Japan Bank Competition & QE Impact

The late 1980s witnessed a remarkable period of high economic growth in Japan, characterized by low risk premiums and nearly zero inflation, according to Yamaguchi (1999). Real GNP per adult growth significantly exceeded the 2% benchmark observed in the United States' long-run growth rate (Hayashi and Prescott, 2002). Throughout the 1980s, robust GDP growth was consistently recorded, starting at 2.82% in 1980 and culminating at 5.37% in 1989, reaching its peak at 7.15% in 1988. This exceptional economic performance led to a surge in asset prices and a rapid expansion of credit. The flourishing economy created an environment ripe for financial expansion and risk-taking.

Concurrent with Japan's economic boom, the country underwent significant financial deregulation and liberalization. This included measures such as the relaxation of interest rate controls, deregulation of capital markets, and the lifting of restrictions on banking activities, as detailed by Kanaya and Woo (2000). These policy changes fundamentally altered the financial landscape, facilitating a dramatic expansion in bank lending. The removal of regulatory constraints spurred rapid credit growth, ultimately contributing to the asset price inflation that would characterize the pre-crisis period and lay the groundwork for the subsequent financial instability. The ease of access to credit fueled economic expansion but ultimately proved to be a precursor to significant challenges.

The rapid economic expansion of the 1980s, driven by strong GDP growth and fueled by financial deregulation, created a climate of seemingly low risk. However, this period of prosperity contained the seeds of future instability. The dramatic rise in asset prices and the rapid expansion of credit, while indicative of a healthy economy in the short-term, created vulnerabilities that would later come to the fore. The seemingly low risk premium masked underlying risks that would eventually lead to the Japanese banking crisis of the 1990s. This period serves as a cautionary tale of the potential downsides of rapid growth fueled by deregulation in the absence of sufficient regulatory oversight and risk management.

The Japanese banking crisis was fundamentally characterized by a dramatic escalation in nonperforming loans. This unprecedented increase in problem loans posed a severe threat to the stability of the entire banking system, forcing numerous banks and financial firms into bankruptcy. The aftermath of the bursting land and asset price bubble saw a sharp rise in problem loans as a significant number of firms faced financial distress or outright failure (Giannetti and Simonov, 2013). The sheer volume of these nonperforming loans created a systemic risk, threatening the solvency of financial institutions and triggering a domino effect of failures. The crisis highlighted a critical weakness in risk management and the interconnectedness of the financial system.

Several high-profile bank failures marked the progression of the Japanese banking crisis. The jusen, specialized housing loan corporations, were among the first institutions to experience significant difficulties, ultimately leading to their liquidation in 1996 (Hoshi and Kashyap, 2010). Hyogo Bank became the first listed bank to fail in 1995. By 1996, the Deposit Insurance Corporation extended its guarantee to cover all deposits in Japan, reflecting the severity of the situation. Subsequent failures included Sanyo Securities in 1997 (triggering the first interbank loan default), Hokkaido Tokushoku Bank, Tokuyo City Bank, and the nationalization of the Long-Term Credit Bank of Japan and Nippon Credit Bank in 1998. These failures underscore the systemic nature of the crisis and the ripple effects of individual bank failures across the financial sector.

The financial cost of the Japanese banking crisis was immense. The cost of bankrupt and restructured loans in 1997 alone reached 30 trillion JPY (Hoshi and Kashyap, 2000), with some estimates placing the true figure well above 100 trillion JPY (Hoshino, 2002). While the level of bad loans eventually decreased after March 2002 due to a reform program led by Heizo Takenaka, the Minister in charge of the Financial Services Agency, the government's response was criticized for its slowness and indecisiveness (Caballero et al., 2008). The initial reluctance of the Ministry of Finance to utilize public funds and statements downplaying the crisis' severity contributed to the crisis' prolonged duration. This highlights the significant costs associated with delayed and inadequate government intervention in addressing financial crises.

The prolonged distress in the Japanese banking sector was partly due to structural issues within the financial system. The bank-centered system, often described as the horizontal keiretsu or main bank system, relied heavily on cross-shareholdings and information monopolies held by banks (Hanazaki and Horiuchi, 2003; Kanaya and Woo, 2000). This “relationship banking” model, while fostering strong ties between banks and borrowers, also impeded objective creditworthiness assessments by third parties. When main banks themselves faced distress, the incentive to address nonperforming loans directly was reduced due to reputational concerns and the added costs of covering losses for other creditors. This resulted in a preference for forbearance rather than decisive action regarding problem loans, thus prolonging the crisis.

The Japanese banking sector is distinguished by its long-term relationships between manufacturing firms and financial institutions within the horizontal keiretsu network. This network emerged after the dissolution of the pre-war zaibatsu, large family-centered conglomerates that controlled a substantial portion of the economy's capital assets (Hadley, 1970). The keiretsu system involves interconnected businesses across various industries, organized around a central main bank holding equity in network firms. Post-war, two main types of horizontal keiretsu emerged: those descended from the pre-war zaibatsu (Mitsubishi, Mitsui, and Sumitomo) and those bank-centered (Fuyo, Sanwa, and Dai-Ichi Kangyo banks). This system created a unique internal capital market and an insurance mechanism between corporate businesses (Nakatani, 1983), offering mutual assistance and reducing information asymmetry, but also potentially hindering efficient allocation of capital and increasing systemic risk.

The Japanese banking system encompasses various bank types, each with distinct characteristics and client bases. City Banks played central roles as main banks in post-war horizontal keiretsu networks, providing funding and board connections to member firms (Nakatani, 1983; Horiuchi et al., 1988). Examples of their influence include the relationships between Sumitomo Bank and Mazda Motors, and Mitsui Bank and Mitsukoshi department stores. In contrast, Regional Banks primarily serve local small and medium-sized enterprises (SMEs), supporting local economic development. Trust Banks combine financing with asset management services, relying on trusts as their main funding source (Drake and Hall, 2003). Finally, Cooperative Banks cater to households and local SMEs, promoting community development with specific membership restrictions and management practices (Barros et al., 2009). This diversity in bank types reflects the unique structure of the Japanese economy and financial system.

The keiretsu system played a significant role in the Japanese banking crisis. While the network's mutual assistance and information sharing features provided risk-lowering advantages (Khanna and Yafeh, 2005), they also contributed to the problem of nonperforming loans. The close relationships between main banks and their client firms, coupled with information asymmetry and the cost of allowing defaults, led to forbearance on problem loans (Kanaya and Woo, 2000). The main banks' roles as both creditors and equity holders created conflicts of interest that hindered timely resolution of financial difficulties. The network effectively insulated weak firms from bankruptcy at the expense of stronger members, creating a systemic vulnerability that was exposed during the crisis. This exemplifies how seemingly beneficial inter-firm relationships can generate systemic risk in a tightly interconnected banking system.

In response to the escalating problem of impaired loans stemming from the burst asset price bubble, Japanese authorities implemented a series of restructuring packages aimed at restoring the financial health of the banking system. A key element of this strategy involved capital injection programs, implemented five times between March 1998 and March 2009. Under the 1998 Financial Revitalisation Plan, a substantial portion of the public funds injected was used to safeguard depositors of insolvent banks and acquire their assets (Montgomery and Shimizutani, 2009). These bailouts and capital injections provided crucial support for struggling banks, helping to prevent further collapse and limit the impact on the wider economy. While necessary to prevent a more devastating crisis, these measures also came under intense scrutiny.

Further measures undertaken by the Japanese government included forcing banks to liquidate poorly performing company shares in 2002. However, the Bank of Japan ended up purchasing many of these shares, highlighting the complexities and challenges in effectively managing the restructuring process. The government also introduced accounting changes that allowed banks to report either book or market values for their stock and real estate holdings. Initially this inflated the value of bank assets but in 2001, the government reversed this, switching back to book values (Hoshi and Kashyap, 2010). These fluctuating accounting practices reveal the challenges in implementing consistent and transparent policies during a period of severe financial stress, ultimately impacting market confidence and the effectiveness of the restructuring efforts.

Despite the government's efforts, its response to the crisis faced significant criticism. The initial slow response and denial of the severity of the nonperforming loan problem significantly prolonged the period of disruption (Giannetti and Simonov, 2013; Hayashi and Prescott, 2002; Hoshi and Kashyap, 2010). Before 1997, the Ministry of Finance resisted using public funds, and even in February 1999, the Vice Minister of International Finance predicted a swift resolution, demonstrating a lack of understanding of the crisis's depth. The government's policy of promoting lending to SMEs after 1998, while intending to boost the economy, was criticized for its support of unprofitable borrowers (“zombie lending”), hindering the necessary restructuring and recovery of healthier firms and banks (Caballero et al., 2008; Hoshi and Kashyap, 2010). This policy inadvertently prolonged the crisis and slowed economic recovery.

Government intervention, while ultimately successful in mitigating the crisis, also led to an unintended consequence. The government's focus on supporting SMEs resulted in a shift in problem loans, moving from a concentration in real estate lending to SME financing. Regional Banks, particularly Regional Banks I and II, had a higher ratio of problem loans to assets compared to City Banks. This underscores how government policies, even those designed to address a crisis, can unintentionally create new challenges and exacerbate existing regional disparities. The data supports the argument that the shift to supporting SMEs, while well-intentioned, exacerbated the problem loan issue in certain sectors of the Japanese banking system.

This study employs a novel approach to measuring bank efficiency in Japan, addressing the limitations of previous research. Unlike earlier studies, this research utilizes a translog enhanced hyperbolic output distance function (Cuesta et al., 2009) to simultaneously consider desirable outputs (e.g., loans) and undesirable outputs (problem loans). This innovative approach allows for a more comprehensive assessment of bank performance by accounting for the negative impact of problem assets. The model incorporates two undesirable outputs: problem loans (consistent with the Financial Reconstruction Law’s classification of problem assets) and problem other earning assets, reflecting the broader investment activities of Japanese commercial banks, which include government bonds, corporate bonds, securities, guarantees, and acceptances. This approach, previously absent in the literature, provides a more accurate measure of bank efficiency by explicitly accounting for the detrimental impact of problem assets.

A key focus of this research is examining the impact of bankrupt and restructured loans on bank efficiency. This is a unique contribution, as prior studies haven't explored this specific relationship in detail. The study leverages a unique database to distinguish between bankrupt loans (loans to borrowers in legal bankruptcy and past-due loans by 6 months or more) and restructured loans (past-due loans by 3 months but less than 6 months and formally restructured loans). These loan categories, disaggregated from the data on risk-monitored loans of Japanese commercial banks, are used to measure the level of risk held within Japanese banks. To account for potential endogeneity concerns, a panel Vector Autoregression (VAR) model is employed, allowing for causality testing between bankrupt/restructured loans, bank-specific variables, macroeconomic factors, and technical efficiency. The study tests four hypotheses: ‘bad luck,’ ‘bad management,’ ‘skimping/moral hazard,’ and ‘risk-averse management’ (Berger and DeYoung, 1997; Koutsomanoli-Filippaki and Mamatzakis, 2009).

This research differentiates itself from prior empirical studies on Japanese bank efficiency in several ways. First, its innovative approach to estimating bank efficiency utilizes a more sophisticated parametric approach that handles both desirable and undesirable outputs. Second, the use of semi-annual data and the inclusion of problem other earning assets, besides problem loans, provides a more comprehensive perspective. Previous studies often treated nonperforming loans as uncontrollable inputs (Drake and Hall, 2003; Hughes and Mester, 2010), quality variables (Hughes and Mester, 1998), or simply undesirable outputs (Assaf et al., 2013; Barros et al., 2012; Berg et al., 1992; Fukuyama and Weber, 2008; Glass et al., 2014; Mamatzakis et al., 2015; Park and Weber, 2006). This research uses a more granular dataset focusing on bankrupt and restructured loans, thus providing a more nuanced understanding of risk and its effects on bank performance. This enhanced data and methodological approach lead to a more precise and comprehensive analysis of bank efficiency in the context of the Japanese banking crisis.

The analysis uses the Boone indicator as a primary measure of bank competition, offering advantages over other metrics like concentration ratios (e.g., Herfindahl-Hirschman Index). Unlike concentration ratios which only capture market share and can be misleading regarding actual competitiveness (Beck, 2008; Schaeck and Cihák, 2014), the Boone indicator accounts for entry barriers and firm interaction (Boone, 2008b). The study also notes limitations of alternative measures such as the Panzar-Rosse H-statistic (requiring assumptions about long-run equilibrium) and the Lerner index (ambiguity regarding product substitutability) (Panzar and Rosse, 1987; Vives, 2008; Mirzaei and Moore, 2014). The choice of the Boone indicator reflects a desire for a robust and nuanced measure that captures the dynamic nature of competition within the Japanese banking sector during the period of study.

Bank risk is measured using multiple proxies: ratios of bankrupt and restructured loans to total assets, and the natural logarithm of the Z-score. Bankrupt loans are defined as loans to borrowers in legal bankruptcy plus past-due loans by six months or more. Restructured loans include past-due loans of three to six months and formally restructured loans. These ratios, derived from risk-monitored loans reported under Japanese banking law, capture credit risk, similar to commonly used nonperforming loan ratios in competition-fragility analyses (Beck, 2008). The Z-score, representing overall bank stability, is incorporated to gauge the likelihood of bank failure (Beck et al., 2013; Laeven and Levine, 2009). Using multiple risk proxies allows for a more thorough investigation of the relationship between competition, quantitative easing, and bank risk, accounting for different dimensions of risk exposure.

The study investigates the impact of quantitative easing on bank risk and competition. Bank-specific lending rates are used as the primary proxy for quantitative easing at the bank level. This microeconomic approach avoids aggregation bias and ensures compatibility with the bank-level Boone indicator and risk measures. The choice of the bank-specific lending rate is justified by the inoperability of short-term interest rates under the zero lower bound policy (Girardin and Moussa, 2011) and the limitations of using aggregate measures like the Bank of Japan loan rate, uncollateralised overnight call rates, total reserves, asset purchase amounts, or government bond yields (Girardin and Moussa, 2011; Lyonnet and Werner, 2012). For robustness checks, the study also employs the 10-year Japanese government bond yield and the Bank of Japan's total assets as alternative proxies for quantitative easing.

To analyze the complex relationships between competition, quantitative easing, and bank risk, the study uses a dynamic panel threshold analysis with Generalized Method of Moments (GMM)-type estimators. This addresses endogeneity concerns (Kremer et al., 2013). The dynamic panel threshold model allows for examining the stability of relationships across different regimes defined by threshold variables, such as the Boone indicator or a quantitative easing proxy. The study also employs a panel vector autoregression (VAR) model to explore the underlying causality and potential endogeneity among these three key variables. This multifaceted methodological approach enables a robust examination of the intricate interactions between these factors, controlling for bank size as an exogenous variable, particularly important given the heterogeneity of the Japanese banking sector.

The findings of the study suggest a positive relationship between bank competition (as measured by the Boone indicator, above a certain threshold) and lower levels of bankrupt and restructured loans, supporting the competition-stability hypothesis. This contradicts the competition-fragility hypothesis found in some studies using Z-scores as a risk proxy (Liu and Wilson, 2013). The study's use of bankrupt and restructured loan ratios, directly reflecting the Japanese banking context, is deemed more reliable than Z-scores, which are subject to limitations (Demirgüç-Kunt and Detragiache, 2011; Cihák and Hesse, 2007; Lepetit and Strobel, 2013). Regarding quantitative easing, the study indicates a complex and potentially negative relationship with technical efficiency but a positive relationship with ROA, suggesting a trade-off between risk and profitability. These findings are discussed in the context of potential moral hazard and the overall impact of monetary policy on the Japanese banking system.

This section of the study delves into the regional variations in bank performance within Japan. The research utilizes convergence cluster analysis and the club convergence test developed by Phillips and Sul (2007) to examine convergence in total factor productivity growth and its components across different regions. This methodology allows for the detection of convergence even within subgroups, providing a more nuanced understanding of regional trends in bank productivity. The analysis considers various factors that might influence regional differences, including the distribution of bank types (City Banks, Regional Banks I & II), the level of competition, and the impact of macroeconomic shocks. The research aims to reveal whether convergence exists in bank productivity across Japan and to identify any regional disparities.

The study employs the club convergence test to analyze whether convergence exists in total factor productivity (TFP) growth among Japanese banks, considering regional variations. This test is particularly useful because it allows for the identification of convergence within subgroups even if overall convergence is rejected for the entire sample. The analysis identifies convergence clubs, representing groups of regions exhibiting similar patterns of TFP growth. The results reveal the presence of convergence clubs, suggesting that certain groups of regions exhibit similar trends in productivity growth, while others diverge. This highlights the importance of considering regional factors when examining the performance and efficiency of Japanese banks.

The analysis explores the characteristics of the identified convergence clubs, relating them to geographic proximity, economic activity, and the structure of the banking sector. The study notes, for instance, the adjacency of Kanto and Chubu regions within a convergence club, suggesting potential spillover effects through the branch networks of Regional Banks (Kano and Tsutsui, 2003). The significant economic contributions of Kanto (with Tokyo as a major center) and Kansai (with Osaka-Kobe) are highlighted, in contrast to other regions like Kyushu (Fukuoka) and Hokkaido (Sapporo). The role of regional stock exchanges in influencing regional convergence or divergence is also considered. Understanding these regional nuances provides valuable insights into the factors driving productivity growth and its dispersion across Japanese banks.