X_i}$ is a vector of control variables. We expect a negative sign for α ₁. We estimate Eq. 7.2 for firms that appear in both surveys (eighth and ninth).²² Table 7.2 reports the results for the Probit estimation of Eq. 7.2, where Des is the dependent variable (dummy).^{23 , 24}

Table 7.2 Credit request and financial indices

The coefficients suggest that the degree of self-reported credit status is statistically correlated with the two ratios. As expected, the coefficients’ sign for the two dummies is negative, so that a firm is less likely to self-report as credit constrained when a threshold is satisfied. The magnitude (of coefficients) does not change with the inclusion of control variables.

Results suggests that the ratios (and thresholds) are correlated with firms’ ability to raise funding. Using QR and ER thresholds, we cluster firms in four different groups, according to the concept of short-term and long-term financial reliability. In our framework, the most constrained firms do not satisfy the conditions for both short-term and long-term financial reliabilities, that is, both QR and ER thresholds are not satisfied, respectively.

Firms in cluster 0 are defined as the most constrained firms, because they report an ER lower than 0.33, and QR smaller than 1. Table 7.3 illustrates how clusters are constructed. Then, we define with Cluster, an indicator variable that takes value 0,1,2, or 3 according to firm’s financial reliability.

Table 7.3 Cluster definition

The cluster should identify (exogenously) whether a firm is constrained or not; it is likely that a firm in group 0 or 1 faces difficulties to finance investments with external resources, because not reliable in the long term.²⁵ We can also think to clusters in Table 7.3 as a financial score. The lower is the score, the lower is the financial reliability of a firm.²⁶

4 Empirical Specification

In this section, we describe the empirical model to test if financially constrained firms largely rely on internally generated cash to overcome sunk costs associated to exports.

Comparing the eighth and the ninth wave, we estimate a discrete choice model (probit) for continuous nonexporting firms and new exporters. We observe 644 firms in 12 different manufacturing sectors: among them 122 firms are reported as new exporter in 2003 (i.e., reported domestic in the eighth survey, and exporter in the ninth survey).²⁷ The empirical model follows the nonstructural approach of Roberts and Tybout (1997) or Bernard and Jensen (1999), namely

$$ {\displaystyle\begin{array}{l}{\mathrm{Entry}}_{i03}=\left\{\begin{array}{c}1\ if\ G\left({\alpha}_0C{S}_i+{\sum}_{c=0}^3{\alpha}_c{X}_c\ast C{S}_i+\boldsymbol{Z}{(n)}_i+\gamma +{\epsilon}_i\right)>0\\ {}0\kern0.5em \mathrm{otherwise}\end{array}\right.\end{array}} $$

(7.3)

where Entry_i03 is the firm i export status in the ninth survey. Variable Entry_i03 takes a value of 1 if a firm starts to export between the eighth and the ninth survey, otherwise it takes value of 0. X _c, with c=0,1,2,3 is a set of dummies that specify cluster membership; for example, if X ₀ = 1, a firm belongs to cluster 0. Our terms of interest are the coefficient of cash stock (α ₀) for log of cash stock Log(CS), and the interactions between liquidity and clusters (α _c).²⁸

The α’s coefficients capture the effect of liquidity on the entry probability, so that a positive sign indicates that the export probability rises when the level of internally generated cash increases. The interaction term is introduced to identity if cash stock has different effect depending on firms’ financial status.

Equation 7.3 also includes a vector of control variables (Z (n)), while ε is the i.i.d. error term. The control variables are retrieved from the Capitalia surveys, or from the associated balance sheet dataset. The former group includes information about the number of banks (Banks), R&D indicator (dummy variable), or product innovation/upgrading dummy (UpProd or NewProd). Balance sheet controls include capital intensity (KL), labor productivity (LabProd), and additional financial ratios as LiqRatio and LevRatio (see Greenaway et al. 2007). The balance sheet controls are defined as averages for the three-year period 2001–2003 (subscript 03). Vector γ includes sector and area dummies (North East, North West, Center, South and Islands). Finally, we cluster the standard error across regions, given that Italian economy is highly regionalized.²⁹

In Table 7.4, we directly report the marginal effects (average marginal effect) obtained by estimating Eq. 7.3. Coefficients can be interpreted as the elasticities of cash with respect to entry probability. Each column represents a different regression, and financial score are defined according to Table 7.3. The average level of cash stock has no effect on the entry probability; instead, the interaction of cash with the dummy X ₀ (and X ₁) has a positive and significant coefficient. In column (1), the effect of cash cancels out across different groups. In the other specifications (from Col.(2) to Col.(7)), an increase by 10% in the level of cash stock raises the entry probability by almost 0.2% for credit-constrained firms belonging to group 0 (i.e., firms without long- and short-term financial reliability). Similarly, an increase by 10% in the level of cash stock for firms in cluster 1 raise their entry probability by 0.1\%.

Table 7.4 Baseline results

The coefficient of Log(CS) is the average marginal effect for all the firms, while interacted terms report the extra gains for firms in groups 0, 1, and 2 compared to group 3. Then, a 10% increase in cash raises the entry probability for constrained firms (in Cluster 0) by an additional 0.2% compared to the entry probability of not-constrained firms.³⁰ The results are statistically more robust for firms in cluster 0 than in cluster 1. It suggests that long-term financial reliability plays a central role in the access to external credit. Finally, coefficients in Table 7.4 are constant across specifications maintaining the same magnitude and sign.

Estimation results suggest that credit access is an important factor to determine the first entry in the export market. If a firm is not reliable from a financial point of view (lack of long-term stability), it has to pay higher price for external financing, and consequently it has to increasingly rely on internal funds. In such a framework, a credit-rationed firm experiences difficulties to overcome sunk cost associated to trade and the entry probability raises with the level of internal liquidity.

4.1 Expansion to New Markets

We demonstrated in the previous section that the entry probability of credit-constrained firms is affected by internal liquidity. Now, we want to understand if trade activity of established exporters is affected by cash stock, and financial reliability too. Therefore, we exploit information about regions served by exporting firm.³¹

We perform three exercises, and in all of them we consider continuous exporters (firms that export in both surveys). We analyze the effect of liquidity on the decision to reach new foreign markets. Compared to previous exercises, sample has changed given that new exporters and domestic firms are excluded.³² In the first two exercises, we estimate a probit model (like Eq. 7.3).

• 1. We estimate the export status in each region in function of cash stock (and interacted values): in this case, the dependent variable is a dummy equal to 1 if a firm exports in a region in 2003; otherwise the dummy takes value of 0.

• 2. In the second exercise, we estimate if cash affects the entry probability in additional markets: here the dependent dummy variable takes value of 1 if a firm adds new regions among its destination markets in 2003 (compared to 2000); otherwise the dummy is equal to 0.

Table 7.5 presents estimations’ results for the first exercise (control variables are not reported for the sake of space). Each column represents an equation for each destination market.³³ Dependent variable takes value of 1 if a continuous exporter (in eighth and ninth surveys) is exporting in a given region in the period 2001–2003, otherwise 0.

Table 7.5 Expansion to new markets

Cash stock coefficient turns to be positive and significant for all destination markets, with the exclusion of EU15 (column 1), while the interacted terms are not statistically significant. Given sample composition, we are just providing correlations among exporting and liquidity, that is, exporters own (on average) a higher liquidity (Greenaway et al. 2007) for each market they serve. Alternatively, a higher in liquidity is associated to a higher probability to serve a foreign market (EU15 excluded).

In the second exercise, the binary-dependent variable describes if an exporter enters in new markets between 2000 and 2003. Also in this case, cash stock coefficient Log(CS) is positive and significant for all the specifications, while interacted term is not. Again, we observe a positive correlation between export activity and liquidity independently from firms’ credit status: an expansion in the extensive margin of trade is associated to higher internal liquidity. It is interesting to note that R&D activity plays an important role to expand regions of destinations rather than to start exporting. Both R&D dummy and new product dummy (NewProd) suggest a positive relationship between firms’ innovation and exporting (Van Beveren and Vandenbussche 2010). Therefore, the development of new products seems important to enter in different destination markets.³⁴

In the last exercise, we estimate the effect of financial variables on the number of new destination markets. We define the dependent variable as a discrete number of new regions served among established exporters (ΔDest_i03). Dependent variable takes value 1, 2, 3, or 4, depending on the number of new added markets.³⁵ Given the nature of the dependent variable (ordered and discrete) we are going to estimate an ordered logit model; compared to Eq. 7.3, the ordered logit model maintains the same vector of independent variables. This last exercise confirms the previous results. First, higher liquidity is associated to a larger number of new regions, independently from credit status; second, innovation activity facilitates the entry in more than one new market.³⁶

We can conclude that the availability of internal resources is particularly relevant for credit-constrained firms that aim to start export activity ex-novo. Internally generated cash are important to increase the extensive margin of export of established exports, but this effect does not vary in function of firms’ financial reliability. The key role of liquidity for new entrants suggests that credit-constrained firms must pay higher cost for external source of financing.³⁷

5 Endogenous Selection of Financial Score

Even if we assume that our clustering process is exogenous (it is exogenous because we are evaluating firms from the external point of view of an investor),³⁸ firms’ selection in groups may be endogenous to the entry in the export market. The endogeneity can be generated by two sources:

• 1. The first source is the omitted variable bias. Whether or not a firm is constrained is likely to be correlated with unobserved firm’s characteristics, even if we include control variables (i.e., from Eq. 7.3, X(i) is correlated with some unobserved characteristics).

• 2. The second type of problem is that credit constraint level and entry decision may be jointly determined; for example, a firm may worsen its financial situation (reduction in ER) because it is using external financing to start export activity. Firms in lower clusters self-select in the export market through anticipated investments. Therefore, financial ratios are endogenous to export status.³⁹

In order to deal with endogeneity, we use an instrumental variable approach. We are going to define an instrument that may explain firm’s ability to obtain financing (or to not be credit constrained), but uncorrelated with export status. Similarly to Minetti and Zhou (2011), we are going information reported in “Struttura funzionale e territoriale del sistema bancario italiano, 1936–1974” (SFT).⁴⁰

In the beginning of 1930s, the Italian regulatory authorities were concerned about financial and banking instability: they thought that an excess of competition has favored this instability. As a result, in 1936 the Comitato Interministeriale per il Credito e il Risparmio (CICR) enacted strict norms for the entry of banks into local credit markets. As a consequence, from 1938 each credit institution could only open branches in an area of competence (one or multiple provinces) determined on the basis of its presence in 1936. Banks were also required to shut down branches outside their area of competence. Guiso et al. (2004) demonstrated empirically that the1936 regulation had a profound impact on the local supply of banking services and credit (creation and location of new branches) and, hence, on firms’ ability to obtain credit.

In this report, SFT are reported several information on Italian banking system in 1936:

• 1. the number of savings by Italian provinces (SavBank);

• 2. the number of cooperative banks by Italian province (CooBank);

• 3. number of overall credit institute by region (NUTS 2) per 1000 inhabitants (RegBank);

• 4. the average number of banks per province by Italian regions (PrBan). We use this information as instrumental variables.

We exploit the variability in the types of banks across provinces in 1936 to predict current level of credit clustering (i.e., the firm’s probability to stay in one of the four clusters). While, territorial distribution of banks in 1936 is unlikely to affect firms’ export decision between 1998 and 2003, it is very likely that the share of different bank types affects credit availability for the Italian firms today.⁴¹

Given that the clustering process is a discrete (and not-ordinal) variable, we are going to estimate a multinomial probit in order to capture the sorting effect (assuming independence of irrelevant alternatives, I.I.A.). Therefore, both the first and the second stage are not linear models, and traditional (linear) instrumental variable approach may not seem adequate. As Terza et al. (2008), we address this issue using the two-stage residual inclusion (2SRI). The 2SRI estimator has the same first stage of a 2-Stage Least Square (2SLS), but in the second stage the endogenous variables are not replaced by their predicted values but by residuals from the first stage are included in addition to endogenous regressors.⁴² Following the 2SRI technique, the main equation in our empirical model is as follows:

$$ {\begin{array}{l}{\mathrm{Entry}}_{i03}=\left\{\begin{array}{c}1\ if\ G\left({\alpha}_0C{S}_i+{\sum}_{c=0}^3{\alpha}_c{X}_c\ast C{S}_i+\boldsymbol{Z}{(n)}_i+{\eta}_n\boldsymbol{Res}\ {\left({\mathbf{X}}_{\mathbf{c}}\right)}_i+\gamma +{\epsilon}_i\right)>0\\ {}0\kern0.5em \mathrm{otherwise}\end{array}\right.\end{array}} $$

(7.4)

where R es (X _c)_i is a vector of residual from multinomial first stage estimation. Given that, in our first stage, we estimate a multinomial probit, we obtain four vectors of residuals, one for each category. To calculate residuals’ vectors, we use the formula for generalized residual for discrete choice models (Vella 1993).

Table 7.6 reports first-stage estimations (we omit exogenous variables). We present the results for the instrumentation of Cluster (as in Table 7.3) considering group 3 as baseline choice. In the first three columns, we use as instruments only credit data for Italian provinces in 1936 (as excluded instruments); in the last three columns we introduce the lagged values of LevRatio and LiqRatio as additional instruments (i.e., lagged averages for period 1998–2000). In this case, we also instrument LevRatio and LiqRatio in 2001–2003 with their lagged values (but we do not report first stage for these two additional variables). The coefficients show that instruments are correlated with endogenous sorting.⁴³ In particular, larger is the presence of saving banks (SavBank) in 1936, and the lower is the probability for a firm (in a given province) to be credit constrained (belonging to group 0)

Table 7.6 First stage (multinomial logit)

Given that, our instruments seem to have very high explanatory power, we include in the second-stage residuals, for alternatives 0, 1, and 2 Eq. 7.4. We estimate the model it with probit (again cluster 3 is omitted for multicollinearity). Finally, to retrieve robust standard errors, we bootstrap the entire two-stage procedure stratifying the sample by regions (Terza 2008; Wooldridge 2008). Table 7.7 presents the second-stage results (marginal effect reported).

Table 7.7 Entry in the export market (second stage)

The estimations confirm the previous intuitions. The coefficients’ sign does not change compared estimations from Table 7.4. The cash stock and interacted terms are jointly significant (Χ ² I° test). For all the specifications, an increase of liquidity raises the entry probability for constrained firms (group 0). More precisely, if cash stock raises by 10%, the entry probability of rationed firms increases by 0.11% (column 1).⁴⁴ Finally, the additional controls (both exogenous and endogenous) have a negligible impact on the entry probability.

Some final comments concern 2SRI approach. In large part of the specifications, the joint significance of the residuals (Res(x)) is rejected (Χ ² II° test): under the null, the coefficients are jointly equal to zero. It suggests that our clustering process is potentially exogenous to the entry decision.

We test if instruments have some explicative power on the main dependent variable (export decision). So, we include instruments from first stage in the second stage (Eq. 7.4). We report in Table 7.7 the p-value of overidentification test (LR test).⁴⁵ The LR test for overidentification suggests that instruments have not additional explanatory power in large part of regressions. Moreover, the test provide evidence that the instruments satisfy the exclusion restriction. This result reinforces also the idea that the sorting process is relatively exogenous.

As last exercise, we implement the 2SRI approach also to analyze expansions of export activity in new regions; we evaluate the effect of financial variables on the export status for a given region, on the binary decision of expanding in new markets. In both cases, we compare firms that report export activity in both surveys.

The results for the second stage show that the coefficients’ signs and statistical significance do not change, when we deal with endogeneity (results remain unchanged compared to Table 7.5). Similarly, to previous analysis, cash stock is positive correlated with exporting. Residuals from first stage are not jointly significant, and the LR Test suggests that instruments have no additional explicative power.⁴⁶

6 Conclusion

Exporting is an activity that entails several costs, and most of them are sunk costs associated with the first entry in the export. In real world, the new exporter faces a well-defined entry costs against an uncertain future profit. If we assume the existence of asymmetric information and imperfect capital markets, not all potential exporters begin export activity. Throughout the chapter, we discuss the impact of financial resources on the probability of entry into the export market, particularly for credit-constrained firms.

In the current chapter, we analyze two important issues. On the one hand, we develop a methodology for identifying a priori the level of a firm’s financial health, borrowing insights from the literature on investments’ sensitivity on cash flows, and using ratios from business economics. On the other hand, we empirically evaluate whether the level of internal resources affects both first entry in the export market and the extensive margin of trade.

We find that the internal resources are an important factor for firms’ internationalization. The level of cash stock is crucial for new entrants which are identified as credit constrained. Moreover, we find that internal liquidity is positively correlated with the extensive margin of trade: an expansion in new destination market is associated to higher liquidity. Findings are robust also to endogeneity concerns.

However, further work is needed to understand the mechanisms through which liquidity affects the internationalization process of medium- and small-sized firms, with a more detailed dataset about export and asset/liabilities.

Appendix

Table A.1 Data description

Table A.2 Descriptive statistics