Can feedback from the jumbo-CD market improve bank surveillance?

We examine the value of jumbo certificate-of-deposit (CD) signals in bank surveillance. To do so, we first construct proxies for default premiums and deposit runoffs and then rank banks based on these risk proxies. Next, we rank banks based on the output of a logit model typical of the econometric models used in off-site surveillance. Finally, we compare jumbo-CD rankings and surveillance-model rankings as tools for predicting financial distress. Our comparisons include eight out-of-sample test windows during the 1990s. We find that rankings obtained from jumbo-CD data would not have improved on rankings obtained from conventional surveillance tools. More importantly, we find that jumbo-CD rankings would not have improved materially over random rankings of the sample banks. These findings validate current surveillance practices and, when viewed with other recent empirical tests, raise questions about the value of market signals in bank surveillance.