I. Introduction

As soon as the Global Financial Crisis of 2008-09 broke out, economists, policy makers and the leaders of the world economy expressed concerns about the adverse effects of financial contraction on world trade, and called for coordinated efforts to prevent a drying-up of trade finance. The G-20 quickly convened and agreed on the need to expand trade finance liquidity, and national governments and international agencies set up programs to supply additional liquidity to finance trade transactions.

Despite these efforts, world trade volume contracted 10.7 percent during 2009, while world GDP contracted 0.6 percent according to the International Monetary Fund (2010). The disproportionately large contraction of trade during the Global Financial Crisis is named the Great Trade Collapse, and many studies attempt to offer explanations. Now there seems to have developed a consensus that most of the fall in trade can be explained by demand contraction caused by the Global Financial Crisis. The recession hit especially hard the demand for consumer durables, manufactured intermediate goods and capital goods, and world trade is concentrated on these kinds of goods.1 However, some observers still find this observation inadequate for explaining trade collapse far larger than GDP contraction. The fall in trade is bigger than most econometric and CGE models would predict, and they argue that there must be supply-side causes, such as a shortage of trade finance or a breakdown of supply chains.

Now we have a large body of literature that examines the role of trade finance during the trade collapse of 2009. Many econometric studies present evidence that trade finance played a significant role in causing trade collapse (Amiti and Weinstein, 2011, Chor and Manova, 2012 and Bricongne et al., 2012), while there are others that disagree (Levchenko, Lewis and Tesar, 2010 and Behrens, Corcos and Mion, 2010). In contrast, informal studies using survey results or direct data on trade finance tend to find the limited role of trade finance in the Great Trade Collapse. For example, many chapters summarized in the World Bank volume edited by Chauffour and Malouche (2011) express this kind of skepticism. We need more country-based studies to obtain a comprehensive picture on the role of trade finance in the trade collapse.

This paper examines the role of trade finance in the trade collapse of 2009 from the perspective of the Korean economy. We use two approaches. Firstly, we make a casual observation on aggregate data to find an indication that trade finance affected trade volume during the crisis. Korea has relatively rich data on trade finance and trade credit. The Bank of Korea publishes aggregate data on “foreign trade loans”, which are loans extended by commercial banks to exporters for the purpose of providing working capital. The Bank also publishes aggregate data on cross-border accounts receivable and payable of private firms, allowing us to track the volume of inter-firm trade credit. The Financial Supervisory Service of Korea reports two sets of data on trade finance: “documentary bills purchased” by commercial banks and “domestic import usances”. We find that the aggregate data do not strongly support the view that trade finance played an active role in causing the trade collapse. The measures of trade finance and the value of trade concurrently dropped, but the ratio of trade finance over trade was stable, and in some cases increased during the crisis period. The aggregate data are consistent with the hypothesis that trade collapse caused trade finance contraction, rather than the other way round.

The second approach of this paper is an econometric examination using firm-level data. Using quarterly data on listed firms in Korea, we conduct panel estimations to test whether exporters with more dependence on external finance or with more reliance on inter-firm trade credit were more vulnerable to the crisis. We adopt firm-level estimations for two reasons. A major problem in estimating the effect of trade finance on trade is to discern whether the fall in trade was due to inadequate trade finance or lower demand. In panel estimation using firm-level data, we can control for industry-specific demand shocks by industry-time dummies. The second reason for using firm-level data is related to the structure of the Korean economy. Most exports from Korea are made by top 30 giant exporters, almost every one of which belongs to a large business group called Chaebol. It is unlikely that these firms were seriously credit-constrained during the crisis. If some firms suffered from a shortage of trade finance, they would be small or medium-sized firms. The effects through SMEs would not show up at industry-level exports where giant firms dominate.

There are a growing number of studies that use firm-level data to test the influence of finance on exports using the difference-in-differences method pioneered by Rajan and Zingales (1998). This paper follows this line of literature, but it can be differentiated in three aspects. Firstly, our data set allows us to check whether the ratio of exports to domestic sales, not the absolute amount of exports, fell more for financially vulnerable firms during the crisis. This approach is required because the challenge here is to find out whether a shortage of trade finance caused a proportionally larger drop of exports than of output. Secondly, we focus on the effect of finance on small exporters, especially on the interaction between the smallness of firms and their financial characteristics. Lastly, we utilize firm-level data on cross-border accounts receivable and payable. Past studies had to rely on data on total trade credit (domestic plus foreign credit), even though data on cross-border credit are required, simply because data on cross-border trade credit were not available.

Our regression analysis suggests that the financial vulnerability of firms, measured by dependence on external fund in investment, the availability of collaterlizable assets, dependence on domestic or international inter-firm trade credit, or dependence on domestic or foreign short-term borrowings, did not contribute to export contraction during the financial crisis. This observation largely applies even to smaller firms, who are usually thought financially more vulnerable. However, our analysis suggests that cross-border trade credit markets did tighten for small exporters that depend heavily on trade credit to import intermediate goods or to export their products. The export contraction of small exporters was intensified, not through the usual channel of bank loans, but by the tightening of international inter-firm trade credits. This paper suggests that policy makers should seriously consider this transmission channel when they design policies for small exporters at a time of financial crisis.

The paper is organized as follows. Section II presents a snapshot of the Korean experience during the crisis based on aggregate data. Section III reports estimation results from the firm-level analysis. Section IV concludes.

1)Eaton et al. (2010) find that more than 80 percent of the decline in the ratio of trade to GDP resulted from an expenditure shift away from manufactured goods, particularly durable goods. Bricongne et al. (2012) find that the impact of finance on the export collapse in France was quantitatively small even though its effect is statistically significant. Asmundson et al. (2011), Behrens, Corcos and Mion (2010) and Committee on the Global Financial System (2014), for example, echo similar views.

II. A Snapshot of Exports and Trade Finance

Figure 1 shows that the won/dollar exchange rate has been rising slowly well before the Lehman shock in September, 2008. The exchange rate then shot up with the Lehman crisis, and by March, 2009, at the peak of financial turmoil, the won depreciated more than 40% from the pre-crisis level. Because of the sharp depreciation, the won prices of Korean exports did not fall much despite a sharp fall in dollar prices, and the won price index actually remained above the pre-crisis level at the height of the Lehman shock. Because of stable export prices, the won value of exports moved in tandem with the quantity of exports during the crisis. Both measures of exports hit the bottom in January, 2009, dropping about 35% from the pre-crisis level. Then they recovered slowly throughout the crisis period.

Figure 2 traces the behavior of foreign trade loans around the crisis period. Foreign trade loans denote credits extended by commercial banks to domestic exporters and export-related firms for the purpose of providing them with working capital necessary for export-related business. The Bank of Korea collects and publishes monthly data on foreign trade loans. These loans are treated as domestic local currency loans of commercial banks because they are denominated in Korean won. We can see from the figure that except for seasonal drops in December, the amount of foreign trade loans was relatively stable during the crisis period.2 The ratio of trade loans to exports did not drop below the downward trend line during the crisis period and even peaked in January 2009, at the height of the crisis. From the graph, it is hard to argue that a sudden contraction in foreign trade loans initiated the export collapse of late 2008 and early 2009.

Figure 3 presents the behavior of documentary bills purchased by commercial banks. They are mostly bills of exchange purchased by commercial banks from exporters.

Domestic exporters are direct recipients of these credits, but because foreign importers or their banks are liable for them, they are treated as external foreign currency assets of banks. Quarterly data are collected by the Financial Supervisory Service of Korea.3 According to Figure 3, both documentary bills purchased by banks and the value of export dropped sharply during Q4 2008 and Q1 2009. However, the ratio of bills purchased to exports remained above the downward trend line and it even peaked in Q1 2009. Again, it would be difficult to argue that the contraction of trade finance in the form of bills purchased by banks initiated the export collapse.

Figure 4 deals with the behavior of import usances. Domestic import usances are banker’ usances extended by commercial banks to domestic importers. These loans are treated as domestic foreign currency assets of banks because domestic importers are liable for them in foreign currency. Quarterly data are collected by the Financial Supervisory Service of Korea.4 Domestic import usances dropped sharply during Q4 2008 and Q1 2009. The value of imports also dropped sharply, but in this case, import usances declined faster, resulting in a slight fall of usances/imports ratio in Q4 2008 and in Q1 2009.

Afterwards, both usances and imports recovered steadily and the usances/imports ratio remained stable. In this case, the initial drop in domestic import usances may have been a factor in causing the sharp drop in imports around the Lehman shock. However, it would be difficult to argue that the lack of import usances was a big drag against the recovery of imports during the crisis period.

Next we investigate the behavior of inter-firm trade credit collected from the Korean International Investment Position. The Bank of Korea records on the external asset side of the International Investment Position cross-border trade credit extended by domestic private firms. We look at only short-term credits. Figure 5 shows that external trade credits extended by domestic private exporters fell from Q3 2008 through Q1 2009. However, the value of exports dropped more sharply, and the ratio of trade receivables to exports increased during the crisis period.

Figure 6 investigates short-term external trade credits extended by foreign exporters to domestic importers in the form of trade payables. It is quite interesting to note that external trade payables of domestic firms increased despite the financial stress and the sharp fall of imports during Q4 2008 and Q1 2009. The ratio of trade payables to imports peaked up in Q1 2009, and was kept at a high level during the crisis period. From figures 5 and 6, it is difficult to argue that a freeze in cross-border trade credit acted as a major impediment to foreign trade expansion during the crisis period.

The data sets above allow us to track the Korean economy both on the path of trade finance and inter-firm trade credit using official statistics. It is repeatedly observed that a measure of trade finance or trade credit sharply dropped during the crisis. However, the value of trade also declined fast such that the ratio of trade finance to trade remained stable and sometimes increased relative to the trend during the trade collapse. This evidence is consistent with the view that trade collapse caused credit contraction, through demand contraction, not the other way round. However, we need a caution in drawing this conclusion. The data are also consistent with the hypothesis that trade finance caused trade collapse, and to explain the rise in the ratio of trade finance to trade noted above, one could argue that a drop in trade finance or credit caused a proportionally larger drop in trade through a multiplier process. We need to make a more rigorous analysis, explicitly recognizing the possibility that causality can run in both directions, by conducting a time-series analysis or a micro-level study. The former approach has been taken by Hwang and Im (2013). They estimate a vector autoregression model and find that the ratio of foreign trade loans to exports and the ratio of documentary bills purchased to exports dropped when variables measuring the degree of financial stress surged during the crisis.5 However, they do not directly estimate the impact of finance on exports. The next section attempts to do so using micro-level data.

2)Because the amount of trade loans is much smaller than the value of exports, Figure 2 and the following figures can give a false impression that trade finance measures were much more stable than the value of trade. Attention should be focused on the movement in the ratio of trade finance to trade.

3)The data do not include credits supplied by the local branches of foreign banks. They might have played a role.

4)Again, the data do not include credits supplied by the local branches of foreign banks.

5)Thus their results are in conflict with Figures 2 and 3 of this paper. The difference seems to come from the fact that they trace the dollar value of foreign trade loans, and use the 12 month average of exports instead of the current value of exports to calculate the ratio of trade finance to exports. Because foreign trade loans are won-denominated loans and the maturity of trade credit is very short, the methods used by this paper seem to be more valid, but we need to make a more careful examination to obtain a reliable conclusion.

III. Firm-level Analysis

In this section, we test using firm-level data whether finance played an independent role in the contraction of Korean exports during the Global Financial Crisis. The econometric method that we will use is a difference-in-differences analysis at a firm level similar to the one used by Behrens et al. (2010) and Bricongne et al. (2012).

We adopt a firm-level approach for two reasons. Firstly, during the Global Financial Crisis of 2008-2009, the predominant concern of the Korean government and economists was its effects on small and medium enterprises. Korean exports are dictated by a small number of giant firms, almost all of which belong to family-controlled business groups called Chaebols. In 2007, top 10 firms exported 60 percent, and top 20 firms exported 74 percent of the total merchandise exports by listed firms in Korea. Almost all these firms belong to Chaebols specially monitored by the Korea Fair Trade Commission, have large cash reserves, and enjoy privileged access to commercial banks and bond markets.6 If financial factors had any independent supply-side effect on the exports of Korean firms, it would show up thorough small and medium exporters. However, the effect on SMEs would be too small to be detectible by industry-level data on exports.7 Indeed, a major response to the financial crisis of the Korean government and public financial institutions was to expand public funds for small and medium firms.8

The second reason that we use firm-level data is that we can control for demand shifts and other time-varying industry-specific effects through industry-time dummies. Most of the contraction in trade finance or trade credit during the crisis is likely to be the one induced by the contraction in demand for Korean exports. A crucial step in detecting the independent role of finance on exports is to purge the influence of these demand shifts from export data. For example, in some industry-level studies, researchers find that export collapse was relatively greater in industries that are more vulnerable to financial shocks, but these industries can be exactly those where external demand dropped more. Carefully-constructed controls for industry-specific demand shocks can avoid this problem, but they are often difficult to obtain. A firm-level analysis allows us to use industry-time dummies to control for demand shifts and other time-varying industry-specific variables.

1. Data

All firm-level data that we use come from the WISEfn database, which compiles the quarterly financial statements of listed firms in Korea published by the Korean Listed Company Association (KLCA).9 The database contains total exports and domestic sales of 1,798 listed firms, along with various income statement and balance sheet variables. For years from 2005 through 2009, on average, 62 percent of listed firms reported positive exports. It is unfortunate that the database does not break total exports by destination and by product code.

A potential problem is dealing with zeroes in exports. If we ignore them, all variations of exports at the extensive margin will be eliminated from data, and bias can be introduced in estimation. To mitigate this problem, we will use mid-point growth rates of exports as in Bricongne et al. (2012).10

g_ikt is the year-on-year mid-point growth rate of exports by i th firm in industry k at quarter t. x_ikt is the won value of exports. The only difference from the standard growth rate is that we put in the denominator the average of the past value and the current value of x instead of the past value alone. Thus the growth rate is well-defined unless both the past value and the current value of x equal zero. The mid-point growth rate moves closely with (log x_ikt - log x_ik(t-4) × 100 when both x_it and x_i(t-4) are positive. It takes the minimum value of -200 when the value of exports drops from a positive number to zero and the maximum value of 200 when it increases from zero to a positive number.

We can classify g_ikt into four categories. Growth at the positive extensive margin occurs when g_it = 200, growth at the negative extensive margin when g_it = -200, growth at the positive intensive margin when g_it is in (0, 200), and growth at the negative intensive margin when g_it is in (-200, 0).

The period that we examine is 20 quarters from Q1 2005 through Q4 2009.11 We restrict our analysis to firms in manufacturing industries. Table 1 reports aggregate growth decomposed into the four margins defined above. We can see that variations at the extensive margin are very small compared to those at the intensive margin.12 We can also note that the export growth rate of listed firms in manufacturing moves closely with the growth rate of goods exports in Korean won as reported by the Bank of Korea (their correlation equals 0.86). However, there are some notable discrepancies between the two series around the Great Financial Crisis period. The total mid-point export growth of listed firms never fell below zero during the crisis period and it recovered quickly from Q3, 2009. In contrast, negative export growth rates were maintained from Q2 2009 through Q4 2009 in the BOK data.

In the analysis below, we will define the crisis period as the two quarters from Q1 2009 through Q2 2009. During this period, the manufactured exports of listed firms grew at near-zero rates, while in Q4 2008, Q3 2009, and Q4 2009, the growth rate was near or above 10 percent. The KLCA classifies manufacturing firms as belonging to one of 80 industries based on the major products of firms. The industry classification is identical to the Korean Standard Industrial Classification at 3-digit level. For expository purpose, we regroup them into 18 bigger industries as shown on the first column of Table 2. The second column shows how the exports from each industry were affected during the crisis period of Q1 2009 and Q2 2009 compared to the non-crisis period. The numbers were obtained by estimating the industry fixed effects of the crisis period from a regression of the mid-point export growth rates of firms on industry dummies and industry dummies interacted with the crisis dummy that takes the value of 1 during the crisis period.13 The three industries that are most adversely affected by the crisis are motor vehicles and parts, basic metal products, and other machinery. The three that are most favorably affected are pharmaceuticals and medical chemicals, domestic appliances, and non-metallic mineral products.

Table 3 shows by industry the summary statistics of key financial variables that we will use in regression. The 9 financial variables that we will investigate are defined as follows.

EXTFIN is a variable that measures the degree to which a firm depends on external finance for physical investment. This measure of financial vulnerability is the most popular in the literature on finance and growth after the influential work by Rajan and Zingales (1998). We may expect that during a credit crunch, a firm that has a higher value of EXTFIN will experience lower export growth because of its higher financial vulnerability. Braun (2003) proposed TANG as a measure of secure access to external finance. We expect that a firm who has a higher value of TANG will suffer less in a tight credit situation because collateralized borrowing will be easier. TPAY measures the dependence of a firm on inter-firm credit offered by its suppliers. Fisman and Love (2003) argue that a firm that has a higher value of TPAY would suffer less in a tight credit market because it depends less on commercial banks for working capital finance. However, during a financial crisis, inter-firm credit can contract more than bank loans, and the opposite situation may develop. We tried alternative ways of calculating TPAY such as the ratio of trade payables to the stock of debt or its flow version. However, they did not work any better. In a similar concept, TREC can be used to measure the dependence of a firm on inter-firm credit, now the firm acting as a credit supplier. One can argue that a seller who has a higher value of TREC can maintain more stable flows of sales during a crisis because its customers have less need to borrow from banks. However, with the increased vulnerability of trade partners during a crisis, sellers can cut credit to their customers more sharply than bankers, and the opposite situation may apply. As in the case of TPAY, we tried alternative ways of calculating TREC, but their performances were not better. INV is added to check whether a firm who maintains a higher stock of inventory relative to sales suffered more during the crisis. This tendency might develop simply because firms with higher levels of inventory become more vulnerable to contraction in short-term finance as they need more working capital during production process. However, there can be more complicated reasons. Alessandria et al. (2010) and Zavacka (2012) argue that firms with large inventories can be made more vulnerable to recessions through inventory dynamics.

In addition to these variables, we test the influence of variables that measure directly the financial vulnerability of firms. One may expect that firms who had borrowed more would suffer more during a crisis because they are more likely to be credit-rationed. There are many measures of debt reliance such as debt-to-equity ratio or interest-expense to sales ratio. Among these financial ratios, we report only results using INTCOST, which is the ratio of net interest expenses to EBITDA (earnings before interest, tax, depreciation and amortization). Other direct measures of debt reliance did not work any better. An attraction of our dataset is that it reports short-term borrowings and trade credits denominated in foreign currencies. STBFOR measures the ratio of short-term borrowings in foreign currencies to total sales. We may expect that a firm that maintains a higher value of this ratio would be hurt more by hikes in interest rates or exchange rates during a crisis. The data on trade credits denominated in foreign currencies are quite valuable as data on cross-border trade credit at firm-level are very rare. TPAYFOR is the ratio of trade credit payable in foreign currencies to exports. Using imports rather than exports in the denominator might be more appropriate, but imports data were not available. However, the ratio of cross-border trade payables to exports can be a more relevant measure for our purpose because we are investigating whether exporters that are dependent on trade credit for importing intermediates to be processed into their exports suffer more during a financial crisis. TRECFOR is the ratio of trade credit receivable in foreign currencies to exports. Again we cannot determine a priori whether exporters that depend more on international trade credit would hurt more or less during a crisis.

Table 3 reports for each of 18 industries the medians and the standard deviations of mid-point export growth rates and the financial ratios of firms to be included in regressions. We calculated the sample-period averages of these variables for each firm, and then obtained the median and the standard deviation of each variable in each industry. For each industry, the upper row shows the medians and the lower row (in parentheses) shows the standard deviations. On the second column, we report export growth rates. The average of industry medians is 6.1 per cent and the standard deviations are quite large. On the third column, we see that the standard deviations of EXTFIN are extremely large in most industries, and this occurs mainly because fixed investments are near zero for many firms. This observation casts doubt about the usefulness of EXTFIN data at firm-level. We find from Table 3 that INTCOST, TPAYFOR and TRECFOR also have high standard deviations relative to medians, but to a far less degree than EXTFIN.

2. Estimation

For our benchmark regressions, we will use the following specification.

VAR is one of the nine financial ratios defined above. We add to this list of financial ratios another variable ASSET, which is a dummy that takes the value of 1 when a firm is in the bottom quintile of its industry in terms of asset size. The small-firm dummy is added to test whether small firms were more adversely affected during the financial crisis.14 CRISIS is a dummy that equals 1 if t is in the crisis period. ln LABORPROD is the natural log of labor productivity calculated as the ratio of value-added to the number of employees. ln EXPSHARE is the natural log of a firm’s export share in its industry.

Note that (2) is written in the form of a difference-in-differences model.15 What we aim to test here is whether firms who had higher values of VAR compared to other firms in the same industry experienced larger drops in export growth during the crisis period.16 VAR is included in the regression equation because firms with difference values of VAR were growing at different rates during the non-crisis period, and what we want to compare across firms is the difference between the growth rate during the crisis period and that during the non-crisis period. What we are most interested in is the sign of β₂. A negative (positive) β₂ implies that firms with higher values of VAR experienced lager (smaller) drops in export growth during the crisis. In this way, we can infer whether the financial characteristics of firms captured by VAR exerted a negative (positive) effect on exporting during the crisis.

A potential problem in applying the difference-in-differences method to a firm-level analysis is that financial factors represented by VAR might be correlated with the error term ɛ. To mitigate the endogeneity problem, we employ two methods. Firstly, we use the past values of VARs in regressions. For example, EXTFIN_ikt for quarter t in year s is calculated by the three-year average of EXTFIN from years (s-4) through (s-2).17 Other VARs are calculated in the same way. Secondly, we add on the right-hand side ln LABORPROD to control for the competitiveness of firms. We may conjecture that firms with higher productivity would have higher export growth. We also include ln EXPSHARE, following Rajan and Zingales (1998). This variable is for capturing the convergence dynamics of firms: young and small firms may grow faster than mature firms, or firms who experienced unusual adverse shocks may recover and grow faster in later periods. Again, we use the past values of these variables in regressions. These two variables could be interacted with the crisis dummy, but we found that the interaction terms are not significant and do not alter our main results.

υ_kt are industry-quarter dummies capturing fixed effects that vary across industries and quarters. They capture variations in price deflators and exchange rates that change with (k, t). However, this dummy variable serves a more important purpose. υ_kt are included to control for demand shifts that vary across time and industries. By doing this, we eliminate the influence of industry-level demand shocks on firm-level exports, and this is a crucial step for our analysis, which tries to separate out the supply-side influence of finance from demand shifts. In our benchmark regressions, we use robust standard errors clustered by industries because export shocks faced by firms in the same industry are likely to be contemporaneously correlated.

Table 4 reports the results of our benchmark regressions. The dependent variable is the mid-point growth rates of firms. Across all regressions, we find that the effects of ln LABORPROD and ln EXPSHARE on export growth are highly significant. A 1 percent increase in labor productivity raises the export growth rate by about 0.1 percentage point. We also find that there is a tendency for convergence. A 1 percent increase in the export share tends to lower the export growth rate by about 0.03 percentage point. In the below, we will not further mention about these two variables as the effects of these two variables are consistently strong across all regressions.

Of the ten variables under investigation, only one variable has significant effects on export contraction during the crisis period: EXTFIN. A higher dependence on external finance in investment increased export growth during the crisis, and the effect is significant at 1 percent. This result is not in line with studies that use data from other countries, which tend to find that industries or firms with higher dependence on external funds for investment suffer more during a financial crisis. Our puzzling result might be explained by the hypothesis that exporters that executed more aggressive investment strategies before the crisis grew more during the crisis, but it could have been caused by data problems.

All other variables are found insignificant. The availability of tangible assets for collateral loans (TANG), total (domestic plus foreign) trade credit received and extended (TPAY and TREC), inventories (INV), interest expenses (INTCOST), short-term borrowing in foreign currencies (STBFOR), and trade payables and receivables in foreign currencies (TPATFOR and TRECFOR) all had insignificant effects on export contraction. It is also interesting to note that smaller firms did not seem to experience additional contraction during the crisis. Our ASSET dummy is found to be insignificant. We find the same result whether we use different percentiles to define small firms or use the number of employees instead of assets to define smallness. This finding runs counter to the expectation of many observers who believe that small firms are discriminated in the Korean financial markets.

We investigate further into the experience of small firms during the crisis because this is an important concern of policy makers. For this purpose, we regress export growth on financial ratios interacted with the asset dummy:

VAR and VAR*ASSET measure the influence of a financial variable VAR on the export growth of non-small firms and small firms, respectively, for normal periods. As in the previous regression, VAR*CRISIS is included to see if VAR intensified export contraction during the crisis, in this case, for non-small firms. Here we are interested in the sign of γ₄ on VAR*ASSET*CRISIS. A negative value implies that VAR caused larger export contraction during the crisis for small firms than for non-small firms. In other words, we are asking whether smallness increased export contraction through its interaction with the financial factor represented by VAR. In Table 5, we can see that smallness significantly intensified export contraction through its interactions with two variables: TPAYFOR and TRECFOR. Regressions (18) and (19) report that non-small firms who use more of international trade payables or receivables did not experience any additional export contraction during the crisis. (The coefficient on VAR*CRISIS is near zero and insignificant.) However, small firms that are more dependent on international trade payables or international trade receivables experienced more export contraction during the crisis. The effects are large and highly significant. From the coefficients on VAR*ASSET*CRISIS, we can infer that a 10 percentage point increase in TPAYFOR causes a 2.99 percentage point drop of export growth for small firms, while a 10 percentage point increase in TRECFOR leads to a 2.51 percentage point drop of export growth for small firms. The other variables do not seem to have any influence on the export experience of small firms.

In the remainder of this section, we do some robustness checks for our benchmark regressions. In Table 6, we check whether EXTFIN, the only variable that we found significant in affecting export growth by itself, is robust, using three different estimation techniques. In regression (20), we use the same specification, but cluster standard errors by firms, instead of industries. This method is more reliable if errors are serially correlated. We still find that EXTFIN has a significant and positive effect on export growth. In the next regression, we regress net export growth on EXTFIN, along with controls. Here net export growth is defined as the mid-point growth rate of exports minus the mid-point growth rate of domestic sales. This check is important because what the literature on trade finance and export collapse tries to answer is the question whether trade finance can help explaining the Great Trade Collapse, namely, trade contraction much larger than GDP contraction during the Global Financial Crisis. If one of our financial variables contributes toward answering this question, it should be found to have reduced exports proportionally more than domestic sales. In regression (21), we find that a higher values of EXTFIN is associated with a lower drop of the ratio of exports to domestic sales. In regression (22), we find a similar result when errors are clustered by firms. EXTFIN had a positive effect both on export and on exports/sales ratio.

In Table 7, we do similar robustness checks for equation (3), by which we estimate the effect of smallness interacted with financial ratios. In regressions (23) through (28), we find that TPAYFOR and TPAYREC interacted with smallness remain highly significant across three different estimation methods. During the crisis, small firms that are more dependent on cross-border trade payables or receivables experienced more contraction in export growth and in the growth of exports/sales ratio, regardless of clustering methods. In addition, though not reported, we ran regressions using a broader industry classification of 18 industries, but we obtained similar results.

To quantity the effects through international trade credit markets, we estimate the joint effects of TPAYFOR and TPAYREC using equation (4):

The estimated values for δ₄ and δ₈ are -0.26 and -0.23, respectively, and both of them are significant at 1%. The coefficients for TPAYFOR*CRISIS and TRECFOR*CRISIS are very small and insignificant, and therefore can be ignored. From these estimates, we can infer that the export growth rate of each firm was lowered by (0.26 TPAYFOR_ikt * ASSET_ikt + 0.23 TPAYREC_ikt * ASSET_ikt) during the crisis period because of the tightening of international trade credit for small firms. By calculating the weighted average of these numbers, where the weight is the export share of each firm, we can estimate how much the growth rate of the total exports of listed firms was lowered by tightened international trade credit. We find that the export growth rate was lowered by 0.008 percentage point in Q1, 2009 and by 0.05 percentage point in Q2, 2009. Some of small firms experienced serious squeeze, but they were too small to affect aggregate exports.

6)The numbers and statements are based on the WISEfn database, which compiles the data from the Korean Listed Company Association (KLCA). During the sample period, listed firms contributed on average 80 percent of total Korean exports reported by the Bank of Korea.

7)These giant exporters depend on small and medium-sized domestic firms for the supply of parts and components, and a financial crisis can affect giant exporters through these supply chains. This is a possibility although we do not believe that it is very likely given the tight control of Chaebols over their supply chains.

8)See Financial Services Commission (2008).

9)We are grateful to the chairman Chulsoon Lee for allowing a discounted access to the data sets.

10)By using this method, we can avoid the problem that arises when we take the log of zero exports. This problem is usually dealt with a Tobit or a non-linear estimation. See Head and Mayer (2013) for the recent development of these techniques. For theoretical foundation on the existence of firms with zero exports, Melitz (2003) and Chaney (2008) are standard references.

11)The quarterly reports of financial statements of listed firms became mandatory from 2004. At the time of writing, there were too many missing data in the financial statements for years 2010 and 2011.

12)An issue is how to treat entries into and exits from stock exchanges by firms. Because the exports of a firm before an entry or after an exit are unobservable to us, we simply treat them as missing. We do not treat them as zeroes because many exporters were exporting before it was listed and kept on exporting after it was delisted. Thus growth at the extensive margin in our data set is mostly due to changes in the exports of small listed firms from and to zero.

13)We estimate γ_k in the following equation: D_k is an industry dummy and CRISIS_t is a dummy that takes the value of 1 in Q1 2009 and Q2 2009.

14)Thus our definition of small firms differs from the official one, which treats as small and medium sized firms those that have less than 300 employees or equity value below 8 billion won.

15)The variable CRISIS was not included in the regression as it is subsumed by υ_kt, industry-time dummies.

16)We can write a simple model based on monopolistic competition in which the exports of a firm with higher dependence on finance fall more during the crisis if they have to pay higher interests or get more credit-constrained. For theoretical foundation on the link between finance and international trade, see, for example, Ahn (2011), Antràs and Foley (2011), and Chaney (2013).

17)For each firm, we calculated the three-year averages of numerators and denominators of VAR, and then took a ratio between them. Firms usually are required to submit the financial statements for the past three years when they apply for bank loans. The value for year (s-1) was not used because exports in that year is used to calculate export growth rates for quarters in year s. Annual values were used to increase the availability of the data.

IV. Conclusion

Official data on trade finance are valuable because they are rare. Aggregate data on various measures of trade finance are available in Korea, and we made a casual observation on their behavior during the financial crisis of 2008-09. We find that these data sets do not convincingly support the view that trade finance played an active role in causing trade collapse.

The results on our firm-level analysis mostly confirm this impression. Our regression analysis suggests that the financial vulnerability of firms, measured by various financial ratios, did not contribute to export contraction that firms experienced during the financial crisis. This observation applies even to smaller firms, who are usually thought financially more vulnerable. However, we find that small exporters that depend heavily on cross-border trade credit to import intermediates or export their products did suffer more drops in export growth, while cross-border trade credit markets did not seem to tighten for non-small firms. Our finding implies that the export contraction of small exporters during a financial crisis can be amplified, not through the usual channel of tightened bank loans, but by the contraction of international inter-firm trade credits. Policy makers should take this possibility into mind when they design policies for assisting small exporters.