Wednesday, April 11, 2012

Safe Withdrawal Rates and Retirement Date Market Conditions

Classic safe withdrawal rates studies such as the works of William Bengen and the Trinity study investigate sustainable withdrawal rates from rolling periods of the historical data, giving us an idea about what would have worked in the past. For a 30-year retirement period, we can learn about the historical sustainable withdrawal rates beginning up to 30 years ago. The question remains as to whether those past outcomes provide sufficient insight about what can reasonably be expected to work in the future.
The general problem with attempting to gain insights from the historical outcomes is that future market returns and withdrawal rate outcomes are connected to the current values of the sources for market returns. Writing in his 2009 book Enough, John Bogle makes this point:
My concern is that too many of us make the implicit assumption that stock market history repeats itself when we know, deep down, that the only valid prism through which to view the market’s future is the one that takes into account not history, but the sources of stock returns. (page 102, original emphasis)
These sources include income, growth, and changing valuation multiples. Future stock returns depend on dividend income, growth of the underlying earnings, and changes in the valuation multiples placed on those earnings. If the current dividend yield is below its historical average, then future stock returns will also tend to be lower. When price-earnings multiples are high, markets tend to exhibit mean reversion and relatively lower future returns should be expected. In case this point is not fully clear yet, John Bogle added in his book:
But no, the contribution of dividend yields to returns depends, not on historic norms, but on the dividend yield that actually exists at the time of the projection of future returns. With the dividend yield at 2.3 percent in July 2008, of what use are historical statistics that reflect a dividend yield that averaged 5 percent - more than twice the present yield? (Answer: None.)
Returns on bonds, meanwhile, depend on the initial bond yield and on subsequent yield changes. Low bond yields will tend to translate into lower returns due to less income and the heightened interest rate risk associated with capital losses if interest rates rise.
Sustainable withdrawal rates are intricately related to the returns provided by the underlying investment portfolio. Current market conditions are much more relevant than historical averages. Past historical success rates are not really the type of information that current and prospective retirees need for making their withdrawal rate decisions.
An alternative way to look at the historical data is to consider not just the past withdrawal rate outcomes, but rather to consider how past withdrawal rates were related to the retirement date values of the underlying sources of returns. When looking at 30-year retirements with historical simulations, we can only consider retirements beginning up to the early 1980s. Because of sequence of returns risk in which early retirement events matter much more, recent market conditions only show up at the end of these retirements and have little bearing on their outcomes. This is a matter of much more than just academic interest, because market conditions have witnessed historical extremes in recent years.
For Robert Shiller’s cyclically-adjusted price-earnings ratio (PE10), the highest value in a January before 1980 was 27.1, which happened in 1929, the year of the stock market crash leading to the Great Depression. Since 1980, though, PE10 was above 27.1 in 8 years, including 1997-2002, 2004, and 2007. A highpoint of 43.8 occurred in January 2000.  Meanwhile, for the pre-1980 period, the dividend yield reached a historic low of 2.7 percent in January 1973. But in every January since 1996 (except 2009), the dividend yield was below this value. Finally, the nominal 10-year bond yield reached a historic low of 1.95 percent in January 1941, and though levels in recent years are not that low, they are also not particularly high and cannot help much to counteract the other factors which may tend to depress sustainable withdrawal rates. With these extreme market conditions, the past may not serve as prologue.
For at least a decade, this issue has been debated ad nauseam at personal finance discussion boards, but it was Michael Kitces’ May 2008 issue of The Kitces Report that brought the issue to wider attention. He divided the historical PE10 values into quintiles and then showed the lowest and highest sustainable withdrawal rates within each quintile. He concluded that retirees should be extra cautious when retiring at times with high PE10 ratios, but his focus was more in the other direction, i.e. that retirees who observe a low PE10 value at retirement (below 12) could safely increase their withdrawal rate to 5.5 percent.
An article I wrote for the August 2011 Journal of Financial Planning, “Can We Predict the Sustainable Withdrawal Rate for New Retirees?” picks up where Michael Kitces left off by combining his idea with a regression technique developed by John Campbell and Robert Shiller in their 1998 article, "Valuation Ratios and the Long-Run Stock Market Outlook," from the Journal of Portfolio Management. Campbell and Shiller found predictive power for market valuations to explain real stock returns over the subsequent 10 years.
I expand that to explain past sustainable withdrawal rates (for inflation-adjusted withdrawals over a 30-year period and using a 60/40 asset allocation of large-capitalization stocks and 10-year government bonds) using retirement date market valuation levels, and dividend and bond yields. I then use the fitted regression model from the past historical data to also predict the maximum sustainable withdrawal rates for retirees in more recent years. Figure 4 shows two things: first that the regression model provides estimates that fit the historical data pretty well, and second the continued predictions from the regression model for the sustainable withdrawal rates of more recent retirees. Again, though we do not know the path of the blue curve in recent years because we don’t have enough data, I can make these predictions with the red curve by including the values of market conditions for more recent retirement dates. Figure 4 tells the story:  

With this approach, the news for recent retirees is not good. The 4 percent withdrawal rate rule cannot be considered as safe for U.S. retirees in recent years when stock market valuations have been at historical highs and the dividend yield has been at historical lows. Despite the peak for PE10 in 2000, I find that sustainable withdrawal rates may continue to decline after 2000 as the continued falling dividend yields and bond yields offset falling earnings-valuation levels. The model predicts sustainable withdrawal rates falling below three percent since 1999, and even below two percent since 2003.

I do hope that withdrawal rates in recent years will not actually fall this low. In the past 15 years, financial markets have really been sailing in uncharted waters. We have never experienced such high valuation multiples and such low dividend yields. This makes it difficult for the model to make predictions for withdrawal rates, as it must make predictions outside the range of historical observation. I think the real lesson from this exercise is that using a 4% withdrawal rate from a portfolio of risky assets is not as safe as the historical outcomes would lead us to believe.

More generally, can retirees have any inclination for whether they are retiring at a time which will allow for a relatively high or low withdrawal rate? I think so. Retirement date market return sources provide a tool to help predict how much retirees can sustainably withdraw from their portfolios. The relationship between these variables and withdrawal rates is stronger than what had first caught Campbell and Shiller’s attention regarding 10-year stock returns. The predictive power is far from perfect, and readers must ultimately judge this matter for themselves, but in my view the model fit is explained by a theoretically sound relationship that is likely to remain relevant in the future.