Tuesday, July 30, 2013

The Power of Diversification and Safe Withdrawal Rates

In the new issue of Advisor Perspectives, Geoff Considine has written an article called, "The Power of Diversification and Safe Withdrawal Rates." It is a response to, "Asset Valuations and Safe Portfolio Withdrawal Rates," written by David Blanchett, Michael Finke, and myself.

Fortunately, he first confirms the general findings from our article for the two asset classes we chose, U.S. large-cap stocks and U.S. government bonds.

But that is just a starting point. The purpose of his article is to show how the 4% rule is still viable in today's market environment by using broader portfolio diversification. Certainly, a limitation which much of the existing research on safe withdrawal rates experiences is that it only uses two simple asset classes. This is an extension of the idea that a lot of the early research was conducted using historical rolling period analysis, and there simply are not that many asset classes for which we have a sufficiently long dataset of historical returns.

In the article, Geoff Considine describes his own model for incorporating broader diversification into the mix to show the viability of the 4% rule.

Last year, I also presented a simpler version that was meant to help break readers free from whatever assumptions are made by the researchers writing any particular article on safe withdrawal rates. My approach is still not super easy, as you must at least know how to make a modern portfolio theory style efficient frontier based on a selection of asset class returns, volatilities, and cross correlations. But, at least if you can do that, you can come up with your own estimates for safe withdrawal rates by plotting the portfolio returns and volatilities on diagrams I made showing sustainable withdrawal rates for different time horizons and probabilities of failure.

The original research article about this was, "Capital Market Expectations, Asset Allocation, and Safe Withdrawal Rates," and further figures showing how this all comes into play for different retirement scenarios are provided at this blog entry.

Thank you to Geoff for emphasizing the important point that portfolio diversification matters, and with better diversification even in the current market environment, it may still be possible to put together a portfolio that will more confidently sustain the 4% withdrawal rate.

Caption: The figure below comes from a past blog entry which shows how users can customize their own portfolios, plot the arithmetic real return and volatility of their portfolio on the diagram, and see a visual representation of the sustainable withdrawal rate for a given failure rate and time horizon.


  1. An excellent post Wade, especially to your past blog entry that shows multiple time frames.

    These time frame illustrations are helpful for a retiree in a couple of ways. First, they can get an idea how to use the failure rates for decision making as my collaborators and I described in our Journal of Financial Planning paper (Mar 2012 – access through link via my name). Secondly, though the shorter the time frame the retiree should adjust the “Sustainable withdrawal rate” they get in your graph to mute the exponential growth nature of the withdrawal rate due to ever-shorter time frames. That adjustment we found is simply multiply the withdrawal rate in your graph by (1 – 1/n) (JFP Dec 2012 paper). Otherwise, the withdrawals grow exponentially too fast, and thus their portfolio value depletes exponentially too fast (exponential decay) with ever shorter time periods.

    Back to the first point: If they use the 5% or 10% failure rate graph (for whatever time frame that matched their expected longevity) as their point of reference they would get a dollar value they can withdraw for the year. In the past, it was hard to decide when the retiree should consider retrenching their spending just a little bit. Clearly, if their portfolio value goes down because of poor market returns during the year (or they spent more for an unplanned expense), and they keep taking the same dollar value out, then they could refer to the graph and see what their new percentage of withdrawal is now higher than they started the year out with (and that the line color may have changed). They could reduce their spending just a little by using the new portfolio value and the old sustainable rate again to ensure they aren’t spending money faster than the reality of their portfolio value suggests. But when is the withdrawal rate on the graph "too high?" Certainly more is better from the retiree's point of view. The next paragraph helps signal when spending may be too high ... it takes the use of two graphs to be fully informed.

    Our Mar 2012 paper determined that failure rates approaching 30% are indistinguishable from higher failure rates for all intents and purposes. Therefore, as we know, people dislike reducing their spending (one behavioral characteristic) and hope things might get better (another behavioral characteristic), especially if they have no clear signal or decision point to tell them they need to change (a third behavioral characteristic). Thus, your 20% failure rate graphs gives them the numerical signal they need to reduce spending back to the 10% (or 5% if that is what they started with) point. Some may say that the 10% graph also says this. It does, but the 20% graph reinforces the message since it also informs the retiree that they are in a higher danger zone (20% fail) while the 10% graph simply says their withdrawal rate is higher without reference to anything else as to what that may mean to their current spending pattern. You can see how this signaling works by using the same standard deviation and same rate of return between the two graphs. PS. Trying to change the allocation due to market sequence is akin to market timing and isn’t going to work (JFP Mar 2012 paper).

    Finally, the time period the retiree uses, as you point out in other posts, should be based on referencing a period life table so they are not simply guessing about how many years they may have remaining.

    An excellent post Wade that shows volatility does have an effect on what is feasible and prudent.

  2. I have some serious doubts about the conclusions of the Considine article. Note that he assumes a base annual arithmetic return of 8.3% for the S&P 500, which seems in the reasonable range, at least. But he assumes higher base returns for many other asset classes, for example: EAFE 11.1%; Emerging Mkt Equity, 13.4%; International REITS, 12.4%; Equity Energy, 12.9%. (Admittedly, all these also have higher estimated volatily than the S&P 500, reducing geometric return.) Still, it's not too difficult to see how it would be possible to get higher returns from a portfolio if one allows the use of asset classes with higher expected returns. The underlying question is whether it is reasonable to assume such high returns for so many asset classes.

  3. I agree with Anonymous. I believe that our profession has unintentionally harmed many retirees because of the use of 'historic' average returns that occurred during the greatest Bull market since the early 1900's. Most of today's planners and advisors haven't experienced a longer term Bear market. I also believe that we are in a New Era of investing that requires us to re-think some of our basic assumptions. Between 1950 and 2000, it was a period of slow, continuous change that lent itself to longer-term planning. Now, I believe we are in a time of rapid, discontinuous change that will reduce the accuracy of longer-term planning. We are all taught that 'past performance is not an indicator of future performance', yet that is what we as planners are taught to base returns on. I realize I am in the minority, but still think it is important to express a divergent opinion. Keep up the good work!