Historical Database Rates with Switching

Research on Safe Withdrawal Rates

Moderator: hocus2004

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

Historical Database Rates with Switching

Post by JWR1945 »

These are Historical Database Rates for optimized portfolios with two thresholds and three stock allocations. The optimal allocations are always 100% below the lower P/E10 threshold and 0% above the higher P/E10 threshold.

Portfolio A: Stocks and Commercial Paper with P/E10 thresholds of 11 and 21. Stock allocations are 100%-40%-0%. The Historical Database Rates with switching range from 5.1% to 12.1% for the years 1921-1980. They range from 4.9% to 12.1% for the years 1871-1980.

Portfolio B: Stocks and 2% TIPS with P/E10 thresholds of 11 and 24. Stock allocations are 100%-30%-0%. The Historical Database Rates with switching range from 5.2% to 10.3% for the years 1921-1980. They range from 4.2% to 10.3% for the years 1871-1980. TIPS can lose money during times of deflation (unless they are held to maturity). This explains the results for the earliest years.

Keep in mind that current stock valuations and dividend yields are outside of the historical range.

Have fun.

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

Post by JWR1945 »

Code: Select all

Year   P/E10   A   B
1871   13.3  10.0%   7.2%
1872   14.5  10.2%   7.3%
1873   15.3   9.8%   6.9%
1874   13.9   9.6%   7.0%
1875   13.6   9.5%   7.1%
1876   13.3   9.9%   7.7%
1877   10.6   9.4%   7.1%
1878    9.7   9.4%   6.9%
1879   10.7  10.7%   7.6%
1880   15.3   7.4%   5.1%
1881   18.5   7.6%   5.4%
1882   15.7   7.6%   5.1%
1883   15.3   7.2%   4.8%
1884   14.4   6.7%   4.7%
1885   13.1   7.0%   5.2%
1886   16.7   6.5%   5.0%
1887   17.5   6.6%   5.2%
1888   15.4   6.4%   4.9%
1889   15.8   6.1%   4.8%
1890   17.2   6.1%   5.0%
1891   15.4   5.9%   4.7%
1892   19.0   6.0%   5.1%
1893   17.7   5.0%   4.2%
1894   15.7   5.2%   4.8%
1895   16.5   5.3%   5.0%
1896   16.6   5.1%   4.9%
1897   17.0   5.2%   5.2%
1898   19.2   5.0%   5.0%
1899   22.9   5.4%   5.5%
1900   18.7   5.3%   4.8%
JWR1945
***** Legend
Posts: 1697
Joined: Tue Nov 26, 2002 3:59 am
Location: Crestview, Florida

Post by JWR1945 »

Code: Select all

Year   P/E10   A      B
1901   21.0   5.1%   4.9%
1902   22.3   5.3%   5.1%
1903   20.3   5.0%   4.6%
1904   15.9   5.6%   5.2%
1905   18.5   5.1%   4.8%
1906   20.1   5.0%   4.9%
1907   17.2   4.9%   4.7%
1908   11.9   5.6%   5.5%
1909   14.8   5.6%   5.5%
1910   14.5   5.0%   4.7%
1911   14.0   5.1%   5.1%
1912   13.8   5.5%   5.6%
1913   13.1   5.6%   5.5%
1914   11.6   5.8%   5.7%
1915   10.4   6.3%   6.2%
1916   12.5   5.8%   5.6%
1917   11.0   6.9%   6.2%
1918    6.6  10.3%   9.3%
1919    6.1  11.4%  10.1%
1920    6.0  10.4%   9.2%
JWR1945
***** Legend
Posts: 1697
Joined: Tue Nov 26, 2002 3:59 am
Location: Crestview, Florida

Post by JWR1945 »

Code: Select all

Year   P/E10   A      B
1921    5.1  11.8%  10.3%
1922    6.3  12.1%  10.3%
1923    8.2  10.8%   9.1%
1924    8.1  11.4%   9.4%
1925    9.7  10.4%   8.4%
1926   11.3   8.9%   7.1%
1927   13.2   8.9%   7.2%
1928   18.8   8.1%   6.8%
1929   27.1   7.1%   6.3%
1930   22.3   6.8%   6.1%
1931   16.7   6.3%   6.4%
1932    9.3   7.0%   7.5%
1933    8.7   8.3%   9.0%
1934   13.0   5.9%   6.4%
1935   11.5   6.5%   6.8%
1936   17.1   5.7%   6.2%
1937   21.6   5.3%   5.9%
1938   13.5   5.4%   6.5%
1939   15.6   5.3%   6.5%
1940   16.4   5.5%   6.9%
1941   13.9   6.6%   8.1%
1942   10.1   7.7%   8.7%
1943   10.2   7.1%   8.1%
1944   11.1   6.3%   7.2%
1945   12.0   6.2%   7.1%
1946   15.6   6.7%   7.8%
1947   11.5   8.1%   8.2%
1948   10.4   8.6%   8.1%
1949   10.2   8.2%   7.7%
1950   10.7   8.1%   7.6%
JWR1945
***** Legend
Posts: 1697
Joined: Tue Nov 26, 2002 3:59 am
Location: Crestview, Florida

Post by JWR1945 »

Code: Select all

Year   P/E10   A      B
1951   11.9   6.7%   6.2%
1952   12.5   6.3%   5.9%
1953   13.0   6.3%   5.9%
1954   12.0   6.5%   5.9%
1955   16.0   5.7%   5.4%
1956   18.3   5.4%   5.3%
1957   16.7   5.6%   5.4%
1958   13.8   6.0%   5.6%
1959   18.0   5.4%   5.2%
1960   18.3   5.4%   5.3%
1961   18.5   5.4%   5.3%
1962   21.2   5.2%   5.2%
1963   19.3   5.3%   5.4%
1964   21.6   5.1%   5.2%
1965   23.3   5.2%   5.2%
1966   24.1   5.3%   5.3%
1967   20.4   5.4%   5.4%
1968   21.5   5.3%   5.4%
1969   21.2   5.5%   5.5%
1970   17.1   5.5%   5.9%
1971   16.5   5.6%   5.9%
1972   17.3   5.6%   6.0%
1973   18.7   5.8%   6.3%
1974   13.5   6.7%   7.1%
1975    8.9   7.8%   7.7%
1976   11.2   6.2%   6.1%
1977   11.4   6.3%   6.1%
1978    9.2   6.9%   6.5%
1979    9.3   6.8%   6.3%
1980    8.9   6.3%   5.9%
Mike
*** Veteran
Posts: 278
Joined: Sun Jul 06, 2003 4:00 am

Post by Mike »

A few random thoughts. Switching certainly would have taken much of the sting out of the two terrible periods last century (1929, 1966). I may be wrong, but increasing the HDR seems to depend to a degree upon mitigating losses during the riskiest periods, which in the past were high P/E years. I am surprised at how favorably commercial paper compares to simulated TIPS during many periods. Which one is best at any given time may depend partially upon how short term interest rates compare to the CPI. I was also impressed by the high HDR for the early 1920's, which shows me how important the returns in the early years of retirement are. Since recent valuations are outside the historical range, toning down early withdrawals may be prudent for those with long projected retirement periods. TIPS will be slowly depleted with mathematical certainty, and history contains no record of the type of downturns that might be expected subsequent to recent valuations. There may not be time to recover this time, if the portfolio gets too low before the market recovers.
User avatar
BenSolar
*** Veteran
Posts: 242
Joined: Mon Nov 25, 2002 5:46 am
Location: Western NC

Re: Historical Database Rates with Switching

Post by BenSolar »

JWR1945 wrote:Portfolio A: Stocks and Commercial Paper with P/E10 thresholds of 11 and 21. Stock allocations are 100%-40%-0%. The Historical Database Rates with switching range from 5.1% to 12.1% for the years 1921-1980. They range from 4.9% to 12.1% for the years 1871-1980.

Portfolio B: Stocks and 2% TIPS with P/E10 thresholds of 11 and 24. Stock allocations are 100%-30%-0%. The Historical Database Rates with switching range from 5.2% to 10.3% for the years 1921-1980. They range from 4.2% to 10.3% for the years 1871-1980. TIPS can lose money during times of deflation (unless they are held to maturity). This explains the results for the earliest years.
I guess the understood withdrawal period is 30 years? Nice results. Interesting how the TIPS model didn't do so great prior to 1921. Do the results change much when you shift the thresholds and allocations a bit? I guess we will need to investigate the behavior over a fairly wide spread of threshholds and allocations so that we can be sure we're not looking at noise. Bumping the lowest HDBR up from about 4% to about 5% is a big jump, if it bears up to further analysis. :great:
"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 »

In response to Mike's comments:

You are certainly right about current valuations. That is why I made an adjustment for valuations in my recent post about what to do at this moment.
http://nofeeboards.com/boards/viewtopic.php?t=2158

I could have pretended that the issue doesn't exist and reported that a withdrawal rate of 5.1% or 5.2% over 30 years is safe when we employ switching. Such a report would have had as strong a foundation as the frequently reported 3.9% or 4.0% rate without switching. Such a report would have been reckless and misleading.

I was happy to see that switching increased the upside when the market is favorable. With 80% stocks and 20% commercial paper, the highest Historical Database Rates were 10.2%, 10.0% and 10.3% (in 1948, 1949 and 1950) for 1921-1980. Using commercial paper and switching bettered these values handily. Using TIPS at 2% (real) interest with switching comes very close with 10.3% in both 1921 and 1922 and 9.4% in 1924.

Of special interest, however, is the fact that switching degraded the performance when compared to maintaining a single allocation for years 1948-1950. Those were the first three years in a prolonged period of multiple expansion that peaked in 1966 with only a couple of brief pullbacks along the way.

Our P/E10 switching algorithm has no memory of momentum. It makes no adjustments for trends (increasing or decreasing multiples). There may be an advantage to using some form of memory.

My impression is that Historical Database Rates of 5.1% and 5.2% are likely to be as high as we can get for 1921-1980 [unless the interest rate for TIPS exceeds 2%]. Most likely, we should start emphasizing sensitivities and how to reduce them.

It is worth remembering that our modeling of securities other than stocks is severely limited (except for commercial paper). Everything is treated as a trading vehicle without any capital gains or losses. We cannot lock in favorable interest rates. We are never stuck with unfavorable interest rates.

Have fun.

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

Post by JWR1945 »

BenSolar wrote:I guess the understood withdrawal period is 30 years? Nice results. Interesting how the TIPS model didn't do so great prior to 1921. Do the results change much when you shift the thresholds and allocations a bit? I guess we will need to investigate the behavior over a fairly wide spread of thresholds and allocations so that we can be sure we're not looking at noise.
Yes. I have standardized on 30 years unless I mention something else explicitly.

I have found that major financial events seem to have occurred at close to 30-year intervals. If the period is 40 years, the data may show the effects of two distinctly different periods. It is not always clear which era is influencing the results. Shorter periods are OK, but some are highly favorable and others are less favorable.

I have looked at TIPS with a variety of thresholds and allocations. I do not know where I have filed my results. (There is a downside to withholding data.) I have presented the best results to this date.

These results are not noise. The middle allocation and the two thresholds cannot be trusted as being entirely accurate going forward. But if we get 4.8% when the model would have produced 5.1% (assuming that history repeated itself exactly), we will have done OK. [Of course, we have to make adjustments for valuations and dividends to get back inside of the historical range.]

The nature of switching (at low Historical Database Rates) is that you either have zero failures or you have several (e.g., 4 to 6). The lower end of the range (of Historical Database Rates) does not consist of individual years in isolation. At the high end, however, quite a few years can be isolated.

I have not made any models for projecting the effects of switching. I have not placed any confidence limits on anything using these results.

Have fun.

John R.
Post Reply