Monday, February 9, 2015

The Sustainable Withdrawal Range

I had an interesting discussion this past week at Adviser Perspectives with two financial advisers who are frustrated by the fact that there are a wide range of recommendations for sustainable withdrawal rates. I sympathize with their frustration, but their suggestion of getting the industry to agree on one specific model of the future that would provide a single, agreed sustainable withdrawal rate isn’t a reasonable solution.

We could get every meteorologist in America to agree that the high temperature in Chapel Hill next Friday will be 42 degrees, but that wouldn’t make it any more likely that the prediction would be correct. In fact, it would be less likely. Different models with different predictions give us a range of possible outcomes to consider. When you can’t accurately predict something, like future market returns or future temperatures, providing upper and lower bounds for the most likely range is the next best information to have.

Let’s look at the current predictions for future sustainable withdrawal rates. The original SWR studies by William Bengen predict a 95%-safe SWR of about 4.4% for a 30-year retirement with a 50% equity portfolio. Wade Pfau et al recently produced a study suggesting that, based on today’s low-return environment, 3.5% might be a better guess. Even Bengen commented that Pfau might be onto something. (If you follow Wade Pfau's blog, by the way, he has a new website at, where you will need to re-subscribe to his email posts.)

Bengens’s approach uses historical market returns, assuming that the future will look like the past. Pfau et al use Monte Carlo simulation based on lower expected returns in the future than we have seen historically. Moshe Milevsky’s formula for probability of ruin using stochastic calculus calculates a 95% safe withdrawal rate of about 3.25%. In his paper, Milevsky notes that his formula often produces withdrawal rates significantly lower than many advisers recommend. Depending on the spending level, Milevsky’s calculation can differ from simulation results dramatically.

There are other studies that predict safe rates both higher and lower than these. The discussion at Adviser Perspectives was about why we can’t just all decide on one approach using the same assumptions and settle on one sustainable withdrawal rate. In other words, which model is right? The reason we can't is that these are all completely justifiable opinions about the future and we can’t know which model will work best. Any of them might turn out to be right.

The safest bet would be that the future 30-year SWR will not be 3.25%, 3.5% or 4.4% precisely. I would bet, however, that the correct answer will turn out to be not much lower than 3.25% and not much higher than 4.5% because that is the range several models suggest. I would plan for the possibility that it will be significantly lower.

If that sounds like a hedge instead of a commitment, that’s exactly what it is. Financial advisers hoping to hear “3.4%" or even “3.3% to 3.5%” would be disappointed.

These are not insignificant differences. If the actual SWR turns out to be 3.25%, a retiree will need to have saved 31 times his retirement income shortfall after Social Security benefits and pensions. If it is 4.5%, he will “only" need to have saved about 22 times that shortfall. If the shortfall is $10,000 a year, those savings requirements would be $307,692 and $222,222.

The wide discrepancy of recommended sustainable withdrawal rates is not a problem with the models that predict them, it is a result of our inability to predict the future of market returns. It is impossible to prove that any of the models are incorrect. . . well, not for 30 years, anyway.

Human beings have a poor record of predicting the future for even a few years, let alone for thirty. A little more than five years ago, there were widespread predictions that by not taking a path of austerity out of the Great Recession we would soon see rampant inflation. The inflation rate last year was 0.8% and deflation seems possible today. The EU took the austerity path and is trying to avoid an existential deflationary spiral. Both predicted their way would be best.

Studies show that “experts” are no better at predicting the future than us non-experts. Investment manager, Ken Fisher, used to project the coming year’s market return by looking at the projections of the same handful of “market experts” every year. He noticed that actual returns usually fell in the gap that no expert had predicted and that there was always such a gap. In other words, he simply chose the return that no one else had chosen. This worked eerily well for several years until others caught on. (Once everyone is playing the same game, no one can win.)

The wide range of projections is a result of our inability to predict market returns and the length of retirement, and thereby SWR’s, not an ability to agree on a model.

Financial risk is defined as the uncertainty of outcomes. Future sustainable withdrawal rates cannot be identified with a great deal of precision, so different models and different assumptions, all reasonable, produce widely disparate estimates of sustainable rates. This is just proof of what we already knew – sustainable withdrawal rates is a risky strategy.

Some advisers at Adviser Perspectives asked how they should communicate this complicated information when a client asks, “How much can I spend each year for the next 30 years and be 95% certain that I won't outlive my savings?" Here is what I would say to a client (or reader):

That amount is impossible to identify with any accuracy because we can’t predict future market returns or know how long you and your spouse will live. The current estimates from a wide range of models and assumptions range from about 3.25% to about 4.5% of your initial portfolio value for the first year, assuming your life expectancy is about 30 years. That percentage, by the way, increases as you age. It could approach 10% of your remaining portfolio balance near the end of your retirement. I would recommend a guess near the low end of the range because that will be safest, but that will also significantly reduce the amount you can spend. I would also recommend that you have a backup plan in case the sustainable rate turns out to be even lower than we expect, because it certainly could. I realize this is a broad estimate, but that’s because SWR is unpredictable, which is the financial definition of “risky.” If that’s more uncertainty than you are comfortable with, there are safer, more predictable spending strategies we can discuss.

Yes, its complicated and probably not what a client wants to hear. But, it is honest and that’s what clients need to hear.


  1. An internet wag says that the answer is easy as pie. 3.14%.

    1. Ha! A fellow punster. And it is entirely possible that will turn out be the number 30 years from now, just not for the reason suggested.

  2. One of the important things to realize about the SWR is that it is still probabilistic with a likelihood of failure. Just as important as coming up with the rate is also understanding what the 5% looks like so that if it is failing, then corrective measures can be taken before it becomes completely unrecoverable.

    So an important question is what are the characteristics of the 30 year rolling returns or the Monte Carlo distributions that cause the 5% probability of failure? Do they have distinguishing characteristics that can be identified in time? Does a certain type of sequence of return in the first 5-10 years cause a much higher probability of failure than the 5% of the total population of returns? If that is the case, then a Plan B can be identified by planners and their clients and put in place if the sequence shows up.

    1. That's correct, no matter how low the probability of failure with SWR, it is still non-zero. Ask yourself, if your savings were depleted at age 85 would you look back and think, "It was a good bet. I just lost it and I'm OK with that," or would you think, "why did I take risk with something so important?" The answer would probably depend on your Plan B, often called the floor portfolio.

      (We could call this the "Pete Carroll Question." Alas, Pete's strategy had no Plan B.)

      Regarding identifying a failing path in time, SWR is not a set-and-forget strategy. You should re-calculate a safe spending rate every year or whenever the key factors change significantly (portfolio value, age and expected returns). By doing so, instead of identifying when you're on a failing path, you will know how much you can spend and avoid the failing path.

      I disagree with your suggestion regarding identifying a failed path "in time." Should your portfolio value drop dramatically in a short time, like 2007-2009, you will have already lost standard of living. Plan B needs to be in effect from the beginning, not put in place when you see your plan is failing. Jumping out of a plane seconds before it crashes isn't a great backup strategy.

  3. Here is an article I wrote recently on the "range" provided for safe withdrawals from 3 different types of withdrawal strategies:


  4. The range you discuss here Dirk relates to concepts in the comments we had for your following week's post (13 Feb) ... the bottom end of the range is probably a good signal for setting the floor, or necessary expenses. The upper end of the range suggests maximum spending FOR THE PRESENT YEAR. The difference is called discretionary spending.

    Should markets decline DURING THE PRESENT YEAR, discretionary spending goes away and the necessary expenses are retained (unless of course the year is particularly bad like 2008 and one's allocation more aggressive than it should be; in which case additional retrenchment may be necessary).

    Having a 3 or 4 year buffer from which spending comes from helps smooth out potential rapid spending reductions too. So the moral of the story is to have a well developed Plan A with specific Decision Rules that research shows make a difference (changing allocations based on which way the wind blows is not one of those).

    As always - we're in agreement saying the same thing differently ... except you're gift for writing says it much better!