Are high failure rates acceptable for retirees? The answer we commonly hear is NO. Rory Terry made the most extreme case I have seen in this regard, as he argued that even a 1% risk of failure is too high given the costs to retirees for living their final years with all of their wealth exhausted. More commonly, withdrawal rate studies generally have an implicit assumption built into them that acceptable failure rates should not be above 10%.
But two recent studies challenge this conventional wisdom. These are “Spending Retirement on Planet Vulcan: The Impact of Longevity Risk Aversion on Optimal Withdrawal Rates” by Moshe A. Milevsky and Huaxiong Huang in the March/April 2011 Financial Analysts Journal, and “Determining Optimal Withdrawal Rates: An Economic Approach” by Duncan Williams and Michael Finke in the Fall 2011 issue of Retirement Management Journal.
Their arguments make a lot of sense. Mostly, I have been researching about “safe withdrawal rates,” which just asks what withdrawal rate a retiree may use to keep their wealth from running out. Implicitly, failure rates are meant to be kept low, hence the “safe” part of the term. But these studies tend to abstract away from two important points. First, retirees generally have other sources of income besides their retirement savings portfolio, such as Social Security or other defined-benefit pensions. Second, people who are thinking rationally and not just acting in a short-sighted manner may be willing to accept higher chances for failure in order to benefit from being able to spend more when they are younger and at least know they are still alive and able to enjoy life.
The Milevsky and Huang study focuses specifically on what they call “longevity risk aversion.” Living a long life is great, but the scary thing about living longer than you anticipate is that you might run out of wealth while still alive. For some people, this fear may be stronger than for other people. For those whom the fear is stronger, the answer is to spend less early in retirement in order to preserve more for later. For those whom the fear is not so strong, a higher spending rate can be used with the acceptance that there is a greater chance of wealth depletion later on. The authors investigate the impacts of these attitudes using a dynamic programming model that simplifies to use constant real asset returns in order to isolate the specific impact of “longevity risk aversion.” The only financial asset is TIPS, and it has a fixed real return. The only uncertain matter that retirees face in their model is that they don’t know how long they will live.
Milevsky and Huang summarize their findings with one sentence: “The optimal portfolio withdrawal rate depends on longevity risk aversion and the level of pre-existing pension income.” (page 49 of their article). The less fearful they are about a long life, and the more guaranteed income sources they have to rely on even if their other wealth is gone, the higher will be the optimal withdrawal rate they are willing to use. I think this is a really important insight! [Note break: Actually, I had read the article by Williams and Finke first, whom I believe honestly did independently discover this insight as well. I was actually visiting with them at Texas Tech University last week when they learned about the earlier article. I applaud both sets of authors in this regard.]
As well, Milevsky and Huang find that retirees will optimally plan to reduce spending as they get older to account for the fact that they have a lower probability of survival to later ages. [Note break: I discussed research here by Ty Bernicke suggesting that retirees spend less as they get older due to becoming less active and traveling less and eating out less, etc. But that is not the argument here. Rather, the argument is that since I know the probability of surviving to age 80 or 90 is low, I should specifically plan to spend less at those ages. I can understand why that is rational, but it seems to me that I would rather plan to be able to continue spending the same amount at higher ages, even though I have a lower probability of making it that long.]
Milevsky and Huang also show how actuarially fair annuities will help to increase spending levels across the retiree’s entire remaining lifespan.
Milevsky and Huang deserve the credit for the important insight I mentioned above, which they summarize again on page 53 of their article, “Planning to deplete wealth by some advanced age is neither wrong nor irrational.”
Moving on now to the article by Duncan Williams and Michael Finke, I do think this article still provides three important contributions beyond those already found in Milevsky and Huang’s work. First, Williams and Finke do also consider the issues of asset allocation, providing guidance about the optimal stock/bond allocation as well as about the optimal withdrawal rate.
Second, unlike the Milevsky and Huang approach that builds in decreasing spending levels over time, Williams and Finke to assume constant inflation-adjusted withdrawal amounts for as long as possible until wealth is gone, and then withdrawals drop immediately to zero. I think this better matches what people will really want to do. Well, of course no one uses precisely constant inflation-adjusted withdrawals, but I think this is closer to the reality.
Third, though the Williams and Finke approach is also based on a utility maximizing framework, they developed a much more simplified approach using what is called “certainty equivalence.” If you think long and hard back to your principles of microeconomics class in college, you already learned what this is and can understand the theoretical background for what Williams and Finke are doing. The dynamic optimization model used by Milevsky and Huang is much more complicated and hard to understand. Another great quotation from Milevsky and Huang’s article is found on page 46, “Unfortunately, much of the financial planning community has ignored these dynamic optimization models, and nowhere is this ignorance more evident than in the world of ‘retirement income planning.’” What I’ve heard from a very well-regarded financial planner whom I’ve had the privilege to discuss this with, is that planners are not keen to get involved with matters unless they can fully explain them to their clients. It seems to me that this is the stumbling block, and so the beauty of the Williams and Finke article is how it can make these important insights with much simpler mathematics than usually found in these sorts of research articles.
These three factors work together to suggest that the framework developed by Williams and Finke will provide more practical and implementable guidelines for financial planners and prospective retirees.
In August when I first learned about the Williams and Finke article, I wrote a post called “Retirement Withdrawal Rates 2.0.” These insights are what I was alluding to at that time. I’ve written a short article called “Retirees and Utility Maximization” in which I attempt to explain the findings of the Williams and Finke article in a completely nontechnical way. I hope to publish this somewhere in the coming days, and I will definitely provide a link when the article is available. So let me hold off saying too much more for now since this is getting rather long anyway. But what their article is doing is to incorporate other income sources and risk aversion into the safe withdrawal rate literature to find the optimal withdrawal rate and asset allocation choices. They find, as well, that more aggressive retirees and retirees with access to larger amounts of guaranteed income sources can go ahead and enjoy a higher withdrawal rate and a higher stock allocation. This may dramatically increase their failure rates as defined in traditional withdrawal rate studies, but that can be OKAY after all.
And that is the main lesson from both of these important research articles.