Pfau's attempt to predict SWR

[fosterscik]

Phau article

This may be of interest to some, but obviously suffers from trying to extrapolate from fitted data.

Predicted Withdrawal Rate = 12.15 - 2.47 x log(PE10) + 0.05 x Bond Yield
This equation informs us that withdrawal rates can be expected to rise when interest rates rise, while they will fall when PE10 rise. The precise numbers in this equation were estimated from the historical relationship between sustainable spending rates and these variables. We take the natural logarithm of PE10 in order to fit a more precise curve for the relationship between market valuations and spending rates.

Exhibit 1 shows the historical sustainable spending rates from 1881 to 1986. A hypothetical 1937 retiree—net of advisory and mutual funds fees—saw a historical SAFEMAX of 3.95%.

The figure also provides the estimates developed through this equation for the entire time period through the start of 2016. Until 1986, we can see that the relationship between the predictions and the actual values is close. The two variables (PE10 and interest rates) can explain 55% of the fluctuations in the historical spending rates.

He "predicts" those who retired in 2000 should limit themselves to ~3%. A few years later he predicts ~4%. Right now,

4.16%. While this is very close to the historical SAFEMAX, it is important to emphasize that this is not an estimate of the “safe” withdrawal rate.

I'm looking to retire mid 2017 so of interest to me.

 

[jebmke]

4.16 - is he sure it won't be 4.15%?

Personally, 4% plus or minus 2% is probably a reasonable range to work with. So 2-6%.

 

[big-papa]

4.16 - is he sure it won't be 4.15%?

Personally, 4% plus or minus 2% is probably a reasonable range to work with. So 2-6%.

Haha. Saw another post somewhere today about precision vs. accuracy.

Still, 4% plus/minus 2% is a +/-50% difference in annual withdrawal. If only I end up in a situation where that I could live with that

 

[Totoro]

+1 on that.

It's even worse actually: a factor x3, going from 2% to 6%.

 

[braumeister]

Measure with a micrometer.
Mark with a piece of chalk.
Cut with an axe.

 

[Dash man]

Measure with a micrometer.

Mark with a piece of chalk.

Cut with an axe.

+1

Followed by Burning in the fireplace!


Enjoying life!

 

[Midpack]

Measure with a micrometer.
Mark with a piece of chalk.
Cut with an axe.

+2. If your retirement income will depend largely on withdrawals - unless you're VERY conservative, no matter what methodology you adopt for withdrawals, you'll have to be prepared to adjust withdrawals/course correct over the course of your (+/-30 yr) retirement. And even then you're unlikely to 'die broke.'

[I was really impressed with the first paper I read from Pfau, I haven't been as impressed with anything tha followed. More and more he seems to promote annuities. YMMV]

 

[donheff]

What am I misunderstanding about this formula. Today's PE/10 is 26.9. log(26.9) = 1.47 so ignoring the small bond rate adjustment you would subtract 3.6309 (2.47*1.47) from 12.15 = 8.52. That is a far cry from 4% so what am I calculating wrong?

 

[big-papa]

What am I misunderstanding about this formula. Today's PE/10 is 26.9. log(26.9) = 1.47 so ignoring the small bond rate adjustment you would subtract 3.6309 (2.47*1.47) from 12.15 = 8.52. That is a far cry from 4% so what am I calculating wrong?

Probably should be natural log. 12.15-ln(26.9)=4.02

Big-papa

 

[athena53]

Pretty hard to plan when they keep flip-flopping. For years 4% was the general rule. After the financial crisis hit the experts started saying well, maybe 2 or 3%. I was 55 back then; not really possible to ramp up savings over such a short period to support 2 or 3%. Fortunately, we've got enough saved that we can get by on 2% in bad years if necessary.

 

[big-papa]

Pretty hard to plan when they keep flip-flopping. For years 4% was the general rule. After the financial crisis hit the experts started saying well, maybe 2 or 3%. I was 55 back then; not really possible to ramp up savings over such a short period to support 2 or 3%. Fortunately, we've got enough saved that we can get by on 2% in bad years if necessary.

And Michael Kitces, Pfau's sometimes collaborator, has interesting info on the original 4% rule after looking back at the financial crisis:
https://www.kitces.com/blog/how-has...he-tech-bubble-and-the-2008-financial-crisis/

Pfau has a dashboard on his website to that estimates, among other things, a withdrawal rate based on current conditions, but he hasn't updated it since Nov. 15th...

 

[donheff]

Probably should be natural log. 12.15-ln(26.9)=4.02

Big-papa

That does it but for slow math types (like me) it would have been preferable if Wade used ln instead of log in the formula. Still at 4% in this over valued world. Makes me feel safe.

 

[Midpack]

And Michael Kitces, Pfau's sometimes collaborator, has interesting info on the original 4% rule after looking back at the financial crisis:
https://www.kitces.com/blog/how-has...he-tech-bubble-and-the-2008-financial-crisis/

Cutting to the chase FWIW...

The bottom line, though, is simply to recognize that even market scenarios like the tech crash in 2000 or the financial crisis of 2008 are not ones that will likely breach the 4% safe withdrawal rate, but merely examples of bad market declines for which the 4% rule was created. In turn, this is an implicit acknowledgement of just how conservative the 4% rule actually is, and how horrible the historical market returns really were that created it. In the end, this doesn’t necessarily mean that the 4% rule is ‘sacred’ and that some future market disaster couldn’t be bad enough to undermine it (and of course, it can/should still be adopted for individual circumstances like a longer/shorter time horizon, the impact of taxes, the impact of fees and other investment costs, etc.). But when the Great Depression and the stagflationary 1970s couldn’t break it, and the crash of 1987 and even the global financial crisis of 2008 were just speed bumps, it will take a lot to set a new safe withdrawal rate below 4%!

 

[big-papa]

That does it but for slow math types (like me) it would have been preferable if Wade used ln instead of log in the formula. Still at 4% in this over valued world. Makes me feel safe.

Yep. Maybe even better if you don't think the world is overvalued. There are those who believe that looking at PE10, for example, based on its long-term average (or even median) doesn't take into account accounting rules changes that happened in the last couple of decades. Not sure there is really any way to know what ahead of time, though...

 

[GalaxyBoy]

That does it but for slow math types (like me) it would have been preferable if Wade used ln instead of log in the formula. Still at 4% in this over valued world. Makes me feel safe.

+1

Although in the text of the article he says it's natural log. Enginerds like me get very annoyed at stuff like this.

 

[FIRE'd@51]

If by explanatory power Pfau means the regression has an r-squared of 55%, it seems like a lot of tracking error (standard error of the estimate) in the predicted SWR from his fitted equation. He doesn't tell us what the standard deviation of the calculated SWR's versus the predicted SWR's was. If we knew that number, we could multiply it by 0.67 (the square root of 1 - 0.55) to estimate the tracking error.

 

[photoguy]

Although in the text of the article he says it's natural log. Enginerds like me get very annoyed at stuff like this.

In all the matrix/stats software I've used, it's common to have the log function mean natural logs. However, I think my old HP calculator explicitly used "log" for base 10 and "ln" for natural logs. Maybe this is a changing custom depending on how old you are?

 

[SumDay]

Not to derail the thread or anything, but we were all told to get "w*rk" Twitter accounts to help promote our company. Several months after I got mine and re-tweeted all the pre-approved, corporate-speak nuggets they tell us to share, Wade Pfau started following me. If he only knew... I found it hilarious.

 

[Taxman59]

Measure with a micrometer.
Mark with a piece of chalk.
Cut with an axe.

+1

 

[big-papa]

This is also being discussed over at bogleheads.
https://www.bogleheads.org/forum/viewtopic.php?t=195191

 

[Big_Hitter]

there is a huge difference between precision and accuracy