A question for those who have access to old issues of the Financial Analysts Journal and might be interested: In his letter to the editor entitled ““Determinants of Portfolio Performance—20 Years Later”: A Comment” (January/February 2006, Vol. 62, No. 1), Mark Kritzman says among other things: “What would the BHB methodology reveal? It would tell us that asset allocation determines 100 percent of portfolio performance and that none of performance is determined by security selection.” I am not sufficiently up to speed on the “BHB methodology”, but I can’t imagine how it could possibly “tell us that asset allocation determines 100 percent of portfolio performance” when the two asset classes in question (which are the only two in Kritzman’s hypothetical scenario) have identical returns. If someone could explain what I’m missing it would be much appreciated.
I think they key is that BHB stated that asset allocation explains 90% of variance not returns. Here is a discussion about that: http://en.wikipedia.org/wiki/Asset_allocation Academic studies In 1986, Gary P. Brinson, L. Randolph Hood, and Gilbert L. Beebower (BHB) published a study about asset allocation of 91 large pension funds measured from 1974 to 1983. They replaced the pension funds’ stock, bond, and cash selections with corresponding market indexes. The indexed quarterly return were found to be higher than pension plan’s actual quarterly return. The two quarterly return series’ linear correlation was measured at 96.7%, with shared variance of 93.6%. A 1991 follow-up study by Brinson, Singer, and Beebower measured a variance of 91.5%. The lessons of the study was that replacing active choices with simple asset classes worked just as well as, if not even better than, professional pension managers. Also, a small number of asset classes was sufficient for financial planning. Financial advisors often pointed to this study to support the idea that asset allocation is more important than all other concerns, which the BHB study lumped together as “market timing”. One problem with the Brinson study was that the cost factor in the two return series was not clearly discussed. However, in response to a letter to the editor, Hood noted that the returns series were gross of management fees. In 1997, William Jahnke initiated debate on this topic, attacking the BHB study in a paper titled The Asset Allocation Hoax. The Jahnke discussion appeared in the Journal of Financial Planning as an opinion piece, not a peer reviewed article. Jahnke’s main criticism, still undisputed, was that BHB’s use of quarterly data dampens the impact of compounding slight portfolio disparities over time, relative to the benchmark. One could compound 2% and 2.15% quarterly over 20 years and see the sizable difference in cumulative return. However, the difference is still 15 basis points (hundredths of a percent) per quarter; the difference is one of perception, not fact. In 2000, Ibbotson and Kaplan used five asset classes in their study Does Asset Allocation Policy Explain 40, 90, or 100 Percent of Performance? The asset classes included were large-cap US stock, small-cap US stock, non-US stock, US bonds, and cash. Ibbotson and Kaplan examined the 10 year return of 94 US balanced mutual funds versus the corresponding indexed returns. This time, after properly adjusting for the cost of running index funds, the actual returns again failed to beat index returns. The linear correlation between monthly index return series and the actual monthly actual return series was measured at 90.2%, with shared variance of 81.4%. Ibbotson concluded 1) that asset allocation explained 40% of the variation of returns across funds, and 2) that it explained virtually 100% of the level of fund returns. Gary Brinson has expressed his general agreement with the Ibbotson-Kaplan conclusions. In both studies, it is misleading to make statements such as “asset allocation explains 93.6% of investment return”. Even “asset allocation explains 93.6% of quarterly performance variance” leaves much to be desired, because the shared variance could be from pension funds’ operating structure. Hood, however, rejects this interpretation on the grounds that pension plans in particular cannot cross-share risks and that they are explicitly singular entities, rendering shared variance irrelevant. The statistics were most helpful when used to demonstrate the similarity of the index return series and the actual return series. A 2000 paper by Meir Statman found that using the same parameters that explained BHB’s 93.6% variance result, a hypothetical financial advisor with perfect foresight in tactical asset allocation performed 8.1% better per year, yet the strategic asset allocation still explained 89.4% of the variance. Thus, explaining variance does not explain performance. Statman says that strategic asset allocation is movement along the efficient frontier, whereas tactical asset allocation involves movement of the efficient frontier. A more common sense explanation of the Brinson, Hood, and Beebower study is that asset allocation explains more than 90% of the volatility of returns of an overall portfolio, but will not explain the ending results of your portfolio over long periods of time. Hood notes in his review of the material over 20 years, however, that explaining performance over time is possible with the BHB approach but was not the focus of the original paper. Bekkers, Doeswijk and Lam (2009) investigate the diversification benefits for a portfolio by distinguishing ten different investment categories simultaneously in a mean-variance analysis as well as a market portfolio approach. The results suggest that real estate, commodities, and high yield add most value to the traditional asset mix of stocks, bonds and cash. A study with such a broad coverage of asset classes has not been conducted before, not in the context of determining capital market expectations and performing a mean-variance analysis, neither in assessing the global market portfolio.
Ibbotson recently covered this in the FAJ didn’t he? A couple of months ago I’m sure there were 2 papers by Ibbotson looking at asset allocation.
newsuper Wrote: ------------------------------------------------------- > Ibbotson recently covered this in the FAJ didn’t > he? A couple of months ago I’m sure there were 2 > papers by Ibbotson looking at asset allocation. I’m pretty sure you are correct. It seems to be a popular topic.
Yes he did. I’ve read those papers. In one of them he mentions the letter to the FAJ editor from Kritzman that I referred to above, in which Kritzman seems to think he makes a point (that is much narrower than the broader debate you guys are talking about) irrefutably but that I don’t understand. Any help with the question I asked in my initial post would be much appreciated.
Captain Windjammer Wrote: ------------------------------------------------------- > A question for those who have access to old issues > of the Financial Analysts Journal and might be > interested: In his letter to the editor entitled > ““Determinants of Portfolio Performance—20 Years > Later”: A Comment” (January/February 2006, Vol. > 62, No. 1), Mark Kritzman says among other > things: > > “What would the BHB methodology reveal? It would > tell us that asset allocation determines 100 > percent of portfolio performance and that none of > performance is determined by security selection.” > > I am not sufficiently up to speed on the “BHB > methodology”, but I can’t imagine how it could > possibly “tell us that asset allocation determines > 100 percent of portfolio performance” when the two > asset classes in question (which are the only two > in Kritzman’s hypothetical scenario) have > identical returns. If someone could explain what > I’m missing it would be much appreciated. I will asnwer your question but you should do your homework yourself. BHB methodology: “BHB calculated the returns ascribable to asset allocation by applying a portfolio’s asset-class weights to asset-class index returns. They attributed the residual return to security selection.2 Next, they regressed the portfolio’s total returns on these component returns to measure the extent to which the components explained variation in the portfolio’s total return.” Effectively, BHB claimed that the R^2 of the regression of the portfolio returns on the corresponding aset-allocation index (defined as the portfolio of asset indices with the same weights as asset weights of the original portfolio) is the explained performance (though it should be explained variance). Now the example considered stock A with returns for each period equal to those of bond A, stock B with returns for each period equal to those of bond B and returns of A are double of returns of B. In other words, if x is time series for stock B. Bond B = Stock B = x, Bond A = Stock A = 2*x. Bond index BI is (x+2*x)/2 = 1.5 x. Stock index SI is (x+2*x)/2 = 1.5x. As a result for any asset weights the corresponding asset-allocation index would be 1.5x. Now if manager A invests in stock A and Bond A, while manager B invests in stock B and bond B, asset allocation has nothing to do with returns because any combination of stock A and bond A gives 2x, any combination of stock B and bond B gives x. However, when you regress x or 2x on 1.5x, the corresponding R^2 is 100%. Here is what the author concludes: “What would the BHB methodology reveal? It would tell us that asset allocation determines 100 percent of portfolio performance and that none of performance is determined by security selection. We may be tempted to conclude that the different performances should be attributed to asset allocation, because the managers have different betas, but this conclusion would be false. BHB did not adjust performance for risk. Moreover, the differences in betas arise from the choice of securities, not asset classes as BHB defined these categories.”
Thanks for the response. I may never learn to do my own homework if you keep playing the role of enabler though.