An actuarial perspective on P/C loss reserves.
Some self insureds establish an aggregate reserve amount on their balance sheets that is equal to the total of the case reserves established by their TPA. Others have an actuarial evaluation performed of their loss reserve position. The reserves established by this second group of self insureds are normally higher and more accurate for financial reporting purposes than reserves established by the first group. Natural questions to ask include: Why is the actuary setting higher reserves than the TPA? In what sense are the actuarial reserves more accurate than case reserves?
Actuaries do not second-guess a claims examiner's individual case reserves. They look only at aggregate loss data and are not trained to evaluate adequacy of reserves on a case-by-case basis. Actuaries find it convenient to organize the aggregate data by accident period. For instance, if we are interested in loss reserves as of the last quarter of 1989, the actuary might segregate the loss data into accidents that occurred during accident year 1989, accidents that occurred during accident year 1988, etc.
Suppose the actuary has segregated the data, valued as of Dec. 31, 1989, as indicated in Table 1. (To simplify, assume losses are fully closed within five years of their occurrence.)
The total of all case reserves set by the TPA is assumed to be $1.9 million. This figure represents the TPAs best estimate of future payments for all claims that were reported as of the last quarter of 1989. For any individual claim, the TPA would set a reserve based on a thorough review of the file and the facts giving rise to the claim. However, for financial reporting purposes, as of Dec. 31, 1989, the self insured generally must book a liability on its balance sheet, which represents all future payments to be made after that period for all accidents that occurred prior to that date, including costs of settlement such as attorney fees.
Suppose that when all claims for accident year 1989 are eventually closed the total losses paid will be estimated at $2 million. Since only $500,000 has been paid as of Dec. 31, the required loss reserves to be booked on the corporation's balance sheet should be $1.5 million, as contrasted with the $500,000 in case reserves established by the TPA. Actuaries are concerned with establishing a reserve that will represent a proper liability on a corporation's balance sheet.
Table I is a snapshot as of Dec. 31, 1989. Suppose a similar snapshot was taken as of Dec. 31, 1988 (see Table 11). Notice that the 1989 accident year is missing from Table 11 since as of Dec. 31, 1988, those accidents have not yet occurred. Table Ill contrasts loss values as of Dec. 31, 1988, with similar values one year later.
For accident year 1988, the number of claims increased from 1,000 to 1,225 from year end 1988 to year end 1989. Undoubtedly, this is because some claims that occurred toward the end of 1988 had not yet been reported by Dec. 31, 1988. In fact, an additional 225 claims that were incurred during calendar year 1988 were actually reported during calendar year 1989. The emergence of the 225 claims during 1989 probably largely contributed to the increase in total reported losses from $1 million to $1.4 million.
Note that accident years prior to 1988 increased in claim count only a small amount from year-end 1988 to year-end 1989, yet reported losses increased by rather large percentages. In fact, accident year 1985 stayed constant at 990 claims, while reported losses increased from $1.4 million to $1.6 million. The cause of this increase in reported losses must be that some of the claims that were open at Dec. 31, 1988, developed adversely during calendar year 1989. This means that some claims were closed during 1989 for higher values than the case reserves established by the TPA at Dec. 31, 1988. It is also possible that some open claims had their case reserves revalued upward during the year because more specific information became available which indicated the claim would settle for a higher amount.
The information contained in tables I and 11 sheds light on what ultimate losses will be for each accident year. As of Dec. 31, 1989, 12 months has elapsed since that year began. Actuaries say the 1989 accident year is 12 months mature as of Dec. 31, 1989. The 1988 accident year reported losses developed upward from $1 million to $1.4 million as the year went from 12 to 24 months maturity.
One might reason that the 1989 year as of 12 months will increase by a similar percentage as it matured an additional 12 months. Similarly, as the 1989 year matures from 24 to 36 months, one might expect its value to change. This is readily apparent as it can be observed that the 1987 year developed from $1.2 million on Dec. 31, 1988, to $1.3 million on Dec. 31, 1989. There would also be possible development on the 1989 year beyond 36 months.
Loss Development Triangles
Each of the accident years as of Dec. 31, 1989, is said to be immature since claims are still open or unreported. As long as this is true, losses for these years can develop over time. To determine how immature periods will develop, until all years are closed, actuaries typically construct loss development triangles (see Table IV) along with attendant loss development factors. Such triangles display and summarize how older years have matured over time and can be used to predict how data for an immature year will develop.
As shown in the loss development triangle in Table IV, the values from Table I as of Dec. 31, 1989, are displayed along the lower most diagonal. Similarly, the values from Table 11 as of Dec. 31, 1988, are displayed along the next lower diagonal. Every other diagonal displays loss values captured at earlier points in time.
The triangle also shows that for 1983 and 1984, losses do not experience any further development from 60 to 72 months due to the simplified assumption that all claims are closed within five years of occurrence and can never reopen. In reality, actuaries spend a lot of time investigating what happens during the "tail" period (past the time there is observable data).
One can summarize what this data reveals about how losses developed by tabulation of loss development factors in Table V. The loss development factor for 1988 from 12 to 24 months is 1.4. This means that when 1988 matured from 12 to 24 months, losses increased by 40 percent from $1 million to $1.4 million. Similarly, the earlier years increased by other percentages from 12 to 24 months.
If one is interested in estimating what will happen to the 1989 year as it develops from 12 to 24 months, it is reasonable to look at how prior years developed during the same time span. Also, if one is interested in knowing how a year will change from 24 to 36 months, he or she can view how other years changed over this period of time. Average loss development factors are tabulated as follows: 1.379 at 12 to 24 month months, 1.196 at 24 to 36 months, 1.152 at 36 to 48 months, 1.073 at 48 to 60 months and 1.0 at 60 to 72 months. Using these average factors, it can be predicted how each year as evaluated at Dec. 31, 1989, will ultimately settle.
For instance, as of Dec. 31, 1989, the 1989 accident year is 12 months mature. Before it ultimately closes, the average loss development factors project that 1989 will develop from 12 to 24 months by 37.9 percent, from 24 to 36 months by another 19.6 percent, from 36 to 48 months by 15.2 percent and from 48 to 60 months by 7.3 percent. From 60 to 72 months, the development factor is unity (1.0), since all claims as of 60 months are already closed. It is therefore reasonable to consecutively multiply the loss development factors listed above by each other to determine how 1989 will ultimately play out.
Similar analysis is appropriate for the other years, as shown in Table VI. The table implies that the appropriate ultimate loss development factor to apply to 1989 losses as of Dec. 31, 1989, or as of 12 months is 2.039. This means losses will more than double as they mature from 12 months to ultimate. Similarly, 1988 losses at 24 months will grow an additional 47.8 percent and so on for the other years.
The actuarial analysis of loss reserves as of Dec. 31, 1989, can now be completed. First, ultimate losses for each accident year must be determined by applying the previously determined ultimate loss development factors (see Table VII). Finally, to determine required loss reserves as of Dec. 31, 1989, losses paid as of Dec. 31, 1989, must be subtracted from the ultimate loss values (see Table VIII). The required reserves based on the actuarial analysis (see Table IX) can then be contrasted with the case reserves as originally demonstrated in Table 1.
Notice the substantial differences in the two estimates of loss reserves, with the actuarial estimates being significantly higher for all years except 1985 where both methods indicated that the answer is zero since all claims are already closed. Approximately half the difference is attributable to the most recent year and that the difference drops over time.
There are three primary reasons for these differences. First, as of Dec. 31, 1989, many claims that have already been incurred prior to this date have not yet been reported. This is especially true for the 1989 accident year, which is only 12 months mature as of Dec. 31, 1989. From 12 to 24 months, significant development in numbers of claims can be expected. These claims are commonly called incurred but not reported (IBNR) claims. It is impossible for TPAs to predict individual values for these claims. Indeed, such a task is generally not within the scope of their job. The actuarial analysis predicts the emergence of these IBNR claims. Note that the 12 to 24 month loss development factor is large, primarily because of the emergence of IBNR claims during this period.
Second, after 24 months of maturity most claims are already reported for the majority of property and casualty lines of insurance; therefore, there are not many IBNR claims. Nevertheless, substantial loss development occurs after 24 months due to the fact that claims adjusters' estimates of losses become more accurate over time as more specific information relating to the claims becomes available. Actuaries generally observe that the initial aggregate estimates of TPAs regarding the claims values are inadequate. This inadequacy, sometimes called reported but not enough (RBNE) claims, is also built into the actuarial analysis.
Third, some claims that are closed may reopen in the future. These reopened claims are especially prevalent in workers' compensation. Once reopened, a claim is likely to be even more costly than it was prior to its original closure. It is impossible for a TPA to predict which claims will eventually reopen, yet a provision for this possibility should be included in the actuarial analysis.
RBNE claims have probably been the biggest source of contention between TPAs and actuaries. Some TPAs are concerned that actuaries are looking over their shoulders and that an actuarial loss reserve analysis will conclude that adjusters are not doing a good job. However, actuaries do not look at individual claims; they are not interested in which claims are under-reserved and have absolutely no way of making that determination. The data simply shows that as an aggregate value, case reserves are normally inadequate.
My research indicates that TPAs are doing a remarkably accurate job on most claims. That is, the vast majority of claims close at values close to or even slightly under their case reserves. However, a small number of claims are reserved for relatively small amounts and have the propensity to become extremely severe cases, which causes the bulk of the aggregate inadequacy in case reserves.
The fact that this might happen is certainly understandable. For instance, in a typical workers' compensation back injury case, after initial investigation, the adjuster might assume lost time of one month. After four weeks, medical evidence reveals the injury is more severe than originally anticipated and the case reserve now assumes six months lost time. The revisions might continue until after two years it becomes clear the claims have developed into a permanent total case worth more than $1 million.
The claims adjuster may recognize that a small percentage of back injury cases will follow this pattern and develop from an initial case reserve of a few thousand dollars to more than $1 million. They have no way of knowing, however, which ones will blow up and obviously cannot assume they all will. The fact is most cases will not blow up.
A proper loss reserving system is a partnership between the client, TPA and actuary. For most cases, the TPA can accurately evaluate the case reserve. These reserves are necessary to monitor progress of the claim settlement. It is critical for a firm committed to maintaining accurate self-insurance reserves to have professional adjusters and actuaries working on its account.
Interestingly enough, one of the first things actuaries learn is never to share the results of their analysis with the claims adjusters who are working on the same account because actuaries rely on the consistency of loss development triangles to determine loss development factors. If the loss development patterns change, this would substantially complicate the actuarial analysis. For instance, a key assumption of the incurred development technique is consistent case reserves over time. If case reserves were suddenly strengthened, as might occur when an adjuster attempts to correct for perceived case reserve deficiencies, the actuary would need to adjust past case reserves to current adequacy levels before proceeding with the analysis.
Although the aforementioned concepts are applicable to property and casualty lines of business, perhaps they are more relevant to casualty lines. For instance, most property claims are reported soon after their occurrence, thus minimizing IBNR losses. Adjusters have an easier time estimating case reserves for many property lines than for casualty lines, because they are not concerned with liability issues and potential jury awards. Finally, property claims do not have the tendency to reopen as some casualty claims do. For these reasons, case reserves as established by TPAs and actuarial reserves tend to be much closer in property lines than they are for casualty lines. Steven M. Visner is director of casualty actuarial consulting at Coopers & Lybrand in New York. TABULAR DATA OMITTED
|Printer friendly Cite/link Email Feedback|
|Author:||Visner, Steven M.|
|Date:||Sep 1, 1991|
|Previous Article:||The alternative market: shopping on a two-way street.|
|Next Article:||CACE spotlights Colorado's capital appeal.|
|Loss reserving: a fresh look: the difficulty in setting reserves and the risk of insolvency are just two of the many reasons to revisit reserving.|