Switching Works

Research on Safe Withdrawal Rates

Moderator: hocus2004

Post Reply
JWR1945
***** Legend
Posts: 1697
Joined: Tue Nov 26, 2002 3:59 am
Location: Crestview, Florida

Switching Works

Post by JWR1945 »

Switching Works

These are some early results from the modified Retire Early Safe Withdrawal Rate calculator that I have described on a previous thread:

The portfolios that I have examined contained stocks (i.e., the S&P 500 index) and (hypothetical) Treasury Inflation Protected Securities (TIPS). I set the withdrawal rate at 5.0% to assure enough failures to produce meaningful data for analysis. I left the expenses at their default rate of 0.20%. I looked at stock allocations of 80%, 50% and 20%. P/E10 levels of 12, 15 and 18 were used for switching allocations. I set the threshold to 50 for collecting baseline data. There is no switching at such a high threshold.

I set the interest rate of the TIPS at 3.0%, which produces 2.8% interest after expenses. Setting this interest rate carefully is critical in making any analysis. During the historical period, even in times of low market returns, stocks yielded 3% or more in dividends. Any investment vehicle with a lower interest rate has been at a severe disadvantage when compared to stocks. When we try to project our Historical Database Rate results onto today's stock market, there is a dangerous mismatch of dividend yields. The key feature of dividends is that they are stable and well behaved when compared to stock prices. High dividend yields mean that you do not have to sell very many shares of stock when prices are low. Low dividend yields mean that you may have to sell heavily at depressed prices and that is what kills retirement portfolios.

By setting the interest rate at 3.0%, I have introduced an investment that would have matched the dividend-only return of stocks when valuations were high but that would not have the price volatility of stocks. It is possible to buy long term TIPS in the secondary market at interest rates close to 2.8%. Since owning them should cost much less than owning a stock fund, I have ignored their expenses. Since today's market offers very little in dividends, this analysis tends to downplay the advantage of switching into TIPS during times of high valuation.

In addition, the 30-year Safe Withdrawal Rate of 2.8% TIPS is slightly above 4.9%. Setting the interest rate of the TIPS at 3.0% (with 0.20% expenses) not only matches dividend yields but it also matches the selected withdrawal rate.

I have restricted my analysis to portfolios beginning in 1921 and later. There is an anomaly associated with earlier years. I have included partially completed sequences because they include portfolio failures. As a practical matter, portfolios started after 1974 have not had enough time to fail.

Data Summary

These are the number of portfolios that failed (starting on or after 1921) within 30 years and within 40 years.

Portfolio allocation: 80% stocks at low P/E10 levels and 20% stocks at high P/E10 levels.

Code: Select all

Threshold    30 years        40 years
12                 0              10
15                11              14
18                 1               8
Portfolio allocation: 50% stocks at low P/E10 levels and 20% stocks at high P/E10 levels.

Code: Select all

Threshold     30 years         40 years
12                 0               12
15                10               14
18                 4               12
Portfolio allocation: 80% stocks at low P/E10 levels and 50% stocks at high P/E10 levels.

Code: Select all

Threshold     30 years         40 years
12                   4               10
15                   8               11
18                   7               11
Baseline portfolio allocations: 20%, 50% or 80% stocks at low P/E10 levels and 20% stocks at high P/E10 levels. P/E10 threshold set at 50.

Code: Select all

Stocks    30 years         40 years
20%            11             32
50%             9             16
80%            13             18
Observations

The middle level threshold (of 15) appears to interfere with switching. You appear to be better off by loading up on stocks when prices are low and they are bargains. Alternatively, you can do very well by holding onto stocks until prices are high and valuations are dangerous. Trying to do both at the same time does not work.

Remember that the calculator has only one threshold and you can only choose between two allocations. Remember as well that the threshold has no memory. It does not consider whether prices are trending up or trending down.

Looking at the baseline results shows what happens when the stock allocation is very low. TIPS could provide almost 30 years of income at a withdrawal rate of 5%, but not any more. It would be slightly less. Stocks are needed to extend the portfolio life span. They provide long-term growth.

Comparing the 80% and 20% portfolio with the 50% and 20% portfolio, it is a good idea to load up on stocks when prices are below threshold, especially at the lowest threshold.

Comparing the 80% and 20% portfolio with the 80% and 50% portfolio, making the bigger change (to 20% instead of 50% stocks) at high valuations looks better.

In terms of whether switching works, look at the data when there is no switching. In every instance there are 9 or more failures in 30 years. If you look at the cases in which you cut back stocks to 20% above threshold, you can find instances with very few failures. Look at the thresholds of 12 and 18, not the intermediate level. In terms how results have been reported in the past, the claim could be made that cutting back to 20% stocks when P/E10 is above 12 produces 100% safety over 30 years. You would have your choice of a stock allocation of either 50% or 80% when P/E10 is lower than 12.

That is, the traditional claim would have been that 2.8% TIPS (after expenses) plus switching (at P/E10 = 12) provides a safe withdrawal rate of 5.0% (with 100% safety) over 30 years. That is certainly the Historical Database Rate result. I would not use it as a safe withdrawal rate because we are outside of the historical range in terms of both valuations and dividends.

Have fun.

John R.
JWR1945
***** Legend
Posts: 1697
Joined: Tue Nov 26, 2002 3:59 am
Location: Crestview, Florida

Post by JWR1945 »

Expanded Data Summary

I have added information about portfolios that would have failed at 35 years or sooner.

These are the number of portfolios that failed (starting on or after 1921) at 30 years (or sooner), 35 years (or sooner) and 40 years (or sooner).

Portfolio allocation: 80% stocks at low P/E10 levels and 20% stocks at high P/E10 levels.

Code: Select all

Threshold    30 years       35 years       40 years
12                 0              2             10
15                11             14             14
18                 1              8              8
Failures occurred in 1956-1973.
Portfolio allocation: 50% stocks at low P/E10 levels and 20% stocks at high P/E10 levels.

Code: Select all

Threshold     30 years        35 years        40 years
12                 0                9             12
15                10               11             14
18                 4               10             12
Failures occurred in 1956-1970.

Portfolio allocation: 80% stocks at low P/E10 levels and 50% stocks at high P/E10 levels.

Code: Select all


Threshold     30 years           35 years        40 years
12                   4                7               10
15                   8               11               11
18                   7               11               11
Failures occurred in 1960-1970.

Baseline portfolio allocations: 20%, 50% or 80% stocks at low P/E10 levels and 20% stocks at high P/E10 levels. P/E10 threshold set at 50.

Code: Select all

Stocks    30 years        35 years        40 years
20%            11             23               32
50%             9             14               16
80%            13             17               18
Failures occurred from 1922-1974.

Observations

The additional data strengthen the case for using a low threshold (P/E10 = 12). It strengthens the case for loading up heavily with stocks when prices are low (i.e., choose 80% instead of 50% when P/E10 is below 12).

We see that the decade of the 1960s was an especially hazardous time to retire (financially). Switching narrowed down the hazardous period to 1956-1973 from 1922-1974.

Have fun.

John R.
User avatar
BenSolar
*** Veteran
Posts: 242
Joined: Mon Nov 25, 2002 5:46 am
Location: Western NC

Re: Switching Works

Post by BenSolar »

JWR1945 wrote:Switching Works

These are some early results from the modified Retire Early Safe Withdrawal Rate calculator that I have described on a previous thread:
Good stuff, John! :D

Interesting how the number of failures is high with the switch set at 15. I don't understand why that is happening. I'm guessing it has something to do with the poor returns from the 60s and 70s when valuations weren't all that high a lot of the time.
"Do not spoil what you have by desiring what you have not; remember that what you now have was once among the things only hoped for." - Epicurus
JWR1945
***** Legend
Posts: 1697
Joined: Tue Nov 26, 2002 3:59 am
Location: Crestview, Florida

Post by JWR1945 »

BenSolar
Interesting how the number of failures is high with the switch set at 15. I don't understand why that is happening. I'm guessing it has something to do with the poor returns from the 60s and 70s when valuations weren't all that high a lot of the time.
BenSolar is correct. I have not yet analyzed cause and effect. I suggest examining this data in light of the real prices, the dividend yields and P/E10s in Professor Shiller's database.

The switching formula is on line 186 of the Retire Early Safe Withdrawal Rate calculator (version 1.61). It starts with this code: IF($F$19<C186,... for the entry in column C. The threshold is found in location F19. The comparisons are with numbers in line 186, which is called the S&P "Real" P/E Ratio. The entries in line 186 are P/E10 values associated with the column. For example, AZ186 is the P/E10 for 1921 (since the column for 1921 is AZ).

These are the years from 1921-2002 in which P/E10 is equal to or greater than the thresholds indicated. They are the years in which the allocations are for high P/E10.

Threshold = 12.0.
1927-1931, 1934, 1936-1941, 1945-1946, 1952-1974, 1987-2002.

Threshold = 15.0.
1928-1931, 1936-1937, 1939-1940, 1946, 1955-1957, 1959-1973, 1989-2002.

Threshold = 18.0.
1928-1930, 1956, 1959-1970, 1973, 1992-2002.

With switching, the failures that occurred on or before 30 years started entirely within the range of 1960-1970. For a stock allocation of 80% / 20%, there was not any failure at a threshold of 12. There were failures in 1960-1970 with a threshold of 15. There was a failure in 1965 with a threshold of 18.

For a stock allocation of 80% / 50%, there were failures in 1965-1967 and 1969 with a threshold of 12. There were failures in 1961, 1963, 1965-1970 with a threshold of 15. There were failures in 1963, 1965-1970 with a threshold of 18.

For a stock allocation of 50% / 20%, there were no failures at a threshold of 12. There were failures in 1960-1969 with a threshold of 15. There were failures in 1963, 1965-1967 with a threshold of 18.

Notice that all of the failed portfolios started in the 1960s. BenSolar was one of the first to emphasize those years instead of the Great Depression. intercst had made the observation, but not with the same degree of emphasis. More recently, peteyperson has pointed to that era, referring to the writings of John Bogle.

Any explanation of the differences at various P/E10 switching levels is found in the years after 1960.

Have fun.

John R.
hocus
Moderator
Posts: 435
Joined: Mon Dec 02, 2002 12:56 am

Post by hocus »

Stocks are needed to extend the portfolio life span. They provide long-term growth.

This comment brought to mind a thread that I started at the Motley Fool board back in May 2002, "Insights Salvaged from a Doomed Thread."

http://boards.fool.com/Message.asp?mid= ... sort=whole

It has long been my view that Joe Dominquez and intercst come at the question of how to invest for early retirement from opposite extremes, and that in both cases the strategy recommended is one that is the result not of dispassionate analysis but of either positive or negative personal experiences with stocks.

Dominguez quit his job in 1969. He says on his tape course that he had expected to achieve FIRE sooner, but that a fall in stock prices caused him to lose a good bit of his accumulated capital. My sense is that that experience caused Dominguez to sour on stocks as an asset class and to recommend that aspiring early retirees go with "safer" investment classes.

In contrast, intercst had done well on his stock investments long before he put together his Safe Withdrawal Rate study. I don't believe that he was doing the analysis to learn anything about which asset class is best for early retirees. I think he "knew" before he started that stocks were always the best asset class, and that the purpose of the study was to produce some spreadsheets that he could point to as "proof" of this particular personal investment preference.

When I was putting together my plan, I did not feel comfortable with the idea of buying into either of these two extremes. I question whether the Dominguez idea of avoiding stocks altogether is as "safe" as it may appear to be on first consideration; my view is that the increase in the size of my FI stash that I will see over time as the result of making prudent stock investments will make my plan safer as I grow older.

On the other hand, I think it is an exercise in self-deception to follow the intercst approach of pretending that changes in valuation levels have zero effect on a plan's safety. The historical data is crystal clear on this point, and any SWR methodology rooted in the presumption that valuation levels have no effect on returns expectations is a dangerously misleading one, in my view.

I think that the middle ground is the way to go with FIRE investing. The middle ground is not to disdain stocks just because there are times when they will cause you setbacks, and not to become so enthused with stocks that you calculate SWRs without regard to valuation levels. The sane course, in my view, is to aim to take advantage of the long-term growth potential offered by stocks while protecting yourself from severe setbacks by including in your financial planning an assessment of the effect that changes in valuation levels are likely to have on your portfolio in the event that the future turns out something like the past.
JWR1945
***** Legend
Posts: 1697
Joined: Tue Nov 26, 2002 3:59 am
Location: Crestview, Florida

Post by JWR1945 »

For BenSolar: I think that the reason for the strange behavior at a P/E10 threshold of 15 has to do with how close the TIPS interest rate (3%) is to the minimum dividend of stocks (around 2.8% or 2.9% and not for long). It does not show up when the TIPS have a clear-cut advantage over stock dividends in times of stress.

For hocus: The writer of the first post in that Motley Fool thread had it right. Subsequent writers were way off the mark.

We have now advanced from a mechanical-but-blind kind of strategy into using cause-and-effect. We do not trust numbers in isolation. Those other writers on that thread did even though the numbers are well outside of the historical range. We can monitor performance. We can design our strategy to ensure safety. We know why things are happening. We can take advantage of opportunities without taking on undue risk.

The other writers insisted upon flying blind into uncharted territory. Yet, they consider themselves to be the smart ones. They were the ones who told us not to look into the subject any more. They already had the answer. There was nothing more to learn.

Have fun.

John R.
JWR1945
***** Legend
Posts: 1697
Joined: Tue Nov 26, 2002 3:59 am
Location: Crestview, Florida

Post by JWR1945 »

I have made several references to the dividend yields of the S&P 500 and the fact that they were almost always larger than 3%. This fills in the gaps.

I listed dividend yields in an earlier thread: Dividend Based Strategies dated Wednesday, July 30, 2003 at 9:46 pm CDT.
http://nofeeboards.com/boards/viewtopic.php?t=1214

I have extracted the relevant section:
Focus on Yields

The most stable relationship that I have found has been the ratio of a year's safe withdrawal rate to its dividend yield. Here is a summary. The list shows the year, dividend yield, the ratio of the year's [Historical Database] Rate for a 50% stock allocation to the dividend yield and the ratio of the year's [Historical Database] Rate for an 80% stock allocation to the dividend yield. (I actually used the rates at which the first failure occurred, which is 0.1% or 0.2% higher than a year's actual [Historical Database] Rate.) [In reference to Historical Database Rates, I used commercial paper as the alternative to stocks.]

Code: Select all

1921    7.11%    1.37    1.65
1922    6.35%    1.35    1.63
1923    5.74%    1.35    1.60
1924    6.02%    1.32    1.59
1925    5.27%    1.40    1.63

1926    4.82%    1.45    1.61
1927    5.19%    1.31    1.42
1928    4.43%    1.30    1.35
1929    3.46%    1.38    1.27
1930    4.47%    1.16    1.11

1931    6.07%    0.85    0.95
1932    9.55%    0.64    0.85
1933    6.98%    0.85    1.23
1934    4.19%    1.19    1.52
1935    4.85%    1.11    1.60

1936    3.50%    1.31    1.60
1937    4.17%    0.95    1.15
1938    7.02%    0.71    0.96
1939    4.10%    1.17    1.51
1940    5.06%    0.94    1.22

1941    6.41%    0.81    1.12
1942    7.87%    0.78    1.16
1943    5.90%    1.15    1.59
1944    5.19%    1.19    1.69
1945    4.77%    1.25    1.76

1946    3.70%    1.45    1.89
1947    4.69%    1.49    2.04
1948    5.69%    1.37    1.93
1949    6.18%    1.29    1.81
1950    6.84%    1.08    1.52

1951    6.98%    1.03    1.37
1952    5.86%    1.19    1.56
1953    5.40%    1.25    1.62
1954    5.73%    1.22    1.60
1955    4.34%    1.38    1.61

1956    3.78%    1.42    1.58
1957    3.82%    1.41    1.62
1958    4.33%    1.38    1.66
1959    3.15%    1.65    1.77
1960    3.21%    1.61    1.68

1961    3.26%    1.59    1.65
1962    2.93%    1.63    1.70
1963    3.28%    1.52    1.58
1964    3.00%    1.60    1.60
1965    2.92%    1.50    1.50

1966    2.93%    1.50    1.43
1967    3.41%    1.34    1.34
1968    3.08%    1.42    1.42
1969    3.01%    1.46    1.46
1970    3.50%    1.42    1.42

1971    3.35%    1.49    2.08
1972    2.98%    1.67    1.67
1973    2.67%    1.79    1.79
1974    3.53%    1.64    1.69
1975    4.99%    1.44    1.68

1976    3.80%    1.78    1.94
1977    3.97%    1.76    1.91
1978    5.22%    1.57    1.80
1979    5.12%    1.71    1.91
1980    5.18%    1.85    2.04
The behavior at an 80% stock allocation is amazingly good. You can spot a problem easily when the dividend yield drops dramatically within the next two or three years. The behavior is good at 50% as well.
The only years from 1921-1980 with dividend yields of 3% or lower were 1962 (2.93%), 1964 (3.00%), 1965 (2.92%), 1966 (2.93%), 1972 (2.98%) and 1973 (2.67%).

Have fun.

John R.
Post Reply