سهم نسبی حقوق صاحبان سهام و سیگنال های تابع بدهی به عنوان پیش بینی پریشانی بانک در طول بحران مالی

Bank supervisors utilize early warning signals to predict which banks are likely to become distressed. Previous research has found that market discipline signals do not significantly improve out-of-sample forecasts relative to accounting-based signals. Most of that evidence, however, comes from periods in the 1990s when the U.S. economy and banking system were healthy, potentially neutralizing an advantage of market signals to incorporate new information quickly. For the period between the fourth quarters of 2006 and 2012, we assess the accuracy of two market signals – expected default frequency (EDF) and subordinated note and debenture (SND) yield spreads – relative to accounting-based signals in forecasting which publicly traded BHCs would become distressed. In 2008, EDF signals were relatively more accurate, but they did not lead to economically significant reductions in missed distress events relative to other signals. Supervisors would have been better off devoting slack resources to monitor BHCs with high commercial real estate concentrations. As the crisis subsided, a failure probability model developed from bank failures in the 1980s and early 1990s was consistently the most accurate signal. For the two dozen BHCs with actively traded SNDs, yield spreads over Treasuries were extremely poor predictors of distress because the spreads were distorted by too-big-to-fail subsidies. The Tier 1 leverage ratio was the most accurate distress signal for these large BHCs. In sum, the evidence to justify systematic reliance on market signals by supervisory agencies to forecast bank distress remains weak.