Can statistics based early warning systems detect problem banks before markets?

摘要

Many statistical early-warning models have proven to have some predictive power. These models involve five basic approaches: logit, discriminant analysis, proportional hazard models, trait, and robust regression.;If markets are at least semi-strong form efficient, then prices must already incorporate any information that could be obtained by using these statistical early warning systems. In this case, either early warning systems do not have special predictive power, or the information they provide is quickly obtained by markets, probably through industry analysts who utilize such models in their analysis. If these systems can be used to earn abnormal profits, then the efficiency of equity markets is called into question.;In this dissertation, I utilize these five early warning systems to find problematic banks using data from 1986 through 2009. A zero cost arbitrage portfolio is formed each quarter by shorting the banks identified by the models as potential problems and going long the remaining non-problematic banks in the sample. The risk adjusted returns on the arbitrage portfolio and its long and short components is compared to risk adjusted returns on a long portfolio of all banks in the sample. If the returns on any of these portfolios are statistically greater than the "all bank" we can infer that the early warning system is able to provide information not available to investors and can conclude that the market is not semi-strong form efficient.;I find that using market returns for portfolios formed by bank EWS is a viable universal standard to judge their ability to discern problematic banks and conclude that newer and/or more complex EWS do not perform better than older and/or simpler models over long periods of time. Only two of the models are able to beat the naive all bank portfolio on a risk adjusted basis over the entire term and none are able to beat the market on a risk adjusted basis, but all are able to form a long portfolio able to screen out some underperforming stocks and so beat a naive strategy on an unadjusted for risk basis. From this, I conclude that the market for publicly traded commercial banks is highly, but not perfectly, semi-strong form efficient.