Retirement Spending Experts

On episode 289 of the Rational Reminder podcast, the guests were retirement spending researchers, David Blanchett, Michael Finke, and Wade Pfau.  The spark for this discussion was Dave Ramsey’s silly assertion that an 8% withdrawal rate is safe.  From there the podcast became a wide-ranging discussion of important retirement spending topics.  I highly recommend having a listen.

Here I collect some questions I would have liked to have asked these experts.

1. How should stock and bond valuations affect withdrawal rates and asset allocations?

It seems logical that retirees should spend a lower percentage of their portfolios when stocks or bonds become expensive.  However, it is not at all obvious how to account for valuations.  I made up two adjustments for my own retirement.  The first is that when Shiller’s CAPE exceeds 20, I reduce future stock return expectations by enough to bring the CAPE back to 20 by the end of my life.  These lower return expectations result in spending a lower percentage of my portfolio after doing some calculations that are similar to required minimum withdrawal calculations.  I have no justification for this adjustment other than that it feels about right.  

The second adjustment is on equally shaky ground.  When the CAPE is above 25, I add the excess CAPE above 25 (as percentage points) to the bond allocation I would otherwise have chosen in the current year of my chosen glidepath.  Part of my reasoning is that when stock prices soar, I’d like to protect some of those gains at a time when I don’t need to take on as much risk.

Are there better ideas than these?  What about adjusting for high or low bond prices?

2. How confident can we be that the measured “retirement spending smile” reflects retiree desired spending levels?

I find that the retirement spending smile is poorly understood among advisors (but not the podcast guests).  In mathematical terms, if S(t) is real spending over time, then dS/dt has the smile shape.  Many advisors seem to think that the spending curve S(t) is shaped like a smile.  I’ve looked at many studies that examine actual retiree spending in different countries, and there is always evidence that a nontrivial cohort of retirees overspend early and have spending cuts forced upon them later.  Both overspending retirees and underspending retirees seem to have the dS/dt smile, but at different levels relative to the x-axis.  Overspenders have their spending decline slowly initially, then decline faster, and then decline slowly again.  Underspenders increase their real spending early on, then increase it slower, and finally increase it quickly at the end.

I don’t see why I should model my retirement on any data that includes retirees who experienced forced spending reductions.  The question is then how to exclude such data.  I saw in one of Dr. Blanchett’s papers that he attempted to exclude such data for his spending models.  Other papers don’t appear to exclude such data at all.  In the end, it becomes a matter of choosing how high the smile should be relative to the x-axis.  If it is high enough, the result becomes not much different from assuming constant inflation-adjusted spending.

Advisors tend to work with wealthy people who save well and may have difficulty increasing their spending to align with their wealth.  So, it’s not surprising that good advisors would embrace research suggesting that retirees should spend more.  However, it’s not obvious to me that all retirees should spend at a high level early with the expectation that they simply won’t want to spend as much later in retirement.  It may be true that healthy people in their mid-80s choose to spend less, but I’ve seen the spending smile results applied in such a way that retirees are expected to reduce real spending each year right from the second year of retirement.

3. How can retirees deal with the gap between annuities in theory and annuities in practice?

The idea of annuitizing part of my portfolio is appealing.  Eliminating some longevity risk brings peace of mind.  However, whenever I compare annuity examples from papers or books to annuities I can actually buy, there is a gap.  Payouts are lower, and inflation protection doesn’t exist (at least in Canada where I live).

In my modeling, I find the optimal allocation to annuities is very sensitive to payout levels.  Further, when I treat inflation as a random variable, fixed payout annuities are unappealing.  It’s possible to buy an annuity whose nominal payout increases by, say, 2% each year, but this is a poor substitute for inflation protection.  If I had bought an annuity before the recent surge in inflation, I’d be looking at a substantial permanent drop in the real value of all my future payouts, and I’d be facing the possibility that it might happen again in the future.


I appreciated the thoughts of the three guests on the podcast.  My guess is that my additional questions are not easy ones.