In a recent issue of this Journal Professor Forbes set out to estimate the effect of access to stock markets on the investment results of a sample of non-life insurers.1 His estimates of the effect of reporting all capital gains and losses on the insurers' rates of return and on the standard deviations of firms' average 1956-72 rates from the group average are both appropriate and interesting. They do in fact reflect only the effect of a change in reporting rather than "access to the stock market", however, and thus do not support conclusions such as "investors in the stock insurers would have fared better on the average if these insurers had not entered the stock market during 1956-72" (631). More serious questions are raised in trying to interpret the regression results reported. While the technique in general of fitting an average "risk-return tradeoff line" for different groups of companies (and then comparing their slopes) is an excellent one, Professor Forbes' definition of risk is quite difficult to interpret. As far as this reader can discern, the measure of risk for each company is the standard deviation of the company's annual rate of return for each of the years 1956-72 from the company's average 1956-72 rate of return. The difficulty is that the average 195672 rate of return is a very poor indicator of the expected rate of return for each of those years, since Professor Forbes notes that the period is "representative of a wide variety of conditions in investment and underwriting markets" (626). What is needed instead is some sort of proxy for "market conditions" in those markets, such as the average rate for all insurers in each year. A simple example will illustrate the data problem this raises. Firm A operates each year so as to earn exactly the rate of interest offered on 10-year (Canadian) government bonds. Firm B manages to operate each year so as to earn at all times the 1956-72 average of that bond yield. Both therefore have the same average yield for the period as a whole. Firm B has a risk measure of exactly zero, despite having earned far more than the bond rate in the early years and far less in the later years. Firm A has a risk measure of 1.25 percent despite having copied the behaviour of what is referred to as the riskless asset.