Author Topic: Kitces Article - Ratcheting SWR - Data Discrepancy?  (Read 11539 times)

brooklynguy

  • Handlebar Stache
  • *****
  • Posts: 2204
  • Age: 38
Kitces Article - Ratcheting SWR - Data Discrepancy?
« on: June 04, 2015, 08:26:30 AM »
Kitces' latest article, "The Ratcheting Safe Withdrawal Rate -  A More Dominant Version of the 4% Rule?", is an interesting read.  In a nutshell, it summarizes the outlandish historical success rate of using a 4% withdrawal plan and highlights the 4% Rule's "problem" of leaving most any retiree who rigidly adheres to it with too much leftover money (as a result of its deliberate status as a worst case planning tool).  None of this should be news to anyone, but Kitces does attempt to quantify some variable spending plans that ratchet up spending under favorable conditions (but never ratchet down spending) and have zero negative impact on the historical success rate.

My specific question, though, is about Kitces' underlying data, which is significantly more optimistic than cFIREsim or FIREcalc.  For example, his data shows a 4% WR as being the absolute historical safe-max for a 60/40 stock/bond portfolio (with 4% succeeding in the historical worst case year of 1966) (see the first chart in the article).  However, cFIREsim shows 3.42% as the absolute safe-max WR using the same parameters.  The discrepancy is not limited to this article; it's consistent across the data presented in all of Kitces' articles.  I'm trying to pinpoint the reason for the discrepancy in the data, but haven't been able to after some quick research (both Kitces and cFIREsim seem to use data from Shiller's website, and cFIREsim's results are still more pessimistic if the investment fees are set to zero, so the discrepancy can't be explained by differences in fee assumptions).  Does anyone happen to know the answer?

arebelspy

  • Administrator
  • Senior Mustachian
  • *****
  • Posts: 27226
  • Age: -999
  • Location: Traveling the World
Re: Kitces Article - Ratcheting SWR - Data Discrepancy?
« Reply #1 on: June 04, 2015, 08:29:55 AM »
I don't know why Kitces' data differs from FIRECalc, but I do know FIRECalc doesn't calculate bond returns correctly so it offers answers too high, thus why cFIREsim looks more pessimistic (but is actually more accurate) than FIRECalc.
We are two former teachers who accumulated a bunch of real estate, retired at 29, and now travel the world full time with two kids.
If you want to know more about me, or how we did that, or see lots of pictures, this Business Insider profile tells our story pretty well.
We (occasionally) blog at AdventuringAlong.com.
You can also read my forum "Journal."

brooklynguy

  • Handlebar Stache
  • *****
  • Posts: 2204
  • Age: 38
Re: Kitces Article - Ratcheting SWR - Data Discrepancy?
« Reply #2 on: June 04, 2015, 08:36:02 AM »
Yeah, I've read about that too.  But I'm curious why Kitces' data differs so significantly from cFIREsim - it's a pretty wild discrepancy (4% vs. 3.42% for absolute-safe WR is a pretty huge difference, and his 60/40 success rate is higher than what cFIREsim reports for 100% equities).  Kitces' stuff is generally very meticulous, so the answer is probably out there if we do enough digging.

skyrefuge

  • Handlebar Stache
  • *****
  • Posts: 1007
  • Location: Suburban Chicago, IL
Re: Kitces Article - Ratcheting SWR - Data Discrepancy?
« Reply #3 on: June 04, 2015, 10:10:01 AM »
I think clearest route to find the answer is to ask why Kitces is different from FIRECalc, not why he's different from cFIREsim. Because we know why cFIREsim is different than FIRECalc (the bond calculation, which is a cFIREsim innovation that I don't imagine Kitces has copied).

The easiest answer would be that Kitces is using 95%-success to determine the SWR, not 100%-success. I agree that his wording somewhat implies that he's using 100%-success, but I don't see him explicitly state that. So he could be using 95% (and maybe even just forgotten that the 95% is implicit in his calculations). If so, 60/40 in FIRECalc with everything else at default gives 4.02%, which perfectly matches his graphs.

dandarc

  • Magnum Stache
  • ******
  • Posts: 3242
  • Age: 36
Re: Kitces Article - Ratcheting SWR - Data Discrepancy?
« Reply #4 on: June 04, 2015, 10:21:47 AM »
Sort of a much more conservative version of the "retire over and over" idea - if I'm reading this right, your portfolio goes up 50%, so you increase your baseline by 10%.  Which results in resetting to < 3% withdrawal rate, which the data seems to indicate is very safe.  If you hit a 2nd ratchet, you're WR is < 2.3%.  So even this plan is potentially unnecessarily conservative.

Maybe I'm just missing this in the article, but are we talking a real or nominal 50% increase in the portfolio?  The above assumes real.

brooklynguy

  • Handlebar Stache
  • *****
  • Posts: 2204
  • Age: 38
Re: Kitces Article - Ratcheting SWR - Data Discrepancy?
« Reply #5 on: June 04, 2015, 10:30:20 AM »
The easiest answer would be that Kitces is using 95%-success to determine the SWR, not 100%-success. I agree that his wording somewhat implies that he's using 100%-success, but I don't see him explicitly state that. So he could be using 95% (and maybe even just forgotten that the 95% is implicit in his calculations). If so, 60/40 in FIRECalc with everything else at default gives 4.02%, which perfectly matches his graphs.

But Kitces shows us the actual maximum safe withdrawal rates broken down by starting year in the first chart in the article (where it shows 4% as being safe for the worst case start year of 1966).  So his data isn't just saying that 4% is the SWR in order to achieve a 95% historical success rate -- his data is saying that in the actual historical worst case year, a 4% WR succeeded (that is, that 4% is the SWR using 100%-success as the benchmark).

skyrefuge

  • Handlebar Stache
  • *****
  • Posts: 1007
  • Location: Suburban Chicago, IL
Re: Kitces Article - Ratcheting SWR - Data Discrepancy?
« Reply #6 on: June 04, 2015, 10:50:52 AM »
But Kitces shows us the actual maximum safe withdrawal rates broken down by starting year in the first chart in the article (where it shows 4% as being safe for the worst case start year of 1966).  So his data isn't just saying that 4% is the SWR in order to achieve a 95% historical success rate -- his data is saying that in the actual historical worst case year, a 4% WR succeeded (that is, that 4% is the SWR using 100%-success as the benchmark).

Duh. Yeah. I even looked at the chart, but apparently had temporarily forgotten everything I know about SWR research!

Ok, next best guess then is that Kitces is using whatever data/methods the old guys used that's different than FIRECalc. For example, the Trinity study (pdf) shows a 100% success for a 75/25 portfolio, while FIRECalc (even with expenses at 0) shows 4 failures.

Or Pfau says that Bengen's calculations gave the 50/50 portfolio SAFEMAX as 4.15% (in 1966), while zero-expense FIRECalc has it as 3.79%.

brooklynguy

  • Handlebar Stache
  • *****
  • Posts: 2204
  • Age: 38
Re: Kitces Article - Ratcheting SWR - Data Discrepancy?
« Reply #7 on: June 04, 2015, 11:13:04 AM »
Sort of a much more conservative version of the "retire over and over" idea - if I'm reading this right, your portfolio goes up 50%, so you increase your baseline by 10%.  Which results in resetting to < 3% withdrawal rate, which the data seems to indicate is very safe.  If you hit a 2nd ratchet, you're WR is < 2.3%.  So even this plan is potentially unnecessarily conservative.

Maybe I'm just missing this in the article, but are we talking a real or nominal 50% increase in the portfolio?  The above assumes real.

Yes, it is a version of the "retire again and again" strategy, but we know there is a glitch in cFIREsim that prevents it from correctly modeling that strategy as you pointed out here (it only ever ratchets up once, not "again and again" as it should).

The article is really unclear about whether the 50% threshold is nominal or real - it starts off talking in nominal terms (which is why the terminal values in the second chart are crazy-high), then switches gears to inflation-adjusted.  My guess is that the 50% threshold is in real terms, because the data in the "Ratched Spending Increases" graph kinda-sorta matches what I would expect after looking at cFIREsim and FIREcalc's output graphs, but it's very hard to tell.

Ok, next best guess then is that Kitces is using whatever data/methods the old guys used that's different than FIRECalc. For example, the Trinity study (pdf) shows a 100% success for a 75/25 portfolio, while FIRECalc (even with expenses at 0) shows 4 failures.

Or Pfau says that Bengen's calculations gave the 50/50 portfolio SAFEMAX as 4.15% (in 1966), while zero-expense FIRECalc has it as 3.79%.

Yeah, that might make sense.  But I always kind of implicitly assumed that the reason for those discrepancies is just that the old guys did their research decades ago and therefore used different data, but the addition of the intervening more recent periods to the data set should not have done anything to lower the safemax as long as the worst case historical periods were all already included in the data sets used in those studies (but is that the case?  Bengen's original 4% rule study was published in 1994, so it couldn't have included the 1966 start year).

brooklynguy

  • Handlebar Stache
  • *****
  • Posts: 2204
  • Age: 38
Re: Kitces Article - Ratcheting SWR - Data Discrepancy?
« Reply #8 on: June 04, 2015, 12:11:58 PM »
Ok, next best guess then is that Kitces is using whatever data/methods the old guys used that's different than FIRECalc. For example, the Trinity study (pdf) shows a 100% success for a 75/25 portfolio, while FIRECalc (even with expenses at 0) shows 4 failures.

Or Pfau says that Bengen's calculations gave the 50/50 portfolio SAFEMAX as 4.15% (in 1966), while zero-expense FIRECalc has it as 3.79%.

I took a closer look at these articles.  In the Pfau post, is he saying that he replicated Bengen's methods to arrive at the 4.15% safemax (in 1966) for a 50/50 portfolio?  (He can't be saying that Bengen included the 30-year period starting in 1966 in his backtest when he published in 1994, right?!)

Maybe when I have a chance I will spend some time looking at the original underlying methodologies of the old guys' research and see if I can determine why they are producing different results than FIREcalc.  I'd be grateful if anyone else with the inclination and the willingness wants to join me (or, better yet, beat me to it!).

arebelspy

  • Administrator
  • Senior Mustachian
  • *****
  • Posts: 27226
  • Age: -999
  • Location: Traveling the World
Re: Kitces Article - Ratcheting SWR - Data Discrepancy?
« Reply #9 on: June 04, 2015, 12:34:24 PM »
Have you tried contacting Kitces (via his blog or email or Twitter or whatever)?  He'll probably have a good idea.
We are two former teachers who accumulated a bunch of real estate, retired at 29, and now travel the world full time with two kids.
If you want to know more about me, or how we did that, or see lots of pictures, this Business Insider profile tells our story pretty well.
We (occasionally) blog at AdventuringAlong.com.
You can also read my forum "Journal."

brooklynguy

  • Handlebar Stache
  • *****
  • Posts: 2204
  • Age: 38
Re: Kitces Article - Ratcheting SWR - Data Discrepancy?
« Reply #10 on: June 04, 2015, 12:42:24 PM »
Have you tried contacting Kitces (via his blog or email or Twitter or whatever)?  He'll probably have a good idea.

Now you're thinking outside the box!  :)

Sometimes I tend to overcomplicate problems so much that the simplest solutions escape me...

I'll ask Kitces and report back here for anyone interested.

dandarc

  • Magnum Stache
  • ******
  • Posts: 3242
  • Age: 36
Re: Kitces Article - Ratcheting SWR - Data Discrepancy?
« Reply #11 on: June 04, 2015, 01:06:42 PM »
Have you tried contacting Kitces (via his blog or email or Twitter or whatever)?  He'll probably have a good idea.
But guessing and speculation is so much more fun!

MDM

  • Walrus Stache
  • *******
  • Posts: 8960
Re: Kitces Article - Ratcheting SWR - Data Discrepancy?
« Reply #12 on: June 04, 2015, 01:16:34 PM »
I'll ask Kitces and report back here for anyone interested.
Following for interest.

forummm

  • Walrus Stache
  • *******
  • Posts: 7358
  • Senior Mustachian
Re: Kitces Article - Ratcheting SWR - Data Discrepancy?
« Reply #13 on: June 04, 2015, 03:32:31 PM »
Some of his other articles people have posted here have also had much higher SWRs than what I expected. There's something different that he's doing in his other articles too.

Manguy888

  • Stubble
  • **
  • Posts: 248
  • Location: Rhode Island
    • EA Mann, Writer
Re: Kitces Article - Ratcheting SWR - Data Discrepancy?
« Reply #14 on: June 05, 2015, 08:18:00 AM »
Following for interest

frugalnacho

  • Magnum Stache
  • ******
  • Posts: 3232
  • Age: 36
  • Location: Madison Heights, Michigan
Re: Kitces Article - Ratcheting SWR - Data Discrepancy?
« Reply #15 on: June 05, 2015, 03:04:41 PM »
I'll ask Kitces and report back here for anyone interested.
Following for interest.
me too

arebelspy

  • Administrator
  • Senior Mustachian
  • *****
  • Posts: 27226
  • Age: -999
  • Location: Traveling the World
Re: Kitces Article - Ratcheting SWR - Data Discrepancy?
« Reply #16 on: June 05, 2015, 03:16:15 PM »
With all these interested folks you might have to do more than leave a blog comment that he apparently doesn't reply to. ;)

He seems fairly active on Twitter...
We are two former teachers who accumulated a bunch of real estate, retired at 29, and now travel the world full time with two kids.
If you want to know more about me, or how we did that, or see lots of pictures, this Business Insider profile tells our story pretty well.
We (occasionally) blog at AdventuringAlong.com.
You can also read my forum "Journal."

brooklynguy

  • Handlebar Stache
  • *****
  • Posts: 2204
  • Age: 38
Re: Kitces Article - Ratcheting SWR - Data Discrepancy?
« Reply #17 on: June 05, 2015, 03:29:02 PM »
With all these interested folks you might have to do more than leave a blog comment that he apparently doesn't reply to. ;)

Was just thinking the same thing.  I can't let down three interested people and counting!

Quote
He seems fairly active on Twitter...

I'm allergic to Twitter.  I have an account that I treat as "read only" to get updates on a handful of people/sites I follow but I've never posted anything and wouldn't even know how to go about doing so.

I'll try emailing him, but if he's not responding to questions about the article in the article's own public comments section I doubt I'll have better luck by email.

Eric, if you're reading this, are you the same Eric that left the blog comment about extra-long retirement periods?  If so, might as well include that question in my email too.

brooklynguy

  • Handlebar Stache
  • *****
  • Posts: 2204
  • Age: 38
Re: Kitces Article - Ratcheting SWR - Data Discrepancy?
« Reply #18 on: June 06, 2015, 09:28:04 AM »
I just emailed Kitces and received the following (nearly instantaneous!) response:

brooklynguy:

Quote
Thanks for the great article on the ratcheting SWR strategy! I had two questions about it I was hoping you could address:

1. Is the 50% threshold for the ratcheting rule based on the nominal value of the portfolio, or the real (inflation-adjusted) value? It wasn't clear from the article.

2. Your historical SWR data in the article (and your other articles) are significantly more optimistic than what is reported by www.firecalc.com and www.cfiresim.com. For example, those sources show the historical safemax withdrawal rate (in the worst case starting year of 1966) as 3.67% and 3.34%, respectively (with the difference between the two, I believe, due to differences in their methodologies for calculating historical bond returns). Do you happen to know why your data produces more optimistic results than either of these sources?

Kitces:

Quote
Thanks for reaching out!
 
The ratcheting rule I wrote there was based on nominal value (nominal value of 50% over original value). I havenít tested yet whether some inflation-adjusted ratcheting target would work better or worse (I hypothesize it might, but in the real world people tend to think in terms of their nominal wealth, not inflation-adjusted it back to a prior point, so I started there).
 
Regarding data differences between my work and FireCalc and CFireSim, I donít know those platforms well enough to respond. But it could well be differences in bond returns. I do my work using short-term government bonds as my bond proxy, as Iíve found separately (see https://www.kitces.com/blog/accelerating-the-rising-equity-glidepath-with-treasury-bills-as-portfolio-ballast/) that adding in duration risk with longer bonds produces longer SWRs (which may be the issue youíre seeing in the other tools). In addition, corporate bonds in general seem to produce lower SWRs than using government bonds (as corporate spreads tend to blow out in recessions, leading corporate bonds to underperform government bonds at the exact time you DONíT want them to because stocks tend to be crashing!), so if the bond proxy in the other tools is using corporate bonds, that too may account for the difference.
 
I hope that helps a little?

On the nominal vs. real point, I'm glad to hear the ratcheting rule was based on nominal values -- it makes the extraordinary historical success of that specific spending strategy that much more impressive!  This brief study actually goes a pretty long way towards addressing the question of how to determine when your portfolio has gotten "too big to fail" (at least, from a historical perspective, like all SWR research in general).

His take on bond returns is also interesting, but I don't know if that is sufficient to explain the discrepancies between his data and FIRECalc's and cFIREsim's.  As I said earlier in this thread, Kitces' reported success rates for a 60/40 portfolio are more optimistic than what FIRECalc and cFIREsim report for even more aggressive portfolios (including 100% equities, with zero investment fees) -- if what Kitces describes above does explain the discrepancy by itself, that would mean adding a substantial short-term government bond allocation significantly increased the historical success rates (which I can't imagine actually being the case, but it's worth exploring in more detail).

DavidAnnArbor

  • Handlebar Stache
  • *****
  • Posts: 1889
  • Age: 53
  • Location: Ann Arbor, Michigan
Re: Kitces Article - Ratcheting SWR - Data Discrepancy?
« Reply #19 on: June 06, 2015, 07:09:41 PM »
Kitces' reported success rates for a 60/40 portfolio are more optimistic than what FIRECalc and cFIREsim report for even more aggressive portfolios (including 100% equities, with zero investment fees) -- if what Kitces describes above does explain the discrepancy by itself, that would mean adding a substantial short-term government bond allocation significantly increased the historical success rates (which I can't imagine actually being the case, but it's worth exploring in more detail).

Long term government bonds would do even better than short term government bonds during a recession.

Eric

  • Magnum Stache
  • ******
  • Posts: 4017
  • Location: On my bike
Re: Kitces Article - Ratcheting SWR - Data Discrepancy?
« Reply #20 on: June 08, 2015, 02:57:09 PM »
Eric, if you're reading this, are you the same Eric that left the blog comment about extra-long retirement periods?  If so, might as well include that question in my email too.

That was me, yes.  Although I commented on it before you started this thread and your questions are better than mine.  I've always enjoyed Kitces as a breath of optimism among a large group of pessimists (or pessimism is probably too strong, but at least conservative by nature), but now I'm wondering if that optimism is slightly misplaced based on the variances mentioned.

MDM

  • Walrus Stache
  • *******
  • Posts: 8960
Re: Kitces Article - Ratcheting SWR - Data Discrepancy?
« Reply #21 on: June 08, 2015, 03:41:51 PM »
His take on bond returns is also interesting, but I don't know if that is sufficient to explain the discrepancies between his data and FIRECalc's and cFIREsim's.
...and the Trinity Study authors', and Bengen's, and Pfau's, etc.

One would first want to know the assumed annual returns for each asset class used in the simulations.  Then, have an explicit list of all other assumptions, e.g., investment fees, rebalancing strategy, withdrawal timing, etc.

Given the same inputs, there is no reason to have different results unless someone simply made a mistake.  Given different inputs (thus the question becomes "what is different?"), getting different results is unsurprising.

Anyone know all the inputs used by the various authors?

brooklynguy

  • Handlebar Stache
  • *****
  • Posts: 2204
  • Age: 38
Re: Kitces Article - Ratcheting SWR - Data Discrepancy?
« Reply #22 on: June 08, 2015, 05:31:10 PM »
That was me, yes.

Sorry, I got an itchy trigger finger and sent Kitces the email before waiting for your response.  But if you email him your question he'll probably get back to you quickly (if not instantaneously -- apparently, the man rivals any member of this forum for deserving recognition as "most likely to be a computer program"; seriously, he got my email, stopped whatever else he was doing (on a Saturday morning, no less), read it, and replied with the detailed response quoted above, all in the space of six minutes).

...and the Trinity Study authors', and Bengen's, and Pfau's, etc.

One would first want to know the assumed annual returns for each asset class used in the simulations.  Then, have an explicit list of all other assumptions, e.g., investment fees, rebalancing strategy, withdrawal timing, etc.

Given the same inputs, there is no reason to have different results unless someone simply made a mistake.  Given different inputs (thus the question becomes "what is different?"), getting different results is unsurprising.

Anyone know all the inputs used by the various authors?

Yeah, this is what I was getting at with my earlier suggestion about examining the underlying methodologies and assumptions of the studies in order to pinpoint the discrepency in the results.

skyrefuge

  • Handlebar Stache
  • *****
  • Posts: 1007
  • Location: Suburban Chicago, IL
Re: Kitces Article - Ratcheting SWR - Data Discrepancy?
« Reply #23 on: June 08, 2015, 08:14:48 PM »
apparently, the man rivals any member of this forum for deserving recognition as "most likely to be a computer program"

Which is unfortunate, actually! I think we were all hoping that he'd be a slower thinker, more humanly curious, and instead say "hmm...I don't know why there are discrepancies in the results. But thank you for pointing them out! For my next research project I will do a meta-analysis of SWR research to determine how and why different sources come to different results."

Because, really, he's probably more suited to do the job than any of us!

And it seems like it's a pretty important topic that deserves more exploration. Yeah, we know that any SWR number that gets spit out needs some error bars around it, but given how often SWR numbers are reported, it would seem important to at least understand how replicable the research is, and to what extent different data sources or methodologies contribute to the size of the error bars.

FIRECalc gives the most insight into the effects of using different sources for "bonds". With a 60/40 portfolio, the safemax rate for each of the four "bond" options is:

Commercial Paper: 3.77%
Long Interest Rate: 3.67%
30 Year Treasury: 3.63%
5 Year Treasury: 3.71%

That's a smaller range than the range between the FIRECalc results and Kitces's results, so it suggests that even if "short-term government bonds" were used instead, they wouldn't fully explain the difference. So there must be more to it, and if different researches use different methodologies, there's a good chance that at least one of them are "wrong".

brooklynguy

  • Handlebar Stache
  • *****
  • Posts: 2204
  • Age: 38
Re: Kitces Article - Ratcheting SWR - Data Discrepancy?
« Reply #24 on: June 09, 2015, 07:18:18 AM »
I think we were all hoping that he'd be a slower thinker, more humanly curious, and instead say "hmm...I don't know why there are discrepancies in the results. But thank you for pointing them out! For my next research project I will do a meta-analysis of SWR research to determine how and why different sources come to different results."

Maybe I should write back to him with that exact suggestion for a future blog post, though I think it's a long shot that he'll take us up on it.

Quote
Because, really, he's probably more suited to do the job than any of us!

That's undoubtedly true, but, if necessary, I bet the brain trust in this forum can also do a decent job assuming the various studies and sources disclose enough information about their respective methodologies and assumptions.  Your own observation about the breakdown of FIRECalc results with differing "bond" categories is already a solid start to the investigation.

Quote
And it seems like it's a pretty important topic that deserves more exploration. Yeah, we know that any SWR number that gets spit out needs some error bars around it, but given how often SWR numbers are reported, it would seem important to at least understand how replicable the research is, and to what extent different data sources or methodologies contribute to the size of the error bars.

FIRECalc gives the most insight into the effects of using different sources for "bonds". With a 60/40 portfolio, the safemax rate for each of the four "bond" options is:

Commercial Paper: 3.77%
Long Interest Rate: 3.67%
30 Year Treasury: 3.63%
5 Year Treasury: 3.71%

That's a smaller range than the range between the FIRECalc results and Kitces's results, so it suggests that even if "short-term government bonds" were used instead, they wouldn't fully explain the difference. So there must be more to it, and if different researches use different methodologies, there's a good chance that at least one of them are "wrong".

We definitely need some massive error bars around any SWR figure when putting it into practice for a forward-looking plan, but, to MDM's point, there should really be no room for error whatsoever in all these backtest studies -- barring outright mistakes, with the given set of inputs, the outputs are precisely determinable.  So we need to figure out the differences in the inputs to understand why there are differences in the outputs.  To Eric's point, it does seem odd that Kitces stands alone in a sea of more pessimistic/conservative peers -- I doubt it's due to simple error, and the stated difference in bond asset class selection doesn't seem to fully explain it, so it would be interesting if he or we can figure it out.

Separately, I think the substantive issue indirectly addressed by the article of "how to know when your portfolio has become too big to fail" also warrants further exploration, but I believe Go Curry Cracker mentioned that he's working on a post on that topic and I'm eager to see what he does with it.

forummm

  • Walrus Stache
  • *******
  • Posts: 7358
  • Senior Mustachian
Re: Kitces Article - Ratcheting SWR - Data Discrepancy?
« Reply #25 on: June 12, 2015, 09:17:52 PM »
Maybe Kitces is using a different source for his ST Treasuries than cFIREsim and FIREcalc? I've seen some sources that differ from each other on historical S&P 500 (including dividends) and Treasury returns. I don't know why that would be.

brooklynguy

  • Handlebar Stache
  • *****
  • Posts: 2204
  • Age: 38
Re: Kitces Article - Ratcheting SWR - Data Discrepancy?
« Reply #26 on: June 15, 2015, 01:22:19 PM »
For some additional data points relevant to this discussion, see Pfau's latest post on his blog (originally appearing in an article on Forbes.com last week).  It is a recreation and update of the Trinity Study using data current through 2014 and using "intermediate term government bonds" for the bond allocation, which produced a 100% success rate for a 4% spending plan using a 50/50 allocation.

skyrefuge

  • Handlebar Stache
  • *****
  • Posts: 1007
  • Location: Suburban Chicago, IL
Re: Kitces Article - Ratcheting SWR - Data Discrepancy?
« Reply #27 on: June 16, 2015, 05:40:19 PM »
For some additional data points relevant to this discussion, see Pfau's latest post on his blog (originally appearing in an article on Forbes.com last week).  It is a recreation and update of the Trinity Study using data current through 2014 and using "intermediate term government bonds" for the bond allocation, which produced a 100% success rate for a 4% spending plan using a 50/50 allocation.

ha, is Pfau reading this thread? I just don't ever recall seeing a discussion of the effect of bond data choice on Bengen's vs. Trinity's results, so it's funny to see it suddenly pop up in this article! Ok, c'mon, Pfau, Kitces, either collaborate or compete to do the full-blown comparison of all the different SWR research methodologies! We're getting so close!

Because it's nice to get some clarity on the bond data used in those two original studies (as well as Pfau's update), but it's still not clear to me what methodology they use(d). Are they like FIRECalc, using only interest-rate data for bond returns, or do they attempt to incorporate price-appreciation as well like cFIREsim does?

Here's the post detailing cFIREsim's "innovation" in bond methodology: http://www.cfiresim.com/phpbb/viewtopic.php?f=2&t=119

arebelspy

  • Administrator
  • Senior Mustachian
  • *****
  • Posts: 27226
  • Age: -999
  • Location: Traveling the World
Re: Kitces Article - Ratcheting SWR - Data Discrepancy?
« Reply #28 on: June 16, 2015, 09:16:31 PM »
For some additional data points relevant to this discussion, see Pfau's latest post on his blog (originally appearing in an article on Forbes.com last week).  It is a recreation and update of the Trinity Study using data current through 2014 and using "intermediate term government bonds" for the bond allocation, which produced a 100% success rate for a 4% spending plan using a 50/50 allocation.

The Ramsey dig was awesome.  :D
We are two former teachers who accumulated a bunch of real estate, retired at 29, and now travel the world full time with two kids.
If you want to know more about me, or how we did that, or see lots of pictures, this Business Insider profile tells our story pretty well.
We (occasionally) blog at AdventuringAlong.com.
You can also read my forum "Journal."

forummm

  • Walrus Stache
  • *******
  • Posts: 7358
  • Senior Mustachian
Re: Kitces Article - Ratcheting SWR - Data Discrepancy?
« Reply #29 on: June 17, 2015, 08:11:32 AM »
For some additional data points relevant to this discussion, see Pfau's latest post on his blog (originally appearing in an article on Forbes.com last week).  It is a recreation and update of the Trinity Study using data current through 2014 and using "intermediate term government bonds" for the bond allocation, which produced a 100% success rate for a 4% spending plan using a 50/50 allocation.

The Ramsey dig was awesome.  :D

He advises an 8% SWR??? How did that not come up in the 2 recent Ramsey advice threads? Or did I just miss that?

arebelspy

  • Administrator
  • Senior Mustachian
  • *****
  • Posts: 27226
  • Age: -999
  • Location: Traveling the World
Re: Kitces Article - Ratcheting SWR - Data Discrepancy?
« Reply #30 on: June 17, 2015, 08:15:30 AM »
For some additional data points relevant to this discussion, see Pfau's latest post on his blog (originally appearing in an article on Forbes.com last week).  It is a recreation and update of the Trinity Study using data current through 2014 and using "intermediate term government bonds" for the bond allocation, which produced a 100% success rate for a 4% spending plan using a 50/50 allocation.

The Ramsey dig was awesome.  :D

He advises an 8% SWR??? How did that not come up in the 2 recent Ramsey advice threads? Or did I just miss that?

Yes, Ramsey says the stock market pretty much always earns 12% and you're safe taking 8% of your money.  That is one of his more egregious money mistakes. I thought it was common knowledge.
We are two former teachers who accumulated a bunch of real estate, retired at 29, and now travel the world full time with two kids.
If you want to know more about me, or how we did that, or see lots of pictures, this Business Insider profile tells our story pretty well.
We (occasionally) blog at AdventuringAlong.com.
You can also read my forum "Journal."

forummm

  • Walrus Stache
  • *******
  • Posts: 7358
  • Senior Mustachian
Re: Kitces Article - Ratcheting SWR - Data Discrepancy?
« Reply #31 on: June 17, 2015, 08:29:40 AM »
For some additional data points relevant to this discussion, see Pfau's latest post on his blog (originally appearing in an article on Forbes.com last week).  It is a recreation and update of the Trinity Study using data current through 2014 and using "intermediate term government bonds" for the bond allocation, which produced a 100% success rate for a 4% spending plan using a 50/50 allocation.

The Ramsey dig was awesome.  :D

He advises an 8% SWR??? How did that not come up in the 2 recent Ramsey advice threads? Or did I just miss that?

Yes, Ramsey says the stock market pretty much always earns 12% and you're safe taking 8% of your money.  That is one of his more egregious money mistakes. I thought it was common knowledge.

Wow. That's major malpractice. For some reason this makes me more upset than him steering people into the 6% load plus 1-2% fee funds he gets a cut of.

But it's nice to know that Ramsey thinks I'm FI!

arebelspy

  • Administrator
  • Senior Mustachian
  • *****
  • Posts: 27226
  • Age: -999
  • Location: Traveling the World
Re: Kitces Article - Ratcheting SWR - Data Discrepancy?
« Reply #32 on: June 17, 2015, 08:33:34 AM »
Yeah, you mostly see disagreement with it online (obviously--it's terrible advice).

https://www.google.com/search?q=dave+ramsey+safe+withdrawal+rate

Some financial planners called him out on it about two years ago, and rather than defend it, he got defensive and attacked them with some pretty rude tweets.

http://www.hullfinancialplanning.com/six-areas-where-i-disagree-with-dave-ramseys-investing-and-retirement-withdrawal-advice/

We are two former teachers who accumulated a bunch of real estate, retired at 29, and now travel the world full time with two kids.
If you want to know more about me, or how we did that, or see lots of pictures, this Business Insider profile tells our story pretty well.
We (occasionally) blog at AdventuringAlong.com.
You can also read my forum "Journal."

brooklynguy

  • Handlebar Stache
  • *****
  • Posts: 2204
  • Age: 38
Re: Kitces Article - Ratcheting SWR - Data Discrepancy?
« Reply #33 on: June 17, 2015, 09:26:22 AM »
ha, is Pfau reading this thread?

I had the same thought.  I still haven't gone back to the various studies to comb their footnotes for disclosures about their methodologies, but I don't ever recall seeing a discussion of the effect of bond data choice on SWR success rates either (outside the discussions that led to cFIREsim's innovation that you linked to above), and certainly not prominently featured in a relatively brief and informal article like this one.

The Ramsey dig was awesome.  :D

Yep, but too bad there's probably close to zero overlap between the subset of Ramsey's audience members who don't already know better and the readership of Pfau's blog (or Forbes.com).

arebelspy

  • Administrator
  • Senior Mustachian
  • *****
  • Posts: 27226
  • Age: -999
  • Location: Traveling the World
Re: Kitces Article - Ratcheting SWR - Data Discrepancy?
« Reply #34 on: June 17, 2015, 09:39:15 AM »
The Ramsey dig was awesome.  :D

Yep, but too bad there's probably close to zero overlap between the subset of Ramsey's audience members who don't already know better and the readership of Pfau's blog (or Forbes.com).

Hah, good point.  I can't imagine anyone reading Pfau's blog who thinks a 8% WR is safe in retirement.  :D
We are two former teachers who accumulated a bunch of real estate, retired at 29, and now travel the world full time with two kids.
If you want to know more about me, or how we did that, or see lots of pictures, this Business Insider profile tells our story pretty well.
We (occasionally) blog at AdventuringAlong.com.
You can also read my forum "Journal."

sol

  • Walrus Stache
  • *******
  • Posts: 7846
  • Age: 42
  • Location: Pacific Northwest
Re: Kitces Article - Ratcheting SWR - Data Discrepancy?
« Reply #35 on: June 17, 2015, 11:23:45 AM »
Hah, good point.  I can't imagine anyone reading Pfau's blog who thinks a 8% WR is safe in retirement.  :D

I can. Like anyone who is only planning for a 10 year retirement, either because they have short life expectancy or because they will draw an adequate pension.

The typical 4 percent rule is too aggressive for some folks, too conservative for others.  I think we don't necessarily do people any favor by blindly repeating it.

arebelspy

  • Administrator
  • Senior Mustachian
  • *****
  • Posts: 27226
  • Age: -999
  • Location: Traveling the World
Re: Kitces Article - Ratcheting SWR - Data Discrepancy?
« Reply #36 on: June 17, 2015, 11:36:37 AM »
Hah, good point.  I can't imagine anyone reading Pfau's blog who thinks a 8% WR is safe in retirement.  :D

I can. Like anyone who is only planning for a 10 year retirement, either because they have short life expectancy or because they will draw an adequate pension.

The typical 4 percent rule is too aggressive for some folks, too conservative for others.  I think we don't necessarily do people any favor by blindly repeating it.

You got me.  There are always exceptions.  For a 30-year ER, I wouldn't expect it to be safe, and for even longer, I think 8% would be suicide.  (But again..exceptions--if you have a pension, etc.)
We are two former teachers who accumulated a bunch of real estate, retired at 29, and now travel the world full time with two kids.
If you want to know more about me, or how we did that, or see lots of pictures, this Business Insider profile tells our story pretty well.
We (occasionally) blog at AdventuringAlong.com.
You can also read my forum "Journal."

sol

  • Walrus Stache
  • *******
  • Posts: 7846
  • Age: 42
  • Location: Pacific Northwest
Re: Kitces Article - Ratcheting SWR - Data Discrepancy?
« Reply #37 on: June 17, 2015, 11:55:11 AM »
You got me.  There are always exceptions. 

It only comes up for me because I'm an exception.  My assets need to last considerably less than 30 years, so I'm planning a higher than 4% withdrawal rate.

brooklynguy

  • Handlebar Stache
  • *****
  • Posts: 2204
  • Age: 38
Re: Kitces Article - Ratcheting SWR - Data Discrepancy?
« Reply #38 on: June 17, 2015, 12:05:06 PM »
You got me.

I don't think so - in the context of this discussion, it was clear that your statement was shorthand for all the bells and whistles normally associated with a SWR-based retirement plan (itself shorthand for a description of the full list of assumptions and caveats that underlie the 4% rule of thumb and its offspring).  In my view, anyone planning an early retirement who relies on blind advice without putting in the time and effort to properly educate themselves about these matters does themselves a disservice, and we actually are doing the community a favor when we avoid hanging an asterisk on every exception-prone statement we make.  Skyrefuge once wisely chastised me for pointing out an obscure exception to an exception to a response to a simple question asked by a novice member of the community, lest every thread in this forum become an encyclopedia-length treatise that scares away many would-be mustachians!

Jeremy

  • Stubble
  • **
  • Posts: 137
  • Location: Taiwan
  • We retired in our 30's and are full time travelers
    • Go Curry Cracker!
Re: Kitces Article - Ratcheting SWR - Data Discrepancy?
« Reply #39 on: June 19, 2015, 09:04:19 PM »
Back when I was reading all of the original 4% rule documents, I noted some deltas between Bengen and the Trinity Trio

I'm not sure which is which, so this isn't really useful, but I seem to recall that one used corporate bonds and the other used US govt bonds (medium?), one did withdrawals at the beginning of the year the other at the end (so first withdrawal done at end of first year or retirement)  and I believe the Trinity study did monthly reallocation as opposed to annual (cFIREsim)

The Shiller data source for stock dividends, bond yield, etc... is all averaged for a time period (monthly), which would cause deltas to any analysis using specific dates from other sources



The Kitces analysis is somewhat interesting: if your original withdrawal rate is now less than 2.67% of the portfolio value, feel free to bump it up a bit

brooklynguy

  • Handlebar Stache
  • *****
  • Posts: 2204
  • Age: 38
Re: Kitces Article - Ratcheting SWR - Data Discrepancy?
« Reply #40 on: June 19, 2015, 09:51:40 PM »
I'm not sure which is which, so this isn't really useful, but I seem to recall that one used corporate bonds and the other used US govt bonds (medium?), one did withdrawals at the beginning of the year the other at the end (so first withdrawal done at end of first year or retirement)  and I believe the Trinity study did monthly reallocation as opposed to annual (cFIREsim)

Yeah, the latest Pfau article linked to above mentions that Bengen used intermediate term gov't bonds and end-of-year withdrawals while the Trinity Study used long term high grade corporates (but it doesn't say which assumption was used for timing of withdrawals).


Quote
The Kitces analysis is somewhat interesting: if your original withdrawal rate is now less than 2.67% of the portfolio value, feel free to bump it up a bit

Yep, and importantly that's on nominal terms, which is good in the sense that the threshold to obtain the same result in real terms is an even higher percentage but frustrating in that divorcing values from inflation renders them much less useful for planning purposes (i.e., you can't really rely on the 2.67% figure because it's not inflation rate-independent).

Was I incorrect when I said above that you have a blog post in the pipeline on how to know when your portfolio has become "too big to fail"?

Jeremy

  • Stubble
  • **
  • Posts: 137
  • Location: Taiwan
  • We retired in our 30's and are full time travelers
    • Go Curry Cracker!
Re: Kitces Article - Ratcheting SWR - Data Discrepancy?
« Reply #41 on: June 20, 2015, 12:53:36 AM »

Was I incorrect when I said above that you have a blog post in the pipeline on how to know when your portfolio has become "too big to fail"?

You are correct, but I have to figure out what the answer is first :)  I have some ideas and a spreadsheet started



Yeah, the latest Pfau article linked to above mentions that Bengen used intermediate term gov't bonds and end-of-year withdrawals while the Trinity Study used long term high grade corporates (but it doesn't say which assumption was used for timing of withdrawals).

If Bengen does end-of-year withdrawals, then Trinity does withdrawals at the beginning of the year.  They were opposites I seem to recall
(cFIREsim also does beginning of year withdrawals)

MDM

  • Walrus Stache
  • *******
  • Posts: 8960
Re: Kitces Article - Ratcheting SWR - Data Discrepancy?
« Reply #42 on: June 20, 2015, 12:08:51 PM »
Yeah, the latest Pfau article linked to above mentions that Bengen used intermediate term gov't bonds and end-of-year withdrawals while the Trinity Study used long term high grade corporates (but it doesn't say which assumption was used for timing of withdrawals).
If Bengen does end-of-year withdrawals, then Trinity does withdrawals at the beginning of the year.  They were opposites I seem to recall
(cFIREsim also does beginning of year withdrawals)

From Pfau's latest:
"Since keeping volatility low is just as important as obtaining high returns, I do think it makes more sense to use intermediate term government bonds than to use corporate bonds. So my re-creation and update of the Trinity study here will follow Mr. Bengenís choice and will also include the assumption that withdrawals are made at the start of the year rather than the end of the year."

brooklynguy

  • Handlebar Stache
  • *****
  • Posts: 2204
  • Age: 38
Re: Kitces Article - Ratcheting SWR - Data Discrepancy?
« Reply #43 on: June 21, 2015, 06:29:08 AM »
Yeah, the latest Pfau article linked to above mentions that Bengen used intermediate term gov't bonds and end-of-year withdrawals while the Trinity Study used long term high grade corporates (but it doesn't say which assumption was used for timing of withdrawals).
If Bengen does end-of-year withdrawals, then Trinity does withdrawals at the beginning of the year.  They were opposites I seem to recall
(cFIREsim also does beginning of year withdrawals)

From Pfau's latest:
"Since keeping volatility low is just as important as obtaining high returns, I do think it makes more sense to use intermediate term government bonds than to use corporate bonds. So my re-creation and update of the Trinity study here will follow Mr. Bengenís choice and will also include the assumption that withdrawals are made at the start of the year rather than the end of the year."

This doesn't say anything different than the quoted posts it responds to, does it?

I just looked back at both original studies, though.  Bengen did use end-of-year withdrawals (see the Appendix box on page 179 of the paper), but the Trinity Study used monthly (not annual) withdrawals, and assumed end-of-month withdrawals (see page 42 of the paper).

MDM

  • Walrus Stache
  • *******
  • Posts: 8960
Re: Kitces Article - Ratcheting SWR - Data Discrepancy?
« Reply #44 on: June 21, 2015, 08:41:38 AM »
Yeah, the latest Pfau article linked to above mentions that Bengen used ... end-of-year withdrawals while the Trinity Study ... doesn't say which assumption was used for timing of withdrawals.
If Bengen does end-of-year withdrawals, then Trinity does withdrawals at the beginning of the year.  They were opposites I seem to recall
(cFIREsim also does beginning of year withdrawals)

From Pfau's latest:
"... So my re-creation and update of the Trinity study here will ... also include the assumption that withdrawals are made at the start of the year rather than the end of the year."

This doesn't say anything different than the quoted posts it responds to, does it?

I just looked back at both original studies, though. Bengen did use end-of-year withdrawals (see the Appendix box on page 179 of the paper), but the Trinity Study used monthly (not annual) withdrawals, and assumed end-of-month withdrawals (see page 42 of the paper).

Many different assumptions being used....