Export Elasticities and Balance Sheet Effects: How Do Depreciations Affect Export Choices? Liliana Varela∗ April 2014

Abstract This paper shows that firm-level export elasticities to exchange rate can be significantly larger than previous estimated if firms are exposed to balance sheet effects. Exchange rate depreciations encourage exporters exposed to currency mismatch to further expand their sales abroad so as to rebalance their assets and liabilities in foreign currency. My empirical analysis uses panel data on firms’ sales and debt in foreign currency over the population of Hungarian firms. As a source of external time variation, I use depreciation of the Hungarian currency triggered by the Great Recession. I construct two indicators -one binary and one continuous- of currency mismatch at firm-level to measure the exposure of firms to exchange rate changes. Using the binary indicator, I find that the export elasticity to exchange rate is four times larger for exporters exposed to currency mismatch than for non-exposed exporters. My continuous measure of currency mismatch shows that firms’ export elasticities increase as a function of their exposure to currency mismatch. Export elasticity increases by ten-fold when moving from the lowest to the highest quartile of currency mismatch distribution. This expansion in exports is parallel to a similar upsurge of their export shares. Results suggest that the reallocation of sales towards foreign markets allowed exposed exporters to reduce their foreign currency denominated liabilities and rebalance their balance sheets. (Keywords: currency mismatch, export elasticity, exchange rate depreciations. JEL: F14, F31, F34)

∗

Paris School of Economics, [email protected].

1

1

Introduction

Trade elasticities to exchange rate have been found to be low at aggregate level. However, micro-level studies report that these elasticities are large at firm-level. Scholars associate this discrepancy to the large degree of heterogeneity found at micro-level, which can create heterogeneity bias in estimates based on aggregate data (Dekle, Jeong, and Ryoo 2010; and Imbs and Mejean 2009). In the last years, a new literature emerged trying to understand the sources of these heterogeneous responses at firm-level. Recent studies attribute part of this heterogeneity to firms’ individual markup elasticity and import intensity (Berman, Martin, and Mayer 2012 and Amiti, Itskhoki, and Konings 2013). One aspect overlooked so far is whether exchange rate shocks affect firm’s export behavior by altering their financing costs. Exchange rate shocks can deeply hit firms’ balance sheets if their liabilities are denominated in foreign currency. As reported by previous studies, firms indebted in foreign currency see larger increases in their financing costs following depreciations than those indebted in local currency. Moreover, the impact on firm’s balance sheet is greater, the larger is the mismatch between its income and debt in foreign currency -thereafter currency mismatch (Kim, Tesar, and Zhang 2012, and Ranciere, Tornell, and Vamvakidis 2010, among others).Does the export elasticity depend on how the exchange rate shock affects firms’ financial cost? If so, how? Do firms exposed to balance sheet effects expand their exports more so as to naturally hedge the increase in the debt burden? This paper investigates empirically these questions by using the currency depreciation in Hungary triggered by the Great Recession. I focus my analysis on Hungary for two reasons. First, in the years prior to the external shock, the Hungarian economy was growing at a steady pace and the exchange rate was stable against main trading currencies. The shock led to a significant exchange rate depreciation, increasing substantially firms’ financial costs (Endrész, Gyöngyösi, and Harasztosi 2012). Second, exporters’ balance sheets were heterogeneously exposed to exchange rate fluctuations. Some firms’ debt in foreign currency was fully hedged by their export incomes. Others, instead, enjoyed large sums of credits denominated in foreign currency, dramatically exceeding their sales abroad. This cross-sectional variation allows me to evaluate firms’ export responses as a function of the initial mismatch between their income and debt in foreign currency. I study the impact of exchange rate fluctuations on exporting firms exposed to currency mismatch using a difference-in-difference strategy. I exploit the depreciation of the Hungarian Forint following 2008 as an external source of time variation, and firms’ initial exposure to currency mismatch as a source of cross-sectional variation. I use a unique database covering

2

the population of firms and reporting information on credits by currency denomination and main outcomes. The database is provided by the National Bank of Hungary (NBH) and combines data from two sources: balance sheet data reported to tax authorities -APEH database-, and all corporate loan contracts with financial institutions in Hungary provided by the credit register dataset -CR. The NBH has combined these two datasets for the period 2005-2010. I use the information on firm’s income and debt in foreign currency to measure of currency mismatch at firm-level. I construct two indicators of currency mismatch: a binary indicator that denotes whether a firm is exposed to currency mismatch; and a continuous indicator that reflects the extent of currency mismatch for each firm. The advantage of this database is twofold. First, precise information about firms’ amount of loans by currency denomination allows to proxy the exposure to currency mismatch for each firm. Second, its extensive coverage allows studying the impact of exchange rate fluctuations on the entire population avoiding sample selection concerns. To my knowledge, this is the first paper analyzing the effect of currency mismatch on firms’ export decisions using such detailed and extensive database. Data reveals that mismatches between income and debt in foreign currency were not uniformly distributed across exporters prior to 2008. In fact, exporters exposed to currency mismatch were larger in size, older and enjoyed greater markups. These exporters tended to be domestically owned, and represented almost a quarter of exporters’ sales before the depreciation. Importantly, they were highly exposed to exchange rate changes: their debt in foreign currency was twice as large as their sales abroad. Estimated results point to a large expansion of exports following the depreciation, particularly for firms exposed to currency mismatch. The binary indicator of currency mismatch reports that the export elasticity of non-exposed firms is 0.7%. Notice that this elasticity is similar to previous firm-level estimates for developed economies. For example, Fitzgerald and Haller (2014) find an elasticity of 0.64% for Irish firms, and Dekle, Jeong, and Ryoo (2010) report an elasticity of 0.77% for Japanese firms. The similarity in these estimates with the Hungarian non-exposed firms is remarkable since both Irish and Japanese firms should neither be exposed to balance sheet effects. In both countries firms are principally indebted in local currency. My empirical results indicate that firm’s export elasticity increases four-fold for exporters exposed to currency mismatch. The continuous indicator of currency mismatch confirms these results. One standard deviation in currency mismatch exposure raises the export elasticity three-fold. Furthermore, export elasticity increases by ten-fold when moving from the lowest to the highest quartile of currency mismatch distribution. To test whether the timing of the

3

increase in firms’ exports really coincides with the exchange rate depreciation, I estimate its effect year by year. Estimated results show that the increase in exposed firms’ exports only starts in 2008, and deepens as the currency depreciates. Importantly, this expansion in exposed firms’ exports is not parallel to an increase in their total sales. Exposed and nonexposed firms’ exports have decreased their sales following 2008 -as is common in economic recessions-, but the drop is equal in magnitude and not statistically significant among these firms. The reallocation of sales towards foreign markets seems to have allowed exposed firms to rebalance their assets and liabilities in foreign currency. The reduction in exposed firms’ foreign currency debt points to that direction. Interesting, while exposed firms do not perform differently in terms of sales, employment, TFP and imports, they present larger decreases in their markups. The conjecture that arises from these results is that exposed firms might be decreasing their margins so as to expand their sales abroad and, with these revenues, decrease the mismatch between their income and debt in foreign currency. There is a long line of research addressing the effect of exchange rate fluctuations on trade. Recent studies document that export elasticities are significant at firm-level (Fitzgerald and Haller 2014; Dekle, Jeong, and Ryoo 2010; and Verhoogen 2008, among others). Berman, Martin, and Mayer (2012) advance this point and show that firms react heterogeneously to exchange rate variations. Using micro-level data on French exporters, they find that export elasticities decrease in firm size. In this paper, I show that firm-level elasticities can exceed those estimated in previous studies, if firms are exposed to mismatches between their income and debt in foreign currency. Following a currency depreciation, exposed firms have greater incentives to increase their foreign income so as to repay their foreign-denominated liabilities. Furthermore, the empirical evidence presented in this paper suggests that export elasticities might not necessarily decrease in firm size. If firms exposed to currency mismatch are larger in size -as in the case of Hungary-, the monotonic decreasing relationship is no longer ensured. This finding could partially explain why export elasticities appear to be so heterogenous across countries (Senhadji and Montenegro 1999). This paper is also related to the literature analyzing the implications of capital flows on economic growth. In most developing and emerging economies, agents do not rise funds in their own currency, but in foreign currency. This foreign borrowing can lead to the mismatch between the currency denomination of assets and liabilities, becoming then a source of financial fragility. Hence, exchange rate depreciation can trigger deep economic downturns as in the case of the Mexico crisis in 1994 and the East Asian crisis in 1997-1998. In the last decades, many scholars have addressed the causes and consequences of currency mismatch,

4

both at macro and micro level (Eichengreen, Hausmann, and Panizza 2007; Calvo, Izquierdo, and Mejía 2008; Brown, Ongena, and Yesin 2009; and Ranciere, Tornell, and Vamvakidis 2010; among others). However, poor data availability on firms’ assets and liabilities in foreign currency limited the extent of previous studies. This paper contributes to this literature by using an extensive database covering the population of firms that allows constructing precise indicators on the degree of currency mismatch for each firm in the economy. This is an advance regarding previous micro-level studies that had to rely on less precise measures of currency mismatch, and were mostly based on samples of large and listed firms. The remainder of the paper is structured as follows. Section 2 outlines the evolution of the Hungarian economy prior to and following the external shock. Section 3 presents the database. Section 4 describes the indicators of currency mismatch and the identification strategy. Section 5 reports the empirical results and Section 6 concludes.

2

Exchange Rate Depreciation and Balance Sheet Effects in the Hungarian Export Market

The Hungarian economy was growing at a steady pace before the downturn triggered by the collapse of Lehman Brothers in 2008. On the external front, the economy faced highly favorable conditions: exports were growing at 15% per year, net capital inflows reached more than 10 billions US dollar per year and the exchange rate against main trading currencies was stable. The change in the external conditions in 2008 substantially hit the Hungarian economy. At the end of year, GDP and exports dramatically slowed down, and dropped 7% and 10% respectively in 2009. Net capital outflows turned the financial account into deficit and the Hungarian currency (HUF) significantly depreciated. By 2010, the HUF depreciation against the Euro had reached 10%, and more than 40% against the Swiss Franc (Figure 1). The depreciation of the Hungarian Forint would entail non-trivial consequences on the economy and, particularly, on the export market. Whilst it would improve firms’ competitiveness in foreign markets, it would also worsen balance sheets of firms indebted in foreign currency. In Hungary, the increase in the financial burden concerned a large part of the export market. By 2005, more than a quarter of exporters -accounting for half of total exportsreported to be indebted in foreign currency. Among them, 60% reported to have greater debt in foreign currency than their sales abroad. Importantly, their balance sheets were highly vulnerable to exchange rate changes: their debt in foreign currency was twice as large as their exports. What was the impact of the exchange rate depreciation in exports? Did it affect 5

differently exporters exposed to currency mismatch? Did it affect their export behavior or financial strategy? In section 5, I assess these questions empirically exploiting an extensive database on firms’ sales and debt in foreign currency.

3

Data

I analyze the effect of the exchange rate depreciation on firms’ exporting decisions using firmlevel census data on firms’ credit and balance sheets. The database is provided by National Bank of Hungary and combines two different datasets: APEH, which contains data on firms’ balance sheets reported to the National Tax and Customs Authority, and the Credit Register, which reports information on all firms’ loans with financial institutions in Hungary. The APEH database covers the population of firms in all economic activities that are subject capital taxation. It includes all economic activities except for public administration and defense, education and health. This database provides information on sales, value added, exports, employment, wages, materials and capital. I use these variables to construct firms’ RTFP and markups. The RTFP is computed using the Petrin and Levinsohn (2011) method following the Wooldridge’s (2009) correction to estimate the parameters of the production function. Following De Loecker and Warzynski (2012) among others, I estimate firm’s markup as a wedge between the firm’s labor share and the labor elasticity of production. To obtain real values, I use price indexes at four-digit NACE activities for materials, investment, value added and production. Firm size varies significantly in the database, spanning from singleemployee firms to corporations employing thousands of workers. Since smaller firms are more subject to measurement errors problems, I retain firms with five or more employees. Even so, the sample accounts for more than 98% of total exports in the economy and covers more than 12,000 firms per year. The Credit Register database reports information on all firms’ credit in the Hungarian financial system by currency denomination. It provides information on each firm’s loans in Hungarian forint, Euros, Swiss Francs and US dollars. In 2005, the first year of the database, credit in foreign currency represented 43% of total credit undertaken by exporting firms, from which 73% was denominated in Euros, 18% in Swiss francs, and 9% in US dollars.

6

4

Identification Strategy

The identification strategy of firms’ exporting behavior following the depreciation of the Hungarian currency is based on their heterogenous exposure to currency mismatch prior to 2008. To proxy for the firm’s financial exposure to changes in the exchange rate, I construct a measure of currency mismatch at firm-level that reflects the difference between the firm’s income and debt in foreign currency. My difference-in-difference strategy employs this measure as a source of cross-sectional variation and uses the exchange rate depreciation triggered by the collapse of Lehman Brothers as a source of time variation. Hence, I estimate the impact of the depreciation on firms’ export and financing decisions in accordance with their initial balance sheets exposure to exchange rate movements. I construct a currency mismatch measure at firm-level in two steps. First, I compute the firm’s solvency ratio in foreign currency (SRF X ), which is defined as the ratio between the firm’s export earnings and total debt in foreign currency, i.e. SRF X = Exports/ Debt in FX This ratio reflects whether firm’s cash flows in foreign currency are sufficient to meet its obligations in foreign currency. The lower this solvency ratio, the more difficult will be for the firm to pay its debt obligations in foreign currency. Next, I define the currency mismatch measure (CM) as follows   0 CM = 1 − SRF X =  (0, 1]

if SRF X ≥ 1 if SRF X ∈ [0, 1)

The currency mismatch measure takes the value of zero if the firm’s exports are at least as high as its debt in foreign currency, and lies between (0,1] if the firm’s debt exceeds its sales abroad. Therefore, a currency mismatch measure of zero indicates that the firm’s solvency is not affected by exchange rate movements. Inversely, when the measure exceeds zero, exchange rate movements affect the firm’s solvency. The higher the currency mismatch measure, the more exposed the firm is to exchange rate movements. This continuous measure of currency mismatch allows capturing the extent to which firms are exposed to currency mismatch. Notice that this measure represents an advantage regarding previous studies, which -due to the lack of data availability- could only rely on binary measures of currency mismatch. Table 1 reports the summary statistics for the currency mismatch measure in the period under analysis. In 2005, the mean currency mismatch for the 12,775 exporting firms was 0.06, but 7

reached 0.49 for firms exposed to currency mismatch. The sample contains a large degree of heterogeneity. The measure has a standard deviation of 0.22 and ranges from a minimum of 0 to a maximum value of 1. Interesting, suggesting that following the depreciation firms are making a higher effort to balance their income and debt in foreign currency, the average currency mismatch decreases. By 2010, its mean had dropped by more than 10%. I employ this measure to construct two indicators of firms’ exposure to currency mismatch prior to the depreciation: a binary indicator reflecting whether the firm was exposed to currency mismatch over the period 2005-07, and a continuous indicator reporting the average level of the firm’s currency mismatch before the depreciation. In section 5, I employ these two indicators to asses the effect of the depreciation on firms’ export and financing decisions. I also use the continuous indicator to create quartiles by currency mismatch and assess whether more exposed firms were differentially affected following the depreciation. As a falsification exercise and to check whether the timing of the effect really coincides with the exchange rate depreciation, I estimate the effect year-by-year. Table 2 reports the main characteristics of firms exposed and non-exposed to currency mismatch prior to the depreciation, where groups are defined using the binary indicator. Exporters exposed to currency mismatch were larger in terms of sales, employment and capital intensity. Interesting, whilst these firms did not seem to be more productive on average than non-exposed firms, they charged higher markups. Exposed firms were also more likely to be domestically owned, and their sales mainly oriented to the local market (the mean export share is 11.5%). Notice as well that these firms were older in age. The differences in means between exporters exposed and non-exposed are statistically significant at one percentage level. To account for these differences as well as other unobserved timeinvariant characteristics, I introduce firm-level fixed effects in all the reduced-form regressions. Firm-fixed effects capture firm and sector time-invariant unobserved characteristics, but they do not account for sectoral patterns that might evolve over time. For example, if sectors were differentially affected by the global economic recession or any other specific demand shock, the estimated coefficients could be subject to omitted variable bias. To address this concern, I introduce sector-year fixed-effects in all regressions. In this way, I compare the evolution of firms differently exposed to currency mismatch within the same two-digit NACE sectors in each year. That is, I compare the manufacturers of plastic products (NACE 22.2) with manufacturer of rubber products (NACE 22.1), but not to manufacturers of basic metals in each sample year that are included in other two-digit sector activity (NACE 24). Hence, if sectoral specific shocks do not have a differential effect across exporters exposed and nonexposed to currency mismatch, I manage to control for them with the sector-year fixed effect.

8

A main assumption of the difference-in-difference strategy is that prior to the depreciation, exposed and non-exposed exporters shared similar growth trends. Figure 2 presents the evolution of the average firm’s export, sales and export shares between 2005 and 2010. Values are normalized to their initial levels in 2005. Figure 2 illustrates that exposed and nonexposed exporters saw similar patterns of growth in sales and exports between 2005 and 2007. However, since the beginning of the depreciation of the Hungarian Forint in 2008, the mean firm exposed to currency mismatch substantially increased its exports. Remarkably, their increase in exports was parallel to a decrease in their total sales, suggesting a reallocation of their activities towards foreign markets. It is interesting that this evolution of exports of exposed firms is absent in non-exposed firms: both their exports and sales decreased following the global economic crisis. More formally, I analyze mean’s growth rates of exposed and nonexposed firms on their main outcomes prior to the depreciation. Table 3 shows that mean growth rates were not statistically significant in sales, employment, RTFP, export, export share, import share, and investment. Next I present the empirical strategy and later the estimation of the impact of the depreciation on the export and financing decisions of firms exposed and non-exposed to currency mismatch.

5

Empirics

This section describes the empirical model and presents the results. First, I present the results on sales, exports and export shares using the binary and the continuous indicators of currency mismatch (sections 5.2.1 and 5.2.2). Second, I report the estimation by quartiles of the currency mismatch distribution (section 5.2.3). Next, I present the falsification test (section 5.2.4) and the effect by the currency denomination of firms’ debt (section 5.2.5). Later, I report the results on the extensive margin (section 5.2.6). Finally, I analyze the evolution of firms’ investment, TFP and markups, and their financial strategy (section 5.3).

5.1

Empirical Model

I estimate the impact of the depreciation on firms’ exports and financing decisions using the following model, Log yit = β Dt ∗ CMi + µi + ξj ∗ φt + εit

(1)

where i, j, t indexes firm, two-digit NACE activities, and time, respectively. y is a vector of firm’s {sales, exports and export shares}. D is a dummy variable equal to one for currency 9

depreciation years (year ≥ 2008) and zero otherwise; CM denotes the currency mismatch index, which is either the binary or the continuous index; µ are firm-fixed effects that capture firm unobserved constant heterogeneity; and ξj ∗ φt are two-digit sector-year fixed effects that are meant to capture any year sectoral shock that affect firms within the same activity in the same way. The coefficient of interest is β that captures whether firms exposed to currency mismatch were differentially affected by the depreciation once any demand specific shock is taken into account. Employing yearly firm-level data can lead to serial correlation in the error terms and to understate the standard deviation of the estimated coefficients as noted by Bertrand, Duflo, and Mullainathan (2004). The two-digit sector-year fixed effects included in equation (1) take out part of this effect, as it considers variations within two-digit sector and year. However, these fixed effect do not average out time variations within two-digit sectors. To account for this possible source of serial correlation, I clustered the standard errors at four-digit sector-year level.

5.2 5.2.1

Results on Sales, Exports and Export Shares Binary Indicator of Currency Mismatch

Table 4 presents the estimated coefficients for log sales, log export and log export share of equation (1) using the binary indicator of currency mismatch. All specifications include firm-fixed effects. Regressors in column 1 include a dummy for the depreciation period and an interaction term with the binary indicator of currency mismatch. The dummy for the depreciation period then captures the effect of the depreciation on exporters non-exposed to currency mismatch, and the interaction term captures (if so) any differential effect on exporters exposed to currency mismatch. The estimated coefficient for the depreciation dummy implies that non-exposed firms saw a decrease of 8.7% (t=-7.25) of their sales between 2008 and 2010. It is interesting that exporters exposed to currency mismatch did not experience a further drop in their sales than their non-exposed counterparts: the estimated coefficient for the interaction term is very small and not statistically significant. In column 2, I replace the depreciation dummy with year-fixed effects. Results remain unaffected by the inclusion of these year dummies. As mentioned above, specific sectoral shocks could affect activities differently and with them the performance of exposed firms. To account for this possibility, I include sector-year fixed effects and report the results in column 3. The estimated coefficient for the interaction term of exposed firms is not statistically significant, confirming that these firms did not see a differential slowdown than their non-exposed competitors.

10

Column 4 reports the results for the total value of exports. As expected, the dummy for the depreciation period, capturing the average variation in exports of firms non-exposed to currency mismatch, reveals an increase of 6.9% (t=4.06) between 2008 and 2010. To understand the size of this expansion, recall that during these years the Hungarian Forint depreciated by 10% against the Euro -which is the main trading currency (more than twothird of total exports are treated in Euros). The firm-level export elasticity to the exchange rate is then 0.7%, which is similar in magnitude to the estimates reported by previous firmlevel studies.1 This increase in exports following an exchange rate shock is not at odd in the standard framework. A currency depreciation decreases the relative costs of exporters and improves their competitiveness in foreign markets. Therefore, if firms are not exposed to currency mismatch in their balance sheet, an exchange rate shock translates in an increase in their exports. Remarkably, exporters exposed to currency mismatch differentially increase their sales abroad. The estimated coefficient in column 4 implies that they expanded their exports 33.7% (t=10.21) more. Columns 5 and 6 show that this coefficient is stable and statistically significant at one percentage point even after controlling for year, and sectoryear fixed effects. The estimated coefficients in columns 1 and 4 suggest that the exchange rate shock led firms exposed to currency mismatch to differentially reallocate their sales towards foreign markets. This reallocation is confirmed in columns 7-9, which show that exposed firms increased their export share by 29% (t=9.06) more. The conjecture that arises from these results is that exposed firms made higher efforts to expand their incomes in foreign currency so as to rebalance the increase in their financial burden. In the next sections, I test this hypothesis by exploiting different sources of cross-sectional variations in terms of firms’ initial exposure to balance sheet effects.

5.2.2

Continuous Indicator of Currency Mismatch

This section employs the continuous indicator of currency mismatch to assess whether firms more financially exposed differentially changed their export behavior following the shock. Columns 1-3 of Table 5 confirm the previously observed decline in sales, similar for both exposed and non-exposed firms. The expansion in total exports is however greater when tak1

For example, Dekle, Jeong, and Ryoo (2010) find that the average Japanese exporter increase its sales

abroad by 7.7% following a 10% depreciation of the Japanese Yen. See also Roberts and Tybout (1997); Verhoogen (2008); and Fitzgerald and Haller (2014) for firm-level evidence on the behavior of exporters following currency depreciations.

11

ing into account firms’ heterogenous exposure to currency mismatch (Columns 4-6). Column 6 shows the estimated coefficient for exposed firms once including firm-level and sector-year fixed effects. The coefficient is 0.741 (t=11.58), implying that firms that are one standard deviation more exposed to currency mismatch increase by more than three-fold their sales abroad. That is, despite the downturn in their sales, exposed firms are dramatically reallocating their activities towards foreign markets. This reallocation becomes clear with the upsurge of their export shares. The estimated coefficient for the export share, after the inclusion of all controls, is 0.728 (t=11.56), implying more than a 300% increase. To see more clearly the implication of these results, consider the median firm exposed to balance sheet effects. Before the depreciation, the median firm exported only 5% of their total sales and had an indicator of currency mismatch of 0.45 (i.e. its exports represented 55% of its total debt in foreign currency). Following the shock, this firm raises three times its export share to 15%. In terms of financing, this reallocation of sales implies -other things equal- that by 2010 the firm would no longer be exposed to currency mismatch.

5.2.3

Effect by Quartile of Currency Mismatch

In this section, I assess whether exposed firms’ responses vary with the initial currency mismatch distribution. That is, I compute quartiles of firms’ initial exposure to currency mismatch, and estimate the impact on the export market for each quartile. Table 6 reports the main results. In line with the previous estimates, I find non statistically different responses of firms’ sales by quartiles (columns 1-3). However, firms’ reallocation of sales towards foreign markets increases with their exposure to currency mismatch. The estimated coefficient for the log exports for the quartile of firms more exposed to currency mismatch, fourth quartile, is 0.714 (t=9.27), and decreases to 0.37 (t=5.29) and 0.221 (t=3.56) for the third and second quartiles (column 6). Notice that firms in the first quartile do not see any further expansion regarding their non-exposed counterparts. In line with these results, the estimated coefficients for log export shares reveal a similar pattern across quartiles of the initial currency mismatch distribution (columns 7-9). In Figure 3, I present graphically the interaction term using quartiles by currency mismatch and the 10% confidence bands. The positive elasticity of export and export shares as a function of the initial level of currency mismatch are clear. The elasticity of exports and export shares increase by three-fold when passing from the second to the fourth quartile of the currency mismatch distribution. That is, firms for which their debt burden in foreign currency increased most, made greater efforts to raise their incomes from abroad. 12

5.2.4

Falsification Test: Effect by Year

So far I have pooled the estimated effect across all year before and after the currency depreciation. To test whether the estimates are capturing the currency depreciation and not something else, I check in this section that the timing coincides with the exchange rate depreciation. With this end, I interact the binary indicator of currency mismatch with year dummies. Results are presented in Table 7. Estimated coefficients for the interaction term on sales in column 1 show that exposed exporters are systematically larger in size than non-exposed exporters, which effect is captured by the year dummies. Column 2 reports the estimated coefficients for the effect on exports. While exposed firms export on average less than non-exposed firms, this relationship decreases after the depreciation. The F test on equality of coefficients confirms this result. The interaction terms for the years 2005 and 2007 are not statistically different, suggesting that exposed firms did not differentially increase their exports before the depreciation. However, since 2008, they systematically expand their exports relative to their non-exposed counterparts. Notice as well that the test on equality of coefficients for 2005 and the depreciation years reveals that the estimated coefficients are all statistically different. Column 3 confirms the relative higher increase in export shares of exposed firms since 2008. This differential expansion in exports and export shares of exposed firms can be clearly seen in Figure 4, where I plot the estimated coefficients and the 10% confidence bands of Table 7. Figure 4 shows that while exposed firms did not differentially expand their export sales between 2005 and 2007, they systematically increased their exports following the depreciation.

5.2.5

Effect by Currency Denomination of the Debt

Exchange rate fluctuations are not uniform across currencies. I have reported in section 2 that, following the collapse of Lehman Brothers, the Hungarian currency depreciated 10% against the Euro, and more than 40% against the Swiss Franc (Figure 1). The depreciation of the Hungarian Forint against these two currencies was critical for firms, as foreign currency loans were mainly contracted under these denominations. In 2005, 44% of foreign currency loans were contracted in Swiss Francs and 54% in Euros. In this section, I use the different degree of depreciation of the Hungarian Forint against the Euro and Swiss Franc to evaluate whether firms’ increase in exports varies as a function of the depreciation of the currency in which the debt was denominated. An interesting feature about Hungary is that, in mid-2000s, firms had undertaken credits

13

in one currency or another. That is, most of the firms were indebted either in Swiss Francs or in Euros. Only a small proportion of firms was indebted in both currencies, and less than 2% of firms was indebted in another currency (mainly US dollars).2 To assess whether the reorientation of sales toward foreign markets increases as a function of the upsurge in the debt burden, I construct a dummy variable indicating whether the firm’s debt was mainly contracted Swiss Franc (Swiss Franc), and one indicating whether it was mainly contracted in Euro (Euro). Then, I interact these dummies with the binary indicator of currency mismatch and the dummy for the depreciation period. Table 8 presents the main results. As expected, firms mainly indebted in Swiss Francs -that saw a larger upsurge of their debt burden- increased their sales abroad 7.4% more than firms whose debt contracts were in Euros. This expansion is parallel to a higher increase in their export shares, 10.9%. Notice that, similarly to previous results, these firms did not see any differential change in their total sales.

5.2.6

Extensive Margin

In this section, I evaluate whether the depreciation changed the extensive margin of exporting firms. That is, I assess whether it increases probability of exporting, and whether it affected the probability of exiting the export market. To test whether after 2008 firms exposed to currency mismatch have greater probability of becoming a new exporter, I construct a dummy variable, Entry, if the firm did not export in year t-1 and started exporting in year t. Next, I regress this dummy on the interaction term of the depreciation with the binary indicator of currency mismatch as in equation (1). Estimated coefficients in columns 1-3 of Table 9 show that firms increased their probability of becoming an exporter by 0.9% following the depreciation. Interesting, exposed firms increased the probability of joining the export market by 1.1% more. Then, I create a dummy, Exit, if the firm exported in year t-1 and did not export in year t. Columns 4-6 report the estimated coefficients for the probability of exiting the export market. In line with the increase in competitiveness induced by the depreciation, after 2008, firms decreased their probability of exiting exporting by 0.3%. The estimated coefficient for the interaction term of exposed firms is marginally higher, which suggests a further reduction of 0.7%. Notice however that the coefficient loses significance when including the sector-year fixed effects. 2

Notice that, for consistency reasons, I focus my analysis on firms that were indebted either in Euros or

Swiss Francs.

14

5.3

Firm-Level Covariates

Last sections reported that, responding to the exchange rate depreciation, firms exposed to currency mismatch differentially increased their exports. The expansion of their export shares suggests that their strategy was to reallocate sales towards foreign markets so as to increase their foreign exchange earnings. A natural question to ask is: what was the firms’ strategy to achieve the increase their exports? I next assess this question by analyzing firms’ main covariates. Finally, I evaluate whether the exchange rate depreciation affected their financing decisions.

5.3.1

Investment, TFP, and Markups

What was the strategy followed by firms exposed to currency mismatch to expand their sales abroad? Did they increase their investments or productivity? Did they reduce their prices? In this section, I assess these questions by analyzing firms’ responses in terms of investment, TFP, markups, employment and imports. Table 10 reports the results in terms of firms’ investments and TFP. Estimated coefficients suggest that the expansion in exports of exposed firms was not spurred by an increase in their investments. Inversely, their investments seem to have been particularly hit by the increase in their financial costs. After the inclusion of firm and sector-year fixed-effects in column 3, the estimated coefficient indicates that their investments dropped by 8% (t=2.03) more than those of exporters non-exposed to currency mismatch. It is also interesting to remark that exposed exporters did not see any expansion of their TFP. The estimated coefficients for TFP reported in columns 4-6 are not statistically significant, suggesting a similar evolution than that exporters non-exposed to currency mismatch.3 What did then exposed firms do to increase their sales abroad? Could them have reduced their prices to become more competitive in foreign markets? Unfortunately, the lack of information on firm’s prices does not allow me to directly test this hypothesis. However, I can compute firms’ markups and test whether exposed firms see a decrease in their markups after the exchange rate depreciation. If exposed firms are reducing their prices to become 3

It is common that investment observations are common zero and, thus, excluded from the analysis in

logs. As a robustness test, I consider that all firms have at least one unit of investment and reestimate the regressions with this new investment variable. Columns 1-3 in Table 11 reports the results that show a similar decrease in exposed firms. As a robustness test of TFP, I compute labor productivity. Results presented in columns 4-6 in Table 11 illustrate that exposed exporters did not see any differential change in their labor productivity.

15

more competitive, their margins per unit of production should also decline. With this end, I estimate firms’ markups following De Loecker and Warzynski (2012) as a wedge between the firm’s labor share and labor elasticity. Results presented in Table 12 show that non-exposed firms see a decrease in their markups of 5.6% (t=4) following the currency depreciation (column 1). Interesting, exporters exposed to currency mismatch present a higher reduction in their markups. These firms differentially decreased their markups by 1.3% (t=2.17) more. Notice that this greater reduction in their markups remains significant even after inclusion of year and sector-year fixed-effects (columns 2 and 3). Next, I assess the evolution of employment for exposed and non-exposed firms. Columns 1-3 of Table 13 present the estimated coefficients, which suggest that in terms of employment exposed and non-exposed exporters evolved similarly. Finally, to test whether the depreciation of the exchange rate affected exposed firms more in terms of imports, I evaluate whether these firms see a differential change in their import shares. Results presented in columns 4-6 of Table 13 illustrate however that the rise in the cost of imports did not affect them differentially. Throughout the paper, I have presented evidence that the exchange rate depreciation in Hungary led exporters to increase their sales abroad. Importantly, exporters facing currency mismatches between their incomes and debt expanded their exports the most. This increase in sales abroad was translated into a parallel upsurge of their export shares. In the next section, I assess whether this strategy was successful in terms of reducing their foreign currency indebtedness.

5.3.2

Financial Strategy

Following the exchange rate depreciation, exporters exposed to currency mismatch expanded their assets in foreign currency through their exports. In this section, I assess the evolution of their liabilities and their foreign-denominated debt. To evaluate the evolution of firms’ liabilities in foreign currency, I use the foreign currency debt-to-total debt ratio. Columns 1-3 in Table 14 present the main results. The estimated coefficient in column 1 suggests that exporters exposed to currency mismatch reduced their foreign-currency debt ratio by 11% (t=-3.43). Notice that the inclusion of sector-year fixedeffects in column 3 reduces the estimated coefficient to -4% (t=-1.71), which still remains significant at 10%. Is this decrease in their foreign-debt associated with a reduction in their leverage? I assess this question using the total debt-to-sales ratio. Remarkably, columns 4-6 in Table 14 do not point to any further reduction in their indebtedness ratio. Overall, these 16

results suggest that, following the depreciation, exposed firms did not cease to use external funds, but their shifted their liabilities towards debt denominated in local currency.

6

Conclusion

How do then exporters respond to exchange rate changes when the currency shock also affect their balance sheets? This paper shows that, by affecting firms’ financial costs, exchange rate depreciations can lead to higher export growth than previously understood. If firms have mismatch between their sales and debt in foreign currency, exchange rate depreciations provide further incentives to expand sales abroad. Hence, firm-level elasticities to exchange rate can be significantly higher than previous estimates. Using for the first time firm-level census data on firms’ sales and debt in foreign currency, I construct different indicators of currency mismatch at firm-level. Empirical results show that the export elasticity to exchange rate is four times larger for exporters exposed to currency mismatch than for non-exposed exporters. A continuous measure of currency mismatch shows that firms’ export elasticities increase as a function of their exposure to currency mismatch. The impact on exports is higher in the upper quantile of the currency mismatch distribution. Interesting, the further reduction of exposed firms’ markups suggests that their strategy might have been to reduce their margins and so to expand their sales abroad. Estimated results suggest that this reallocation of sales towards foreign markets have allowed exposed exporters to reduce their foreign-denominated debt ratio and rebalance their balance sheets.

References Amiti, M., O. Itskhoki, and J. Konings (2013): “Importers, Exporters and Exchange Rate Disconnect,” American Economic Review, forthcoming. Berman, N., P. Martin, and T. Mayer (2012): “How do Different Exporters React to Exchange Rate Changes?,” The Quarterly Journal of Economics, 127(1), 437–492. Bertrand, M., E. Duflo, and S. Mullainathan (2004): “How Much Should We Trust Differences-in-Differences Estimates?,” The Quarterly Journal of Economics, 119(1), 249– 275. 17

Brown, M., S. Ongena, and P. Yesin (2009): “Foreign Currency Borrowing by Small Firms,” CEPR Discussion Papers 7540, C.E.P.R. Discussion Papers. Calvo, G. A., A. Izquierdo, and L.-F. Mejía (2008): “Systemic Sudden Stops: The Relevance Of Balance-Sheet Effects And Financial Integration,” Working Paper 14026, National Bureau of Economic Research. Dekle, R., H. Jeong, and H. Ryoo (2010): “A Re-examination of the Exchange Rate Disconnect Puzzle: Evidence from Japanese Firm Level Data,” IEPR Working Papers 06.46. Eichengreen, B., R. Hausmann, and U. Panizza (2007): “Currency Mismatches, Debt Intolerance, and the Original Sin: Why They Are Not the Same and Why It Matters,” in Capital Controls and Capital Flows in Emerging Economies: Policies, Practices and Consequences, NBER Chapters, pp. 121–170. National Bureau of Economic Research, Inc. Endrész, M., G. Gyöngyösi, and P. Harasztosi (2012): “Currency mismatch and the sub-prime crisis: firm-level stylised facts from Hungary,” MNB Working Papers 2012/8, Magyar Nemzeti Bank (the central bank of Hungary). Fitzgerald, D., and S. Haller (2014): “Exporters and Shocks: Dissecting the International Elasticity Puzzle,” Working Paper 19968, National Bureau of Economic Research. Imbs, J., and I. Mejean (2009): “Elasticity Optimism,” CEPR Discussion Papers 7177, C.E.P.R. Discussion Papers. Kim, Y. J., L. Tesar, and J. Zhang (2012): “The Impact of Foreign Liabilities on Small Firms: Firm-Level Evidence from the Korean Crisis,” Working Paper 17756, National Bureau of Economic Research. Loecker, J. D., and F. Warzynski (2012): “Markups and Firm-Level Export Status,” American Economic Review, 102(6), 2437–71. Petrin, A., and J. Levinsohn (2011): “Measuring Aggregate Productivity Growth Using Plant-Level Data,” NBER Working Papers 11887, RAND Journal of Economics. Ranciere, R., A. Tornell, and A. Vamvakidis (2010): “Currency mismatch, systemic risk and growth in emerging Europe,” Economic Policy, 25, 597–658. Roberts, M. J., and J. R. Tybout (1997): “The Decision to Export in Colombia: An Empirical Model of Entry with Sunk Costs,” The American Economic Review, 87(4), pp. 545–564. 18

Senhadji, A. S., and C. E. Montenegro (1999): “Time Series Analysis of Export Demand Equations: A Cross-Country Analysis,” IMF Staff Papers, 46(3), 2. Verhoogen, E. A. (2008): “Trade, Quality Upgrading, and Wage Inequality in the Mexican Manufacturing Sector,” The Quarterly Journal of Economics, 123(2), 489–530.

19

Figures and Tables 280 270 260 250 220 200 180 160 HUF/Euro 2005 2006 2007 2008 2009 2010 HUF/Swiss Exchange Franc Rate Evolution Exchange Rate Evolution

HUF/Swiss Franc

250

160

260

180

270

200

220

280

HUF/Euro

2005

2006

2007

2008

2009

2010

2005

2006

2007

2008

2009

2010

Figure 1: Depreciation of the HUF Forint 1Exposed .99 Exports 1.04 1.03 1.02 1.01 2005 2006 2007 2008 2009 2010 Note: Sales Export Non-Exposed .01 DecemberMarket 2005=1

Export Market

1.03 1.02 1.01 1 .99

.99

1

1.01

1.02

1.03

1.04

Sales

1.04

Exports

2005

2006

Note: December 2005=1 Note: December 2005=1

2007

2008

2009

2010

Non-Exposed

2005

2006

2007

2008

2009

2010

Exposed

Figure 2: Evolution of the Average Firm’s Export Market

20

.2uartiles of 0 -.2 Exports .8 .6 .4 .2 1 2 3 Q 4 Quartiles Export Share Currency Mismatch

.6 .4 .2 0 -.2

-.2

0

.2

.4

.6

.8

Export Share

.8

Exports

1

2 3 Quartiles of Currency Mismatch

4

1

2 3 Quartiles of Currency Mismatch

4

Figure 3: Effect of the RER Depreciation on Exports by Quartiles of Currency Mismatch

-.4 -.6 -.8 -1 -1.2 -1.4 2005 2006 2007 2008 2009 2010 Year Export Share

-.6 -.8 -1 -1.2 -1.4

-1.4

-1.2

-1

-.8

-.6

-.4

Export Share

-.4

Export

2005

2006

2007 2008 Year

2009

2010

2005

2006

2007 2008 Year

2009

Figure 4: Falsification Test: Effect by Year

21

2010

Table 1: Currency Mismatch Measure Year

Mean

Standard

Minimum

Maximum

Deviation

Number of Firms

2005

0.0635

0.2202

0

1

12,775

2006

0.0640

0.2199

0

1

13,161

2007

0.0689

0.2278

0

1

14,515

2008

0.0769

0.2397

0

1

15,134

2009

0.0686

0.2273

0

1

15,196

2010

0.0563

0.2088

0

1

16,183

Source: APEH and CR

22

Table 2: Descriptive Statistics Exporters Non-Exposed to CM Log Sales

Log Employment

Log Labor Productivity

Log Capital Intensity

Log Markup

Log Age

Export Share

Import Share

Share of Foreign Firms

N. of Firms

Exposed to CM

Difference in Means

12.6420

13.0414

-0.3994***

(0.0105)

(0.0187)

(0.0250)

3.0867

3.1375

-0.0507***

(0.0067)

(0.0135)

(0.0159)

8.1571

8.1104

0.0467*

(0.0106)

(0.0199)

(0.0253)

7.6973

8.4169

-0.7195***

(0.0088)

(0.0157)

(0.0202)

0.05529

0.1221

-0.0669***

(0.0033)

(0.0070)

(0.0077)

1.877

2.0569

-0.1796***

(0.0032)

(0.0069)

(0.0085)

0.3453

0.1148

0.2305***

(0.0020)

(0.0026)

(0.0046)

0.4128

0.4224

-0.0096

(0.2038)

(0.1902)

(0.4627)

0.3315

0.1894

0.1421***

(0.0026)

(0.0048)

(0.0061)

10,614

1,965

12,579

Notes: *, **, *** significant at 10, 5, and 1 percent. Std. errors in parenthesis. The firm is exposed to CM if the measure of currency mismatch exceeds zero before the depreciation (binary measure=1). Source: APEH and CR

23

Table 3: Growth Rates Before the Depreciation Exporters

Sales

Exports

Non-Exposed to CM

Exposed to CM

Difference in Means

0.1206

0.1324

-0.0118

(0.0054)

(0.0121)

(0.0140)

0.2430

0.2634

-0.0204

(0.0130)

(0.0466)

(0.0369)

Export Share

0.1224

0.1311

-0.0085

(0.0119)

(0.0447)

(0.0340)

Employment

0.0756

0.0803

-0.0047

(0.0041)

(0.0094)

(0.0105)

-0.1364

-0.1387

0.0023

(0.0198)

(0.0455)

(0.0499)

0.0367

0.0138

0.0230

(0.0064)

(0.0149)

(0.0160)

0.0093

0.0336

-0.0242

(0.0190)

(0.0430)

(0.0472)

10,614

1,965

12,579

Import Share

TFP

Investment

N. of Firms

Notes: *, **, *** significant at 10, 5, and 1 percent. Std. errors in parenthesis. The firm is exposed to CM if the measure of currency mismatch exceeds zero before the depreciation (binary measure=1). Source: APEH and CR

24

Table 4: Binary Indicator of CM: Export, Sales and Export Share

Log Sales

D*CM

D

Firm FE

Log Export

(1)

(2)

(3)

(4)

(5)

-0.009

-0.007

0.005

0.337***

(0.014)

(0.013)

(0.014)

(0.033)

Log Export Share (6)

(7)

(8)

(9)

0.342***

0.295***

(0.034)

(0.033)

0.346***

0.349***

0.290***

(0.034)

(0.033)

(0.032)

Yes

Yes

-0.087***

0.069***

0.156***

(0.012)

(0.017)

(0.013)

Yes

Year FE

Yes

Yes

Yes

Yes

Yes

Sector*Year FE

Yes

Yes

Yes Yes

Yes Yes

Yes

R2

0.965

0.967

0.967

0.889

0.890

0.890

0.854

0.855

0.856

N

82,271

82,271

82,271

82,271

82,271

82,271

82,271

82,271

82,271

Notes: *, **, *** significant at 10, 5, and 1 percent. Std errors are clustered at year and 4-digit NACE industries. Source: APEH and CR

Table 5: Continuous Indicator of CM: Export, Sales and Export Share

Log Sales

D*CM

D Firm FE

Log Export

(1)

(2)

(3)

(4)

(5)

-0.015

-0.011

0.012

0.805***

(0.026)

(0.024)

(0.024)

(0.064)

Log Export Share (6)

(7)

(8)

(9)

0.819***

0.741***

(0.064)

(0.064)

0.819***

0.829***

0.728***

(0.066)

(0.065)

(0.063)

Yes

Yes

-0.089***

0.060***

0.150***

(0.012)

(0.017)

(0.013)

Yes

Year FE

Yes

Yes

Yes

Yes

Yes

Sector*Year FE

Yes

Yes

Yes Yes

Yes Yes

Yes

R2

0.965

0.966

0.966

0.890

0.891

0.891

0.855

0.856

0.857

N

82,271

82,271

82,271

82,271

82,271

82,271

82,271

82,271

82,271

Notes: *, **, *** significant at 10, 5, and 1 percent. Std errors are clustered at year and 4-digit NACE industries. Source: APEH and CR

25

Table 6: Effect by Quartiles of Currency Mismatch: Export, Sales and Export Share

Log Sales

D*Q4

D*Q3

D*Q2

D*Q1

D

Firm FE

Log Export

(1)

(2)

(3)

(4)

(5)

-0.013 (0.026)

-0.011

0.001

0.762***

(0.024)

(0.024)

(0.076)

-0.015 (0.025)

-0.012

0.001

0.413***

(0.024)

(0.023)

(0.070)

0.022

0.025

0.027

(0.019)

(0.019)

(0.020)

0.260*** (0.062)

-0.016

-0.019

-0.015

0.071

(0.018)

(0.018)

(0.019)

(0.054)

Log Export Share (6)

(7)

(8)

(9)

0.771***

0.714***

(0.077)

(0.077)

0.775***

0.781***

0.713***

(0.077)

(0.077)

(0.077)

0.425***

0.370***

(0.069)

(0.070)

0.428***

0.437***

0.368***

(0.069)

(0.068)

(0.066)

0.270***

0.221***

0.238***

0.245***

0.194***

(0.062)

(0.062)

(0.060)

(0.059)

(0.059)

0.066

0.026

0.088*

0.085*

0.040

(0.054)

(0.054)

(0.051)

(0.051)

(0.050)

Yes

Yes

-0.090***

0.067***

0.157***

(0.013)

(0.017)

(0.013)

Yes

Year FE

Yes

Yes

Yes

Yes

Yes

Sector*Year FE

Yes

Yes

Yes Yes

Yes Yes

Yes

R2

0.967

0.968

0.968

0.893

0.894

0.894

0.859

0.860

0.861

N

82,271

82,271

82,271

82,271

82,271

82,271

82,271

82,271

82,271

Notes: *, **, *** significant at 10, 5, and 1 percent. Std errors are clustered at year and 4-digit NACE industries. Source: APEH and CR

26

Table 7: Falsification Test: Effect by Year Log Sales

Log Exports

Log Export Share

(1)

(2)

(3)

0.093***

-1.114***

-1.207***

(0.023)

(0.061)

(0.062)

0.089***

-1.155***

-1.245***

(0.023)

(0.054)

(0.046)

2007*CM

0.064***

-1.121***

-1.183***

(0.019)

(0.055)

(0.051)

2008*CM

0.086***

-0.815***

-0.901***

(0.019)

(0.054)

(0.051)

0.098***

-0.653***

-0.751***

(0.023)

(0.061)

(0.060)

0.069**

-0.610***

-0.678***

(0.029)

(0.059)

(0.054)

2005*CM 2006*CM

2009*CM 2010*CM y2006 y2007

0.098***

0.171***

0.073***

(0.016)

(0.028)

(0.026)

0.179***

0.298***

0.118***

(0.015)

(0.026)

(0.024)

y2008

0.199***

0.289***

0.091***

(0.015)

(0.027)

(0.025)

y2009

0.022

0.122***

0.099***

(0.016)

(0.029)

(0.027)

y2010

0.051***

0.187***

0.136***

(0.019)

(0.031)

(0.027)

Firm-Level Controls

Yes

Yes

Yes

Sector FE

Yes

Yes

Yes

R2

0.756

0.418

0.230

N

82,271

82,271

82,271

F Tests on Equality of Coefficients F-stat 2005-07

0.99

0.01

0.08

pvalue

0.3191

0.9312

0.7713

F-stat 2007-08

0.73

16.44

15.75

pvalue

0.3944

0.0001

0.0001

F-stat 2007-09

1.28

32.89

29.40

pvalue

0.2578

0.0000

0.0000

F-stat 2007-10

0.03

40.63

47.01

pvalue

0.8728

0.0000

0.0000

Notes: *, **, *** significant at 10, 5, and 1 percent. Std errors are clustered at year and 4-digit NACE industries. Firm-level controls are employment, labor productivity and age in the initial year (2005). All regressions include four-digit sector fixed-effects and a constant. Source: APEH and CR

27

Table 8: Effect by Currency Denomination of the Debt: Export, Sales and Export Share

Log Sales

D*CM*Swiss Franc

D*CM*Euro

D*Swiss Franc

D*Euro

D Firm FE

(1)

(2)

Log Export (3)

(4)

(5)

Log Export Share (6)

(7)

(8)

(9)

-0.028

-0.024

-0.012

0.517***

0.532***

0.492***

0.544***

0.556***

0.504***

(0.021)

(0.020)

(0.019)

(0.056)

(0.057)

(0.056)

(0.056)

(0.056)

(0.054)

0.007

0.008

0.022

0.443***

0.453***

0.434***

0.436***

0.445***

0.413***

(0.023)

(0.022)

(0.022)

(0.052)

(0.051)

(0.052)

(0.046)

(0.046)

(0.046)

-0.001

-0.003

-0.001

-0.207***

-0.216***

-0.213***

-0.206***

-0.213***

-0.212***

(0.014)

(0.013)

(0.017)

(0.035)

(0.035)

(0.036)

(0.033)

(0.033)

(0.032)

0.018

0.016

0.010

-0.209***

-0.224***

-0.219***

-0.226***

-0.240***

-0.228***

(0.015)

(0.014)

(0.017)

(0.027)

(0.025)

(0.028)

(0.022)

(0.022)

(0.020)

Yes

Yes

-0.090***

0.124***

0.213***

(0.012)

(0.018)

(0.016)

Yes

Year FE

Yes

Yes

Yes

Yes

Sector*Year FE

Yes

Yes

Yes

Yes Yes

Yes Yes

Yes

R2

0.965

0.966

0.966

0.888

0.889

0.889

0.854

0.855

0.856

N

81,681

81,681

81,681

81,681

81,681

81,681

81,681

81,681

81,681

Notes: *, **, *** significant at 10, 5, and 1 percent. Std errors are clustered at year and 4-digit NACE industries. Source: APEH and CR

Table 9: Extensive Margin: Entry and Exit in the Export Market Entry

D*CM

D

Firm FE

Exit

(1)

(2)

(3)

(4)

0.011***

0.010***

0.008***

-0.007***

-0.006***

-0.003

(0.003)

(0.002)

(0.003)

(0.002)

(0.002)

(0.002)

Yes

Yes

0.009***

-0.003***

(0.001)

(0.001)

Yes

Year FE

Yes

Yes

Yes

Yes

Sector*Year FE

(5)

(6)

Yes Yes

Yes

R2

0.265

0.268

0.267

0.263

0.266

0.264

N

709,968

709,968

709,968

709,968

709,968

709,968

Notes: *, **, *** significant at 10, 5, and 1 percent. Std errors are clustered at year and 4-digit NACE industries. Source: APEH and CR

28

Table 10: Effect on Investment and TFP Log Investment

D*CM

D

Firm FE

(3)

Log TFP

(1)

(2)

(4)

(5)

-0.119***

-0.123***

-0.079**

-0.024

-0.023

-0.016

(0.039)

(0.038)

(0.039)

(0.021)

(0.020)

(0.013)

Yes

Yes

-0.285***

-0.073

(0.025)

(0.057)

Yes

Year FE

Yes

Yes

Yes

Yes

Sector*Year FE

(6)

Yes Yes

Yes

R2

0.825

0.828

0.827

0.844

0.846

0.849

N

58,955

58,955

58,955

70,984

70,984

70,984

Notes: *, **, *** significant at 10, 5, and 1 percent. Std errors are clustered at year and 4-digit NACE industries. Source: APEH and CR

Table 11: Robustness Tests: Effect on Investment and Labor Productivity Log Investment

D*CM

D

Firm FE

(3)

Log Labor Productivity

(1)

(2)

(4)

-0.113***

-0.117***

-0.071*

-0.004

-0.003

0.003

(0.040)

(0.039)

(0.040)

(0.012)

(0.012)

(0.013)

Yes

Yes

-0.313***

-0.078***

(0.027)

(0.010)

Yes

Year FE

Yes

Yes

Yes

Yes

Sector*Year FE

(5)

(6)

Yes Yes

Yes

R2

0.832

0.835

0.834

0.832

0.835

0.837

N

60,073

60,073

60,073

72,475

72,475

72,475

Notes: *, **, *** significant at 10, 5, and 1 percent. Std errors are clustered at year and 4-digit NACE industries. Source: APEH and CR

29

Table 12: Effect on Firms’ Markups Log Markup

D*CM

D

(1)

(2)

(3)

-0.013*

-0.014*

-0.013*

(0.006)

(0.006)

(0.007)

Yes

Yes

-0.056** (0.014)

Firm FE

Yes

Year FE

Yes

Sector*Year FE

Yes

R2

0.774

0.775

0.778

N

57,266

57,266

57,266

Notes: *, **, *** significant at 10, 5, and 1 percent. Std errors are clustered at year and 4-digit NACE industries. Source: APEH and CR

Table 13: Effect on Employment and Import Share Log Employment

D*CM

D

Firm FE

Log Import Share

(1)

(2)

(3)

(4)

(5)

(6)

-0.002

-0.001

0.002

-0.012

-0.016

-0.037

(0.008)

(0.008)

(0.008)

(0.038)

(0.038)

(0.037)

Yes

Yes

0.015***

-0.236***

(0.005)

(0.017)

Yes

Year FE

Yes

Yes

Yes

Yes

Sector*Year FE

Yes Yes

Yes

R2

0.965

0.965

0.966

0.870

0.870

0.872

N

76,913

76,913

76,913

37,951

37,951

37,951

Notes: *, **, *** significant at 10, 5, and 1 percent. Std errors are clustered at year and 4-digit NACE industries. Source: APEH and CR

30

Table 14: Effect on Indebtedness and Foreign Currency Debt Log FX Debt-to-Debt Ratio

D*CM

D

Firm FE

(1)

(2)

(3)

-0.113***

-0.117***

-0.041*

(0.033)

(0.033)

(0.024)

Log Debt-to-Sales Ratio (4)

(5)

(6)

0.013

0.012

0.031

(0.031)

(0.031)

(0.029)

Yes

Yes

0.025

-0.079***

(0.024)

(0.019)

Yes

Year FE

Yes

Yes

Yes

Yes

Sector*Year FE

Yes Yes

Yes

R2

0.781

0.781

0.783

0.799

0.800

0.801

N

22,834

22,834

22,833

51,310

51,310

51,308

Notes: *, **, *** significant at 10, 5, and 1 percent. Std errors are clustered at year and 4-digit NACE industries. Source: APEH and CR

31