How do you help a client who is rapidly approaching retirement figure out how much to spend on a monthly basis? In other words, they want to know what they can safely budget to spend each month in retirement without running out of money.
This is a question that has challenged financial advisors for decades, and it’s become all the more important now that, other than Social Security, the typical retiree has no locked-in lifetime income at retirement. The Secure and Secure 2.0 laws make it clear that for the foreseeable future, Americans will need to do their retirement planning in a defined contribution world. As a result, most retirees are saddled with the task of assessing their retirement capital and figuring out how much they can take in income each month.
So, again, how can a retirement advisor help in this daunting task? At the American College of Financial Services, we recently took a new look at this age-old question during a Q&A webinar, “New Trends in Retirement Income Planning: Beyond the Use of Monte Carlo.”
In recent years, many advisors have used Monte Carlo modeling to project a retirement income strategy for their clients. But this tool has its limits, and with the incredible power of modern software, we wanted to explore what possibilities exist beyond using Monte Carlo.
What brought us to Monte Carlo?
When I first became a financial advisor in the 1980s, my trusty HP 12C calculator (which I affectionately still use) assisted me with the retirement challenge. I calculated present values and internal rates of return to make my best guess at a reasonable retirement income for my clients. In the 1990s, we started using more sophisticated computing and presentation powers available from personal computers. With this exciting new tool, we were able to make static assumptions for growth of an investment portfolio and show it to our client in either a printed or screen version. We no longer had to say, “Trust me, here’s the number.” We could disclose our assumptions and show them the results.
The new century saw the introduction of software that could actually model retirement portfolios. We were no longer stuck with using static assumptions such as the investor earning a fixed rate of return annually or the retiree taking a fixed monthly income for life. With access to massive amounts of data and the ramped-up computing speed to analyze this data, an advisor could run hundreds, even thousands, of simulations using variable inputs. This was the beginning of the Monte Carlo revolution.
The Monte Carlo method uses a random sampling of information — including historical returns, asset correlations or other relevant assumptions — to demonstrate a possible strategy. It allows an advisor to present the retiree with a range of outcomes under random conditions. The advisor can report that, “After running 1,000 trials using a randomized set of scenarios, in 85% of the scenarios, it was found safe to withdraw 4% of your retirement capital each year and still have money left over after 30 years.”
Monte Carlo limitations
The Monte Carlo method is a wonderful tool to have, and it enables us to provide clients with far more information than simply a non-variable, one-time answer. Not only can the model assume varying rates of returns, but it can also test the sequence of returns. However, as with any financial tool, challenges, there are challenges with this approach.
In my mind, the two big issues with using Monte Carlo in retirement planning are statistical and behavioral.
The statistical challenge is “garbage in; garbage out.” This tool runs simulations based on inputs, and the inputs may be flawed. As a simple example, let’s say the modeling assumptions for investment returns by industry identify the automobile industry as negatively correlated with the oil industry. The scenarios assume that as the price of oil goes up, car sales go down — and vice versa. What if the growing trend toward purchasing electric vehicles makes this assumed negative correlation wrong? What if there is no longer a direct correlation between these industries? The Monte Carlo outcomes become suspect because they can’t be any better than the inputs being used.
The behavioral challenge with Monte Carlo is that advisors often show, and consumers often interpret, the Monte Carlo output to reflect probabilities. In other words, when the advisor says the model reports that in 85% of the runs it was safe to withdraw 4% from retirement capital each year, the consumer assumes, “I have an 85% probability that taking 4% a year will be safe; I can live with that!”
Remember that the output is based solely on the inputs. These inputs can be historical, relate to asset correlations or anticipate market trends. It’s accurate to indicate that in 85% of the trials, using these inputs on a random basis, the 4% withdrawal rate was safe. This doesn’t mean, however, that in real life there’s an 85% probability of success. For example, if the runs are based on historical investment returns from the 20th century, and 21st century doesn’t perform anywhere near like the past, the Monte Carlo model may prove grossly inaccurate.
The other behavioral challenge is that this approach tends to put the retiree in the uncomfortable situation of being in a pass/fail mindset. A consumer can’t live on averages. Their own results, not average results, are what matter.
A pandemic-related example
In our webinar, one of our experts used this example:
Say that in 2020, the Monte Carlo-generated strategy for the client reported an 85% success rate (or, stated differently, a 15% failure rate). Then, Covid hits and the stock market sours in 2021. A new Monte Carlo simulation shows the client’s failure rate has increased to 30%. The client, understandably, is going to wonder, “What did I do wrong? I’m taking the 4% withdrawal we agreed on, but now my chances of failure have doubled!”
A related behavioral issue associated with Monte Carlo is, what is a “good” probability of success, and is there a grade associated with a percent? For example, will a client assume an 80% success rate is “good,” but a 21% failure rate is too risky? It is difficult for consumers without backgrounds in statistics to interpret what the Monte Carlo output is saying.
A recent Prudential research survey found that approximately 80% of financial advisors use Monte Carlo in one way or another. Given the challenges with this approach, are there tools that can take us beyond Monte Carlo? Can we help the client figure out a retirement strategy without deciding purely based on the Monte Carlo probabilities? The answer to both of these questions is yes.
Beyond Monte Carlo
More and more often, retirement planners who use Monte Carlo simulations to develop a strategy are also employing other tools to help the client determine and manage withdrawals in retirement. A popular example is to create a withdrawal strategy and then include “guardrails” to tweak the actual drawdowns.
Setting limits on withdrawals reflects the boots-on-the-ground realities of changes people experience in retirement: markets change, income needs change. Rather than measure the strategy as either a success or a failure, the idea is to create guardrails so that the retiree can stay on course.
For example, if the stock market tanks by 20%, have your client agree in advance to reduce spending by, say, 10%. If, instead, stocks are in a sustained bull market, agree to put an upper limit on spending — even though the client’s retirement account is flush on paper.
These guardrail adjustments can reflect the retiree’s personal reality as well. If one year involves taking the kids to Disneyworld, the client’s withdrawal percentage from savings can be temporarily increased as long as the plan reflects an offsetting decrease in the future — to the extent needed. Monte Carlo can still be used behind the scenes to stress test these scenarios, but the client no longer has to make the decision based on an X% failure rate.
Working in ‘what ifs’
The newer breed of software tools seeks to present these strategies in ways the retiree can better see and use to make decisions. The retiree and advisor can work in “what ifs” as time passes. For instance, if the stock market plummets, the tools look beyond the headlines and address the retiree’s long-term concerns. It addresses, “If the market recovers, how will I look? If this poor performance sustains for years, will I need to adjust my spending?” Similarly, these kinds of tools can address the current issue of, “What happens if interest rates stay high?”
At one level, the software helps with the revenue side because it addresses how a targeted investment strategy can increase available income in retirement.
The software can also address the expense side. If the client wants to buy a house, it models what this drain on capital means for the future viability of the retirement portfolio; for example, will the client need to lower expenses to be safe in the future? Similarly, if the client has a change in expected life expectancy, whether good or bad, it’s helpful to have a tool that can project the effect this has on retirement spending. A shortened life expectancy may mean spending can increase
An example of software that seeks to go beyond Monte Carlo is Income Labs. An advisor can still use Monte Carlo outcomes to design an income strategy, but also use this kind of software to help clients set up guard rails, test “what if” scenarios, adjust investment allocations, and realistically look to future outcomes.
In the 1980s, a well-educated and dedicated advisor who used his HP 12C to calculate a possible income strategy could do a good job for a client planning to retire. By accumulating enough during the working years, creating a drawdown strategy in retirement, and monitoring the plan for internal and external changes, that 1980s retiree may still be flourishing today. It’s just that the migration from primitive tools like calculators to using advanced software has the potential to make the quality advisor all the more successful.
The advantage for advisors of using these kinds of tools is they help analyze data to project possible outcomes for different income strategies. They also help the advisor transition the client from an accumulation mindset to dealing with the realities of decumulating a retirement portfolio. Perhaps more important, an advisor can present these results so the client becomes a participant in the process rather than a mere observer.
In sum, these sophisticated, yet flexible software tools help both the advisor and client agree on how and when to make changes, whether the changes are because of market conditions or deviations in the client’s personal situation. As long as clients are living in this defined-contribution environment, an advisor will need powerful software tools to create workable solutions. And these tools must progress beyond just using Monte Carlo outputs.
To watch the webinar, click here. Steve Parrish, JD, RICP, CLU, ChFC, AEP, is the co-director of the Center for Retirement Income at The American College of Financial Services, where he also serves as an adjunct professor of advanced planning. With over 45 years’ experience as an attorney and financial planner, Parrish frequently addresses the financial challenges of individuals, business owners, and executives nationwide.