Thanks for the link. Excellent suggestion to discuss this! 

I can't really agree with the methodology: Monte Carlo and then targeting a 90% success rate (=10% failure rate). I would not accept a 10% failure rate in my personal finances.

But then again, Monte Carlo gives results that are consistently too conservative because you can't reliably simulate the strong bull markets after a market blowup. That might explain in the end the numbers are not too far away from what I would use. The net effect of a) worse than reality results from MC and b) a failure probability that's too high.

From page 20: This quote didn't age well. They should have at least written a short caveat about the most recent inflation spike. It's probably transitory. But for how long?

Somewhat counterbalancing these concerns is the good news about inflation. Over the past century, the U.S. inflation rate has never before been this low for this long. Nor has it been so stable. Previously, periods of low inflation were punctuated by spikes—or even bouts of deflation. Yet since the early '90s, inflation has almost always been under 5%, while gradually trending downward. Should the inflation rate remain steadily low, as the bond market anticipates, sustainable withdrawal rates should be higher.

Agree with your points! A few others:

1. Seems like they are taking a sort of "hybrid Monte Carlo" approach in that their model uses historical data but then shuffles it up within a band of time. Still don't love it as you don't get the pattern of market recovery after drawdown:

From 1950 through 1980, a portfolio that consisted of 50% stocks and 50% fixed-income securities returned an average of 8.20% per year, with an annualized standard deviation of 7.77%. The average rate of inflation was 4.01%. The calculation for the safe withdrawal rate for the 1950-79 time period creates 1,000 hypothetical return patterns that cluster around these averages.

2. I like to see the following as we still so often read advice that portfolios should be more conservative in retirement:

For real withdrawal rates over 30 years, opting for more stocks has never been a mistake.

3. I found this one a little provocative:

In hindsight, we know that each equity bear market was followed by a recovery that allowed retirees to continue with their plan. At the time, no such certainty existed. Nor does it in the future. Although stocks have always recovered from their ills during the 70 years covered by this study, that doesn’t mean that they inevitably must.

4. A point you have made many times:

The primary reason why safe withdrawal rates have been unstable has not been the effect of economic recessions but instead that of changing inflation rates.

5. On current inflation trending, they did add the slightest of caveats:

Third, inflation is low—rising at the time that this paper is written but, nevertheless, over the past several years as low as it has been over the past 70 years.

6. A real indictment on the good ol' 4% rule here. Even over 30 years they have it failing a quarter of the time!

According to the model, the 4.0% withdrawal rate would survive the full time period,
while making all scheduled payments, on 74% of occasions.

7. I found this one interesting. Does this run counter to your findings?

The highest safe withdrawal rates come from portfolios that hold 30% to 60% stock positions. This occurs not because fixed-income securities are expected to perform particularly well but instead because of stocks’ volatility.

8. They really like Guyton-Klinger-style guardrails. Much is said about it, this is just one snippet:

The guardrails approach was among the most efficient we tested, meaning that periodic course corrections help the retiree consume more of the portfolio in up markets but not too much in bad ones.

So with all that said, here are the real money quesitons:

1. They are saying 3.2% over 30 years at 70/30 allocation with 10% failure and full capital depletion (page 21). The SWR Toolbox, with those same parameters, yields 4.03%, even at CAPE>20 and SPX at all-time high. (Note when they say 70/30 it seems their 30 is split into 20% bonds and 10% T-bills)

2. They have 40-year 70/30 with 10% failure at 2.7% (same page). Toolbox has it at 3.71%

3. They have a 30-year 50/50 failsafe at a brutal 1.9%! (page 45). Toolbox has it at 3.77%

Tell me if I got the Toolbox wrong, but these are pretty big differences. How do we reconcile it? Or should we not bother based on their approach?