PP, Retirement, and Safe Withdrawal Rates

[Tyler]

This doesn't cover anything that hasn't already been discussed here in one thread or another, but after writing a long post on another forum I thought it would also be good to cross-post it here for posterity.

*****

All portfolios composed of only stocks and bonds are vulnerable to periods of high inflation; thus, they are vulnerable to portfolio failure. Reducing the withdrawal rate will lessen, but not remove this vulnerability.  The Permanent Portfolio addresses this vulnerability by adding more asset classes; although, I am not familiar with any studies trying to ascertain a safe withdrawal rate for the Permanent Portfolio.

I've studied this pretty extensively. I'll try to keep it brief, but it's a complicated topic. 

First, it's very important to understand that retirement calculators like CFireSim do not model the Permanent Portfolio well at all.  There are two basic reasons for this: 1) the PP only applies to the time period after 1972 when the US came off the gold standard, and 2) the bonds in CfireSim and Firecalc are not long-term treasuries and the cash is not short-term treasuries.  So only 25% of the portfolio is accurately modeled for PP purposes. I know these calculators are fun, but for the purpose of studying the PP throw them out. 

Second, because of point #1 above, it's difficult to run the same 30-year analysis of the PP and compare apples-to-apples to the Trinity study.  Starting in 1972, you just won't get enough full 30-year periods to have a statistically significant sample for "percent success" numbers.

Third, as an early retiree I've always been dissatisfied with the Trinity study's basic assumption that only having one dollar left in your account after 30 years would constitute a "success".  I would feel pretty desperate having only $1 in investments 30 years from now.  So even the basic "percent success" numbers are meaningless to me. 

Knowing all of this, I still wanted to know whether the PP was a good retirement portfolio.  So I modeled it myself using excel (adjusting the initial WR for inflation each year, withdrawing from cash, and rebalancing by the traditional method when a 15/35 band was crossed by one of the assets).  Instead of 30-year periods, I used 10-year periods.  To compensate for the shorter timeframes, I used a much more conservative assumption that "success" means having the same inflation-adjusted balance in ten years as you did when you started.  That means that if you set your initial WR to the lowest point on the chart, you will have the same inflation-adjusted balance in ten years and more than that in every other 10-year period.  Then I ran a bunch of optimizations.  Here are the results:



The blue line is the PP.  The red is a typical 50/50 Stock/Bond Boglehead portfolio.  The BH portfolio looks really nice in this timeframe because of the stock run in the 80's and 90's, but remember that the Trinity study still only recommended a 4% WR.  Why is that?  Because of multiple historic rolling periods like the left side of the chart where the real CAGR was negative for more than 10 years at a time.  In contrast, this shows how the PP consistently generates real returns of 4-6% in all kinds of economic environments, including ones where stocks and bonds struggle.

I have lots of other charts and tables if anyone is interested, but the basic takeaway is that while a traditional stock/bond portfolio is generally assumed to accommodate a 4% SWR and not go broke in 30 years, the PP has historically handled the same 4% SWR and maintained principal with a much smoother ride along the way. 


Note that this image shows the CAGR for your investments after expenses. 

So yes, I do have 100% of my investments in the permanent portfolio.  And no, I don't follow my calculations blindly in the same way I don't recommend people put all their eggs in the Trinity Study basket.  If you're not willing to adjust to adversity, you're not ready to retire.  But hopefully this helps people understand the PP as a viable alternative to traditional retirement portfolios.

[dualstow]

I appreciate the cross posting. It'll be easier to refer people to this thread.

[Libertarian666]

On a related topic, I'm making pretty good progress on my program to model retirement withdrawal optimization including the effects of taxes. I'll probably be ready for beta testers sometime this fall.

Maybe earlier if I get laid off from my job if/when the project I'm on crashes and burns, which there is every indication it will.

[Jake]

Tyler your research is great. Did you ever think of publishing it all in a book?

[MachineGhost]

When backtesting, I only go back to Jan 1 1975, when legal restrictions on private ownership of gold were repealed in the U.S.  The returns on gold in '72, '73 and '74 were 49%, 71% and 72%.  The returns for those three years, when ownership of gold was still restricted, have a large effect on calculated return and risk effects of a gold allocation. 

That was just for domestic ownership.  Owning gold physically in Switzerland was doable if you were wealthy enough.  A lot of the brilliant uncirculated pre-1933 gold coins are from caches and hoards held outside the country.

[Tyler]

When backtesting, I only go back to Jan 1 1975, when legal restrictions on private ownership of gold were repealed in the U.S.  The returns on gold in '72, '73 and '74 were 49%, 71% and 72%.  The returns for those three years, when ownership of gold was still restricted, have a large effect on calculated return and risk effects of a gold allocation. 

Excellent point.  I always prefer analysis that studies all rolling periods versus ones that just start in a single year.  For clarity, be sure to think of the line graph above not as retiring at the left end of the chart and tracking balances as you go, but as retiring every year along the way and plotting the original WR that would (in retrospect) have sustained your inflation-adjusted principal for the next ten years.  If you want to discount the early part because you feel the gold performance was anomalous, just ignore the first three data points.  FWIW, the interesting thing data-wise is that gold taking off in the early 70's did not affect the portfolio the way you'd expect.  The stock market in '73 and '74 was brutal, and PP returns actually went up after 1975. 

Tyler, are you planning a 4% withdrawal rate for more than 30 years with 100% in the PP? 

I did not make my early retirement decision based solely on a SWR.  I needed to line up other things first and also be confident in our expenses, and in the process we saved even more.  My current WR is under 3%.  Our investments are 100% in the PP, and I'm as comfortable as I can be about our financial plan without knowing the future.

Let me put it this way -- I don't believe any magical withdrawal percentage will save you from being responsible for monitoring your finances as you age.  For that reason you'll note I also don't make any statements about what withdrawal rate is "safe" with the PP.  However, I would personally be more comfortable withdrawing 4% from the PP for 30+ years than I would be in a traditional stock-heavy portfolio that most people use as the baseline for retirement calculations.  And I'm happy to be the guinea pig. 

[Tyler]

Tyler your research is great. Did you ever think of publishing it all in a book?

Thanks!  I haven't really thought about it.  But it is something I think I can contribute that would be helpful to people, so I guess I wouldn't rule it out.  I feel like I'll need to read a lot more to be qualified to write a whole book on the topic.  Craig and MTex have set an intimidatingly high bar. 

[Tyler]

Can you remind me what age you retired?

37.  It's actually less of a retirement and more of a new beginning.  I consider being secure enough in a financial plan to walk away from one's previous career and embrace new opportunities regardless of the financial reward to be an exciting starting point rather than a finish line. 

[iwealth]

Some of my favorite discussions on this subject:

http://gyroscopicinvesting.com/forum/pe ... ic-graphic
http://gyroscopicinvesting.com/forum/pe ... -pictures/
http://gyroscopicinvesting.com/forum/pe ... in-the-pp/
http://gyroscopicinvesting.com/forum/pe ... awal-rate/
http://gyroscopicinvesting.com/forum/ot ... or-a-myth/

[Tyler]

Just a quick follow-up.

Since writing the OP I discovered a retirement calculator mentioned on the forum that I believe models the Permanent Portfolio pretty darn well (wish I had built it myself!).  PortfolioVisualizer.com uses the same data from the Simba spreadsheet, so it includes all of the appropriate PP assets since 1972 only.  The Monte Carlo simulator uses a novel method of sampling that data to create simulations up to 50 years out, and you can read how it works and play with the settings. 

Using the default sampling method with a 4% WR and a 30-year retirement:
60/40 Boglehead portfolio: 92% success
Permanent Portfolio (with short term treasuries): 99% success

The PP stacks up very well. 

I do not recommend making retirement plans based solely on calculators like these (life is more complicated than that), but this is a nice tool to compare various retirement portfolios.  Note that you can also study way more than just PP and Boglehead options.

Enjoy.

[barrett]

Thanks for posting that link, Tyler. It's very easy to use and there is only a 3% chance that you nice folks are going to have to support me when I am 96!

[mathjak107]

actually 90% of the time following a 4% swr you hit 30 years with more money than you started with  and 67%  had at least 2x what you started with..

the 4% swr was based on the 4 worst 30 year time frames 1907 1929 1937 and 1966.

when you do your spread sheets keep in mind the common denominator to the 4 worst periods  ever were the average real return fell below 2% over the first 15 years of a 30 year time frame.  if it wasn't for those 4 time frames the average swr could be 6.50%.  but the swr is based on worst case scenario's only .

no matter how good things got after the first 15 years  it was to little to late and the time frame failed.

suppose you were so unlucky to retire in one of those worst time framess ,what would your 30 year results look like :  standard 60/40 mix

1907 stocks returned 7.77% -- bonds 4.250-- rebalanced portfolio 7.02- - inflation 1.64--

1929 stocks 8.19% - - bonds 1.74%-- rebalanced portfolio 6.28-- inflation 1.69--

1937 stocks 10.12 - - bonds 2.13 - rebalanced portfolio -- 7.24 inflation-- 2.82

1966 stocks 10.23 - -bonds 7.85 -- rebalanced portfolio 9.56- - inflation 5.38

for comparison the 140 year average's were:

stocks 8.39--bonds 2.85%--rebalanced portfolio 6.17% inflation 2.23%

so what made those time frames the worst ? what made them the worst is the fact in every single retirement time frame the outcome of that 30 year period was determined not by what happened over the 30 years but the entire outcome was decided in the first 15 years.

so lets look at the first 15 years in those time frames determined to be the worst we ever had.

1907--- stocks minus 1.47%---- bonds minus .39%-- rebalanced minus .70% ---inflation 1.64%

1929---stocks 1.07%---bonds 1.79%---rebalanced 2.29%--inflation 1.69%

1937---stocks -- 3.45%---bonds minus 3.07%-- rebalanced 1.23%--inflation 2.82%

1966-stocks minus .13%--bonds 1.08%--rebalanced .64%-- inflation 5.38%

it is those 15 year horrible time frames that the 4% safe withdrawal rate was born out of since you had to reduce from what could have been 6.50% as a swr down to just 4% to get through those worst of times.

while 6.50% to 4% does not sound like a lot 1 million at 4% is an initial draw rate of 40k , at 6.50% you could have had 65k . that is a whopping 60% more .

so what it boils down to is any time you fall below a 2% real return average over the first 15 years you run the danger of 4% not holding. but even a 1/2% cut in spending will make you whole again over the next 15 years or longer.


which is why i always say if you have little discretionary spending you can cut from then you should not be in equities in retirement .

although the odds of being a worst case is very very low it can happen and the y2k retiree may be one as stock and bond returns have fallen below 2% real return the last 15 years so spending cuts may be in order.


so the bottom line is you can't just run any time frames , to stress test things you need to run the 1st 15 years of a 30 year period and must include those identified as the worst.

the problem is the permanent portfolio couldn't exist back then to back test it  since we were on the gold standard.

[mathjak107]

tyler , just making sure you realized that the 4 worst scenario's i mentioned above that are what the 4% safe withdrawal rate is based on cannot be back tested by the PP.

the fact the PP  made it through the 1970's really does not mean much as far as success rate  except for those who started retirement in 1966  no other group would have hit a failure period passing through the 1970's ..

[Tyler]

Understood. 

Also, the original SWR studies (both the Bengen and Trinity versions) were published in the 90s.  The subsequent crashes in 2001 and 2008 caused guys like Wade Pfau to reevaluate some of the assumptions and do more research, and his outlook is a lot more pessimistic today.

I personally don't believe backtesting any portfolio even for hundreds of years guarantees success looking forward at all.  The best you can do is to make an intelligent decision based on the best available data and go with the flow.  A high historic success rate does not absolve you from keeping your eyes open and making good decisions with your money in the future. 

[mathjak107]

i would assume they had to use some kind of monte carlo simulations  since you can't use historical data like firecalc  or the trinity.

[mathjak107]

Understood. 

Also, the original SWR studies (both the Bengen and Trinity versions) were published in the 90s.  The subsequent crashes in 2001 and 2008 caused guys like Wade Pfau to reevaluate some of the assumptions and do more research, and his outlook is a lot more pessimistic today.

I personally don't believe backtesting any portfolio even for hundreds of years guarantees success looking forward at all.  The best you can do is to make an intelligent decision based on the best available data and go with the flow.  A high historic success rate does not absolve you from keeping your eyes open and making good decisions with your money in the future.

the usefulness is not in predicting but in analyzing the numbers.

thanks to  research by michael kitces we now know  how to turn  that data in to useful  useable information.

all failure periods happened when the 15 year real return  average over the first 15 years of a 30 year period  fell below 2%  .

that is all we need to know mathematically now.

so take the y2k hypothetical retiree with a 60/40 mix.  15 years later  his 15  year average is just under 2% real return.  we now know that is a danger sign and withdrawals have to be cut or mathematically you are on track to fail.

this is where the back testing is useful , because it helps us identify the failures and once we can quantify those failures we have real world numbers we can benchmark against .

[mathjak107]

these  swr rules are only ball barks to get you to the gate of retirement with a number. since it is a laboratory number with no human spending patterns or intervention  you can see what worked and what didn't more often than not.

work by michael kitces has shown that every time a portfolio is up 50% from the starting point every 3 years another 10% can be taken on top of the standard 4% inflation adjusted withdrawal.

so if you are up 50% from the start you can take another 10%.  3 years later if you are still up at least 50% take another 10% and so on and so on.


i intend to use bob clyatts withdrawal method.

each year i take 4% of the portfolio value .  if markets fell i take 4% or 5% less than i was drawing , which ever is more.

it is nice because it is dynamic , it has never failed and inflation adjusting is not needed  as it is built in just by being dynamic.

[Tyler]

I like the different perspective.  Thanks!

Unfortunately we can't backtest every possible portfolio back to 1900.  But at least we can learn from danger signs in the past and keep a lookout for similar patterns today. 

Since its inception, the worst 15-year real CAGR for the PP over any rolling period is 3.7%.  If that ever starts to creep down close to 2%, my warning light will start blinking.

[mathjak107]

exactly, and  knowing that the alarm should go off at the 2% real return level is all the back testing  ever done in this area boils down to.

it isn't about patterns repeating , or averages or events. it all boils down to just identifying the common denominator to those time frames that did fail.

[MachineGhost]

all failure periods happened when the 15 year real return  average over the first 15 years of a 30 year period  fell below 2%  .

that is all we need to know mathematically now.

Using silver pre-1968:

Code: Select all

1928	1942	2.79%
1929	1943	2.16%
1930	1944	2.72%
1931	1945	4.59%
1932	1946	3.77%
1933	1947	2.67%
1934	1948	0.38%
1935	1949	0.56%
1936	1950	0.03%
1937	1951	0.07%
1938	1952	1.06%
1939	1953	0.42%
1940	1954	1.68%
1941	1955	2.47%
1942	1956	3.22%
1943	1957	2.74%
1944	1958	3.08%
1945	1959	2.99%
1946	1960	1.68%
1947	1961	3.28%
1948	1962	4.19%
1949	1963	4.80%
1950	1964	4.54%
1951	1965	4.39%
1952	1966	3.89%
1953	1967	4.86%
1954	1968	5.09%
1955	1969	3.27%
1956	1970	2.84%
1957	1971	3.46%
1958	1972	4.80%
1959	1973	4.53%
1960	1974	4.41%
1961	1975	4.19%
1962	1976	3.90%
1963	1977	3.64%
1964	1978	3.37%
1965	1979	4.68%
1966	1980	4.59%
1967	1981	3.96%
1968	1982	4.02%
1969	1983	3.69%
1970	1984	4.41%
1971	1985	5.32%
1972	1986	5.97%
1973	1987	5.07%
1974	1988	4.68%
1975	1989	5.34%
1976	1990	5.09%
1977	1991	5.28%
1978	1992	5.41%
1979	1993	5.91%
1980	1994	3.93%
1981	1995	5.01%
1982	1996	6.07%
1983	1997	5.26%
1984	1998	6.07%
1985	1999	6.19%
1986	2000	5.08%
1987	2001	3.81%
1988	2002	3.69%
1989	2003	4.50%
1990	2004	4.04%
1991	2005	4.63%
1992	2006	4.63%
1993	2007	5.20%
1994	2008	4.71%
1995	2009	5.24%
1996	2010	4.89%
1997	2011	5.32%
1998	2012	5.20%
1999	2013	4.18%
2000	2014	4.78%

[mathjak107]

as you see , the failures at 4% swr  would have been low to moderate  based on the 15 year average falling below 2% . i think i counted 6x.

true ,we don't know what gold would have done  exactly ,but the fact is from what you show it would have failed  a few times when using the actual worst time frames.

from the previous charts i saw  posted i realized the thinking was that by just figuring in the 1970's early 80's for the PP  that  would have been enough  to be considered a worst case but in reality it was not really nailing it.

it was those who retired in 1965-1966 who had A very different scenario play out from just including the time frame in other  periods.


what i find most interesting looking at your silver based portfolio is the worst case time frames are not the same traditional worst case time frames for the most part , but rather different years.

In the years after 1965, the perfect storm of retirement killing conditions took place.  Inflation grew rapidly over the following decade, exceeding 10% in several years in the 1970’s and averaging 6% a year from 1965 to 1985.  Interest rates rose rapidly, from ~4% in 1965 to ~8% in 1970, up to 15% in 1982, causing bonds prices to plummet.  The combo of fast rising high inflation and rising interest rates destroyed bonds.

Stocks also performed horribly adjusted for inflation.

1965-1966 is the all time worst period so far based on more than 146 years of history. it is the one the 4% swr is modeled after.  it is also why 90% of the time following a 4% swr left you with more than you started with  30 years later and 67% of the time left you with 2x or more than you started with.


wow i see i went from junior member to associate member in a day    ha ha ha . 

but the permanent portfolio has always been something i believed in even though i got bored with it decades ago and found it like watching paint dry.

now that i am retiring boring is good.


based  on 1965-1966 conditions here is how  a conventional portfolio did.

anything under a 90% success rate is unacceptable.

[mathjak107]

so traditional investments worst cases were 1907-1929-1936 -1966 and potentially 2000

the PP worst case's using silver  were 1935  1936 1937-1939-1946 , no more modern times.

this is very interesting .....  i think craig needs to look in to this a bit more .  no modern day worst cases , at least with silver anyway using michael kitce's 15 year real returns as a guide for success rates.,

[mathjak107]

all failure periods happened when the 15 year real return  average over the first 15 years of a 30 year period  fell below 2%  .

that is all we need to know mathematically now.

Using silver pre-1968:

Code: Select all

1928	1942	2.79%
1929	1943	2.16%
1930	1944	2.72%
1931	1945	4.59%
1932	1946	3.77%
1933	1947	2.67%
1934	1948	0.38%
1935	1949	0.56%
1936	1950	0.03%
1937	1951	0.07%
1938	1952	1.06%
1939	1953	0.42%
1940	1954	1.68%
1941	1955	2.47%
1942	1956	3.22%
1943	1957	2.74%
1944	1958	3.08%
1945	1959	2.99%
1946	1960	1.68%
1947	1961	3.28%
1948	1962	4.19%
1949	1963	4.80%
1950	1964	4.54%
1951	1965	4.39%
1952	1966	3.89%
1953	1967	4.86%
1954	1968	5.09%
1955	1969	3.27%
1956	1970	2.84%
1957	1971	3.46%
1958	1972	4.80%
1959	1973	4.53%
1960	1974	4.41%
1961	1975	4.19%
1962	1976	3.90%
1963	1977	3.64%
1964	1978	3.37%
1965	1979	4.68%
1966	1980	4.59%
1967	1981	3.96%
1968	1982	4.02%
1969	1983	3.69%
1970	1984	4.41%
1971	1985	5.32%
1972	1986	5.97%
1973	1987	5.07%
1974	1988	4.68%
1975	1989	5.34%
1976	1990	5.09%
1977	1991	5.28%
1978	1992	5.41%
1979	1993	5.91%
1980	1994	3.93%
1981	1995	5.01%
1982	1996	6.07%
1983	1997	5.26%
1984	1998	6.07%
1985	1999	6.19%
1986	2000	5.08%
1987	2001	3.81%
1988	2002	3.69%
1989	2003	4.50%
1990	2004	4.04%
1991	2005	4.63%
1992	2006	4.63%
1993	2007	5.20%
1994	2008	4.71%
1995	2009	5.24%
1996	2010	4.89%
1997	2011	5.32%
1998	2012	5.20%
1999	2013	4.18%
2000	2014	4.78%

i think your work here may be opening up some very interesting facts about the pp and retirement success rates.    thanks

[barrett]

mathjak,

I don't know what time zone you are in but please get some sleep so that we don't lose you to THS (Tired Human Syndrome).

The low returns for 15 years is really helpful info. My question would be, how is knowing that actionable in terms of a safe withdrawal rate? I mean, we don't know in advance what any 15-year period will look like. You've posted about being able to withdraw extra $ if the PP is outperforming. What would be your strategy when it is underperforming? How soon would you cut spending and to what extent? Thanks.

[mathjak107]

actually i fall asleep by 8:30 pm every night.    60 may be the new 45 but 9:00pm  is the new midnight  ha ha ha

the  2% is actionable because all you need to save the time frame is a slight downward adjustment is spending.  you still can salvage the next 15 years .

that is why i find michael kitce's work quite informative based on data that up to now really did us little good going forward