A line-by-line case study of two high-priced portfolios, comparing fixed, variable and capital costs against forecasts of regional market prices.

A multi-billion-dollar wave of utility...

Ratings Reveal Risk

In an interesting article (Oct. 15, 1994), Professor Brooks Marshall suggests that bond ratings are a poor predictor of equity risk, based on a regression of utility risk measures on measures of utility bond ratings. For utility variables, he used 1) stock beta and 2) risk premiums based on analysts' expected forecasts. For bond rating measures, he used 1) average yields for various classes of bonds and 2) a numerical ranking. I suggest that Professor Marshall's measures are misspecified, causing him to understate the strength of the bond rating/equity risk relationship.

With the recent demise of the capital-asset pricing model (see Journal of Investing, Fall 1994, p. 10), investment professionals no longer consider beta a reasonable measure of total equity risk. Value Line, for example, prefers its Safety Rank measure. Interestingly, there is no statistically significant relationship between beta and Safety Rank for utility stocks. Therefore, testing the degree to which bond ratings explain variation in beta tells us little about bond ratings' ability to explain total equity risk.

Is Professor Marshall's risk premium a reasonable measure of equity risk? Unfortunately, the answer here is also no. Investment professionals and financial academics find risk premium calculations, especially those based on analysts' forecasts, of questionable reliability (see Forbes, Oct. 10, 1994, p. 154). Again, Value Line's Safety Rank is a far superior measure.

Further, good econometric practice argues against Professor Marshall's use of an index variable (that is, 1=AA+ to 9=BBB-) to measure bond rating. Such an index imposes an overly restrictive linear constraint on the bond rating variable. Professor Marshall's approach assumes that the difference in risk between AA and A rated utilities is the same as that between A and BBB rated companies. That is not necessarily the case, nor does financial theory suggest such a relationship.

To improve upon Professor Marshall's analysis, I conducted the following analysis based on data for Value Line's electric utility stocks. For the total equity risk measure, I used Value Line's Safety Rank. (This measure ranges from 1 for lowest-risk stocks to 5 for highest-risk stocks.) For the bond rating measure, I used a simple two-variable indicator system based on utility bond ratings. (Statistical methods permit two independent (x) variables to define three categories.)

Bond Rating Dependent Variable X1 X2

AA or above Safety Rank 0 0

A Safety Rank 1 0

BBB or below Safety Rank 0 1

By using only three bond rating categories, instead Professor Marshall's nine, I actually make it more difficult for my model to explain the utility's Safety Rank. Nevertheless, I find this approach a more reasonable reflection of the precision of bond ratings. By separating the categories into two distinct variables (instead of Professor Marshall's one), however, I significantly improve the regression model's ability to find the true relationship between bond ratings and equity risk as one moves between rating categories.

Fitting the regression model above to the electric utility data reveals that the bond rating explains 45 percent of the variation in electric utility stocks' total equity risk. Professor Marshall's analysis suggested that the