ریسک نقدینگی و مدیریت پرتفولیو بانک در یک سیستم مالی بدون بیمه سپرده: شواهد تجربی از قبل از جنگ ژاپن

Using data from prewar Japan, this paper investigates the impact of a liquidity shock induced by depositors' behavior on bank portfolio management during financial crises in a system lacking deposit insurance. It is found that banks reacted to the liquidity shock sensitively through an increase in their cash holdings not by liquidating bank loans but by selling securities in the financial market. Moreover, banks exposed to local financial contagion adjusted the liquidity of their portfolio mainly by actively selling and buying their securities in the financial market. Finally, there is no evidence to conclude that the existence of the lender of last resort mitigated the liquidity constraints in bank portfolio adjustments.

One of the most important roles of banks is to perform the maturity transformation of assets by issuing short-term liabilities and holding long-term assets. Therefore, given a system where depositors are insufficiently protected, banking systems are potentially exposed to the probability of bank runs. In fact, during prewar periods, bank runs occurred frequently in the U.S., Europe, Japan, and contemporary emerging countries where deposit insurance was nonexistent. Since Diamond and Dybvig (1983), a considerable number of studies have been conducted on bank runs and panics. The majority of these studies have focused on the origin and causes of bank runs, which can be classified into two alternative views. The first is the random withdrawal theory, which considers bank runs as self-fulfilling phenomena (e.g., Chang and Velasco, 2000, Chang and Velasco, 2001, Postlewaite and Vives, 1987 and Waldo, 1985). The other is the information based theory, which considers bank runs as phenomena induced by the market discipline of depositors under asymmetric information (e.g., Chari and Jagannathan, 1988, Calomiris and Gorton, 1991 and Gorton, 1985). Recently, there has been renewed interest in portfolio management with respect to economies where runs are possible (Cooper and Ross, 1998, Ennis and Keister, 2006, Franck and Krausz, 2007 and Peck and Shell, 2003). These studies, further developing Diamond and Dybvig (1983), analyze how banks manage the liquidity of their portfolios, taking into account the strategic behavior of depositors. Cooper and Ross, 1998 and Ennis and Keister, 2006 examine the relation between the probability of a bank run and the level of liquid assets held by banks. Peck and Shell (2003) investigate how restrictions on the holding of illiquid assets affect the level of liquid assets chosen by banks. Franck and Krausz (2007) analyze the impact of the stock market and the presence of a lender of last resort (LLR) on the portfolio allocations of banks when they are faced with random withdrawals by depositors. Despite such theoretical developments, empirical evidence is lacking on the liquidity management of banks in a system devoid of deposit insurance—at least, since Friedman and Schwartz (1963). These authors demonstrated that the deposit–reserve ratio is inclined to decrease during periods of banking crisis (as in 1907 and the early 1930s). However, they analyzed the time-series behavior of banks on the basis of aggregate data. Moreover, they did not conduct a detailed statistical analysis. Therefore, few details are available on the behavior of individual banks that are exposed to the liquidity shock induced by depositors' behavior in a system without deposit insurance. To investigate portfolio management with respect to banks exposed to the liquidity shock induced by depositors' behavior, this study uses micro-level data pertaining to the prewar Japanese banking industry. In prewar Japan, since numerous small banks operated without deposit insurance, the banking sector was subject to liquidity risk due to the frequent occurrence of bank runs. In fact, nationwide bank runs occurred in the late 1920s and early 1930s, which resulted in numerous bank closures. This paper focuses on bank behavior during the banking crisis from 1927 to 1932. To the best of our knowledge, the first time bank-level analysis is used to examine the liquidity management of banks operating without deposit insurance during past periods of financial crises. This paper examines how banks adjusted the liquidity of their asset portfolios in response to depositors' behavior during periods of banking crises. We specifically analyze the short-term effect of the liquidity shock induced by withdrawals of deposits on the change in the proportions of three assets in its asset portfolio—loans, cash, and securities. Any bank in the economy without deposit insurance is likely to perceive a liquidity shock induced by depositors' behavior, even if it is a small shock, to an increase in its own liquidity risk. Therefore, such banks are expected to immediately raise the cash in hand. In this process, however, the banks will have to choose to liquidate either their loans or securities. In the very short run, it may be difficult to liquidate non-marketable assets such as loans. Furthermore, this paper investigates the impact of financial contagion on bank portfolios. In other words, we examine how banks reacted to the failure of a neighboring bank through portfolio management. Interestingly, this analysis can also be interpreted to capture the effect of a temporary deposit shock on bank portfolio. Finally, this paper examines the influence of the central bank as the lender of last resort (LLR) on the portfolio management of private banks. The existence of the LLR is likely to affect the portfolio management of private banks when the central bank guarantees the supply of loans to them in times of emergency. In prewar Japan, the Bank of Japan (BOJ), which was the central bank of Japan, exercised the role of the LLR by providing loans to private banks during periods of financial crises. Moreover, the BOJ had a well-known tendency to provide LLR loans to banks with which it had an established transactional relationship (Ishii, 1980, Shiratori, 2003 and Okazaki, 2007). The role of the LLR is herein further examined through an analytical comparison of the portfolio management of the BOJ's correspondent banks and non-correspondent banks. This paper is in line with the recent policy discussions on the need to create financial safety nets. It is generally recognized that deposit insurance suffocates market discipline, although it is effective in preventing self-fulfilling depositor runs that induce bank panic. Many financial economists point out the importance of market discipline and propose various safety net designs to guarantee, among other things, its proper functioning (Demirguc-Kunt and Huizinga, 2004, Demirguc-Kunt et al., 2006 and Kane, 2000). However, the cost impact on banks that operate without deposit insurance is not thoroughly understood, particularly in terms of bank behavior.1 In light of these arguments, this paper will attempt to provide useful evidence. The remainder of this paper is organized as follows. Section 2 discusses the historical background of the Japanese banking system. Section 3 describes the study's data and methodology. Section 4 presents the study's main results. Section 5 considers the impact of financial contagion on bank portfolios. Section 6 addresses the role of the LLR and the BOJ. Section 7 concludes the study.

This study empirically examined the portfolio management of the banking industry in response to depositors' behavior during the interwar years in Japan where deposit insurance schemes were inexistent. Concretely, we analyzed the short-term effect of a liquidity shock induced by depositors' behavior on the liquidity of bank portfolios. Regression analyses confirmed the negative effect of a liquidity shock on the cash–asset ratio and the positive effect on the security–asset ratio, which suggested that since the banks perceived a negative liquidity shock to be an increase in their own liquidity risk, they attempted to procure cash quickly through the sale of government bonds or other securities, which could be relatively easily sold in the financial market. By contrast, we could not confirm that such banks adjusted the liquidity of portfolio by liquidating their loans. Furthermore, we examined the effect of local financial contagion on bank portfolios to capture the effect of a temporary deposit shock. Consequently, the banks susceptible to the local contagion adjusted the liquidity of their portfolios by actively selling and buying their securities in the financial market. The effect of the central bank as the LLR on the liquidity adjustments in a bank's portfolio was also examined. However, there was no evidence indicating that the existence of the LLR mitigated the liquidity constraints in bank portfolio adjustments (guarantee effect). From these results, it is safe to say that the security market exercised a significant role in the short-term adjustment of liquidity in bank portfolios in a system without deposit insurance schemes. In other words, the security market acted as a buffer against a banking crisis when deposit insurance schemes were lacking. These results are consistent with Franck and Krausz (2007) who state that security markets are more significant for the liquidity of banks than the LLR. Moreover, our results are considered to have important implications in the design of financial safety nets. For instance, while policymakers have recently demonstrated an interest in creating a financial safety net in which the discipline of depositors functions efficiently, this paper suggests that the development or the reform of the security market should be simultaneously dealt with, in particular with respect to those countries with underdeveloped financial infrastructures. Finally, this paper does not determine the long-term effect of a liquidity shock on a bank portfolio. It is important to examine it in that we can understand the larger picture of liquidity policy in banks. This remains an aim of future research.