This board has been migrated to our new platform! Check out the new home page at discussion.fool.com or click below to go directly to the new Board on the new site.
<< <<One of the problems I have with the Trinity study is that it assumes an average market rate of return.>> >><< No, it does no such thing. The study compiled by three Trinity University professors, Philip Cooley, Carl Hubbard, and Daniel Walz, examined this issue by looking at historical annual returns for stocks and bonds from 1926 through 1995.>>What I should have said, and meant to say, is that for the stock component of the tested portfolios the Trinity study uses the average market rate of return. I am looking forward to reading your study.I actually think that we are in substantial agreement. I have read the references you cite, and have done some of my own analysis for a time period similar to yours.Essentially, the FF must be balanced with at least one other, preferably non-correlated, asset in order to obtain the the highest possible indefinitely sustainable rate of withdrawal. The computations necessary for balancing with two other assets are much more numerous than for one, so most studies limit themselvs to stocks plus one other asset, typically bonds. It is not as obvious to me as it is to the authors of the cited works that the best other asset is bonds, especially if we substitute the FF for the market as a whole. There are mainstream US-based assets other than bonds which are less correlated with the FF and seem to improve the sustainable withdrawal rates to something on the order of 7%. Not an inflation-adjusted 7% of the initial portfolio, but 7% of the current value of the portfolio each year. Unfortunately, I have only tested this against historical data. While it is easy to test against historical data, if we truely believe that market performance is a continuing upward trend with superimposed random fluctuations, we should use a more sophisticated analysis technique. For instance, we can use the historical data to estimate the statistical parameters and then use these parameters to generate a large number of equally-probable scenarios. Testing against all of these scenarios would give a much higher level of confidence than just looking at the single historical data set.
Best Of |
Favorites & Replies |
My Fool |