Q: Why do actuaries always come up with such high numbers?

A: I am personally aware of many situations where different consulting actuaries, including myself, have provided estimates that later proved to be too low. This awareness makes me less inclined to think all actuarial projections of reserves and rates are biased on the high side.

On the other hand, as in any profession, there is a wide variation among actuaries in terms of the degree of conservatism (or absence thereof) in the projections they provide. Some have a tendency generally to take a pessimistic view of how things may turn out.

Your opinion could be due to a difference in perspective between clients and consultants. A consultant's experience is with a broad range of clients, including those for which the ultimate losses have turned out both higher and lower than their projections. Clients' experiences, either good or bad, are with their own programs.

Those with only good experience usually wonder how the consultant could justify putting in a provision for the possibility of bad experience, since they haven't seen that in their own data. Those with bad experience are less inclined to criticize actuaries for high projections, though they might if they believe that one bad year of experience was a fluke.

Many factors influence an actuary's selection of ultimate loss numbers, and most of these tend to push the numbers higher.

Although claims history may not show any large losses in recent years, the actuary generally puts in a provision for a large loss or bad overall year once every five, 10 or 20 years or so, depending on the chances of a large loss occurring. Obviously, it is not humanly possible to pinpoint in advance which year is going to develop into a particularly bad one. Therefore, the actuary's goal is one of being correct over the longer term, rather than trying to pinpoint exact losses in advance each year. This also has the beneficial result of smooth loss projections, rather than ones that bounce around substantially each year.

For instance, suppose claims history reflects losses of around $2 million for each of the last five years, with the exception of one year that had losses of $3.5 million. For the next few years, it would be perfectly reasonable for the actuary to include a provision of a portion of that $1.5 million in extra losses (say one-fifth, or $300,000). If the pattern stays the same, the actuarial projection would develop downward by $300,000 for four out of five years. However, for one out of those five years, the projection would develop upwards by $1.2 million. By including a provision for large losses each year, total reserves include a cushion to provide for the year that develops adversely.

As touched on earlier, there is an understandable tendency of clients to view a large loss as a fluke. In contrast, actuaries are likely to see that large loss as something more likely to happen sooner or later. From a client perspective, future large losses may be viewed as being controllable-maybe there are plans to implement a loss control program or to defend claims more aggressively. Thus, the chances of large losses in the future are often viewed in an optimistic light.

An actuary, who deals with a much larger current database of claims, has heard these positive forecasts repeatedly. Yet, as time goes by, the actuary will often see large losses or adverse overall years continue to emerge, regardless of the best efforts of management. An actuary must use a certain level of skepticism in evaluating management projections, until data emerges to support them. For this reason, an actuary will generally be more cautious about how favorable future loss levels might be than a claims manager or administrator.

A claims manager may be thinking, "We won't have any claims larger than $150,000, because that's the largest one we had in the past." One problem with this assumption is that it ignores the effects of inflation. Even with the current low level of interest and inflation rates, loss cost trends play a major factor in the projection of future claims severities. Beyond simple Consumer Price Index comparisons, loss cost trends are affected by medical inflation, changes in litigiousness, changes in claims handling and a host of other factors. Anytime past claims costs are analyzed in relation to current costs, the effects of loss cost trends must be factored in. If that $150,000 loss occurred 10 years ago, today it might be a $300,000 loss-given 7% inflation in claims costs per year.

The other problem with this thinking is that the size of the largest claim of the recent past is subject to tremendous variation-so that using it as a basis for a forecast of the size of future large losses is highly unreliable.

An organization may see a favorable trend in loss rates emerging and want this emphasized in future funding rate selections. The actuary, however, is concerned that this recent trend could be merely the result of good luck in the most recent years or could be caused by decreases in case reserve adequacy, more slowly emerging losses, or other complex factors.

In addition, the provision for incurred but not reported losses-IBNR-for the most current years must be much larger than that for earlier years. The observed favorable trend could be caused by nothing more than inadequate IBNR provision for the most recent past years. For these reasons, the actuary generally is cautious about using newly emerging trends heavily in the projections and may tend to take a longer term average.

Several components go into an actuary's projections of reserves:

* Case reserves, or the total of claim-specific reserves.

* Adverse development on case reserves. Although most cases develop downward, a few large claims generally turn out to be much worse than expected. In fact, the adverse development on these few claims often outweighs the positive development on the rest.

* True IBNR. These are reserves for claims for which the loss has been experienced but for one reason or another the claim hasn't been reported yet.

* Reopened claims. Often, an actuary will include IBNR reserves for an older year for which all claims have been closed. This is because for certain lines of business, most commonly workers compensation, it is common for claims to reopen-sometimes years after they were originally settled.

* Unallocated loss adjustment expenses. Reserves for the general expenses incurred during settlement of claims, such as claims adjuster salaries, office rent, overhead and so forth.

A risk manager's opinion regarding adequate reserve levels or ultimate losses may have been formed with consideration of only some of these components, rather than of all five.

For some previous years, the client may assume that no more loss development is possible-and therefore that no more reserves are necessary. However, the actuary may turn to evidence from outside sources regarding the length of time during which claims may continue to develop. Ignoring the potential of old claims to develop could lead not only to inadequate reserves but to future inadequacy in funding rates. Assumptions regarding additional development beyond what is included in current data are incorporated into what is known as a tail factor. It is not a popular concept but one that is often appropriate.

Credibility may be a problem for organizations with a small claims database. Until a sufficient database of claims has built up in the claims history, an actuary will be reluctant to give full weight to the organization's data, considering it too sparse to be an accurate predictor of future costs. This is another case where the actuary will turn to outside sources or industry data. Often, using industry data will result in higher projected losses than using the organization's past data. However, this may be because the organization's losses are simply too immature to have experienced that inevitable large loss yet. Industry data will have the average large loss for that type of business factored in.

Actuarial estimates may be stated at a confidence level well above expected. If reserves are stated at a 75% confidence level, they were provided with the explicit goal of being high enough that they will develop downward three times out of four. Higher confidence levels are especially important when evaluating self-insurance funds or loss sharing pools, which may not have significant surplus set aside to absorb adverse deviations from expected. If such a pool is funded based on expected loss estimates rather than estimates at a higher confidence level, its current-year funding would be short approximately half the time.

Another potential source of difference in perspective is the "savings on closure" fallacy. An organization may have an average savings on closure of 20%-in other words, cases are settled for an average of 20% percent less than their final reserve. Yet, the actuary may indicate that those reserves are deficient. The problem is that the 20% statistic doesn't look at reserves over the life of the claim. Those redundant reserves are often set up just before settlement. If one compares initial reserve to final cost, or average reserve over the life of the claim to final cost, the answer is often that the reserves were deficient much of the time that the claim was open.

An actuary is in the difficult position of having to prepare unbiased estimates without being influenced by the many conflicting priorities with which he or she is surrounded. I am reminded of an old metaphor: An insurance company is like a car on a treacherous mountain road. The vp of marketing has his foot on the gas. The vp of underwriting has his foot on the brakes. And the president has his hands on the wheel as he takes directions from the actuary in the back seat who has just made a map by looking out the back window.

The problems caused by slightly high projections are generally easier to deal with than those caused by projections on the low side. Low ultimate loss projections normally result in low rate projections, which eventually can cause erosion of a program's fund balance.

After considering all these factors, you may change your opinion regarding whether the actuary is providing high estimates.

But given all the factors above, it is difficult to ascertain at first glance whether a set of projections are too high-a substantial amount of analysis is generally required.

Loss estimates perceived as being on the high side are usually unpopular. However, it is an actuary's task to make sure that a reasoned provision for both good years and bad is included in projections. The future financial health of your program may depend on it.

Would you like advice from an experienced colleague on a risk management, benefits management or actuarial problem? Four quarterly features in the Perspective section of Business Insurance can give you some answers.

Ask A Benefit Manager, Ask A Risk Manager, Ask A Benefit Actuary and Ask A Casualty Actuary answer written questions from readers on risk and benefits management issues and actuarial problems.

This month's column on actuarial issues in the casualty field is written by Richard E. Sherman, president of Richard E. Sherman & Associates Inc. in Ashland, Ore. Dennis J. Nirtaut, managing director of compensation and benefits for Arthur Andersen L.L.P. in Chicago, answers questions for benefit managers. Christopher E. Mandel, director of risk management at Tricon Global Restaurants Inc. in Louisville, Ky., answers questions on risk management issues. William J. Miner, an actuary with Watson Wyatt Worldwide in Chicago, answers actuarial questions on benefits issues.

Address your questions to ASK, Business Insurance, 740 N. Rush St., Chicago, Ill. 60611. Please give us your name, title and employer; however, Business Insurance will consider unsigned letters.