Credit channel of monetary policy in Japan: resolving the supply versus demand puzzle

Abstract

The credit view is that a monetary tightening affects the real economy by shifting the supply schedule of bank credit left. While bank credit typically contracts following a monetary tightening, the financial contraction does not necessarily mean a shift of the supply schedule. Testing the credit view requires the identification of the shifts of the demand and supply schedules of credit. Using an original approach, this study shows that the credit view is supported for Japan. The credit view is, however, composed of two different views, namely the lending view and the balance-sheet view. While the balance-sheet view implies that the cutback of lending has no impact on the real economy, the lending view implies independent impacts of the cutback. Given the acceptance of the credit view, this study further attempts to test the balance-sheet view against the lending view.

Acknowledgements

An earlier version of this paper was written as part of my PhD thesis at the Australian National University. I would like to thank Graeme Wells and Mardi Dungey for their constructive comments and discussions. I am grateful to the Bank of Japan for the provision of the survey data.

Notes

 See Gertler and Gilchrist (1993) and Oliner and Rudebusch (1996) for the ambiguities concerning the interpretation of the Kashyap et al. (1993) results.

 A difficulty lies in measuring the price of bank loans. The construction of the price of bank loans will be discussed in Section III.

 While imposing a recursive structure is conventional, non-recursive identifying restrictions have been used in the literature. See, for example, Kasa and Popper (1997), Kwon (1998) and Shioji (2000) for structural VAR models of the Japanese monetary policy with non-recursive identifying restrictions.

 This common practice may be contentious. Examining the information set of the Federal Reserve Bank, Rudebusch (1998) argues that a standard equation of the federal funds rate in a VAR model for the USA does not correctly model its reaction, and that the VAR approach is subject to issues such as time invariance, linearity and variable selection. Sims (1998) criticizes critiques made by Rudebusch as unconstructive quibbles, however.

 The Bank of Japan (BOJ) now officially announces that the overnight call rate is its operating target. See minutes of the Monetary Policy Meeting, which are available in English. Not until the late 1990s did the BOJ disclose how it implemented monetary policy. Nevertheless, there have been economists who have argued that the BOJ always attempted to control the overnight call rate. See, for instance, Okina (1993), Ueda (1993) and Yoshikawa (1995).

 In Japan, banks are required to maintain reserves, which are the product of the reserve ratio and average deposits outstanding in each calendar month, during the period from the 16th of that month to the 15th of the next month.

 The model (1) is estimated with different orderings. The conclusion, which is discussed later, is essentially robust to the ordering of the variables. The estimation results with different orderings are available from the author on request.

 The impulse response functions shown in this study are calculated from the estimated model with two lags. The model was also estimated with lags 3 and 4. As for identifying the supply and demand schedules in the bank loan market, however, the results were robust to the choice of lags. Those results are available from the author on request.

 The model (1) was re-estimated with the loan rate replaced by the spread between the loan rate and the call rate. The calculated impulse response function showed that the spread between the two interest rates significantly shrank over quarters in response to a hike of the call rate.

 In the survey, the Bank of Japan asks non-financial firms whether financial institutions’ lending attitude is ‘accommodative’, ‘not so severe’, or ‘severe’. The BOJ calculates the diffusion index by subtracting the percentage of the firms answering ‘severe’ from the percentage of those answering ‘accommodative’. In the estimation, the diffusion index is multiplied by −1, so that it may be positively correlated with the unobservable price of loans. See Appendix for further discussion on the diffusion index.

 A variety of robustness checks were conducted, and the results are essentially robust. Those results are reported on the earlier version of this paper, which is available as a working paper.

 The results are essentially robust to the choice of the loan price. When the loan price is measured by the interest rate on loans, the persistent negative responses of real GDP and the loan quantity are significant at a 10% significance level.

 Ogawa (2000) estimates a six variable VAR (inflation rate, the call rate, sales of firms, the market value of land, total loans outstanding, and fixed investment) for Japan, finding that a positive shock to land stock is followed by significant increases of total loans and fixed investment. These results are consistent with those in the present study. Without showing that monetary policy has impacts on the market value of land, however, he concludes that the balance-sheet channel of monetary policy is operative in Japan.

 Kwon (1998) estimates a nine variable VAR model for Japan over the period 1963:Q1 to 1993:Q4. The variables in his model are essentially the same as the ones used in this section except for the credit variables. He finds that a persistent fall of land price following a monetary tightening is significant. His finding complements the results of this study.

 The address of the English web of the Japanese Real Estate Institute is www.reinet.or.jp/study/index-e.htm.