The Too-Big-to-Fail Doctrine vs Bageoht (1873)'s Rule : the case of the panic of 1927
I have uploaded my new discussion paper to SSRN.
The paper examine determnants of bank bailouts by the bank of Japan in response to the panic of 1927 using a data set consisting of observations on 1364 ordinary banks. The results by tobits how that the Bank of Japan provided liquidity support not for lenders but for borrowers in the call market, and that rescued banks had higher closure risk and more deposits. Bank bailouts in response to the panic of 1927 was to practice the "too-big-to-fail" doctrine on the oremise that lenders in the call market would recover liquidity performance by the self-help efforts.
The major challenge in recent studies on liquidity shocks is the lack of detailed data on interbank activities during the crises. I notice a handbook for bankers at the time. While there is no centralized record-keeping on bank activities in the interbank market during the panic of 1927, the handbook has listed banks for three categories: dashite (lenders in the call market), torite (borrowers), and dashite-oyobi-torite (borrowing-lenders). Based on these three lists, I have created group dummy variables as indicators of activities in the call market during the panic.
There is another problem of exploraring determinants of the LLR (lender-of-last-resort) assistance. Decision process of the LLR assistance maight cause the sample selection problem. As a initial selection mechanism, banks had to decide whether to apply for the LLR support. In the next stage, the LLR would decide whether to rescue. There could be three bank groups: (Y1)banks that did not apply for the LLR assistance; (Y2)banks that were declined the assistance; and (Y3)rescued banks. If the empirical studies ignore an initial selection mechanism, they might lead to misspecification since the untreated group might comprise more healthy and more unhealthy banks. In this situation, the treatment-effects regression could be useful (Vossmeyer 2016). On the other hand, the data on applicants for toku-yu are not available. The empirical studies need other alternatives. I use the propensity score to close during the panic period as the proxy on applying the LLR assistance.
The Results are not consistent with those of empirical studies by Okazaki (2007). My paper points out that Okazaki (2007) has faced the sample-selection problem. Okazaki (2007) emphasized that the Bank of Japan adopted Bageoht (1873)'s rule after the panic of 1927. This conclusion is wrong. The Bank of Japan did not provide liquidity support for sound banks.
