There is much to be learned by taking a deep dive into analyzing variable universal life (VUL) as a potential investment. I recently purchased an accumulation-oriented VUL policy for my 21-year-old son in a no-load chassis with no AUM fees. My goal was to optimize investment performance within a life insurance policy, without any gimmicks.
Creating a Monte Carlo Simulation
The first step was to minimize the death benefit ($155,501). To demonstrate the impact of return volatility on the performance of a policy, I ran a Monte Carlo simulation.
Here were the assumptions:
- We used the Guideline Premium Test for the definition of life insurance.
- A 100% investment in an S&P 500 index fund, even though we diversified across a basket of low-cost equity (mostly index) funds.
- The death benefit option would be increasing during the premium-paying period (through age 60) and then would switch to level thereafter.
- Although the S&P 500 index with dividends has averaged more than 11% since 1928, we are conservatively using an 8% mean return with historical S&P 500 volatility as the baseline assumption for Monte Carlo testing.
- Premiums of $5,000 are assumed to be paid annually for 39 years (until age 60), though they could continue to be paid for life if desired.
To perform a Monte Carlo simulation, I had to custom build an illustration tool. In order to evaluate the policy correctly, the tool needed to able to match the baseline static sales illustration (with no volatility). It also had to be capable of producing dynamic illustrations based on randomly generated monthly returns consistent with the chosen mean and volatility (standard deviation).
The Results
Here are the high-level takeaways from this analysis:
Early cash surrender values and liquidity can be extremely high on a no-load VUL policy. If the gross investment return is 9% or higher, then the first-year cash surrender value is greater than $5,000 (the first premium). That’s not an artificial value that the policyholder cannot access – that is the value available upon surrender. Having a breakeven year of 1 (or year 2 if returns were say 6-9% annually) is extraordinary.
The lifetime drag at life expectancy in both a static illustration and in the mean dynamic illustration can be as low as 50 bps. That is, if the gross return assumption is 8%, the illustrated death benefit IRR at life expectancy (age 90) can be as high as 7.5% (net of income tax, which should be zero). Compare that to investing in the stock market, which would be subject to fund fees, potential management fees, and potential taxes. In an 8% gross return world for a taxable account, the net after-tax return could range anywhere from 4.5% to 7%, depending on expense, tax, and potential distribution specifics.
The median outcome with a Monte Carlo simulation is substantially worse than the mean. In the baseline scenario, while the mean outcome is a death benefit at age 90 of $9.9 million and an IRR of 7.5%, the median outcome is a death benefit at age 90 of $5.3 million and an IRR of 6.3%.
- The median, in this case, winds up way lower than the mean.
- In fact, roughly 2/3 of all scenarios end up worse than the mean.
- Some might argue that the median is a better measure of central tendency due to the asymmetric nature of the death benefit distribution.
- A drag of 170 bps (median) portrays a different comparison with a taxable investment than a drag of 50 bps (mean), and the actual drag will be a function of actual returns.
The VUL can still show very favorably in certain situations, particularly inside trusts which are taxed at a higher rate (and may not have a step-up in basis at death) or in distribution scenarios where taxation would be unavoidable in a taxable account. Note that this asymmetry is not an issue specific to VUL, but rather a Monte Carlo phenomenon. However, it is slightly exacerbated by the VUL chassis and some negative synergy that occurs within a life insurance wrapper.
There is a tremendous gap between the ends of the spectrum. For instance, the 95th percentile death benefit at age 90 is $28.8 million whereas the 5th percentile death benefit at age 90 is $0.8 million.
- This demonstrates the folly of planning for a predetermined level of income distributions from a variable policy many years down the road.
- If income distributions late in life are sought, one needs to be flexible based on investment performance both before and after distributions begin. Changing the investment mix to be more conservative during the distribution phase may be worth considering.
- Most importantly, one can never take out so much from a life insurance policy that the ability of the policy to be in force at the time of the insured’s death is in jeopardy.
The risk of policy failure is high if not managed properly. Under our baseline dynamic case, there is a 14% probability of failure at age 100.
- If we only pay 10 premiums instead of 39, the probability of failure increases to 19%.
- If the true investment mean going forward is only 5% instead of 8%, the probability of failure increases to 48%.
- If the policy is underfunded (10 premiums) and investment mean returns are poor (5%), the probability of failure increases to 66%.
These are unacceptably high failures of probability, but they reflect the inherent risk of a variable life policy that is simply left on cruise control, even one that was originally designed to be accumulation-oriented rather than the more inherently risky protection-oriented designs.
Reducing the death benefit can possibly eliminate the probability of failure. At age 60, which is the age where I assume that premiums stop and where the death benefit will switch from increasing to level, if we annually lower the death benefit to the minimum amount allowed by the corridor test of the definition of life insurance, then we can virtually eliminate the probability of failure.
Perhaps some would conclude that receiving a very low death benefit and having a low IRR was akin to a failure, but it’s not a failure in the sense that the policy has lapsed. We modeled all scenarios described previously and incorporated an annual death benefit adjustment starting at age 60. There would hypothetically be no failures with this approach, including the grim combination of only 10 premiums and 5% mean investment returns.
Policies must be actively managed by the policyholder. It would be an industry game-changer if automatic death benefit adjustments were available in the same manner that automatic rebalancing is available for investment options.
8% Mean Returns
Assuming we employ the annual death benefit adjustment starting at age 60 and that 8% mean returns are indeed appropriate (and that we pay 39 premiums up to age 60), here are the relevant statistics at age 90 for the Monte Carlo simulation:
| 5th percentile DB | $0.8 million (2.8% IRR) |
| 50th percentile DB (median) | $5.5 million (6.4% IRR) |
| Mean DB | $9.5 million (7.4% IRR) |
| 95th percentile DB | $30.2 million (9.6% IRR) |
| Zero scenarios failed at age 100 |
5% Mean Returns
If we do the same but instead use an investment mean of 5%, here are the relevant statistics at age 90 for the Monte Carlo simulation:
| 5th percentile DB | $0.2 million (0.0% IRR) |
| 50th percentile DB (median) | $1.2 million (3.5% IRR) |
| Mean DB | $2.0 million (4.5% IRR) |
| 95th percentile DB | $6.1 million (6.6% IRR) |
| Zero scenarios failed at age 100 |
11% Mean Returns
If we do the same but instead use the historical mean of 11%, here are the relevant statistics at age 90 for the Monte Carlo simulation:
| 5th percentile DB | $5.7 million (3.7% IRR) |
| 50th percentile DB (median) | $25.5 million (9.2% IRR) |
| Mean DB | $48.0 million (10.4% IRR) |
| 95th percentile DB | $153.2 million (12.5% IRR) |
| Zero scenarios failed at age 100 |
Conclusion
Every client situation is different. While the analysis we did showed favorable results for our son, in no way does it imply this policy would work for anyone else. Our objective in sharing our analysis with you was to demonstrate the high degree of technical prowess that goes into understanding how these policies may behave over time.
Don’t buy a cash value insurance policy for yourself, a client, or anyone else without conducting thorough research, as we did. In the process, it’s important to understand not only the insurance policy features, but also how investment performance and volatility affects these types of vehicles.
As you can see, there are vast differences in outcomes that can result from policy management decisions. These are not to be taken lightly. If you are unable to carry out this analysis for yourself or for your client, it’s best to consult with a professional who can.
Does your client need a policy review?
I am an Actuary and a Consumer Advocate (not an insurance agent) who helps high-net-worth individuals with $100,000 or more invested in cash value life insurance or annuities to maximize the value of their policies.
High-net-worth individuals and their advisors hire me to help:
- Analyze existing life insurance policies and annuities to provide customized recommendations for optimizing value
- Consider impact of objectives, longevity, tax considerations, and opportunity cost on life insurance and annuity decisions
- Access/design life insurance policies and annuities that eliminate (or reduce) agent compensation and maximize policy value
- Provide an unbiased perspective free from any conflicts of interest
- Avoid making critical mistakes and provide peace of mind
- Provide a qualified appraisal on life insurance policies
To learn more, please contact me.
Disclaimer
Witt Actuarial Services does not guarantee any specific level of performance, the success of any strategy that Witt Actuarial Services may use, or the success of any program. Information contained herein may become out of date; Witt Actuarial Services is under no obligation to advise users of subsequent changes to statements or information contained herein. There is no guarantee that the information contained herein is accurate. This information is general in nature; specific advice applicable to your current situation is only available through an engagement. Any perceived similarity with persons living or deceased is entirely coincidental.