The fifth dynamic test (change in the reserving ratio) is derived from a static ratio between reserve increases and renewal and single premium. The dynamic test result is equal to the change in this static ratio between the two years. Therefore, it would be possible to report the static ratios as well as the dynamic test result. We recommend, however, that these static ratios not be included in the early warning report, for two reasons. First, the static ratio does not have, and was never expected to have, the ability to distinguish between troubled and sound companies. We feel that inclusion of this ratio in the report would be likely either to mislead the user into thinking the ratio itself is significant or to confuse him as to the purpose of the report. Second, the user does not need the static ratio in order to understand and interpret results on the dynamic test. If, for example, the dynamic test result is plus 12 percentage points, the user knows that increases in reserves in the current year were greater than the amount they would have been if the reserving ratio had not changed by 12 percent of renewal and single premium. He can even calculate this amount and determine the impact on the company's net gain of the change in the reserving ratio. Thus, the user need not know the ratio of reserve increases to premium in order to understand and interpret results on the test of the change in the reserving ratio.

Four of the recommended tests are static in nature: net gain to total income, investment yield, commissions and expenses to premium, and nonadmitted to admitted assets. For these tests, we recommend that the changes in the ratios from one year to the next not be printed in the early warning report. Printing the changes would not provide any additional information the changes can easily be calculated from the test results themselves. More importantly, the dynamic versions of these tests are significantly less helpful in identifying troubled companies than the tests themselves. Printing the dynamic versions of these tests would be likely to distract the user's attention from the more helpful test results, whereas omitting them would not deprive him of any information.

REPORT FORMAT

A suggested format for reporting test results during 1974 is illustrated in Exhibit XVII. Key features include:

Printing all results for each company on three lines, one for each year

Showing test names and cutoffs for exceptional values

Identifying exceptional values with an asterisk

Rounding results to nearest whole percent except for the investment yield, change in product mix, and change in asset mix

tests

Showing state of domicile and premium.

10

Eleven insolvents would not have been priority companies in one or more of their last four

3 RECOMMENDATIONS FOR 1975

In order to be most useful, the Early Warning System should clearly indicate which companies are most likely to require in-depth analysis and on-site examination. Simply providing the results for each company on the nine tests leaves the user uncertain as to how these results should be interpreted together and how an overall opinion on a company's condition should be formed. Our interviews and questionnaire responses indicate that this problem has in many states been an obstacle to achieving the maximum benefit from the Property and Liability Early Warning System during the past two years. Therefore, we have developed a method by which each life company's results on the nine recommended tests can be used to place that company in either a priority or a nonpriority group. Priority companies would be those most likely to require further analysis and examination.

RECOMMENDED SYSTEM

We recommend that, beginning with test results calculated in 1975, the priority designation be given to companies with exceptional values on three or more of the nine recommended tests. Under this system, all but one of the insolvent companies studied would have been given the priority designation at some time during their last four years. The single exception was Century Life, of Texas, for which test results were available for only the last two years. Ten other insolvent companies would not have received the priority designation in one or more of their last four years; Exhibit XVIII indicates the years in which these companies would and would not have been priority companies. All of the remaining insolvents would have been priority companies in all years for which test results could be calculated. About 88 percent of the insolvent companies would have been priority companies in each of the four years prior to insolvency; 17 percent of the solvent company sample would be priority companies (Exhibit XIX).

Exhibit XX compares this "three-of-nine tests" rule with two- and fourtest rules. If all companies with exceptional values on two or more tests received the priority designation, the percentage of solvents so classified would almost double. If only those companies with exceptional values on four or more tests were priority companies, the percentage of insolvents

EXHIBIT XXI

The recommended system is significantly more effective than the current system....

Half the improvement in effectiveness comes from dropping the ineffective tests

correctly identified would drop as low as 70 percent in the second year prior to insolvency.

Before the recommended priority company system is implemented, we suggest the task force determine exactly how many and which companies would have received the priority designation in 1974. Because our solvent company sample was screened on several criteria, it is not possible to conclude that the percentage of all companies that would be identified would be equal to the 17 percent of the solvent sample. However, we expect that the percentage of all companies identified should be less than 20 percent. Calculation of the results on the nine recommended tests during 1974 will permit the task force to determine whether this estimate is accurate and will also allow the companies that would have been classified as priority companies this year to be identified.

If this analysis reveals that a signíficantly larger percentage of companies than expected (say, 25 percent or more) would be classified as priority companies, the task force should fine-tune the screening system to reduce priority companies to a manageable number. This fine-tuning could be accomplished by: (1) raising the number of exceptional values required for priority identification, or (2) altering the cutoffs for establishing exceptional values. Of course, the best approach would be that which reduces the percentage of all companies identified to a manageable number while maintaining the highest possible percentage of insolvents identified. To assist the task force in this analysis, should it prove necessary, we will provide a tabulation of the test results for the insolvent companies in our sample.

COMPARISON WITH

CURRENT SYSTEM

The discriminating effectiveness of a group of tests can be measured by a graph similar to the graphs used to measure the effectiveness of individual tests. Such a graph would be based on the number of exceptional values out of the group of tests, rather than on the test result on a single test.

Exhibit XXI uses such a graph to compare the recommended system with the current system of 26 tests with cutoffs for determining exceptional values established at the tenth percentile of solvent company results. In order to identify as many of the insolvents in the third year prior to insolvency as the recommended system (87 percent), the screening criterion under the current tests and bench marks would have to be set at two or more exceptional values; this would also throw about 60 percent of the solvent sample into the priority group, compared with only 17 percent for the