Identifying the demand and supply effects of financial crises on bank credit--evidence from Taiwan.

• Author: Chen, Nan-Kuang

1. Introduction

   The credit market in Taiwan underwent rapid development in the first half of the 1990s. As a step toward financial liberalization, the government lifted the ban on new establishments of commercial banks in 1991, pushing 18 new domestic banks to set up by the end of 1996. The 1997 Asian financial crisis, however, inflicted a serious blow to the market. The annual growth rate of bank loan supply was on average 17% from 1991 to 1996, but the figure dropped to 11% between 1997 and 2000. At the same time, the economy suffered the worst recession since 1950. Although it is widely acknowledged that declining bank credit is one of the major causes of the recession, the sources of the credit contraction in the aftermath of the financial crisis are subject to debate.

   The purpose of this paper is to provide a novel empirical approach to discern the demand and supply effects of the changes in credit growth. The approach is based on the short-side rule of market transactions to infer the relative shifts of demand and supply. The novelty is particularly shown in our proposed micro data model in which only the demand- or supply-side data is required for the analysis.

   We use econometric models to examine whether the slowdown of the credit growth following the crisis came from a shift in the demand or a shift in the supply. In particular, we seek to answer the following questions: How did the demand and supply of bank credits change after the financial crisis, and what were the contributing factors? How did the changes explain the dramatic credit slowdown? What types of firms were affected the most in the credit slowdown? Results of this study should shed light on the causes and consequences of a dramatic change in the bank credit market in the event of a large and systematic shock.

   Declines in bank lending can occur either because firms demand less credit or because banks cut back on loans. As for the demand side, firms may cut back investment because of pessimistic perspectives on the economy, and thus reduce their demand for bank loans. More importantly, an unusual slowdown in demand could be due to deterioration in borrowers' balance sheets, which in turn are the consequences of lowered collateral values, declining earnings, and debt overhangs in the postfinancial crisis era. Therefore, firms can restructure their balance sheets by reducing effective demand for external finance, particularly bank credit in a bank-based financial system.

   A shift of bank loan supply by contrast can occur if banks' ability and willingness to extend loans are affected in a financial crisis. In the wake of a financial crisis, banks face higher default risk because of the weakness of borrowers' balance sheets. The increased risk of loan portfolios, which can be shown in the rise of overdue loans, would adversely affect banks' willingness to lend. When the initial overdue loans result in default, the erosion of bank capital constrains banks' ability to lend. The ability to lend is further affected if capital outflow that followed from the crisis causes deposit drain in the local banking system. The supply-side effect is particularly damaging to bank-dependent firms because it is difficult for these firms to replace bank credit with other sources of funds. (1)

   Correctly identifying the cause of declines in bank credits has important implications for policy makers. For example, if the credit decline is on account of weak demand, then economic policies that aim at stimulating aggregate demand might be effective. If, on the other hand, the decline is due to weakened willingness and ability to lend on the supply side, then an easy monetary policy would simply raise excess reserves held by banks and have little effect on raising bank lending and investment. (2)

   A vast amount of literature tries to disentangle the shift of loan supply from the shift of loan demand. One major thread has devoted much effort in identifying the loan supply shocks by providing evidence for the significance of the "bank lending channel" (see Kashyap, Lamont, and Stein [1994] and Bernanke and Gertler [1995] for surveys of this view). Two strategies are commonly used in the literature. One is to compare the relative changes of aggregate variables, such as bank loans, commercial papers, interest rate spreads, and loans, with commitment to identify shifts in loan supply. For example, Kashyap, Stein, and Wilcox (1993) identified the change in loan supply by looking at the compositional changes in firms' bank and nonbank finances after a tightened monetary policy. They found that monetary tightening reduces the ratio of outstanding bank debt to the sum of bank debt plus commercial paper for the U.S. economy, which indicates support for the bank lending view. The other strategy identifies the loan supply effect by analyzing the differential responses of small versus large firms based on firm-level data (Gertler and Gilchrist 1994; Oliner and Rudebush 1995, 1996b). Results indicated that small firms account for a disproportionately large share of the manufacturing decline after a tightening monetary policy. Because small firms tend to be bank dependent, the result provides evidence for the inward shift of bank loan supply that follows a tightening of monetary policy. Both of the above approaches, however, are likely to suffer from criticism for not adequately controlling for loan demand factors (Oliner and Rudebusch 1996a, b; Kashyap and Stein 2000; Peek and Rosengren 2000). (3)

   In this paper, two different yet complementary econometric models are used to identify the cause of credit contraction. First, we use aggregate data to estimate a model that measures the degree of credit tightness as the extent of excess demand in the market. Excess demand can occur for various reasons. For instance, if the loan supply is not infinitely elastic, then interest rates might not adjust quickly enough and sufficiently to clear the market. The persistence of excess demand could also arise when banks do not raise the interest rate they charge to clear the market because doing so might reduce their expected rate of return (e.g., Stiglitz and Weiss 1981; Williamson 1987).

   The empirical model with aggregate data is essentially a disequilibrium model that has been developed and applied in empirical studies since the 1970s (e.g., Fair and Jaffee 1972; Maddala 1986; Quandt 1988). Studies adopt this empirical framework often because disequilibrium is known to the markets under research. We adopt the disequilibrium model not only for the nature of the market but also for an important advantage: The model enables us to unambiguously determine the relative shifts of demand and supply during an episode of significant credit decline. If the fall in bank credit is mainly caused by shifts in the supply, then the market would exhibit excess demand; if it is due to shifts in demand, then excess supply is more likely.

   Our second model is a newly proposed disequilibrium model for disaggregate data that accommodates the short-side rule of market transactions in a demand-based or a supply-based model. Unlike the aggregate model shown in the first approach, the new model requires only borrowers' or lenders' data, alleviating the need for matched demand and supply data, which is difficult to come by at the disaggregate level. This advantage enables us to make use of the wealth of disaggregate data, on the basis of which we are able to provide more precise estimates of the demand and supply parameters. We are also able to investigate whether the credit decline is borne disproportionately by certain types of firms.

   As will be shown later, the econometric model of the micro-based disequilibrium analysis resembles a stochastic frontier model of the production efficiency literature (e.g., Kumbhakar and Lovell 2000). It differs only in interpretation. In particular, the model of Wang and Schmidt (2002) is adopted in this study. A stochastic frontier model is most often applied to estimate the technical efficiency of firm production. The application to the disequilibrium model of demand and supply is new to the existing literature.

   Estimation results from both the aggregate and disaggregate data clearly indicate that the credit growth decline in the post-1998 period is mainly caused by a large inward shift of supply. We then identify deposit outflows and the increase in overdue loans as the most important factors leading to that event. Evidence also suggests a possible capital crunch effect in the data. The deposit drain coincided with the large-scale capital outflow in Taiwan after the Asian crisis, and the high-rising overdue loans not only increased the overall credit risk perceived by banks but also raised the concern of capital loss. Furthermore, substitutability between bank credits and other types of financing (particularly the public debt) was impaired after the financial crisis, which could also have contributed to the tightness in the bank credit market. Finally, we find that smaller firms were more likely to have experienced unsatisfied loan demand, particularly after the financial crisis. This indicates that the credit decline in the aftermath of the crisis reflects the "flight to quality" behavior of bank lending, which is widely documented in the literature (Bernanke, Gertler, and Gilchrist 1996), in order to restructure banks' loan portfolios amid the problem of adverse selection.

   The organization for the rest of the paper is as follows: Section 2 uses a descriptive analysis of key macro variables to provide a general background of the recent decline in bank credits in Taiwan. Section 3 estimates a demand and supply model with the use of aggregate data. The estimated probabilities of excess demand are presented and their implications are discussed. The subsection on disaggregate data derives a novel model that accommodates the short-side rule of market transactions in a...