In My Opinion

Elections 2000: An Actuarial Perspective by C.K. "Stan" Khury

The events surrounding the final resolution of the 2000 presidential election are widely acknowledged to be unique in the history of the American republic. The twists and turns leading to the ultimate resolution are well known and require little additional comment. However, one aspect of the story deserves special consideration—especially when viewed from an actuarial perspective.

During the long five weeks following the election it became increasingly clear that, despite public expectation and despite all appearances to the contrary, the act of counting votes and deciding the presidential victor is a process that is based on approximations. This reality, clearly present and operating for generations, somehow had never focused in our consciousness, notwithstanding a great deal of evidence plainly pointing to it. Consider, for example, that absentee ballots are not counted unless they could change the outcome of the election and that millions of ballots are incapable of being counted by the counting equipment (for a host of reasons including voter error, voting machine malfunction, and counting machine malfunction). The "problem" does not emerge fully unless and until there is a very close race such as the Florida voters produced, and election officials seek to replace an approximation with a precise result. This desire, in the case of Florida, spawned a web of "protests," "contests," and countless invocations of local election laws, federal voting rights laws, and constitutional arguments. And, of course, as we all saw, none of these methods of converting an approximation into a precise final result was particularly satisfying to either the contestants or their supporters.

Which brings us to the actuarial reserving paradigm. A reserve is an estimate—and at that an estimate of some fixed but unknown ultimate amount. Such an estimate is often accompanied by a statement along the following lines: "This estimate is subject to numerous uncertainties and errors of estimation—and actual development may turn out to be higher or lower than this estimate and may be much higher or much lower than this estimate." Imagine the Florida officials describing the Florida results using the actuarial jargon: "The ballots have been counted and recounted and it appears that Bush has won the election. When every ballot has been recounted and the clear intention of voters has been ascertained, the results may be different (for example, Bush still wins the election but by a different margin than that of the first approximation) or the results may be very different (for example, Gore wins the election)."

If an actuary who produced a reserve estimate is asked to be more precise, the actuary may be able to resort to different and more elaborate methods of estimation that may or may not improve the precision of the estimate. But, at the end of the day, we can be certain that the actuary could never remove the uncertainty inherent in the estimation process.

Several attempts have been made to remove the election uncertainty, but after many legal battles on numerous fronts, the approximation had to stand as an approximation, much to the dismay of the Gore supporters. Thus a (political) life and death situation was decided on the basis of an approximation.

The contemporary mathematician John Allen Paulos (author of Innumeracy) commented on the impossibility of determining the victor in the Florida election because the potential margin of error (of estimation) is greater than the margin of victory. In other words, no matter what is done to determine the victor, the process is so imprecise that whoever was declared the winner, by whatever process, cannot be the absolute winner as the margin of error can overwhelm the result of any counting one may undertake.

The actuarial equivalent of this commentary is something that is relatively obvious to each of us as a result of our training and experience: an actuarial estimate, given enough uncertainty, can be overwhelmed by the magnitude of the variability inherent to the estimation process. We do our best to disclose this condition by the use of appropriate caveats. However, at the end of the day, a single number is entered on the financial statement of an insurer (or other risk-bearing entity). The caveats, however, remain ensconced in the supporting actuarial report. The typical reader of the published financial results has only the vaguest notion, if any indeed, of the amount of variability that can be attributed to the loss reserve estimate.

Given the importance of the margin of (estimation) errors, a case can be made that the caveats that accompany a loss reserve estimate should be lifted and carried forward as a note in the published financial statements that rely on such reserve estimates. The errors of estimation may or may not be reported in the history books that tell the story of the 2000 elections, depending on authorship. We can be certain that such caveats are going to appear in the history books that tell the story of the 2000 elections according to the vanquished. Of course the history books written by the victors will simply fail to note the caveat altogether—and why shouldn't they, they were the victors, after all.

One of the perennial challenges of casualty actuarial practice is finding ways to quantify the potential errors of estimation and their associated probabilities, thus reducing our reliance on vague caveats that may be state of the art but are hardly helpful to the user. This is a fertile question capable of continuously challenging each of us individually as well as the casualty actuarial profession at large.

The Florida elections and the attendant angst in reaching resolution have brought many issues into a sharp focus—along with a number of actuarial issues. More specifically, an actuarial issue that the Florida election has brought into sharp focus highlights the need for casualty actuaries to double and redouble their efforts to reduce the size and import of caveats in reserve analyses or to clarify the caveats by quantifying the potential errors of estimation. The Florida problem had to be resolved because the clock was ticking and a resolution was needed by a certain date. The system did what it could to move to a resolution. In the actuarial case, we do not have a ticking clock forcing the profession into action. But we do have the opportunity, as well as ample time, to improve our estimation procedures and our ability to describe the potential estimation errors to the benefit of all participants: the casualty actuary, the user, the regulator, the investor, as well as the casualty actuarial profession at large. Are we up to the challenge?