Tuesday, December 10, 2013

New Column: How Much Can Clients Spend in Retirement? A Test of the Two Most Prominent Approaches

My new column, "How Much Can Clients Spend in Retirement? A Test of the Two Most Prominent Approaches" is now available at Advisor Perspectives. In the column, I simulate and describe the spending paths created by two of the best known variable withdrawal strategies: Jonathan Guyton's Decision Rules and the actuarial approach developed by Larry Frank/John Mitchell/David Blanchett as it culminated in David's September article providing a spreadsheet to calculate sustainable withdrawal rates on a year-by-year basis throughout retirement.

In other news, yesterday's Wall Street Journal webcast panel discussion with Bill Bengen, David Blanchett, and I went well and has over 10,000 views already. The replay is available. I must say, 30 minutes is just not enough time to build any momentum for the discussion, but a lot was covered nonetheless. Hopefully there can be more of these in the future. 

Let me clarify one point, as I know some people will be confused about why Bill Bengen speaks of the 4.5% rule instead of the 4% rule. The answer is based, again, on worst-case scenarios in US history. When the historically much more volatile small-cap stock index is added (and overweighted as about 40% of the overall portfolio) to the investment mix along with large-cap stocks and intermediate term government bonds, the worst-case withdrawal rate in history was 4.58%. However, I am less confident than him about whether it is appropriate to consider this as a forward-looking safe withdrawal rate for today's retirees.
It's a snow day here in Philadelphia and please take care wherever you are.


  1. A very good summary Wade.
    A couple of questions/points as lead researcher on the papers Blanchett developed the formulas:
    1) Blanchett’s simplified formulas did NOT adjust the withdrawal rate as we proposed in our “superannuation” paper (JFP Dec 2012 http://www.fpanet.org/journal/transitionthrougholdage) where I think Blanchett’s formual should be restated wr% = 1/n * (1-1/n) to incorporate those results. This would temper the dramatic exponential increase in the withdrawal rate and resulting exponential decrease in portfolio values. Did you make such and adjustment?

    2) Our paper (JFP Mar 2012 http://www.fpanet.org/journal/AnAgeBasedThreeDimensionalDistributionModel/) discussed a method of using the life tables to “pull” or “push” consumption into earlier or later years. This method provides a rational way to use life tables as a method to measure and monitor spending impacts on later years. This allows the spending of retirees to be customized to their specific needs rather than rote application of a rule of thumb.

    3) The volatility of spending assumed in using a constant POF/variable WR approach can be easily tempered through using lower (10% POF) as calculation points for a consistent dollar amount that year, and decision rules to change spending if POF goes up (signaled by rising WR) to 20 to 30%, or down (signal potential extra funds at that moment in time).

    Basically, it is hard to compare nuances of dynamic processes. It is good you do compare and discuss them so practitioners may get a better understanding - an understanding they need to better use the processes. Another great article Wade!

    1. Thanks Larry. Let me answer:

      (1) I'm aware of that nice 1/n * (1-1/n) rule and will look at it more in the future, but as this was a test of Blanchett's spreadsheet rule, I didn't include it

      (2) I need to have another look at this!

      (3) Yes, I think this is a good idea for smoothing spending. I'm still working on these issues in the background.

      Thanks for the kind words.

    2. 1) This adjustment would improve the final years in all figures for "Blanchett" results and have a small effect in the early years, if applied there too. In practice I don't apply the rule until late 70's or so, depending on specifics of health and client desires or concerns.

      Both 1) and 2) combined provide a method that provides interesting conversation points with clients, then realize their spending decisions, spend more now vs save it to spend later, have nuances to outcomes for cash flow and portfolio balances (that can be monitored and measured as your graphs illustrate nicely). There is no free lunch like spend more now and spend more later (unless the markets and economies are kind enough in the future to support that - an unknown today).

      Great stuff Wade as you put it all together.

  2. Character limit on comment wouldn't let me incorporate a link that illustrates point 3 application of POF based decision rule for retrenched spending based on pre-calculated portfolio values for current monthly spending; and what that spending may need to be reduced to if portfolio values reach decision points (to get spending back to 10% POF level).

    In practice, monthly spending remains pretty consistent applying spending rules instead of changing the spending implied by a constant POF approach (where a constant WR is applied at any given time to the portfolio balance at that time).

    The annual review that practitioners do with clients would re-evaluate everything each year.

  3. And then I forget the link anyway! http://blog.betterfinancialeducation.com/sustainable-retirement/how-much-is-too-much-taking-too-much-in-retirement/

  4. Oh Snap! David Blanchett's recently proposed actuarial withdrawal strategy has already become more "prominent" than mine? Sadly this is probably true. But, the result of your comparison between the two strategies that you selected is not surprising to me, as you did not incorporate an algorithm for the actuarial approach to smooth experience results from year to year. See my post of October 11 of this year for a description of a recommended algorithm.


    1. Ken,

      I'm sorry, I don't pick titles and my suggested "Two of the Most Prominent" was morphed into "The Two Most Prominent"

      I did recently review your approach again and appreciate that you have a smoothing algorithm built in. That's important!

  5. Interestingly analysis, but this seems to be mixing a tad apples and oranges. G-K, when used with sensible parameters, tends to conserve the portfolio value 'forever' (I did a ton of back testing on it, so I am quite familiar with its dynamics). While the Blanchett method (which I am less familiar with) seems to aim at spending most of the portfolio by the time of death. Very different goals for retirees, and rather different withdrawal profile to be rightfully expected. It doesn't seem quite right to compare those two methods side to side?

    This actually begs a higher level question. How to define criteria to compare withdrawal methods? This isn't an easy question. I have my own opinion, but would be curious to hear your thoughts on this topic (as I suspect you're going to do more work on this general theme of withdrawal methods).

    1. Siamond,

      I'd agree with your assessment of G-K in the sense that it is a version of a safe withdrawal rate strategy which is meant to work in worst cases, and therefore allows wealth to grow on average. The same is true for the 4% rule.

      And you are right that Blanchett focuses more on spending down wealth.

      But I don't think this characteristic of G-K was ever explicitly stated by them. I do think the comparison is still useful because it can give people a better idea of what approach they will be more comfortable with. And if G-K does preserve wealth, then that is important to know.

    2. I would add that the rote application of Blanchett's formulas would tend to consume wealth by design, especially using his wr=1/n rule, without some tempering adjustment like 1/n(1-1/n), because of the exponential nature of later spending due to shorter distribution periods.

      However, not compared in Wade's assessment in this article is the use of life tables to push consumption (inheritance or bequest goal of retiree while still spending some of their portfolio during their lifetime) into later years where the retiree may be longer lived. The paper that discusses this is referenced in my point 2 above.

      Using both longevity percentiles combined with 1-1/n adjustment tends to also preserve wealth for potential later consumption. A knowledgeable practitioner into current methods is like today's surgeon who can customize surgery based on what they're seeing with each patient.

      I applaud Wade's foray into comparisons since it broadens knowledge and insight through thought and discussion that wouldn't otherwise happen. The very purpose and outcome of research in my opinion.

    3. Wade, fair enough, one could indeed use G-K in a somewhat aggressive or somewhat conservative manner (pick your initial SWR accordingly), hence map to various wealth spend, conservation or growth goals. Your 5.5% choice was clearly somewhat aggressive for the AA you selected. I guess I was essentially pointing out that a meaningful comparison should probably be constrained by the context of predefined high-level goals.

      What about my 2nd question? Did you give some thoughts to comparison criteria?

    4. Sorry for overlooking your second question.

      I have a couple of articles which include comparison criteria:

      Most recently:
      “An Efficient Frontier for Retirement Income.” Journal of Financial Planning (February 2013) [P] [WP]

      and, earlier, these two which go together:

      “Choosing a Retirement Income Strategy: A New Evaluation Framework.” Retirement Management Journal (Fall 2012). [P] [WP]

      “Choosing a Retirement Income Strategy: Outcome Measures and Best Practices.” Retirement Management Journal (Fall 2012) [P] [WP]

      You can find links on the tab above for "articles and columns". That's still how I'm thinking about things.

      Also there is Moshe Milevsky's Stochast Present Value concept which I am very recently digging deeper into as well.