Market Capitalization Weighting

With most U.S. equity market indicators flashing "over-valued" warnings, this is a good time to review the impact of market capitalization weighting on index fund performance.

Most indexes utilize "market capitalization" weighting – that is, the weighting of a company or country included in the index depends on the formers’ relative market value. During most periods, this weighting approach works reasonably well. Problems can arise, however, when a small number of companies, sectors, or countries become valued at much higher multiples than most of the other index constituents. When this happens, index investors find themselves systematically increasing their exposure to a smaller number of companies, while decreasing their exposure to the broad market. In other words, the answer to the riddle "when is an index fund not an index fund?" is "when one part of it is substantially overvalued."

Here is an example of what we mean. At year-end 1999, the top ten companies (by market capitalization) in the S&P 500 index accounted for 25.5% of the total value of the index. Ten years earlier, at year-end 1989, the top ten companies in the index accounted for only 20% of its total value. Almost by definition, when markets become overvalued, total market index funds (e.g., based on either the S&P 500 or the Wilshire 5000) will have a larger than normal tilt toward large cap value stocks, and when they become overvalued, the tilt will shift toward large cap value stocks.

The key question for index investors is whether or not they could improve their performance relative to the market as a whole by investing in a stable mix of style or region-based sub-index funds. To answer this question, we looked at recent U.S. and global data.

Let’s start with the best proxy for the U.S. equity market as a whole – the Wilshire 5000 index. Between January, 1978 and December, 1999, the WLSH had an average annual return of 18.5% per year, with a standard deviation of 18.04 and a simple Sharpe Ratio (which, in effect, shows how much return you got per unit of risk you took) of 18.50/18.04 = 1.03. As we discussed, the effective weights put on small, mid, and large cap stocks, as well as value and growth stocks varied over time as the relative popularity and market capitalizations of these six "style types" changed. Could an investor have gotten more risk per unit of return if, instead of buying the index as a whole, he or she had held a fixed percentage of different style types?

We approached this two ways. First, we simply allocated 25% of our portfolio to each of large and small cap growth and large and small cap value. Over the 1/78 to 12/99 period, this simple approach produced an average annual return of 18.75% with a standard deviation of 18.12% and a Sharpe Ratio of 1.04. Pretty much the same as the market index portfolio.

Next, we optimized our allocation between all six asset class options (including midcap value and growth), with the caveat that no more than 33.34% of the portfolio could be invested in any one class. We then constructed two portfolios: one to match the 18.5% average return achieved by the market portfolio, and one to match its standard deviation of 18.04%. The first of these achieved the targeted 18.5% average annual return with a standard deviation of only 15.86% and a Sharpe Ratio of 1.17. This portfolio was composed of 33% small cap value, 24% large cap growth, 22% large cap value, and 21% midcap value.

The second portfolio matched the market’s standard deviation of 18.04%, while achieving average annual returns of 19.16% and a Sharpe Ratio of 1.07. This portfolio was composed of large cap growth, 33%, small cap value, 33%, midcap growth, 25%, and midcap value, 9%.

We also conducted an international test. This time, our base portfolio was 60% MSCI All Country World equity index (a capitalization weighted mix of developed and emerging markets) and 40% Salomon Brothers world 1+ year maturity government bond index. Between January, 1985 and December, 1999, this portfolio generated average annual returns of 11.24%, with a standard deviation of 9.82% and a Sharpe Ratio of 1.15.

In our alternative portfolio, the mix of equities was based on their relative GDP weights rather than their relative market capitalization weights. In this portfolio, Japan had a 6% weight, other pacific countries 15%, Europe 15%, Latin America 6% and the U.S. and Canada 18% for a total equity weight of 60%. Over the 1/85 to 12/99 period, this portfolio had average annual returns of 14.28%, with a standard deviation of 11.19% and a Sharpe Ratio of 1.28. Once again, using a fixed mix of "sub-classes" delivered superior performance when compared to use of a single capitalization weighted market portfolio.

Our overall conclusion from this analysis is that while a simple "buy the whole market in one investment" approach has the virtue of simplicity, it may come at a price – You might be able to obtain better risk/return trade-offs by investing in a fixed mix of sub-asset classes. For this reason, our model portfolios continue to be based on the latter approach, though with the obvious caveat that the past cannot be relied upon to be an accurate predictor of the future. At best it is an estimate of what may happen, not a guarantee that it will.