For any federal worker looking to maximize revenue for retirement from annuities, Social Security and investments, deciding when to retire is a complex calculation.
There’s no basic rule of thumb or simple algorithm to make the decision for you. Even methods traditionally used by professionals can lead you to a decision you may regret later. That’s because traditional “static” analysis — analysis that assumes a fixed time horizon and regular annual investment returns — fails to account for the fact that lifespans and annual investment returns vary from those expected. Fortunately, today there are more sophisticated analytic techniques available that can stresstest your retirement plan against potential bad-luck scenarios.
To illustrate how the planning process should work, the sorts of information and decisions that are required, and the results that are produced, I’ll present a retirement study I completed recently. Names and some details that do not affect the study results have been changed to protect the clients’ privacy.
In this example, the clients are a couple in their mid-50s — the husband, Bob, a GS-13 covered by the Federal Employees Retirement System with investments in the Thrift Savings Plan; and his wife, Linda, who has IRA and 401(k) accounts. Bob and Linda were recently advised by a stock broker that they could safely retire in a year. They told him they thought they could do fine in retirement if they could contain spending at $50,000, indexed for inflation, each year as long as either of them survived. They also estimated they’d need an additional $20,000 for their daughter’s wedding in three years. They figured they’d need another $10,000 every five years for a car, beginning in three years. They have about
$420,000 saved up, mostly in Bob’s TSP account and Linda’s IRA and 401(k) accounts. Since the risk-tolerance questionnaire that the stock broker had Bob and Linda complete indicated they were conservative investors, he proposed managing their assets in mutual funds comprised of 60 percent stocks and 40 percent bonds. He predicted a 10 percent annual rate of return on their portfolio before any investment expenses or taxes were assessed. He also used their statistical life expectancies of 83 for Bob and 87 for Linda in his analysis.
Based on their needs and his analysis, their stock broker predicted that they should expect to end their lives with more than $800,000 in their investment accounts and that this would provide a deep cushion against any unexpected events. Based on his presentation, which included a rather impressive leather-bound report, Bob and Linda felt quite comfortable that retirement next year would be a safe choice.
The couple came to me, as many do, to get another check and make sure they hadn’t missed anything.
Testing the plan
The first analysis conducted was to test the broker’s plan using a technique called Monte Carlo probability analysis. This technique simulates a large number of hypothetical lifetimes for the couple as they draw their income, invest assets, take withdrawals, pay taxes and account for inflation. Unlike static analysis, however, which relies on one simulated lifetime ending at a fixed point in time, my analysis drew Bob’s and Linda’s ages for their hypothetical deaths from a statistically appropriate pool of possibilities before each simulation was begun. And, instead of assuming a constant annual investment rate of return, my analysis allowed the annual returns to vary from year to year in each simulation. So, for each 1,000 lifetimes that were simulated, there was a uniquely selected time horizon and a unique sequence of annual investment returns. This allowed undesirable events to coincide and place stress on the plan. After thousands of simulations, the results are tallied and used to produce a probability of success, or confidence level in the plan.
Because of the way the analysis is conducted, confidence levels of between 75 percent and 90 percent are adequate. Confidence levels below 75 percent indicate an unacceptable likelihood of plan failure. Confidence levels above 90 percent may be subjecting the investor to undue sacrifice in lifestyle or estate objectives. I usually shoot for 90 percent if possible, but I’m comfortable with anything over 79 percent.
The analysis revealed some surprising results. If Bob and Linda were to retire in a year, their plan had only a 62 percent confidence level — or a 38 percent probability of failure. Failure in this case was defined as exhausting their investment assets while at least one of them was still alive.
Also, the broker’s plan had failed to account for the costs that would be incurred for his services and management of the mutual funds. I calculated these costs to be about 1.5 percent of invested assets annually.
Bob and Linda were understandably concerned about the results, so we looked at how we could increase plan confidence. We discussed what compromises they could make. They set priorities on changing certain aspects of the plan — the key planning variables — where they may have significant control and ability to make changes.
They were most willing to make changes to their investment strategy and then to their retirement ages, within reason. They were reluctant to commit to spending less than $50,000 per year and would work as long as necessary to avoid this, including part-time work in retirement.
Based on this, I began to modify their plan. I recommended that they find lower-cost investment management. Using low-cost index funds and avoiding high management fees, they should be able to lower their annual advisory and management costs from 1.5 percent to 0.5 percent. This alone raised plan confidence by 18 points to 80 percent. This was good, but they agreed that getting to 90 percent would be worth some compromise. So, I showed them that delaying their retirement by one year would raise plan confidence to 95 percent — five points higher than we considered optimal.
Bob and Linda were very comfortable working another year in exchange for the increased confidence in their plan. Waiting another year to retire also made Bob eligible for the special retirement supplement under FERS. With a 95 percent probability of plan success, the couple could consider adjustments that would lower confidence a little. This is not considered risky, because Monte Carlo simulation allows certain combinations of events to occur that would not be reasonably likely in the real world, such as two or three catastrophic bear markets in succession. It is reasonable to allow, without concern, for 10 percent or more of the simulations to fail based on these types of event combinations.
So, with this in mind, I identified a few alternatives that they could consider that would maintain plan confidence of about 90 percent:
- Increasing annual retirement spending to $54,000, adjusted for inflation.
- Moving to a more aggressive growth investment strategy and raising the ending savings balance target to $400,000.
After some discussion and consideration, Bob and Linda decided that moving forward with the growth strategy while attempting to keep retirement spending at $50,000 per year would be the smartest move. The projected ending savings balance would allow for adjustments later if they found they needed to exceed their spending budget or accommodate unforeseen events.
They were also attracted to the fact that the growth strategy alternative offered less downside risk over the longer time period with significantly higher upside potential.
Written by Mike Miles
For the Federal Times
Publication June, 20, 2005