The State of Casualty Actuarial Science Today
Actuarial Roundtable Discussion
Arthur J. Schwartz
In February, four distinguished actuaries and I got together by conference call to discuss the state of casualty actuarial science today. Our panelists include:
Glenn G. Meyers, with Insurance Services Office in New Jersey, has written many papers on risk loads, catastrophe ratemaking, and capital allocation, among many other topics. He has served the CAS on the Examination Committee for several years and on a number of CAS research committees.
Howard C. Mahler currently lives in Boston, teaches actuarial exam seminars, and consults. He has written papers on a variety of topics including workers compensation, credibility, experience rating, retrospective rating, and underwriting profit models. He served for a dozen years on the CAS Examination Committee, including three years as chair of the committee.
Sholom Feldblum works with corporate financial models for Liberty Mutual in Boston. He has written numerous papers explaining actuarial concepts and helping students learn these topics efficiently.
Stephen W. Philbrick is with Conning, a division of Swiss Re, in Baltimore. His paper on credibility concepts won the Woodward-Fondiller Prize. He has been active on numerous CAS committees, including chairing the Committee on Principles, which seeks common principles with other actuarial organizations. He writes the “Brainstorms” column for The Actuarial Review, which discusses interesting new applications of actuarial science.
Schwartz: What are the most important technical challenges in casualty actuarial science today? These can be in pricing, reserving, financial, or other areas. Please briefly outline any methods or schools of thought on how to solve each technical challenge.
Meyers: There are two main problems. The first is the area of quantifying uncertainty. This needs to be considered on a prospective and retrospective basis. The second area is in mak- ing use of new methods and new information for what we’ve traditionally done in ratemaking and reserving. For example, we’re considering how to use credit information and the new data mining techniques. Also, we’re considering how to predict differently than the traditional approaches, which are usually some form of generalized linear modeling.
Philbrick: I’d echo that. A third key area is communicating what we do. Actuaries have a problem communicating. We need to learn to say things in the language of financial theorists. A fourth area is challenges to the traditional principles of ratemaking and reserving; specifically, how do we use new classes of information?
Feldblum: There are two main problems. First, we are not sufficiently aware of the developments in related fields such as economics and finance. Actuaries seek to quantify risk, but practicing actuaries have little knowledge of the financial theory of risk. Pricing actuaries estimate needed rate changes, but few of us are aware of the economic work on the elasticity effects of product prices on market share and profitability.
A second problem is the actuarial syllabus. Students spend too many years memorizing details. Our students spend years memorizing the details of life contingencies and statutory accounting, both of which are important topics, because the exams require this memorization of details.
Mahler: I would agree with Glenn that one key technical challenge is quantifying uncertainty. This relates to, among other things, parameter uncertainty, risk loads, allocating capital, and determining the cost of capital. However, there are also levels of uncertainty. For example how do we quantify the possibility of unprecedented events? Gary Blumsohn has an excellent discussion of levels of determinism in his paper in the 1998 Proceedings. In addition, there have been continued improvements in solving some of the same problems that began the CAS in 1914. Specifically, there continue to be new advances in experience rating, retrospective rating, class ratemaking, and reserving techniques. These new advances may occur in mundane, ho-hum areas, rather than some hot new area of actuarial science, yet these advances are also important.
Philbrick: In addition to all these new areas, we are making steady improvement in our traditional tools for ratemaking and reserving. This is similar to a “six sigma” approach. We may not be closer to a final answer but our tools are becoming better and better. It’s similar to a think tank at General Electric. We are not only tackling new things, we are doing what we do, better.
Feldblum: We must examine how our actuarial problems are handled in other fields. The cost of capital is much discussed in finance, and we should take account of the financial costs of capital when discussing risk loads. The relative benefits of equity and debt have been discussed by financial analysts ever since the publication of the Miller and Modigliani propositions. Actuaries discuss these same issues without being aware of the financial literature.
Meyers: Most actuaries have a reasonable grasp of the financial issues. Where I believe our methods fall short is in the area of quantifying risk. It’s the biggest challenge we face. Getting the insurer’s distribution of losses by analyzing data (i.e., quantifying the underwriting risk) is the tough part. Compared to this, the financial stuff is rather easy.
Mahler: I have worked on some of these issues at the intersection of actuarial science and economics for over 15 years. Yet adapting what we do to the economists’ and finance theorists’ framework is very difficult. Sholom is right that we should see if an already existing application of finance or economics may have already solved our problem. We also need to ask does their work make sense in an insurance context, in the context of a business whose sole purpose is to take on risk that other businesses don’t want?
Meyers: You can apply their methods, but it takes a lot of work. The key is to get the right input for these models.
Mahler: So far, people have been unable to quantify many of these things accurately for use in a practical application.
Meyers: What’s in the financial literature on handling catastrophes? Here I think we need to get away from the mean-variance analysis that is prevalent in the financial literature.
Feldblum: That is a complex issue.
Meyers: I wanted to throw that in since I believe many classical financial theories have been inadequate for the insurance (risk-assuming) paradigm.
Philbrick: My impression is that the classical financial models make the assumption that catastrophes do not occur. We need to do a better job, see what the economists and finance theorists have done, and build upon that.
Feldblum: Financial theorists deal with systematic risk; shareholders can diversify their risk. The actuarial view is to look at total risk. Since shareholders can diversify catastrophe risk, financial theory does not treat it separately.
Mahler: If an insurance company can buy protection against catastrophes in the financial markets, then if there is a relatively stable market price, the actuary can use this as a tool to help price risk taken on by an insurer. The insurance company and regulators desire protection against the insurer having insufficient funds to pay losses. If protection against catastrophes can be readily bought by insurers, and if the sellers of that protection are using financial theories based on diversified investors to price that protection, then this could link actuarial science and the financial markets. Also, securitization of risks leads one to believe that the market can price risk, to some extent. If securitization of risk continues to expand, in the following decades this may be a fruitful path connecting financial measures of risk to appropriate insurance risk loads.
Feldblum: In effect, the market may be able to transform insurance risk to shareholder risk.
Mahler: Even if we didn’t get that close to a final stage; if we simply got closer to connecting something we want to quantify, to something we see in the markets, we could make progress. For example, actuaries currently make some use in our work of the market-based “risk-free rate” and the increment of return required by buyers of bonds other than U.S. treasury securities.
Meyers: Insurance company management wants to keep the company viable in the face of catastrophes. That’s why regulators and investors require surplus. The shareholders’ views are, they’ll just buy a small portion of the risk, so their maximum loss is limited. The insurer must pay claims; the insurer must have sufficient resources to provide the coverage that they are promising.
Mahler: From the shareholder’s view, if my risks are diversified, then it’s better to have no surplus. As has been mentioned by many others, if one were not constrained by regulations, one could form 50 insurers, one per state, each with $1 in surplus; then one would write high-risk auto insurance. Perhaps 40 insurers would fail, the other 10 though would make money, and the net return to shareholders, as opposed to society, would be substantial.
Meyers: It’s heads, I win. Tails I lose, but I lose only a little.
Mahler: Consider a line like automobile liability insurance that many people must buy if they wish to drive a car. Only if the buyer or regulator insists on capital, is there a need to put up capital, otherwise there’s no requirement to have substantial capital in order to write insurance. In contrast, if I am a manufacturer, I need capital to build a factory and get supplies.
Philbrick: The risks to insurance company solvency are interesting. I have seen an ISO study on insolvency that came to rather discouraging conclusions. It seemed to show that most insurers do not fail because of issues related to the law of large numbers, i.e., diversification of their risks. The study concluded that most insolvencies, about 80 percent, result from fraud or mismanagement. The challenge is to stay solvent. Our tools for preventing insolvency, such as risk-based capital, deal with about 20 percent of the insolvencies. The other 80 percent are caused by the two problems mentioned above.
Meyers: To some extent, fraud and mismanagement are the auditor’s problem, not an actuary’s problem. Getting back to our previous discussion on learning from other fields, a generalized linear model is the traditional way of setting rates or estimating reserves. A new way of setting rates or estimating reserves may come from approaches as diverse as data mining, neural networks, artificial intelligence, and clustering.
Schwartz: The syllabus is our most precious possession for transmitting casualty actuarial science to future generations of actuaries. Are there any technical or business skills that the current CAS syllabus either does not cover or does not cover adequately? Are there any technical or business skills that the current CAS syllabus covers more thoroughly than necessary (out of proportion to their usefulness)?
Mahler: Yes and yes.
Meyers: When I first entered the field, actuaries worked primarily on pricing and reserving. It was possible to cover most of what actuaries did in our exam syllabus. Today we are working in areas we could not have conceived of, even a few years ago. Now it is impossible to cover all aspects of actuarial work on the exam syllabus. We have to decide what to include, and what to leave off the syllabus.
Philbrick: We need to cover more on finance. The current syllabus is much improved in this area, but some of us who finished our exams decades ago probably need more continuing education in this area. By the way, is there a favorite or recommended text on finance? I’ve heard from students that the readings on the finance exam, Exam 8, are very good. Specifically, what is important on the cost of holding capital?
Feldblum: The Modigliani and Miller (M&M) propositions1 are important. There are three streams of thought to solve the M&M anomalies: Miller’s explanation, Myers’s explanation, and an explanation relying on agent-principle differences. This material is not on the Exam 8 syllabus, but it is the foundation of financial theory on the capital structure of corporations.
Feldblum: Another issue to consider is that the actuarial profession is competitive with other financial professions. The amount of study time required for the actuarial designation compared to the study needed for other professions that offer equally good financial and career rewards is an important consideration. We have to rethink our syllabus in that light.
Meyers: I agree. Allocating space on the syllabus should be seen as a task similar to allocating any scarce resource.
Philbrick: I also agree. At one time we had 28 lines of business on the annual statement. It was thought important to have a paper on the syllabus on each of those 28 lines of business. I believe the syllabus needs to be improved in the area of investments and finance as they affect the insurance industry.
Feldblum: I agree that it’s ideas and not specialized applications that we need to study. We must consider how much time is required to prepare for these exams and to what extent we can fail students. We have to make sure that the exams do not dissuade potential students from the profession.
Philbrick: The problem is not only in the syllabus but the methods of testing. I know a bright student who says when he’s taking exams, “I do not need to know the material, I only need to know how to answer the questions.” That’s the kind of attitude that the testing method encourages.
Part Two of the roundtable discussion will appear in the next issue of The Actuarial Review. Questions or comments can be sent to email@example.com.
1Editor’s note: According to Investments, Bodie, Kane, and Marcus, page 551: “M&M claim that if we take as given a firm’s future investments, then the value of its existing common stock is not affected by how those investments are financed. Therefore neither the firm’s dividend policy nor its capital structure should affect the value of a share of its equity.” This is true for a world without taxes. When dividends are taxed fully but debt payments are tax deductible, debt is cheaper than equity. In practice, debt seems to be much cheaper (in theory) than equity, yet firms in many industries hold large amounts of equity. The issue of why insurers hold capital is a general financial question: Why do firms hold such large amounts of equity when debt is the theoretically preferred method of financing?