Why Did Japanese Bank Mergers Boost the Markets?

In 2005 Japan saw the birth of Mitsubishi UFJ Financial Group, one of the country’s biggest bank mergers. Another milestone came in 2018 when Resona Holdings joined forces with other banks. But what really happened on the markets? Researchers examined these two events using four different tools.

1. **Event study**: by comparing stock prices with models like CAPM and the Fama‑French three‑factor model, they calculated cumulative abnormal returns (CAR). The results show a significant price jump right after the mergers.

2. **VAR and impulse response**: these models tested whether changes in one bank affect others. The response graphs indicate that the positive effect spread for months, hinting at synergies among institutions.

3. **Propensity score matching**: by pairing merged banks with non‑merged ones, they estimated the average treatment effect (ATT). Again, the outcome favored the merged banks.

In short, the mergers sparked a wave of optimism: investors rewarded the new entities, and non‑participating banks benefited from a more stable climate. For finance and policy watchers, the data suggest that, at least in Japan, large M&A deals can boost market efficiency without obvious systemic risks.

The analyses in the study rely on advanced econometric methods. The event study uses the market model to estimate a stock’s expected return, subtracting it from the observed return to obtain the abnormal return. Extending to CAPM adds the market beta, while the Fama‑French three‑factor model incorporates size (SMB) and value (HML) dimensions, capturing effects not explained by beta alone.

The VAR (Vector Autoregression) model examines temporal interdependencies among multiple time series, here the returns of involved and non‑involved banks. Impulse response functions (IRF) show how a shock to one bank propagates to others over time, highlighting the duration and intensity of spillover effects.

Propensity score matching (PSM) is a quasi‑experimental technique that reduces selection bias. It calculates the probability (propensity score) that a bank participates in the merger based on pre‑treatment characteristics (size, market cap, liquidity). Banks with similar scores but not merged form the control group, allowing a more credible estimate of the average treatment effect (ATT).

Results show significant positive CARs (p < 0.05) for both mergers, IRFs indicating positive spillovers up to 12 months, and a positive ATT with confidence intervals excluding zero. These indicators suggest that the mergers generated value not only for the merged entities but also for the broader Japanese banking system, without raising systemic instability risk. However, models assume linearity and homoscedasticity; violations could affect the robustness of the findings.