Financial innovations and excesses revisited: the case of auction rate preferred stock - Security Design Special Issue

Financial Management (Financial Management Association), Summer, 1993 by Michael J. Alderson, Donald R. Fraser

It is reasonable to expect that the time series of CP percentages for any particular issue would be influenced by changes in both the perceived risk characteristics of the shares and the explicit tax levied on their dividends. To control for changes in the CP percentage brought about by tax changes, we constructed an adjusted CP percentage variable which measures the ratio of the observed CP percentage to the no-arbitrage CP percentage implied in a risk-neutral framework.

Exhibit 3 contains the results from the logit regressions relating firm characteristics to the decision to redeem an TABULAR DATA OMITTED issue of auction rate preferred stock. For each model, the coefficient estimates from the logit regression are displayed along with their respective asymptotic (large sample) p-values, the p-value for the model and the pseudo R-square measure (which is based upon the differences between the actual frequencies in each subgroup and the estimated frequencies obtained from the regression model). Overall, the estimated coefficients are consistent with the hypotheses that redemption activity in the auction rate market was encouraged by both investor aversion to high-risk issues and by the thrift reform process.

For redeemed issues, the cost of preferred equity was measured by the last available CP percentage just prior to redemption; for issues still outstanding, it was estimated by the latest available CP percentage in the first quarter of 1991. The first set of coefficient estimates (model 1) in Exhibit 3 indicate that the decision to redeem was more likely the higher the cost of preferred equity. The estimated coefficient for the final dividend percentage was 0.031, with an asymptotic p-value of 0.048. Similar results were obtained in the second model, where the cost of preferred equity was measured by the adjusted final dividend percentage.

To test for the influence of the thrift reform process on the redemption decision, we included a dummy variable to identify limited purpose subsidiaries in the logit analysis (model 3). The dummy variable was assigned a value of one if Moody's Bond Survey or other sources of information such as the Wall Street Journal indicated that the shares were issued by a bankruptcy-remote, limited-purpose subsidiary created specifically for the purpose of issuing auction rate preferred stock. Separate consideration of the limited-purpose subsidiary issues was necessitated by the fact that their collateralized structure generally commanded a lower dividend yield than comparable (Aaa-rated) non-credit-enhanced issues.

As shown in Exhibit 3, each of the coefficient estimates in this expanded model are significant at the five percent level. The estimated coefficients were also significant when the adjusted dividend yield at issue was employed to measure the cost of auction rate equity (model 4). This result is particularly intriguing, because it appears to indicate that the market has selected against issues that were identified as high risk at the time they were sold. The results from both specifications provide evidence consistent with the hypothesis that the decision to redeem was also more likely for issues which were affected by the thrift reform process.