Monte Carlo Failures Aren't Plane Crashes

headshot of David Blanchett of PGIM

The probability of success is the most common outcome metric for a Monte Carlo projection. The success rate is simply the number of trials (or runs) the respective goal is accomplished (e.g., retirement income) divided by the total number of trials.

Some financial advisors, most commonly those who are selling a product with some type of explicit guarantee (e.g., an annuity), may use an analogy to imply that any chance of failure is simply unacceptable. Some examples include:

Would you board an airplane that has a 10% chance of failure?
A brain surgeon with a 95% success rate means 5% of his patients die.
Closing windows on a house to ensure no birds or unwanted intruders can enter.

The analogies all imply that an unsuccessful trial is somehow a cataclysmic failure. In real life, the impact of “failure” is likely to be significantly less severe. For example, a significant flaw in success rate metrics is that they ignore the magnitude of failure. For example, falling $1 short in the 35th year of a projection with a $100,000 would be treated as failure, despite the fact the person would have accomplished 99.999%+ of their goal.

In reality, people are unlikely to “fail” (as conveyed by a success rate). They will likely have to make some kind of adjustment to their plan during retirement instead. This adjustment, often in the form of a cutback in spending in later years, can be relatively minor or potentially more significant, but what’s important is that it is effectively impossible to somehow remove all the uncertainties with a single product or solution.

See also  Helping Affluent Clients Understand the Risk Management Landscape

For example, even products that provide guaranteed (or protected) lifetime income (e.g., an annuity, of which I am typically a fan) do not ensure a successful outcome across all future possible states. For example, annuities can introduce inflation risk to a strategy (since they are not typically explicitly linked to inflation) as well as liquidity risk (e.g., those that require an irrevocable election). So while some products have the potential to reduce or mitigate certain risks, it’s not possible to eliminate them all.

While beyond the scope of this piece, I do think our industry should move away from success rates as an outcome metric. I recently published some research walking through a more realistic, implementable retirement income planning model.

For those financial advisors who are stuck using existing tools, focusing on the income generated in retirement for certain percentiles (e.g., the worst 1 in 5 trials) at various ages (e.g., age 95) is a way to provide more useful context than simply suggesting an individual has a 75.234% success rate.

Conclusions

Unlike a plane trip, retirement is not a binary outcome, where there is complete failure or success. In reality, retirees have the ability to adjust over time and are going to do so as situations warrant. Understanding this nuance is important when conveying the pros and cons of Monte Carlo projections and ensuring investors make the best planning decisions possible!