Deposit insurance and depositor discipline: direct evidence on bank switching behaviour in Japan

Abstract

As Japan's financial system becomes more market oriented, depositor discipline is playing a larger role in the monitoring of banks. Matching household survey data with banks’ financial data, we examine households’ response to bank risk and different deposit insurance schemes. We find that bank switching in response to risk increased between 1996 and 2001 and households’ choice of bank adequately reflects banks’ financial health. We also examine the determinants of households’ knowledge of the deposit insurance scheme and how this affects switching behaviour. The results suggest that depositor discipline works and could play an important supplementary role in bank monitoring.

Acknowledgements

We would like to thank Naohito Abe, Masahiro Hori, Kaoru Hosono, Tokuo Iwaisako, Keiko Murata and Ralph Paprzycki for their useful comments. The main parts of this project were completed while Inakura and Shimizutani were visiting researchers at the Economic and Social Research Institute at the Economic and Social Research Institute, Cabinet Office. An earlier version of this article was published as Inakura and Shimizutani (2005) and presented at the Japanese Economic Association Meeting in June. We would also like to thank the Institute of Economic Research, Hitotsubashi University, for providing us with the micro-data from the NEEDS-RADAR Financial Behaviour Survey. The views expressed in this article are our own.

Notes

¹ For details on the timing of the reinstatement of the ‘pay-off’ scheme for the different types of deposits, see NLI Research Institute (2004). Tsuru (2003) also provided a brief history of recent deposit insurance reforms.

² Basel Committee on Banking Supervision (2004).

³ See Flannery (1998) and Demirguc-Kunt and Kane (2002) for detailed reviews of the literature.

⁴ For example, rather than reflecting bank switching behaviour in response to risk, a decrease in the amount of deposits may be the result of banks’ attempt to improve their capital adequacy ratio to meet BIS standards or household decisions unrelated to changes in financial institutions’ riskiness.

⁵ The Public Survey on Household Financial Assets and Liabilities by the Central Council for Financial Services Information also contains information on households’ responses to the change in the pay-off scheme, but it does not allow us to match the data with banks’ financial information.

⁶ The survey thus includes households from parts of Tokyo, Kanagawa, Chiba and Saitama prefectures. The 2001 survey also covers parts of Ibaraki prefecture. The sampling design is based on two steps. First, municipalities are chosen using stratified sampling based on location, population and other factors. Second, based on household registration, persons aged between 25 and 69 are randomly chosen. Then, each person in the sample is visited and asked to complete a questionnaire.

⁷ It is important to note that although the survey was conducted in the autumn of 1996, the actual survey question still referred to the system in effect before the introduction of the blanket insurance of deposits in the spring of that year.

⁸ If households switched deposits from more than one ‘old’ bank to more than one ‘new’ bank (i.e. from bank A to bank B and from bank C to bank D), they were asked to provide information on the transfer involving the largest amount of deposits.

⁹ The cap on ordinary deposits was scheduled to be reimposed a year later, in April 2003, but this was subsequently postponed to April 2005.

¹⁰ To save space, we omit the table showing bank switching behaviour categorized by the amount of household financial assets.

¹¹ These indicators are the same as those employed by Murata and Hori (2004).

¹² Risk-adjusted nonperforming loans are officially defined as the sum of loans to legally bankrupt borrowers, loans in arrears by 6 months or more, loans in arrears by 3 to 6 months, and restructured loans. Data on risk-adjusted nonperforming loans are not publicly available and we need to estimate these figures. Due to a lack of all the necessary data, risk-adjusted nonperforming loans for 1995 and 1996 are calculated as the sum of loans to legally bankrupt borrowers and loans in arrears by 6 months or more.

¹³ In this analysis, all except one of the shinkin banks are excluded because they cannot be identified. Also excluded are credit cooperatives, labour credit associations as well as several types of financial institutions which do not fall under the deposit insurance scheme, such as foreign banks, agricultural co-operatives, securities companies, life insurance companies and Japan Post. To protect deposits, agricultural co-ops have to join the Agricultural and Fishery Cooperative Savings Insurance Corporation, securities companies are members of the Investor Protection Fund, and life insurance companies join the Life Insurance Policyholders Protection Corporation of Japan.

¹⁴ Concretely, the averages for ‘old’ and ‘new’ banks were calculated as follows: Using the information on households’ ‘old’ and ‘new’ banks, e.g. Household 1: OLD=A, NEW=B; Household 2: OLD=D, NEW=A; Household 3: OLD=C, NEW=A, the average values for ‘old’ banks consist of the financial data for A, D and C, while that of the ‘new’ banks consist of the data for B, A and A.

¹⁵ In contrast with the 1996 survey, the 2001 survey also offered respondents the option to answer they were ‘uncertain’ whether they would switch banks. We removed households that chose this answer from the sample used for this regression.

¹⁶ Since it is possible that households that knew about the deposit insurance reform may have decided not to switch banks simply because they already were with safe banks, it would more sense when examining the effectiveness of the deposit insurance scheme in terms of market discipline to isolate households that were with risky ‘old’ banks and consider the relationship between their switching behaviour and their knowledge with regard to the deposit insurance scheme. However, we obtain information ‘old banks’ only if a household switched banks in response to the change in the deposit insurance scheme. We need information on ‘old banks’ for households that did not but this information is not available. Our dataset contains information on each households’ ‘main bank’ but we cannot discern whether this information refers to households’ ‘old’ or ‘new’ bank.