The Procyclical Leverage Effect of Collateral Value on Bank Loans-Evidence from the Transaction Data of Taiwan

Article excerpt

I. INTRODUCTION

In the financial contract literature, collateral has been identified as serving as a screening device and sorting mechanism, as in Berger and Udell (1990) and Bester (1985). The practical significance of collateral is recognized in recent studies, for example, Bernanke and Gertler (1989), Chen (2001a), and Kiyotaki and Moore (1997), on financial contracts in securing repayments, which consider collateral as a primarily important factor in determining external financing and investment. In particular, Kiyotaki and Moore (1997) investigated exogenous shocks' transmission mechanism of collateral channel through the interaction of credit constraints and the value of collateralized assets. Fluctuations in asset prices change the value of the collateral and affect the firm's ability to obtain external financing, thus magnifying business fluctuations. Chen (2001a) extended the connection of collateral value and bank loans by taking banks' financial characteristics into consideration, emphasizing the role of banks in affecting the amount of collateral-secured loans. Their models hence provide a theoretical link among fluctuations of collateral value, firms' credit constraint, and the capacity of bank lending. Their models further imply that the leverage per dollar value of collateral is positively correlated with the (expected) value of the collateral.

The significance of firms' collateral value on bank lending has been empirically examined in the literature, and most of the analyses are based on macrodata. For instance, Kiyotaki and West (1996) and Ogawa et al. (1996) used macro time series data to explain the investment behavior of Japanese firms in the 1990s and found that land value significantly affects investment behavior. Among those few researchers who applied microdata in their analysis, Ogawa and Suzuki (1998, 2000) used Japanese firm-level data and found that Japanese land value has a significant effect on the degree of borrowing constraint. These works, however, do not fully control for bank heterogeneity, which can be important in explaining observed lending behavior and the collateral requirement.

In this study, we used a panel transaction data set from Taiwan to investigate the empirical relationship between a firm's collateral value (1) and land-/real estate-secured loans (2) (henceforth land-secured loans) over asset price cycles. In particular, along the line of Kiyotaki and Moore (1997) and Chen (2001a), we examined whether collateral's leverage on bank credits exhibits procyclicality to asset price cycles. The unique data set used in this study combines detailed loan transaction information with the borrowers' and the lenders' financial profiles. Based on the data set, we estimated a simultaneous equation model of demand and supply to investigate determinants of land-secured loans from the demand side and also from the supply side. Among the determinants, we were most interested in the collateral effect and the procyclicality therein. The transaction-level data set allows us to control for bank heterogeneity by taking into account the financial stances of the lending banks and by including unobserved bank-specific effects in the model.

Collateral-based lending for the business investment closely relates to the issue of lending procyclicality, that is, lending increases significantly during business cycle expansions and then falls considerably during subsequent downturns, occasionally so drastically to be considered a "credit crunch," for example, as noted by Berger and Udell (1994); Bernanke and Lown (1991); and Hancock, Laing, and Wilcox (1995). A large body of evidence has shown that changes in bank credits have strong impacts on aggregate fluctuations via the credit channel, for example, Bernanke and Lown (1991), Kashyap and Stein (2000), and Oliner and Rudebusch (1996), and the impacts are particularly amplified by movements of asset prices and collateral values, for example, Chen (2001b), Goyal and Yamada (2004), Iacoviello (2005), and Peek and Rosengren (2000). …