Current Research
Previous Investigations
Current Research A: Fundamental Breakthroughs associated with TIPS
Latest update: June 16, 2005.
Current Research A: Fundamental Breakthroughs associated with TIPS
Current Research B: Keeping In Tune with the Human Element
Updated August 21, 2005.
Current Research B: Keeping In Tune with the Human Element
Current Research C: True Buy-and-Hold Investing
Dated: September 12, 2005.
Current Research C: True Buy-and-Hold Investing
Current Research D: Expanded Switching Algorithms
Dated: October 23, 2005
Letter from Mike
I received an interesting letter from Mike.
Subject: Growth-Value Switching
In a recent Forbes [October 17, 2005 page 124], Ken Fisher indicated that he had positive results with a switching model between value and growth based upon the yield curve. Value companies need loans to thrive, and a flat yield curve dries up their loans, which makes value do poorly several months after the yield curve flattens. I thought you might find this interesting.
Gummy 04A01SwModA Calculator (October 14, 2005)
I have added several new switching algorithms to my Gummy 04A01 calculator.
The basic Gummy 04A01 calculator replaces the basic S&P500 stock holdings with a portfolio built from Gummy's asset classes. The weights for this portfolio are listed starting from cell A3626. The calculator replaces commercial paper with a second portfolio built from Gummy's asset classes. The weights for this portfolio are listed starting from cell A3666.
Gummy's database has annual returns for Large Capitalization Growth, Large Capitalization Value, Small Capitalization Growth, Small Capitalization Value, S&P500 (different from Professor Shiller's data), T-Bills, 5-Year Treasury Notes and Government Long Bonds.
Gummy's annual returns include capital gains and losses. This improves the realism of how the calculator treats fixed income investments. It is limited nonetheless. For example, you cannot lock-in high interest rates. Nor can you build a bond ladder.
The two portfolios that replace stocks and commercial paper maintain constant allocations among Gummy's eight asset classes within each individual portfolio. The calculator allocates between these two portfolios in the same way that it would normally allocate between stocks and commercial paper. For example, you can start with a specified fixed allocation between the two portfolios and rebalance the two annually. You can start with a specified allocation between the two portfolios and allow them to grow independently. You can vary allocations in accordance with the switching algorithm.
Gummy 04A01SwModA, with Switching Modification A, has seven choices of switching criteria. Choice A is P/E10. Choice B is the interest rate of commercial paper. Choice C is the return of T-Bills from Gummy's database. It includes capital gains and losses. Choice D is the interest rate of Treasury Bonds (30-year treasuries). Choice E is the return of the Government Long Bond from Gummy's database. It includes capital gains and losses. Choice F is the difference between the interest rates of Treasury Bonds and commercial paper (T.Bond-C.Paper). Choice G is the difference in the annual returns of Government Long Bonds and T-Bills (L.Bond-T-Bills).
The selected algorithm (identified in cell B3) puts numbers into row 2548. Row 186, which previously contained P/E10 information, now contains whatever is on row 2548. Only if the letter A or a is in cell B3 will it now contain P/E10. If none of the letters A, B, C, D, E, F or G (or a, b, c, d, e, f or g) appear in cell B3, all of the entries in row 186 become zero.
Everything that the calculator has previously used as P/E10 is replaced with whatever has been placed into row 186 via the letter entered in cell B3. This includes the two special algorithms (that I call Gummy Algorithm 1 and Gummy Algorithm 2).
Notice that I only change commercial paper among the original non-stock investments. If you choose TIPS or I-Bonds, for example, the calculator uses the special portfolio from Gummy's database that replaces stocks, but it does not use the other special portfolio. You use Gummy's second portfolio only if you select commercial paper (entry 1 in cell B7 or switching FI type 1 in cell F22).
I have placed a self-extracting zip file of the new Gummy 04A01SwModA Calculator (October 14, 2005) in my Yahoo Briefcase jwr19452000. It is available to the public as a free download.
Yahoo Briefcase
Growth-Value Switching using P/E10
I examined this previously. The upside potential was not a good as switching allocations between stocks and commercial paper (or TIPS or I-Bonds).
If you switch between Growth and Value in accordance with P/E10, treat your Value stocks as the traditional stock component (i.e., S&P500 in studies). Treat your Growth stocks as your substitute for commercial paper.
Growth-Value Switching
Growth-Value Switching Data
Growth-Value Switching based on choice F (T.Bond-C.Paper)
This is my initial survey. It is based upon Ken Fisher's observation, which Mike has reported. It shows promise.
You can combine this algorithm with our traditional switching algorithms of stocks and TIPS (or commercial paper or I-Bonds). P/E10 changes separately. You could not do that with my original Growth-Value switching approach. Otherwise, I might have continued with the earlier investigation.
Conditions
I constructed a portfolio of 80% Large Capitalization Value stocks and 20% T-Bills as my stock holding. I constructed a portfolio of 80% Large Capitalization Growth stocks and 20% T-Bills as my substitute for commercial paper. I set expenses equal to 0.00%. I turned off Gummy Algorithms 1 and 2. I varied the withdrawal rate in increments of 0.1% (of the initial balance plus inflation).
I observed that the difference in interest rates fell between plus and minus 2% for most years but not all.
I ended up using an upper threshold of plus 2% and a lower threshold that varied. This shows up as threshold entries of varies-2-79-80. I set the allocations to varies-0%-0%. Since the lowest allocation is 100% and the highest allocation is 0%, the overall allocations are 100%-varies-0%-0%-0%.
I determined the lowest withdrawal rate that resulted in one or more failures, five (or more) failures and ten (or more) failures for 30-year sequences beginning in the years 1928-1980.
Basic Survey Results
With a lower threshold of minus 2% and a 20% stock allocation:
The first failure occurred at 4.6%.
The fifth failure occurred at 4.9%.
The tenth failure occurred at 5.5%.
With a lower threshold of minus 2% and a 50% stock allocation:
The first failure occurred at 4.4%.
The fifth failure occurred at 5.3%.
The tenth failure occurred at 5.7%.
With a lower threshold of minus 2% and a 80% stock allocation:
The first failure occurred at 4.2%.
The fifth failure occurred at 5.6%.
The tenth failure occurred at 6.5%.
With a lower threshold of minus 1% and a 20% stock allocation:
The first failure occurred at 4.6%.
The fifth failure occurred at 4.9%.
The tenth failure occurred at 5.3%.
With a lower threshold of minus 1% and a 50% stock allocation:
The first failure occurred at 4.4%.
The fifth failure occurred at 5.2%.
The tenth failure occurred at 5.9%.
With a lower threshold of minus 1% and a 80% stock allocation:
The first failure occurred at 4.2%.
The fifth failure occurred at 5.6%.
The tenth failure occurred at 6.5%.
With a lower threshold of zero 0% and a 20% stock allocation:
The first failure occurred at 3.9%.
The fifth failure occurred at 5.1%.
The tenth failure occurred at 5.8%.
With a lower threshold of zero 0% and a 50% stock allocation:
The first failure occurred at 4.0%.
The fifth failure occurred at 5.3%.
The tenth failure occurred at 6.3%.
With a lower threshold of zero 0% and a 80% stock allocation:
The first failure occurred at 4.0%.
The fifth failure occurred at 5.6%.
The tenth failure occurred at 6.7%.
Analysis of the Basic Survey Results
Looking at when the first failures occurred:
With a 20% stock allocation, first failures occurred at 4.6%, 4.6% and 3.9%.
With a 50% stock allocation, first failures occurred at 4.4%, 4.4% and 4.0%.
With an 80% stock allocation, first failures occurred at 4.2%, 4.2% and 4.0%.
The first failures favor allocations of 20% and 50%.
Looking at when the fifth failures occurred:
With a 20% stock allocation, fifth failures occurred at 4.9%, 4.9% and 5.1%.
With a 50% stock allocation, fifth failures occurred at 5.3%, 5.2% and 5.3%.
With an 80% stock allocation, fifth failures occurred at 5.6%, 5.6% and 5.6%.
The fifth failures favor an allocation of 80%.
Looking at when the tenth failures occurred:
With a 20% stock allocation, the tenth failures occurred at 5.5%, 5.3% and 5.8%.
With a 50% stock allocation, the tenth failures occurred at 5.7%, 5.9% and 6.3%.
With an 80% stock allocation, the tenth failure occurred at 6.5%, 6.5% and 6.7%.
The tenth failures favor an allocation of 80%.
I have been able to identify WHY the first failure results differ from the fifth failure results and the tenth failure results. In every instance, the first failures occurred in 1929 or 1930. The first four failures occurred during the years of the depression. The fifth (and sixth) failures consistently came from the 1960s.
This is useful information. Growth-Value switching based on the yield curve works better (i.e., has higher Historical Surviving Withdrawal Rates) when there is inflation than when there is deflation.
Other Conditions
I looked at thresholds of plus and minus 1%.
With thresholds of plus and minus 1% and a 20% stock allocation:
The first failure occurred at 4.2%.
The fifth failure occurred at 4.6%.
The tenth failure occurred at 5.1%.
With thresholds of plus and minus 1% and a 50% stock allocation:
The first failure occurred at 3.6%.
The fifth failure occurred at 4.6%.
The tenth failure occurred at 5.2%.
With thresholds of plus and minus 1% and a 80% stock allocation:
The first failure occurred at 3.1%.
The fifth failure occurred at 4.6%.
The tenth failure occurred at 5.3%.
It is better to set the upper threshold at plus 2% than plus 1%.
I swapped portfolios. That is, I treated the Large Capitalization Growth portfolio (which has 20% T-Bills) as my stock holding. I used my Large Capitalization Value portfolio (which also has 20% T-Bills) as my substitute for commercial paper.
With thresholds of plus and minus 2% and a 20% stock allocation:
The first failure occurred at 3.7%.
The fifth failure occurred at 4.7%.
The tenth failure occurred at 5.6%.
With thresholds of plus and minus 2% and a 50% stock allocation:
The first failure occurred at 3.9%.
The fifth failure occurred at 4.4%.
The tenth failure occurred at 5.0%.
With thresholds of plus and minus 2% and a 80% stock allocation:
The first failure occurred at 4.0%.
The fifth failure occurred at 4.3%.
The tenth failure occurred at 4.6%.
It is better to place the Large Capitalization Value stocks into the stock holding. It is better to use the Large Capitalization Growth stocks as the substitute for commercial paper.
Growth-Value Switching with Algorithm D (Treasury Bond Interest Rates)
Our investigation of Growth-Value Switching based on Algorithm D (Treasury Bond Interest Rates) puts Algorithm F’s (T.Bonds-C.Paper) suitability in doubt.
Growth-Value Switching with Algorithm D
Algorithm G (L.Bond-T-Bills)
I have made an extensive survey using Algorithm G. It subtracts the year-to-year (percentage) total return of Treasury Bills from the total return of the Government Long Bond. It uses numbers from Gummy’s database.
Once again, it was unclear which portfolio should be treated as stocks and which should be the substitute for commercial paper. This time, I collected a full set of data for both.
The Value portfolio consists of 80% Large Capitalization Value Stocks and 20% T-Bills, rebalanced annually. The Growth portfolio consists of 80% Large Capitalization Growth Stocks and 20% T-Bills, rebalanced annually.
These are the Best Results when treating the Value Portfolio as stocks:
Lower Threshold = minus 100 percent
Higher Threshold = plus 300 percent
Allocations of the Value portfolio: 100%-80%-0%
Withdrawal rate at first failure: 4.1%
Withdrawal rate at fifth failure: 5.6%
Withdrawal rate at tenth failure: 6.4%
Lower Threshold = plus 100 percent
Higher Threshold = plus 300 percent
Allocations of the Value portfolio: 100%-80%-0%
Withdrawal rate at first failure: 3.7%
Withdrawal rate at fifth failure: 5.5%
Withdrawal rate at tenth failure: 6.7%
These are the Best Results when treating the Growth Portfolio as stocks:
Lower Threshold = zero percent
Higher Threshold = zero percent
Allocations of the Growth portfolio: 100%-0%. There is no intermediate allocation since the two thresholds are equal.
Withdrawal rate at first failure: 4.6%
Withdrawal rate at fifth failure: 5.0%
Withdrawal rate at tenth failure: 5.7%
Lower Threshold = minus 200 percent
Higher Threshold = zero percent
Allocations of the Growth portfolio: 100%-50%-0%
Withdrawal rate at first failure: 4.5%
Withdrawal rate at fifth failure: 4.9%
Withdrawal rate at tenth failure: 5.7%
Always Insist on a Baseline
The Value portfolio consists of 80% Large Capitalization Value Stocks and 20% T-Bills, rebalanced annually. The Growth portfolio consists of 80% Large Capitalization Growth Stocks and 20% T-Bills, rebalanced annually.
Growth Portfolio
I set the Algorithm G thresholds at plus 698-699-700-800 percent. This forces the selection of the Growth portfolio at all times.
Growth Portfolio
Withdrawal rate at first failure: 3.7%
Withdrawal rate at fifth failure: 3.9%
Withdrawal rate at tenth failure: 4.3%
Value Portfolio
I set the Algorithm G thresholds at minus 902-minus 901-minus 900-minus 800 percent. This forces the selection of the Value portfolio at all times.
Value Portfolio
Withdrawal rate at first failure: 3.9%
Withdrawal rate at fifth failure: 5.6%
Withdrawal rate at tenth failure: 6.8%
Assessments
My overall assessment is that these special algorithms add very little in general. However, they have boosted the results under the conditions that existed in 1929 and 1930.
I find it interesting that adding the Growth portfolio helps. Its baseline is horrible. Yet, it makes a positive contribution.
Assessment of Algorithm F
Algorithm F (my original assessment, T.Bond-C.Paper) boosts the withdrawal rate at first failure. Otherwise, its results are similar to those of the Value portfolio by without switching.
Algorithm F allocates 100%-80%-0% to the Value portfolio (with the remainder of 0%-20%-100% going to the Growth portfolio) using thresholds (of the Treasury Bond interest rate minus the commercial paper interest rate) of plus and minus two percent.
Algorithm F results:
The first failure occurred at 4.2%.
The fifth failure occurred at 5.6%.
The tenth failure occurred at 6.5%.
Assessment of Algorithm D
Algorithm D (T.Bond) boosts the withdrawal rate at first failure. Otherwise, its results are similar to those of the Value portfolio by without switching.
I placed too much emphasis on first failure results when I investigated Algorithm D (T.Bond interest rates by themselves). It is possible that I could have improved the fifth and tenth failure withdrawal rates when treating the Value portfolio as the stock holding.
I selected a 20% (Growth) allocation with Treasury Bond interest rate thresholds of 2% and 6%. Under such circumstances, the withdrawal rates are:
The first failure occurred at a 4.1% withdrawal rate.
The fifth failure occurred at a 5.7% withdrawal rate.
The tenth failure occurred at a 6.7% withdrawal rate.
Assessment of Algorithm G
Algorithm G (L.Bond-T-Bills) boosts the withdrawal rate at first failure when treating the Growth portfolio as stocks. It reduces the withdrawal rates at the fifth and tenth failures when compared with the Value baseline.
Lower Threshold = zero percent
Higher Threshold = zero percent
There is no intermediate allocation since the two thresholds are equal.
Allocations of the Growth portfolio: 100%-0%
Withdrawal rate at first failure: 4.6%
Withdrawal rate at fifth failure: 5.0%
Withdrawal rate at tenth failure: 5.7%
Algorithm G (L.Bond-T-Bills) produces withdrawal rates similar to those of the Value portfolio by itself when treating the Value portfolio as stocks.
Lower Threshold = minus 100 percent
Higher Threshold = plus 300 percent
Allocations of the Value portfolio: 100%-80%-0%
Withdrawal rate at first failure: 4.1%
Withdrawal rate at fifth failure: 5.6%
Withdrawal rate at tenth failure: 6.4%
Expanded Baselines
I have calculated the 30-Year Historical Surviving Withdrawal Rates of the four baselines. They are almost identical.
Expanded Baselines
The Data versus Ken Fisher
I have compared the single-year returns of Value portfolios and Growth portfolios to Algorithm F, D and G inputs. I am unable to find the correlations that he has mentioned.
I am not sure why Ken Fisher associates Value companies with high debt loads and Growth companies with low debt loads. The amount of leverage is high in some industries with stable revenues such as utilities. But I do not believe that many people would focus directly on leverage as an indicator of Value.
When I think of Value companies, I think of low price to earnings ratios, low price to book values, low price to sales ratios and high dividend levels. I do not look for high levels of debt.
When I think of Growth stocks, I think in terms of expansion to grow revenues and market share. This is often supported by acquisitions, often (but not always) through the issuance of new debt.
I have summarized my data and put it into a Microsoft Word document in my Yahoo Briefcase. It is available for viewing by the general public. It is in the My Documents folder. I call it Growth Value Algorithms.
Yahoo Briefcase
More to Come?
It is not clear that we should choose Growth-Value switching.
It is not clear to me that Growth-Value switching is anything more than a short-term trading approach. It is not clear to me whether it really works.
Have fun.
John Walter Russell
October 23, 2005