Studying the Effects of Household and Firm Credit on the Trade Balance: The Allocation of Funds Matters Berrak Buyukkarabacak and Stefan Krause∗ April 2005 Abstract One of the most widely used indicators of financial development in the empirical literature is the Private Credit to GDP ratio. A key shortcoming of this measure is that it does not distinguish between the share of credit extended to households visà-vis firms. It is our contention that this distinction is crucial to analyze the effects of financial development on the trade balance: Changes in the composition of private credit should have an impact on the foreign trade deficit. Our empirical findings show that: 1) private credit to households is negatively and significantly correlated with net exports; 2) private credit to firms is not significantly correlated with net exports; and 3) the allocation of credit matters; a higher proportion of firm credit is positively and significantly correlated with net exports. A key implication of these results is that, whenever there is a sizeable trade deficit or a high risk of a currency crisis, policy makers should limit the growth of household credit while, at the same time, encourage further allocation of funds to firms. JEL classification: F32, F41, G21 Keywords: Firm and household credit, financial development, trade balance ∗ Department of Economics, Emory University. We thank Neven T. Valev for useful discussion. For comments, please contact us at bbuyukk@emory.edu or skrause@emory.edu. 1 Introduction A well functioning financial system plays a key role in the economic performance of a country, so factors that perversely affect the intermediation process will negatively impact the economy. Bernanke and Gertler (1990, 1995) argue that a reduction in the supply of bank credit is likely to increase the external finance premium and therefore reduce real economic activity. Turning to the empirical evidence, Roubini and Sala-i Martín (1992) show that, after controlling for other determinants of economic development, various measures of financial repression affect growth negatively. Rajan and Zingales (1998) find that financial development decreases the costs of external finance thus facilitating economic growth. More recently, Beck, Levine, and Loayza (2000), and Levine, Loayza, and Beck (Levine et al, 2000) find strong statistical evidence that a larger participation of financial intermediaries causes an increase in economic growth, at least in the long-run.1 However, other studies find inconclusive evidence, or even suggest detrimental effects that stem from financial development. For countries with very low levels of financial development, Rioja and Valev (2004a,b) observe that the effect of additional improvements in financial development on output growth is uncertain. Aghion, Bacchetta, and Banerjee (2004) show that economies at an intermediate level of financial development are more unstable than either very developed or very underdeveloped economies and argue that countries which are going through a phase of financial development may become more unstable in the short run. Demirgüç-Kunt and Detragiache (1999, 2002) argue that the relationship between financial liberalization, financial development and growth should be analyzed cautiously whenever prudential regulation and supervision are not fully developed. De Gregorio and Guidotti (1995) find evidence suggesting negative effects of financial development on growth for twelve Latin American countries, and argue that the negative effect is the result of financial liberalization in a poor regulatory environment. Finally, Demirgüç-Kunt and Detragiache (1998) find that a high share of credit to the private sector and elevated past economic growth may 1 See King and Levine (1993a,b) and Levine (1997) for an exhaustive literature review. 1 be associated with a higher probability of a crisis. A widely used indicator of financial development in this rich literature is the relative importance of loans issued by commercial banks and other financial intermediaries to the entire private sector; i.e., the Private Credit to Gross Domestic Product (GDP) ratio, following the definition by King and Levine (1993a,b). However, this measure has one major shortcoming: It does not distinguish between the fraction of credit extended to households vis-à-vis firms. It is our contention that this distinction is crucial to analyze the effects of financial development on several key macroeconomic performance variables, such as growth and inflation among others. Out of the several possible macroeconomic variables of interest, in this paper we concentrate our efforts on a particular one: The trade balance. Our rationale is that for small, open economies the external sector plays a pivotal role in determining overall macroeconomic performance. We show empirically how the allocation of loans to the private sector between consumers and producers affects a country’s foreign trade position. While an increase in household credit, all other things equal, results in an overall rise in consumption and, therefore, in the purchase of imported goods, an increase in direct lending to firms will affect investment and eventually contribute towards the production and export capacity of a country. Therefore, changes in the composition of private credit should have an impact on the trade deficit. The main objective of our study is to analyze the effect of the allocation of credit to the private sector on the trade balance. As argued by Wood (1997), "[...] the [International Monetary] Fund continues to place most emphasis within its programmes on the stabilisation objective and therefore the core of IMF programmes continues to be based upon the restriction of domestic credit creation [...]." The practice of limiting credit growth has been adopted by several developing countries; in particular Mexico during the 1994-1995 crisis (Gruben and McComb, 1997) and Brazil in 1999 (International Monetary Fund, 2003). However, even though the distinction is made between credit to the public versus the private sector, there 2 is no mention as to different outcomes resulting from restricting household or firm credit. More recently, Chan-Lau and Chen (2002) find that in the developing stages of a financial system, the country is more exposed to exchange rate and financial crises. In their extensive and detailed review of the causes and consequences of the East Asian Crisis, Ito and Krueger (2001), found one common underlying explanation: the expansion of private domestic credit, combined with weak regulation, is considered the main catalyst of the currency crises and trade imbalances in East Asian countries during the latter 1990s.2 Again, these studies focus on total lending to the private sector and do not empirically distinguish between the separate effects household and firm credit have on currency crises and trade imbalances.3 The remainder of the paper is organized as follows: We first describe our sample selection and data in Section 2. Clearly, we are restricted to analyzing only countries and time periods for which data on household and firm credit was, not only available, but reliable as well. These restrictions lead us to examine an unbalanced panel of nine countries.4 As we elaborate below , we focus exclusively on developing and transition economies for two main reasons: 1) these countries have undergone a financial liberalization process more recently than most industrialized economies; and 2) for all of them, the main source of funds for consumers and producers is bank lending and financial intermediation, with stock markets playing a relatively minor role. In Section 3 we detail the instrumental variable estimation of the household and firm credit ratios, which we use in order to overcome potential endogeneity issues. We then utilize the first-stage estimates of these two main components of private credit in Section 4 to perform several dynamic panel data analyses. Here we model the trade surplus as a function of private lending to consumers and producers, and the relative importance of each 2 See Corsetti, Pesenti, and Roubini (1999) for the derivation of a model of financial crisis that focuses on moral hazard as the common source of overinvestment, excessive borrowing, and current account deficits in an economy with a poorly supervised and regulated financial sector. 3 One exception in the theoretical literature is the work by McKinnon and Pill (1997). They develop a model that shows how, whenever domestic consumer credit is insufficiently restrained, the effect of moral hazard in banks becomes more damaging, generating financial instability and exchange rate crises. 4 These countries are: Argentina, Brazil, Bulgaria, Costa Rica, Hungary, Mexico, Poland, Turkey, and Uruguay. 3 component, alongside some control variables. Our results show that: 1) private credit to households is negatively and significantly correlated with net exports; 2) private credit to firms is not significantly correlated with net exports; and 3) the composition of credit to the private sector matters; a higher share of firm credit is positively and significantly correlated with net exports. These findings are robust to alternative specifications and estimation methods. We perform several comparative statics exercises in Section 5 and provide some policy prescriptions based on our main observations. A key implication of our findings is that in the presence of a sizeable trade deficit or a high risk of a currency crisis, policy makers should limit the growth of household credit only, rather than restricting the expansion of total domestic credit to the private sector. Section 6 concludes and elaborates on possible extensions. 2 Sample Selection and Data Description As we point out above, we look only at developing and transition economies for which data on lending to consumers and producers is both available and reliable. Nine countries fit these criteria: Argentina, Brazil, Bulgaria, Costa Rica, Hungary, Mexico, Poland, Turkey, and Uruguay. We list the maximum number of sequential quarters, between 1990:I-2004:IV, for which information on household and firm credit was obtainable and display this information in Table 1. The data periods coincide with the ones provided by the national central banks in their respective websites with three exceptions. First, for the case of Argentina there exists information beyond 1999:IV; nevertheless, we found a change in the reporting methods that renders the data on the decomposition of credit to the private sector starting 2000:I impossible to compare with the information reported for the 1990s. 4 Table 1: Data Description Country Period Variable used for Household Credit Variable used for Firm Credit Argentina 1993:I-1999:IV Family & Individual Loans Primary, Industry & Services Sector Loans Brazil 1995:I-2004:II Housing & Individuals' Loans Industry, Rural & Commerce Sector Loans Bulgaria 1998:I-2004:II Household Credits Firm Credits Costa Rica 1996:I-2004:I Consumption Loans Agricultural, Industry & Services Sector Loans Hungary 1995:I-2004:II Households' Balance Sheet Firms' Consolidated Balance Sheet Mexico 1995:I-2004:III Credit to Housing & Consumption Credit to Primary, Industry & Services Poland 1995:I-2001:IV Claims to Household Sector Claims to Corporate Sector Turkey 1990:I-2003:IV Private Credit to Firms Private Credit to Consumers Uruguay 1990:I-2001:IV Credit to Consumption Credit to Agriculture, Industry, Commerce & Service Second, Brazil does have information on lending to consumers and producers between 1991:I and 1994:IV, while Bulgaria has data on these two categories between 1996:I and 1997:IV. However, Brazil experienced an average annual inflation of nearly 1,500% between 1991 and 1994; meanwhile, Bulgaria had an inflation rate close to 700% on average in 1996 and 1997. The outcome of these extremely inflationary periods is that, since these two central banks report nominal values, and we are unable to determine during what time of 5 the month or quarter the information is collected (and whether or not an adjustment has been made), we cannot rely on this data for our analysis.5 Another important criterion for our data selection was to consider countries that have undertaken measures to eliminate financial repression either slightly before or during the 1990s. The nine countries in our sample fit this description. Also, some of these countries have been either directly or indirectly affected by recent balance-of-payments and foreign exchange crises. The Mexican currency crisis of 1994/1995 had consequences Argentina, Brazil, Costa Rica and, to a lesser extent, Uruguay (through the "Tequila Effect"); the East Asian and Russian Crisis of 1997/1998 impacted the economies of former communist countries, Turkey and several South American countries; while the Brazilian and Argentinean fiscal troubles and banking panics, at the end of the 1990s and in 2002, respectively, caused severe problems for those two countries and some of their trading partners. One final criterion is that all of these countries had relatively small stock markets during most of the period of study, as shown in Table 2. With the exception of Mexico, Brazil and Argentina, who have experienced a steep development of external financing since the latter part of the 1990s, stock market capitalization does not even exceed ten percent of GDP during the past decade. This implies that the private sector in these economies heavily relies on banks and financial intermediaries as a source of funds for consumption and investment. 5 Fortunately, no other country or period in our sample has such high inflationary outcomes. The closest is Turkey in 1994, where average inflation was about 100%; this number is more manageable. 6 Table 2: Stock Market Size (1990-1999) Country Avg. stock market capitalization to GDP (in %) /a Avg. stock market total value traded to GDP (in %) /b Argentina 12.71 4.03 Brazil 18.92 12.98 Bulgaria 2.76 0.19 Costa Rica 6.60 0.29 Hungary 9.11 9.66 Mexico 26.93 11.81 Poland 3.97 3.60 Turkey 9.86 20.87 Uruguay 1.06 0.03 Average 10.21 7.05 /a: Source: Own computation based on data from the Standard and Poor's Emerging Market Database (and Emerging Stock Markets Factbook) and the World Development Indicators (2004) /b: 3 Source: Own computation based on data from the World Development Indicators (2004) First Stage: Instrumental Variable Estimation We now turn our attention to the measurement of the household and firm credit de- composition. We argue that the distinction between these two components of total private credit is of great value for many empirical applications since changes in the allocation of funds between consumers and producers will have an important effect on macroeconomic variables, including output growth. We choose to focus our efforts on the effects of loans to 7 households and firms on the trade balance, given the crucial link that exists between macroeconomic outcomes and the external sector for small, open economies. This way, we hope to provide some insight into preventing future currency and foreign trade crises in developing and transition economies.6 Before we proceed to estimating the relationship between private credit allocation and the trade balance, we need to deal with potential endogeneity issues. Clearly, the level of aggregate spending is directly related to bank lending and conversely. Also, by national accounting, both exports and imports depend on the level of aggregate demand and viceversa, which implies that net exports and domestic credit are determined simultaneously. Therefore, we require a first-stage estimation to obtain instrumented estimates of the credit to households and credit to firms variables. To do this we run the following two autoregressive models: ct + εht ht = α0h + α1h ht−1 + α2h rt−1 + α3hb ft = α0f + α1h ft−1 + α2f rt−1 + α3fbit + εf t , (1) (2) where h and f represent, respectively, household and firm credit (as defined in Table 1) as ratios of GDP; r is the (effective) real interest rate; b c is the instrumental variable (IV) predicted value of aggregate consumption over GDP; and bi represents the IV-estimate of the aggregate investment-GDP ratio. b c is the outcome of modeling the consumption-GDP ratio as a linear function of its one-period lagged value, one-period lag of credit-to-household-GDP ratio, imports and time; while the regressors used to obtain bi are its autoregressive term, the lagged value of credit-to-firm-GDP ratio, exports and time. We report the R-squared and the estimates of the α1 ’s, α2 ’s, and α3 ’s for equations (1) and (2) in Tables 3a and 3b, respectively. Prior to describing the results of this first stage process, it is imperative that we justify 6 For example, Mendoza and Uribe (2000) show how devaluation risk leads to distortions in real prices and economic activity through the increase in real interest rates resulting from a higher risk premium. 8 the choice of our instruments. Employing the lagged real interest rate in equations (1) and (2) is relatively straightforward; credit to the private sector, whether it is used to finance spending or production, depends on the cost of funds, albeit not immediately but with at least a few months of delay. Also, as argued by Bernanke and Gertler (1990, 1995) and others, the (real) interest rate first affects credit and then spending (through the lending channel), which justifies our exclusion of this variable in the consumption and investment IV equations. We also find it reasonable to instrument consumption through lagged imports. The reasoning for this is that a higher level of imports makes more goods available in the economy and, as a consequence, all sectors (in particular households) will purchase more goods. This effect need not be immediate, but eventually (a three-month lag is our assumption) consumers’ spending will increase as a result of a rise in imports. As an outcome of the higher level of consumption, credit to households will also rise through the increased spending effect. Finally, we can apply a similar logical argument to the export-investment relationship: A boost in exports could indicate a growing foreign market with more opportunities for domestic firms. These firms will have to invest in order to expand production capabilities, resulting eventually in the need for more bank loans. Table 3.a describes the regression results of the household credit ratio to GDP. In general, the results are as expected: consumer credit is positively and significantly correlated with its own past value and either negatively or uncorrelated with the lagged real interest rate. Finally, for all countries we find that the IV estimate of consumption is either positively and significantly correlated or uncorrelated with household credit. In general, the goodness of fit criterion also shows that these regressors provide an important explanatory power (over 63% for all countries) in accounting for the variance of consumer credit. 9 Table 3.a.: First Stage: IV regression (Dependent Variable: Household Credit /GDP) Country Lagged household credit Lagged real rate Consump. (IV-Estim.) R2 Argentina 1.0307 -0.0026 -0.0275 (0.000) (0.941) (0.952) 0.9036 Brazil 0.4956 -0.0111 0.4901 (0.030) (0.045) (0.158) Bulgaria 1.1780 -0.0613 0.3248 (0.000) (0.003) (0.023) Costa Rica 0.8692 -0.1766 -0.2298 (0.000) (0.448) (0.253) Hungary 1.7164 -0.2533 -0.4759 (0.000) (0.746) (0.346) Mexico 0.5597 -0.0167 -0.3443 (0.002) (0.346) (0.154) Poland 0.9570 -0.0482 0.3802 (0.000) (0.662) (0.088) Turkey 1.0047 -0.0321 -0.0970 (0.000) (0.013) (0.652) Uruguay 0.9851 0.0171 -0.0457 (0.000) (0.112) (0.349) 0.6341 0.9722 0.9422 0.7613 0.6827 0.9594 0.7749 0.9152 Turning to the equation of firm-credit-to-GDP ratio in Table 3.b, the lagged dependent variable once again has the expected sign and is always significant. Not too surprisingly, we encounter no correlation between the real interest rate and loans to producers. Consistent with our expectations, we do find that the effect of the IV estimate for investment is positive and significant for six countries, and the coefficient is only not distinguishable from zero for Bulgaria, Costa Rica and Mexico. The goodness of fit criterion shows that the explanatory variables account for well over 80% of the variance with the exception of three countries: Argentina, Brazil and Turkey. Nevertheless, we note that for these countries the coefficients 10 on lagged firm credit and the IV estimate for investment are significant and have the expected positive sign, which brings further support to our specification. Table 3.b.: First Stage: IV regression (Dependent Variable: Firm Credit /GDP) Country Lagged firm credit Argentina 0.3760 0.0426 0.2735 (0.021) (0.111) (0.016) Brazil 0.5191 0.0004 1.7026 (0.001) (0.971) (0.045) Bulgaria 1.0351 0.1004 -0.2341 (0.000) (0.190) (0.549) Costa Rica 0.9566 0.1461 1.1793 (0.000) (0.431) (0.113) Hungary 0.9376 0.7171 1.5813 (0.000) (0.349) (0.005) Mexico 1.0251 -0.0146 -0.1317 (0.000) (0.289) (0.129) Poland 0.8069 -0.1147 1.0035 (0.000) (0.684) (0.000) Turkey 0.6307 -0.0405 0.6105 (0.000) (0.436) (0.074) Uruguay 0.9919 0.0081 0.5897 (0.000) (0.766) (0.068) Lagged Investment real rate (IV-Estim.) R2 0.3132 0.5263 0.8514 0.9555 0.8192 0.9609 0.8089 0.4933 0.8203 Summarizing the results from this section, we believe that the first-stage adjusted values for both household and firm credit are warranted and can therefore be employed to analyze the implications that the composition of private credit has on the trade balance. 11 4 Second Stage: Panel Data Analysis Our main objective is to determine whether or not the particular allocation of private credit between consumers and producers has an effect on the trade balance. We believe that this composition is important: household credit will have a different effect than firm credit on net exports. Formally, our main hypotheses can be summarized as follows: • Hypothesis 1: A rise in household credit causes a decrease in the foreign trade surplus through its effect on consumption spending. • Hypothesis 2: Firm credit has no immediate effect on net exports.7 • Hypothesis 3 : Controlling for total credit to the private sector, a higher share of credit extended to producers (vis-à-vis credit extended to households) increases the foreign trade surplus . We now proceed to study how the trade balance, as a fraction of GDP, is associated with household and firm credit. We perform an unbalanced dynamic panel analysis, controlling for the current value and the one-quarter lag of the real devaluation rate. The models we estimate are represented by the following equations: f + γ 11 di,t + γ 12 di,t−1 + ν i + ni,t = β 11 hfi,t + β 12 fi,t 1i,t , (3) ni,t = β 21 sfi,t + β 22 pfi,t + γ 21 di,t + γ 22 di,t−1 + η i + 2i,t , (4) where: t represents the quarter; n represents exports minus imports divided by GDP; αi is the country-specific effects; hf and f f are the fitted values of household-credit-to-GDP ratio and household-credit-to-GDP ratio, obtained from equations (1) and (2), respectively; sf (≡ ff ) hf +f f represents the share of total private credit that is directed to firms; pf is the amount of total private credit relative to GDP; and d is the real exchange devaluation rate. 7 Eventually, an expansion in firm credit should lead a rise in net exports by boosting capital accumulation and production. However, since it is quite difficult to empirically determine how long will it take for this full effect to take place, we will refrain from testing this implication in our present analysis. 12 Hypothesis 1 and Hypothesis 2 imply that β 11 < 0 and β 12 = 0 in equation (3). We test Hypothesis 3 by estimating β 21 in equation (4). Regarding our control variables, we expect a positive impact of the real devaluation rate (γ 11 > 0 and γ 21 > 0) on net exports. We use two econometric techniques to estimate the relationship between trade balance and credit levels: (1) an uncorrected random effects model; and (2) a random effects model with AR(1) disturbances. We note that in 11 out of the 12 specifications we analyze, the Hausman (1978) test does not reject the hypothesis that individual country effects are uncorrelated with other regressors at the 5% significance level, which leads us to conclude that the random effects model is indeed the best choice. Table 4.a: Panel Data Analysis Household and Firm Credit Ratios Explanatory variable Specification 1 Specification 2 Specification 3 -0.0670 -0.0674 -0.0668 (0.003) (0.002) (0.003) -0.0237 -0.0223 -0.0220 p-value (0.113) (0.126) (0.129) Real Devaluation 0.0020 0.0100 p-value (0.899) (0.342) Lag. Real Devaluation 0.0085 0.0097 (0.482) (0.239) Household Credit /a p-value Firm Credit /a p-value Chi2-statistic 22.56 21.61 22.02 p-value (0.000) (0.000) (0.000) /a: First stage estimates of household credit to GDP and firm credit to GDP. 13 Finally, the reason we do not employ a dynamic panel generalized method of moments (GMM) approach is that our cross-section sample size is restricted with nine countries. Since we deal with endogeneity problems by instrumenting our credit variables, we believe that the two techniques above will provide us robust estimators. To verify the robustness of our results, we employ an alternative estimation method; namely, a random effects model with autoregressive (AR(1)) disturbances (Baltagi and Wu, 1999). Table 4.b: AR (1) Random Effects Household and Firm Credit Ratios Explanatory variable Household Credit /a p-value Firm Credit /a Specification 1 Specification 2 Specification 3 -0.0557 -0.0560 -0.0560 (0.015) (0.014) (0.014) -0.0056 -0.0055 -0.0056 p-value (0.678) (0.679) (0.674) Real Devaluation 0.0129 0.0158 (0.415) (0.201) p-value Lag. Real Devaluation p-value 0.0036 0.0097 (0.766) (0.303) Chi2-statistic 10.37 10.30 9.82 p-value (0.035) (0.036) (0.044) /a: First stage estimates of household credit to GDP and firm credit to GDP. Our results from Tables 4.a. and 4.b. support both Hypothesis 1 and Hypothesis 2 : the ratio of household credit to GDP has a negative and (at the 1% and 2% level, respectively) significant impact on the trade balance, whereas firm credit has no significant effect on net 14 exports. These results are robust under the three different specifications and across the two alternative estimation models. Finally, we note that the coefficients on current and lagged real devaluation are positive but insignificant under both estimation procedures. Table 5.a: Panel Data Analysis Share of Total Private Credit Directed to Firms Explanatory variable Specification 1 Specification 2 Specification 3 Share of Credit to Firms/a 6.1828 6.1475 6.1106 p-value (0.040) (0.037) (0.038) Total Private Credit/a -3.2748 -3.1696 -3.1541 p-value (0.001) (0.001) (0.001) Real Devaluation 0.0035 0.0118 p-value (0.822) (0.259) Lag. Real Devaluation 0.0087 0.0108 (0.469) (0.186) p-value Chi2-Statistic 25.04 23.65 24.19 p-value (0.000) (0.000) (0.000) /a: Share of Credit to Firms and Total Private Credit are computed using first stage estimates of household credit to GDP and firm credit to GDP. We present the estimation results of (4) in Tables 5a and 5b. The data strongly favors Hypothesis 3 : controlling for total credit to the private sector, the share of firm credit to total private credit has a positive impact on net exports, and this effect is significant in all three specifications and in both alternative estimation procedures (with the exception of the specification with only lagged real devaluation with the AR-correction model, where it becomes not-significant at the 10% level). 15 It is worth noting that the coefficient on total private credit over GDP is negative and significant at the 5% level, and this observation consistent with the policy prescription recommended by the IMF; namely, that restricting total domestic lending will indeed result in a more favorable trade balance. Nevertheless, our initial claim that the composition of credit matters is validated by our results. Finally, we again fail to encounter any effect of current and lagged real devaluation on the trade balance. Table 5.b: AR (1) Random Effects Share of Total Private Credit Directed to Firms Explanatory variable Specification 1 Specification 2 Specification 3 Share of Credit to Firms/a 5.1609 5.1880 5.0596 p-value (0.100) (0.098) (0.106) Total Private Credit/a -1.8587 -1.8662 -1.8707 p-value (0.041) (0.040) (0.040) Real Devaluation 0.0139 0.0171 (0.382) (0.167) p-value Lag. Real Devaluation p-value 0.0039 0.0106 (0.747) (0.263) Chi2-Statistic 10.33 10.24 9.68 p-value (0.036) (0.037) (0.046) /a: Share of Credit to Firms and Total Private Credit are computed using first stage estimates of household credit to GDP and firm credit to GDP. Overall, the empirical evidence provides strong and robust support for our claim that consumer credit and producer credit have sharply different impacts on the trade balance. We should also note that the above results are robust to employing the actual data on the 16 household and firm credit ratios.8 In the next section we look more closely at the implications of these empirical outcomes and provide a few comparative statics exercises. 5 Comparative Statics and Implications Exactly how big are the effects of changes in the allocation of funds to consumption and the composition of private credit on the trade balance? Table 6.a runs a comparative statics exercise for the two countries for which we have the longest time series: Turkey and Uruguay. From 1990:I until 2003:IV, Turkey experienced an increase of household credit from 0.53 percent of GDP to 16.52 percent of GDP. This dramatic change, all other things equal, translates into an average of nearly one percent rise in the foreign trade deficit. The numbers for Uruguay are smaller, but nonetheless still important: as a ratio of GDP, consumer credit rose from 2.46 percent in 1990:I to 8.33 percent in 2001:IV, resulting in an average contribution of 0.37 percent increase to the external deficit. For all other countries except Brazil, private lending to households also increased as a percentage of GDP. We should point out that the Brazilian economy has been experiencing trade surpluses that average over 3 percent of GDP since 2002; larger than any of the other eight countries in our sample. Table 6.a.: Comparative Statics: Effect of Change in Household Credit Ratio on Net Exports Period of Interest Actual Increase in Household Credit Predicted Range of Change on Surplus Turkey From 1990:I to 2003:IV 15.99% of GDP -1.08% to -0.89% of GDP Uruguay From 1990:I to 2001:IV 5.87% of GDP -0.40% to -0.33% of GDP Country 8 The results of these alternative tests are available upon request from the authors. 17 The measured effects of the composition of credit to the private sector on the foreign trade account are even more astonishing, as can be seen in Table 6.b. Looking once again at the case of Turkey, the participation of firm credit in total credit to the private sector fell from 0.987 during the first quarter of 1990 to 0.562 for the last quarter of 2003, resulting in an average increase in the trade deficit of 5.37 percent of GDP. This change in the composition of credit alone accounts for two-thirds of the reported foreign trade deficit of 2003! Table 6.b.: Comparative Statics: Effect of Change in Composition of Private Credit on Net Exports Period of Interest Change in Share of Firm Credit Predicted Change in Trade Surplus Turkey From 1990:I to 2003:IV -0.425 -2.63% to -2.15% of GDP Uruguay From 1990:I to 2001:IV -0.133 -0.82% to -0.67% of GDP Country For Uruguay, the comparative statics exercise indicates that the decline in the portion of private credit extended to firms from 0.884 in 1990:I to 0.751 in 2001:IV is responsible for an average of 1.68 percent of the trade deficit: Close to two-fifths of the reported value of 2.04 percent for 2001. As stated before, all countries but Brazil experienced a decrease in the relative importance of credit to producers vis-à-vis consumers over the period of analysis. We therefore should expect that their foreign trade deficits would have been smaller if it were not for the increase in both the relative size of consumer credit and its share in the total funds destined to the private sector. There are several key implications that stem from our results in Section 4 and the comparative statics analyses: • private credit to firms has no significant effect on the trade balance; • a decrease in net exports is associated with both an increase in household credit and 18 a larger share of funds allocated for consumption expenses; • the effects are sizeable: a one percent rise in the foreign trade deficit may result from either an increase in consumer credit as a fraction of GDP in the order of only less than 14 percent, or a larger participation of loans to households in total lending in the order of merely 0.15; • as a direct consequence, policy makers should pay more attention to the composition of domestic credit rather than the absolute size of it. 6 Conclusions We have argued and presented empirical evidence showing that analyzing the effects of the particular distribution of funds between households and firms is more important for explaining foreign trade imbalances than the size of domestic credit per se. The measured effects are quite large; therefore, policy makers, faced with either an unmanageable trade deficit or the risk of a currency crisis, should focus their efforts on restricting consumer credit, instead of total private credit. We do not encounter any evidence suggesting that a larger amount of funds allocated to firms has any significant effect on foreign trade deficit. Our findings support the hypothesis that a higher relative share of credit to producers is both significantly and in a sizeable manner associated with a boost to net exports. The discussion on the effects of firms’ credit on foreign trade deficit can be extended by analyzing the imports of goods and raw materials. Contingent on data availability, we would be interested in studying the likely outcomes of the allocation of funds to the private sector by way of considering the value added from exports relative to the purchase of (imported) raw materials and capital. Nevertheless, we emphasize that in the present study we fail to find any direct positive effect of firm credit on the trade balance. 19 Limitations of our study are mainly related to data availability, both in terms of time series and the number of countries. We look forward to the possibility of obtaining additional reliable data to broaden the scope of our analysis. On a final note, making the distinction between private credit to households versus firms can have enormous implications in potential empirical studies that involve analyzing the effects of financial development on other key macroeconomic variables such as growth, inflation, interest rates, and exchange rates. We hope to extend our line of research into this direction in the very near future. 7 Appendix: Data Sources The countries included in the sample are: Argentina, Brazil, Bulgaria, Costa Rica, Hungary, Mexico, Poland, Turkey, and Uruguay. Firm Credit and Household Credit data was obtained from the individual central banks’ websites for each individual country, while information for the remaining variables was obtained from the International Monetary Fund’s International Financial Statistics database. Gross Domestic Product is line 99b; Private Consumption is line 96f. Gross Fixed Capital Formation, our measure for Investment is on line 93e. Inflation is given by the percentage change in the Consumer Price Index with respect to the corresponding quarter of the previous year, and it is taken from line 64. Devaluation is the percentage change in the exchange rate of national currency per U.S. dollar at the end of the quarter, with respect to the corresponding period of the preceding year. The information for Exports and Imports of goods and services was obtained from lines 90c and 98c, respectively. The Interest Rates used differed from country to country depending on data availability: the money market rate for Argentina, Brazil and Poland; the discount rate for Hungary and Costa Rica; the interbank rate is for Bulgaria and Turkey; the time deposit rate for Uruguay; and the bankers’ acceptances rate for Mexico. 20 References [1] Aghion, P., P. Bacchetta, and A. Banerjee (2004). "Financial Development and the Instability of Open Economies", Journal of Monetary Economics, 51, pp. 1077-106. [2] Beck, T., R. Levine, and N. Loayza (2000). "Finance and the Sources of Growth", Journal of Financial Economics, 58, pp. 261-300. [3] Bekaert, G., C. R. Harvey, and C. Lundblad (2001). "Does Financial Liberalization Spur Economic Growth?", NBER Working Paper No. 8245. [4] Bakaert, G., C. R. Harvey, and C. Lundblad (2004). "Growth Volatility and Financial Liberalization", NBER Working Paper No. 10560. [5] Baltagi, B. H., and P. X. Wu (1999). "Unequally Spaced Panel Data Regressions with AR(1) Disturbances", Econometric Theory, 15, pp. 814-823. [6] Bernanke, B., and M. Gertler (1990). "Financial Fragility and Economic Performance", Quarterly Journal of Economics, 105, pp. 87-114. [7] Bernanke, B., and M. Gertler (1995). "Inside the Black Box: The Credit Channel of Monetary Policy Transmission", Journal of Economic Perspectives, 9, pp. 27-48. [8] Chan-Lau, J., and Z. Chen (2002) "A Theoretical Model of Financial Crisis", Review of International Economics, 10, pp 53-63. [9] Corsetti, G., Pesenti, P., and N. Roubini (1999). "Paper Tigers? A Model of the Asian Crisis", European Economic Review, 43, pp. 1211-36. [10] Demirgüç-Kunt, A., and E. Detragiache (1998). "The Determinants of Banking Crises in Developing and Developed Countries", IMF Staff Papers, 45, pp. 81-109. 21 [11] Demirgüç-Kunt, A., and E. Detragiache (1999). "Financial Liberalization and Financial Fragility", in B. Pleskovic and J. Stiglitz (eds.), Proceedings of the World Bank Annual Conference on Development Economics, Washington, DC. [12] Demirgüç-Kunt, A., and E. Detragiache (2002). "Does Deposit Insurance Increase Banking System Stability? An Empirical Investigation", Journal of Monetary Economics, 49, pp. 1373-406. [13] De Gregorio, J., and P. Guidotti (1995). "Financial Development and Economic Growth", World Development 23 (3), pp. 433-48. [14] Díaz-Alejandro, C. (1985). "Good-Bye Financial Repression, Hello Financial Crash", Journal of Development Economics, 19, pp. 1-24. [15] Gruben, W. C., and R. McComb (1997). "Liberalization, Privatization, and Crash: Mexico’s Banking System in the 1990s." Economic Review: Federal Reserve Bank of Dallas, First Quarter, pp. 21-30. [16] Hausman, J. (1978). "Specification Tests in Econometrics", Econometrica, 46, pp. 125171. [17] International Monetary Fund (2003). "The IMF and Recent Capital Account Crises Indonesia, Korea, Brazil" Report by the Independent Evaluation Office (IEO) (September). [18] Ito, T. and A. O. Krueger, eds. (2001). Regional and Global Capital Flows: Macroeconomic Causes and Consequences (NBER- East Asia Seminar on Economics, vol. 10), Chicago: University of Chicago Press. [19] Kaminsky, G., and S. Schmuckler (2003). "Short-Run Pain, Long Run Gain: The Effects of Financial Liberalization", NBER Working Paper No. 9787. 22 [20] King, R. G., and R. Levine (1993a). "Finance and Growth: Schumpeter Might be Right", Quarterly Journal of Economics, 108 (3), pp. 717-37. [21] King, R. G., and R. Levine (1993b). "Finance, Entrepreneurship, and Growth: Theory and Evidence", Journal of Monetary Economics, 32 (3), pp. 513-42. [22] Klein, M. W., and G. Olivei (1999). "Capital Account Liberalization, Financial Depth and Economic Growth," NBER Working Paper No. 7384. [23] Levine, R. (1997). "Financial Development and Economic Growth: Views and Agenda", Journal of Economic Literature, 35, pp. 688-726. [24] Levine, R., N. Loayza, and T. Beck (2000). "Financial Intermediation and Growth: Causality and Causes." Journal of Monetary Economics, 46, pp. 31-77. [25] McKinnon, R. I. (1973). Money and Capital in Economic Development, Washington, DC: Brookings Institution. [26] McKinnon, R. I., and H. Pill (1997). "Credible Economic Liberalizations and Overborrowing", American Economic Review Papers and Proceedings, 87, pp. 189-93. [27] Mendoza, E., and M. Uribe (2000). "Devaluation Risk and the Business-Cycle Implications of Exchange-Rate Management", Carnegie-Rochester Conference Series on Public Policy, 53, pp. 239-96. [28] Morisset, J. (1991). "Does Financial Liberalization Really Improve Private Investment in Developing Countries?", Journal of Development Economics, 40, pp. 133-50. [29] Rajan, R., and L. Zingales (1998). "Financial Dependence and Growth", American Economic Review, pp. 559-86. [30] Rioja, F., and N. Valev (2004a). "Finance and the Sources of Growth at Various Stages of Economic Development", Economic Inquiry, pp. 127-40. 23 [31] Rioja, F. and N. Valev (2004b). "Does One Size Fit All? A Re-examination of the Finance and Growth Relationship", Journal of Development Economics, 74, pp. 429-47. [32] Roubini, N., and X. Sala-i-Martín (1992). "Financial Repression and Economic Growth", Journal of Development Economics, 39, pp. 5-30. [33] Shaw, E. S. (1973). Financial Deepening in Economic Development, New York: Oxford University Press. [34] Wood, A. (1997). "The International Monetary Funds Enhanced Structural Adjustment Facility: What Role for Development?", Bretton Woods Project (September). 24