In an article in this journal, Hoyt and Trieschmann (1991) investigated mean returns for life-health, property-liability, and diversified insurers. Their stated goal was the assessment of whether identifiable risk-return advantages existed for investors in each of the three insurer segments during the period 1973 through 1987. their goal is appropriate and their study expands the focus of previous risk-return analyses, it suffers from a basic methodological fault, making the results and conclusions questionable. Focus of the Hoyt-Trieschmann Study According to Hoyt and Trieschmann, their study differs from previous risk-return investigations in several ways. Their study focuses upon a comparison of risk and return for all three segments of the insurance industry: life-health, property-liability, and diversified insurers.(1) Their analysis addresses whether investors gained by purchasing the stock of a diversified insurer rather than separately buying the stock of a life-health insurer and a property-liability insurer. By simultaneously examining all three segments of the insurance market over a similar time period, Hoyt and Trieschmann broadened the scope of previous risk-return examinations. Hoyt and Trieschmann's study examines risk-return issues, applying both the capital asset pricing model and mean-variance approaches, and uses both accounting and stock data. Because they use Value Line data, the market data and the risk measures they used were known by investors. Finally, the Hoyt-Trieschmann analysis incorporates the underwriting cycle for property liability insurers and of rising and falling interest rates, a type of underwriting cycle proxy for life-health insurers, by using time periods that represent complete cycles. Weaknesses Inferences Not Valid The problems with Hoyt and Trieschmann's study arise from an abuse of the sampling process in making inferences about the population of insurers. Hoyt and Trieschmann's analysis of risk-return relationships for life-health, property-liability, and diversified insurers is based upon inadequate sample sizes (eleven life-health insurers, ten property-liability insurers, and nine diversified insurers). Their conclusions regarding the industry segments are speculative, since convenience in data collection seemingly outweighed unbiased and accurate scientific inquiry. Unfortunately, this small sample problem is not unique to Hoyt and Trieschmann. As Cox and Griepentrog (1988, p. 614) stated in this journal: |Hill and Modigliani~ cogently demonstrate the data limitation problems related to P-L insurer market returns when they cite their ten-firm sample of monthly returns for the 1967-1980 period as having exhausted the possibilities of using data from financial markets to infer the risk of underwriting. One should not cast aside the steps in the sampling process necessary for accuracy of measurement and the ability to generalize the results to a larger population simply because the data are not readily available. Hoyt and Trieschmann acknowledge that the samples are small: Although the number of insurers included in the study is relatively small, notice that Standard & Poor's, Dow Jones, A. M. Best, and Value Line all use similar stocks in their indexes for the three insurer segments. Such an explanation is inadequate and akin to the argument, All of the other kids do it. In a published academic analysis, one expects an unbiased approach based upon the necessary requirements of the methodology employed. Researchers should not force through a method or model because of data constraints. insurance researchers are often tempted to utilize mainstream finance approaches, one must remember that financial return data are quite limited because of the confounded ownership of many stock insurers and the mutual form of ownership. There are not enough insurers for which share price or return data are available to make inferences about the overall insurer population. …