The Alfred M. Best organization has commented similarly on the 1968 trends in insurance stocks as follows: "The spectacular gains in the property/liability 11"Insurance Stock Trends - 1968," Best's Weekly News Digest, Property/ Liability Insurance Edition, January 13, 1969, pp. 2-3. HM-14 VI. SAMPLING TECHNIQUE Hofflander and Mason state that "unfortunately no mention is made in the report of either the sampling technique used or the representativeness of the sample" (p. 3). They refer here to the sample of 43 property and liability company groups we used to construct the risk/ return point in Figure 8 of the ADL Report (p. 34). In fact, Appendix C of the ADL Report presents a 13-page discussion of the sampling technique and representativeness of the sample. The reviewers have missed a more fundamental point, however. The results we reported about the earnings of the property and liability stock insurance industry were not based primarily on the sample value for the period studied. On the contrary, they were based on industrywide statistics. The use of samples almost always casts some doubt on the validity of econometric results. For this reason, we used the sample of property and liability stock firms only for those statistics not obtainable from industry-wide data. Specifically, the insurance point on the risk/return chart is comprised of a sample of 43 groups 143 companies. However, we based our conclusions about the insurance industry's rate of return not primarily on this sample's average, but on the value for the entire stock industry. Table 4 of the ADL Report (p. 39), to which the reviewers referred, presents all measures of insurance rates of return and is constructed from A.M. Best's total industry population; it is not based on a sample of the industry. We are reassured of the representativeness of our 43-group sample by the fact that its average total rate of return for the 1955-1965 period was 4.4%, while for the same period the Best's industry average value was 4.2%. VII. TIME PERIOD Hofflander and Mason point out that the particular time period we used, 1955 through 1965 (the most up-to-date period for which data were available to us), could have influenced our measure of risk. Once again, they are confounding a temporal with a spatial measure of risk. Temporal measures of risk are severely affected by the choice of time periods. However, there is minimal effect on the interspatial variance of the time period chosen. Professor Webb in his analysis of the Hofflander and Mason review addressed another of their objections to the period studied: "Hofflander and Mason question whether the period of the (Webb's footnote: "For those few who are interested, Hofflander and Mason may have also been addressing themselves to the fact that we used a 16-year period to construct our economy-wide risk/return charts and only an 11-year period for the insurance industry point. In our study when discussing the risk/return charts we noted that: "Lack of necessary data prevented our using the 1950-65 period for the insurance companies. However, analysis of available industry data indicates that the results are insensitive to this difference in time periods" (ADL Report, p. 35, n. 8). That is, 1950-1954 data on individual insurance companies (needed for the risk/return analysis) were not available, overall industry data were; analysis of the available industry data did not change our conclusions. More important, our conclusions were not based primarily on the risk/return analysis of these sample companies. They were based on the total industry data where the most recent 11-year period appeared to be the most appropriate basis for comparison. It was the same basis as that used for the other financial industries. HM-16 VIII. CORRECT MEASURE OF PROFITABILITY In discussing the appropriate measure of rate of return, Hofflander and Mason state that the ADL Report presents "several alternative measures of return, allowing the reader to pick that definition which best meets with his approval. (p. 2) This is a puzzling misreading of our report. In the Summary Report (p. 15) we stated that "the selection of the proper rate of return is of key importance in insuring the economic meaningfulness of this investigation....We feel it mandatory to concentrate on the totality of return to the totality of assets employed in the economic endeavor...." Further, on the very table in the Summary Report which presents the eight rates of return for the insurance industry (Table 3, p. 24), we had the footnote "This measure of return [N4/D2, the one upon which we draw our conclusions] is the one which evaluates overall economic earnings on total economic resources employed. The other measures reported depart from this concept in varying degrees.' We reported all possible rates of return in the interest of good scholarship and with the objective of saving any other researchers the trouble of constructing them from the raw data. We clearly stated that only one of these reported rates of return was appropriate for the study at hand. In the full report we expanded on this point as follows: "The total rate of return (N4/D2) is the only measure It is clear that our study does not allow "the reader to pick that definition which best meets with his approval." 39-280 71 pt. 17A 33 IX. MEASURE OF RISK The reviewers present four arguments against our cross-sectional measure of risk. They prefer a temporal, or time-series, measure. We strongly disagree with all of their arguments. Our results about the property and liability insurance industry did not depend on the precise way we chose to measure risk. We note that the reviewers themselves admit the minor importance of this argument to the discussion at hand. The reviewers' objections to our risk measurements are discussed in greater detail in the Appendix. HM-18 |