1 Copyright 2012 by Glenn S. Daily. All rights reserved. Version: April 12, 2012 (original: March 7, 2012) Should You Invest in Life Settlement Funds? A life settlement is the purchase of a life insurance policy by a third party as an investment. A life settlement fund pools settled policies to create a diversified portfolio, with the goal of providing an attractive risk-adjusted return that has low correlation with the returns provided by other types of assets. Consider these developments: Most observers think that the life settlement market is now a buyer s market. More policyholders are becoming aware of the life settlement option, which increases supply, while investors are focused on buying distressed portfolios of already-settled policies, and new capital remains scarce. Life expectancy estimation is improving. Life settlement underwriting firms are committed to auditing their results and adjusting their methods as needed. In April 2011 Life Expectancy Providers, a group of major underwriting firms, released a document describing best practices. Performance appraisal methodology is being discussed by industry insiders (e.g., A. Hasan Qureshi and Michael V. Fasano, Measuring Actual to Expected Accuracy for Life Settlement Underwriting, Reinsurance News, July 2010) and outsiders (e.g., Robert Shavelle, Rating the Raters: Evaluating the Predictions from a Life Expectancy Rating Service, Journal of Insurance Medicine, 2009; 41: ). Alternative approaches to mortality modeling, such as the Longevity Cost Calculator, are also emerging. In November 2011 the U.K. s Financial Services Authority warned investment firms that life settlement funds are rarely suitable for retail investors. In the unlikely event that, after conducting extensive research, a firm considers that [life settlements] might be suitable for a particular retail investor, or wants to include them within another investment, like a fund of funds, that firm must be able to provide detailed and robust justification for its reasoning, the regulator stated. In March 2012 the A.M. Best Company issued a revised criteria for rating life settlement securitizations. Their research confirms that life expectancy estimates have lengthened and converged, possibly indicating improved quality, but they do not expect to rate many securitizations in the near future, due to problems with adequate scale for cash flow stability, insurable interest and acquisition costs. Researchers are examining life settlement funds. A paper in the March 2012 issue of the Journal of Risk and Insurance (Alexander Braun, Nadine Gatzert and Hato Schmeiser, Performance and Risks of Open- End Life Settlement Funds ) looks at the risk-adjusted returns for a sample of funds from December 2003 to June The authors conclude that life settlement funds performed well in relation to other asset classes, but they caution that the attractive returns should be viewed as compensation for significant risks. The due diligence process for life settlement funds is time-consuming and expensive. It should include these areas: the manager s industry experience and past performance, policy selection, medical underwriting and mortality modeling, financial underwriting, premium optimization, expenses, financing and reserving arrangements, longevity risk mitigation, liquidity, fair treatment of persisting and departing shareholders, taxation, probability distribution of returns, financial reporting, and operational, legal and regulatory risks. The quantitative work involves four steps: Determine the return profile of the fund for a base-case set of assumptions. Determine the risks that can negatively affect performance. Determine the return profile for alternative sets of assumptions with adverse deviations. Examine model risk. Here s an example to show what I have in mind. One policy with no expenses
2 Assume that a purchased life insurance policy has these characteristics: It is nonguaranteed universal life with a $3 million level death benefit. Premiums stop at age 100 but the full death benefit continues for life. The grace period test is based on cash surrender value, which is close to $0. The insured is a male nonsmoker, age 84. Also, there are three life expectancy estimates: 73, 73 and 76 months. For the policy valuation, let s use minimum monthly premiums, an 18% discount rate and the longest life expectancy estimate to calibrate a mortality table based on the 2008 Valuation Basic Table (Primary, Select & Ultimate, 0.5% annual improvement). The adjusted life expectancy is 80 months. With these assumptions, the policy is worth about $480,000. Of course, the policy will be a spectacular investment for the buyer if the insured dies quickly, but it will be a poor investment if the insured dies long beyond life expectancy. The table below shows the probability of achieving the selected internal rates of return, based on the $3 million death benefit, the projected premiums and the assumed mortality rates. All expenses, other than premiums, are ignored. Single Policy 25% 34.3% 34.3% 20% 7.5% 41.8% 18% 2.6% 44.4% 16% 4.0% 48.4% 14% 3.6% 52.0% 12% 4.4% 56.4% 10% 5.9% 62.3% 8% 5.4% 67.7% 6% 6.2% 73.9% 4% 6.5% 80.4% 2% 6.2% 86.6% 0% 3.9% 90.5% -100% 9.5% 100.0% There is a 34.3% chance of achieving an IRR of at least 25%. There is a 7.5% chance of achieving an IRR between 20% and 25%, and so on. Note that there is a 9.5% chance of losing money; that is, the $3 million death benefit will be less than the $480,000 purchase price plus the premiums paid after purchase. Even though the $480,000 price is based on an 18% discount rate, this policy is clearly a risky investment. Many investors would reasonably decide to walk away from this proposition. Policyholders face a similar problem when they are deciding whether to keep or drop a policy, and it can lead them to view a life settlement as their most attractive option, even in a buyer s market. Portfolio of similar policies with no expenses A single policy is risky, but what happens if we have a portfolio of policies? Some insureds will die early and some will die late, so that should reduce the overall risk.
3 Let s assume 100 similar policies and similar insureds. We can use Monte Carlo simulation to estimate the investment results for the portfolio. Each insured s month of death is determined randomly in accordance with the assumed mortality rates. A death benefit is received when each insured dies, and premiums are paid on the remaining policies. All of these cash flows are aggregated, and the IRR is calculated. This is just one trial of the simulation. To get a reliable picture, we need to do many trials, so that we include the results for many possible sequences of deaths. Here are the results for a simulation with 100,000 trials: Portfolio With No Expenses 25% 2.4% 2.4% 20% 23.4% 25.8% 18% 23.7% 49.5% 16% 26.4% 75.9% 14% 17.3% 93.2% 12% 6.0% 99.2% 10% 0.8% >99.9% 8% <0.1% >99.9% 6% <0.1% 100.0% 4% 2% 0% -100% As expected, a portfolio of policies is much less risky than a single policy. You have a much lower chance of getting an IRR above 25%, but you also have a much lower chance of getting an IRR below 10%. Now let s take the next step of adding expenses to turn the portfolio of policies into a life settlement fund. Portfolio of similar policies with expenses Assume a closed-end fund with a total initial investment of about $51 million. Assume these expenses, which do not match those for any actual fund: The provider s fee for obtaining the policies for the fund is 0.5% of face amount. Distribution costs are 2% of the initial investment. Other start-up expenses are $300,000. The annual management fee is 1.0%, modeled for convenience as 1.0% of the policy purchase price rising at 12% per year. There is no performance fee. The annual servicing fee is $25,000 plus $1,500 per policy. Other expenses are $100,000 per year.
4 Here are the initial costs to set up the fund: Policy purchase $48,000,000 Provider's fee 1,500,000 Distribution costs 1,016,327 Other start-up costs 300,000 Total $50,816,327 For convenience, assume no financing or reserving activities; additional money is provided monthly by investors as required to pay premiums. Also for convenience, assume that distributions are made monthly to zero out the fund balance when death benefits are received. The table below shows the probability distribution of IRRs for this base case. (I am choosing the probability distribution of IRRs as the performance measure for the fund because it is familiar, not because it is the most theoretically sound measure, especially when I have ignored financing and reserving activities. For a thorough discussion of performance measures, see Harold Bierman, Jr. and Seymour Smidt, The Capital Budgeting Decision, Ninth Edition, 2007.) Portfolio With Expenses IRR Base Case % of total Cumulative 25% 0.4% 0.4% 20% 8.1% 8.5% 18% 13.5% 22.0% 16% 24.3% 46.3% 14% 28.6% 74.9% 12% 18.5% 93.4% 10% 5.9% 99.3% 8% 0.6% >99.9% 6% <0.1% >99.9% 4% <0.1% 100.0% 2% 0% -100% The fund s raw material the life insurance policies has a median return of about 18%, and the various expenses reduce the net pre-tax return to investors to a little less than 16%. Of course, expenses affect the entire probability distribution of returns, not just the median return, as shown by this chart:
5 The next chart shows that the fund s expenses reduce the investor s IRR by about 2.2% for all probabilities. For example, if there is a 50% chance of getting at least 18.0% with no expenses, there will be a 50% chance of getting at least 15.8% with expenses, and if there is a 70% chance of getting at least 16.5% with no expenses, there will be a 70% chance of getting at least 14.3% with expenses. The impact of expenses is quite consistent throughout the cumulative probability distribution. You can think of 2.2% as the all-in expense ratio for this fund. It includes the cost of obtaining the policies for the fund (the provider s fee), the cost of managing the fund (the manager s fee) and all other operating costs.
6 Longevity risk What is the impact on fund performance if the insureds live longer than expected, because of flawed underwriting or health improvement? Suppose the insureds are the equivalent of one table rating healthier than expected (about a 25% lower multiplier applied to standard mortality rates), and suppose the annual rate of mortality improvement is much higher than expected, leading to a life expectancy of 92 months rather than 80 months and a mortality curve with a different shape. This shock is actually less severe than what occurred in the life settlement market a few years ago, leading some institutional investors to suffer losses of hundreds of millions of dollars on their portfolios. Using these mortality rates, the policy would be worth only $302,000 rather than $480,000, so the fund would be paying far too much for the 100 policies. Here are the results for this scenario: Portfolio With Expenses If Insureds Live Longer 25% <0.1% <0.1% 20% 0.1% 0.1% 18% 0.3% 0.4% 16% 1.3% 1.7% 14% 5.6% 7.3% 12% 15.5% 22.8% 10% 29.6% 52.4% -100% 8% 30.9% 83.3% 6% 14.3% 97.6% 4% 2.3% >99.9% 2% <0.1% >99.9% 0% <0.1% 100.0% The median return has dropped from a little less than 16% to a little more than 10%, although there is still only a tiny chance of getting less than 4%. Litigation risk Life settlement funds operate in a hostile environment. Life insurance companies do not like life settlements, and they especially dislike stranger-originated life insurance (STOLI) programs. Some companies have refused to pay death claims when they suspect that the insured participated in a STOLI program. The outcomes in litigation have been mixed. Suppose each death claim has a 10% chance of not being paid, for whatever reason. Here are the results, assuming the base-case mortality rates:
7 Portfolio With Expenses If Death Claims Aren t Paid 25% <0.1% <0.1% 20% 0.6% 0.6% 18% 1.7% 2.3% 16% 5.6% 7.9% 14% 14.0% 21.9% 12% 24.7% 46.6% 10% 28.4% 75.0% 8% 18.6% 93.6% 6% 5.6% 99.2% 4% 0.7% >99.9% 2% <0.1% 100.0% 0% -100% The median return has dropped from a little less than 16% to a little less than 12%, although there is still only a tiny chance of getting less than 4%. Premium increase risk The optimal premium schedule from the buyer s perspective is usually the minimum premium to keep the policy in force, and that is mainly based on the cost-of-insurance charge. For no-lapse universal life, that charge is guaranteed in the contract and cannot increase, except possibly if the insurance company becomes insolvent. For nonguaranteed universal life, however, the insurance company usually charges much less than the maximum rate that is guaranteed in the contract, and it has the right to increase the current rate. Cost-of-insurance rate increases are not common, and the life settlement market has generally ignored this risk. If more rate increases occur in the future, the minimum premiums will also increase.
8 Suppose the cost-of-insurance rates increase by 10% every five years. Here are the results, assuming the base-case mortality rates: Portfolio With Expenses If Premiums Increase 25% 0.1% 0.1% 20% 2.2% 2.3% 18% 5.3% 7.6% 16% 12.5% 20.1% 14% 23.1% 43.2% 12% 28.3% 71.5% 10% 20.0% 91.5% 8% 7.3% 98.8% 6% 1.1% >99.9% 4% <0.1% >99.9% 2% <0.1% 100.0% 0% -100% The median return has dropped from a little less than 16% to about 13.5%, and there is only a tiny chance of getting less than 6%. Combination of risks What happens if several bad things occur together? Here are the results if the insureds live longer than expected (as above) and there is a 10% chance of nonpayment of the death claim (as above) and the cost-of-insurance rates increase by 10% every five years (as above): Portfolio With Expenses If Insureds Live Longer and Death Claims Aren t Paid and Premiums Increase 25% 20% 18% 16% <0.1% <0.1% 14% 0.1% 0.1% 12% 0.2% 0.3% 10% 0.9% 1.2% 8% 4.2% 5.4% 6% 12.5% 17.9% 4% 24.7% 42.6% 2% 29.6% 72.2% 0% 20.0% 92.2% -100% 7.8% 100.0%
9 Now the median return has dropped to less than 4%, and there is almost an 8% chance of losing money. This is certainly not pretty, but you have to wonder if this multiple-risk scenario is going too far. At some point a stress test crosses the line and becomes homicide. Model risk It is a good practice to approach financial modeling with humility and to consider if there are omissions in the model that could significantly affect the results. (For humility lessons, see Francis X. Diebold, Neil A. Doherty and Richard J. Herring, editors, The Known, the Unknown, and the Unknowable in Financial Risk Management, 2010.) Here are some areas where there is certainly room for improvement: Portfolio composition. I have assumed a portfolio of similar policies and insureds, but real portfolios have different policies and insureds. Correlation of deaths. I have assumed that deaths occur independently. If deaths are correlated, then there would be less diversification than expected. Premium prepayment and delay in receipt of death benefits. I have implicitly assumed that the fund manager pays all premiums just before they are due and that death benefits are collected about two weeks after death. In practice, a prudent manager might want to pay premiums at least one month early to avoid entering the grace period, and it can take up to three months to collect death benefits. With those assumptions, the policy would be worth only about $425,000 rather than $480,000; we can infer that the impact on returns would be less severe than the impact of the mortality rate shock in the longevity risk example. Of course, the fund can take these factors into account when it makes a purchase offer for the policies. Financing, reserving and distribution procedures. Contrary to my assumptions, life settlement funds do not collect money from their investors each month to pay premiums, and they do not immediately pay out death benefits when they are received. Funds have various procedures in place to make sure that money is available to pay premiums, and they have various rules for making distributions. These activities can help or hurt returns, depending on the details. Operational risks. The fund manager, a key employee or a service provider could turn out to be a thief, or they might just be incompetent or preoccupied with other things. Fund termination. A closed-end life settlement fund is in run-off mode from the day it begins. How does it actually end? I have assumed a natural termination when the last insured dies. For my hypothetical fund, that is likely to be within 20 years, but it could exceed 30 years. In reality, it will probably make sense for a fund to sell its remaining policies and liquidate long before its natural termination. Software errors. I did these calculations using Microsoft Excel 2003 and Excel 2007, which have known deficiencies for statistical work. In particular, Excel s random number generator may fail randomness tests in large simulations (see B.D. McCullough, Microsoft Excel s Not The Wichmann-Hill random number generators, Computational Statistics and Data Analysis, June 2008). In my calculations, repeated simulations always produced results within a few tenths of a percent and often within one-tenth of a percent.
10 Conclusion If you think that everyone will soon live to be 120 years old, you should stay away from life settlement funds. But if you have a more mainstream view of human mortality, you may be able to find a life settlement fund that can survive a conscientious due diligence process. As shown above, a fund can take some nasty hits and still provide a decent return. Here are a few ways to increase your chances of being satisfied with your investment in a life settlement fund: By Glenn S. Daily Look for low expenses. Of course, that s good advice for all investing, but it is especially important for exotic investments where the prospect of high returns can make high expenses seem more tolerable. There is currently no standardized measure of total expenses, similar to the expense ratio for mutual funds, so you have to review each expense item. Look for diversification. The portfolio should contain policies issued by different insurance companies, and the insureds should have different kinds of health impairments. There can also be diversification by face amount, gender, life expectancy, region and type of policy (for example, nonguaranteed versus guaranteed universal life). Look for manager skill. In what ways will the manager try to enhance the fund s return and reduce the risk of bad outcomes? Look for alignment of interests. Does the compensation of the manager and the provider create incentives that are aligned with or in conflict with the interests of the investors? This is not always easy to determine; for example, performance fees may seem like a good idea, but some performance fees have been criticized for creating a moral hazard to play games with valuation. Look for cushions in the financial projections to offset bad surprises. Is the manager assuming a higher purchase price for the policies than may actually occur? Are the assumed mortality rates lower than what the underwriters estimated? (One fund manager assumes that no one will die during the first 18 months, no matter what the mortality table says.) Are other assumptions carefully chosen to be conservative? Look for fair treatment of all shareholders. Life settlements are not liquid, and their current market value may not be known. When a shareholder wants to pull money out of the fund, how is the manager making sure that the amount distributed represents the fair value of the shareholder s piece of the pie? What is the risk that the distribution procedures will lead to future litigation? Look for informative financial reporting. The periodic reports should explain the fund s recent performance, the reasons for any deviations from expected performance, suggested revisions to expectations about future performance, and the actions that the manager is taking to achieve performance goals.