Page 1 of 1
SCV Switching
Posted: Fri May 21, 2004 2:48 am
by Mike
SCV seems to respond well to switching based upon S&P P/E 10. Setting the P/E parameters at 0% equity for less than P/E 14, 40% up to P/E 22, and 0% thereafter yields large gains in terminal values plus improved portfolio survival at a 4.1% withdrawal rate. The resultant gains in terminal portfolio values are startling, especially over longer periods.
Re: SCV Switching
Posted: Fri May 21, 2004 8:30 am
by BenSolar
Mike wrote:SCV seems to respond well to switching based upon S&P P/E 10. Setting the P/E parameters at 0% equity for less than P/E 14, 40% up to P/E 22, and 0% thereafter yields large gains in terminal values plus improved portfolio survival at a 4.1% withdrawal rate. The resultant gains in terminal portfolio values are startling, especially over longer periods.
0% then 40% then 0%? Can you spell out a little more what the allocations and switches are?
Posted: Fri May 21, 2004 9:34 am
by JWR1945
BenSolar wrote:0% then 40% then 0%? Can you spell out a little more what the allocations and switches are?
I suspect that Mike means 100%-40%-0% with the lower threshold equal to 14 and the higher threshold equal to 22.
I suspect that the switching is between SCV and commercial paper, possibly with 0.2% in expenses and over a 30-year period. In any event, I am collecting data with those numbers.
These are interesting results. These results indicate that P/E10, which is based on the S&P500 and which is dominated by large capitalization stocks, still helps in the Small Cap Value segment. Evidently, the effects P/E10 are not limited to a single sector.
I am collecting Historical Database Rates for the conditions indicated. There is at least one special effect: 1980 shows an early failure, but that is because of the dummy data after 2002 (?). I will have to assess carefully what to report in terms of partially completed sequences. I would prefer not to stop in 1972 just because that is the last complete sequence involving real data.
Have fun.
John R.
Posted: Fri May 21, 2004 10:40 am
by BenSolar
JWR1945 wrote:These are interesting results. These results indicate that P/E10, which is based on the S&P500 and which is dominated by large capitalization stocks, still helps in the Small Cap Value segment. Evidently, the effects P/E10 are not limited to a single sector.
Some of the effect will just be correlation between SCV and S&P500: i.e. S&P 500 is way up and overvalued, SCV will tend to be too, to some extent. I wonder if there is any other linkage going on.
I believe historical valuation for SCV is available through Ken French's site in the form of price/book or it's inverse. These would seemingly be better than PE-10 of the S&P 500. Even better would be to track down or coallate a database of reported historical earnings and then we could do a PE-10 of the SCV sector.
Regards,
Posted: Fri May 21, 2004 11:38 am
by Mike
Using SCV valuation would indeed yield more precise results, and is a future goal. John is correct in his interpretation of what switching algorithms I used.
Posted: Fri May 21, 2004 12:14 pm
by JWR1945
Here are the Small Cap Value (SCV) Historical Database Rates for 1921-1980. The Small Cap Value data itself runs from 1927-2003. Portfolios for retirements begun in 1921-1926 started out with the S&P500 index and switched over to SCV in 1927.
The calculator has dummy data in 2004-2010, with a stock market return of (18.62%), a big loss each year, and with inflation at 3.0% and with the commercial paper interest rate at 1.76%.
Because of the dummy data, I would disregard the 1979 and 1980 Historical Database Rates. I would use 5.3% (from 1945 and 1961) to make comparisons with the 4% mentioned using the conventional approach (without switching allocations).
That is, without switching, the minimum Historical Database Rates were around 3.9% or 4.0% for the S&P500/commercial paper portfolio, 5.1% with switching and 5.3% for the SCV/commercial paper portfolio with switching. SCV had a much, much greater upside.
Look at the high end in 1929-1932. I checked the data. P/E10 got you out of the market during these years. SCV stocks lost tremendously large percentages in each of these years. Being out of the market was a boon. In addition, commercial paper still produced positive nominal returns during deflation.
Have fun.
John R.
Code: Select all
Year P/E10 SCV
1921 5.1 16.4
1922 6.3 17.7
1923 8.2 17.0
1924 8.1 19.4
1925 9.7 19.4
1926 11.3 18.5
1927 13.2 16.7
1928 18.8 14.3
1929 27.1 19.6
1930 22.3 21.7
1931 16.7 24.2
1932 9.3 28.8
1933 8.7 14.1
1934 13.0 12.1
1935 11.5 9.2
1936 17.1 7.4
1937 21.6 10.0
1938 13.5 9.1
1939 15.6 8.9
1940 16.4 9.8
1941 13.9 12.8
1942 10.1 13.3
1943 10.2 10.5
1944 11.1 8.2
1945 12.0 5.3
1946 15.6 7.6
1947 11.5 9.2
1948 10.4 11.0
1949 10.2 9.6
1950 10.7 8.6
1951 11.9 8.3
1952 12.5 8.3
1953 13.0 9.1
1954 12.0 6.8
1955 16.0 6.3
1956 18.3 6.6
1957 16.7 7.6
1958 13.8 6.2
1959 18.0 5.4
1960 18.3 5.6
1961 18.5 5.3
1962 21.2 6.1
1963 19.3 6.2
1964 21.6 6.2
1965 23.3 5.6
1966 24.1 5.8
1967 20.4 5.9
1968 21.5 5.5
1969 21.2 6.5
1970 17.1 7.4
1971 16.5 7.2
1972 17.3 7.7
1973 18.7 10.4
1974 13.5 13.7
1975 8.9 9.9
1976 11.2 8.0
1977 11.4 7.4
1978 9.2 6.9
1979 9.3 5.2
1980 8.9 3.6
Here are the SCV Historical Database Rates for 1921-1980 in a form suitable for copying.
Code: Select all
16.4
17.7
17.0
19.4
19.4
18.5
16.7
14.3
19.6
21.7
24.2
28.8
14.1
12.1
9.2
7.4
10.0
9.1
8.9
9.8
12.8
13.3
10.5
8.2
5.3
7.6
9.2
11.0
9.6
8.6
8.3
8.3
9.1
6.8
6.3
6.6
7.6
6.2
5.4
5.6
5.3
6.1
6.2
6.2
5.6
5.8
5.9
5.5
6.5
7.4
7.2
7.7
10.4
13.7
9.9
8.0
7.4
6.9
5.2
3.6
Posted: Sat May 22, 2004 10:59 am
by Mike
Have you found a way to automate determining HDBR for each individual year?
Posted: Sat May 22, 2004 12:20 pm
by JWR1945
Mike wrote:Have you found a way to automate determining HDBR for each individual year?
No, but my early data reduction software helps a lot.
I summarized the number of failures at 10, 20, 30, 40, 50 and 60 years for a specified range of years. It took the calculator's original data summary and converted the nominal balances to 1's for positive and 0's for negative and placed those results in columns BA through BG. Those modifications were in the Special Versions of mid-November 2003, but not in the September 2003 calculators. I have a special summary in columns L through P and rows 1 through 9. The summary is what makes determining Historical Database Rates a reasonable task.
I increment withdrawal rates (from low to high) and check the number of failures, usually at 30 years and either for 1871-1980 or 1921-1980 for the start years. Every time I detect a new failure or failures, I find out what year or years they correspond to. I write the withdrawal rate that caused the failure minus the increment (0.1%) into a table on another Excel spreadsheet. I also use a pen and paper to help me keep track of how many failures there have been already, the withdrawal rate that I am looking at and the years that have failed.
It still takes two to four hours to construct a table. But that is about a day or two (or a week?) faster than it would have been without the data summary modifications.
Have fun.
John R.
Posted: Sat May 22, 2004 11:31 pm
by Mike
Thank you John.
Posted: Sun May 23, 2004 11:29 am
by JWR1945
Here are some references.
For an overview, read
A Reference Guide for Modifying Calculators dated Thursday, Dec 25, 2003 at 3:34 pm CST
http://nofeeboards.com/boards/viewtopic.php?t=1907
Here are the instructions for making the data reduction improvements that I have just mentioned. They are a Godsend.
Modifications to Improve Data Summaries dated Monday, Dec 15, 2003 at 5:56 pm CST
http://nofeeboards.com/boards/viewtopic.php?t=1841
Have fun.
John R.
Posted: Fri Jul 16, 2004 12:24 pm
by JWR1945
If you look closely at the numbers, you will notice that switching produced tremendous results for portfolios just before and during the Great Depression. Avoiding the 1929 crash and some of the swings in the early 1930s helped a lot.
Small Cap (Capitalization) Value with switching results in the 1960s was still excellent but less dramatic.
I made a plot of the Small Cap Value with switching for portfolios starting in 1926-1975. It had an R-squared value of 0.0703, which is low. Switching has taken advantage of most of the effects of earnings yield. The formula is y = 0.6718x + 5.3563, where x is the percentage earnings yield 100E10/P or 100% / [P/E10] and y is the calculated rate. The line has a slight upward slope.
Of interest, data below the line are close together. Data above the line include quite a few points with exceptionally high Historical Surviving Withdrawal Rates (or Historical Database Rates). Checking the table, those conditions came from the 1920s and 1930s, not later on.
Have fun.
John R.