One of the oldest and most common ways to classify stocks is by size. Size is usually measured by market capitalization - the total value of all outstanding shares of a stock. Conventional wisdom holds that smaller stocks offer higher returns but more risk when compared to larger stocks. In this article I will use data from Ken French's website to examine the relationship between size and performance, and then look at some index-based ETFs from a size perspective.
French's data covers US stocks from July 1926 through the present, currently November 2013. He sorts firms based on market equity, aka market capitalization, at the end of June of each year, then groups stocks of similar sizes into portfolios. For what follows, I will look at the portfolios constructed by arranging stocks into ten different groups, or deciles. (They are not literally deciles; the breakpoints are formed using only NYSE data but the portfolios include stocks from the NYSE, the Nasdaq, and the AMEX. Nasdaq/AMEX stocks have tended to be smaller than NYSE stocks, so the small 'deciles' usually have more stocks than the large ones. I will continue to refer to them as deciles for convenience). I denote the group containing the largest stocks by D10, the group with the next largest stocks by D9 and so on, with D1 representing the group with the smallest stocks.
Here is the performance of several deciles over the entire period. To keep the graph clean, I did not include all ten. To give an idea of the scale, each dollar invested in D10 at the beginning of July of 1926 would have been worth $2,520.36 at the end of November 2013, while a dollar invested in D1 would have been worth $40,615.65
Below are all the results presented numerically. I also checked just the last 25 years to see if there is a noticeable difference in the more recent data. 'Excess' returns refer to difference between the portfolio returns and the risk-free rate, which is the 1-month T-bill (also found on French's site).
The conventional wisdom holds in some regards. D10, and to a lesser extent D9 and D8, produced lower returns with less volatility relative to smaller stocks. As one moved from D7 to D1, volatility rose but returns did not always rise along with it, especially in recent years. The notable anomaly is D2 which offers much of the volatility of D1 but without much of the excess returns. The Sharpe ratios suggest very large and small stocks offer less attractive combinations of returns and volatility, with D5-D9 composing the 'sweet spot'.
Changing over time
Since WWII, D10 has generated the lowest returns. To highlight the size premium over time, I plotted the ratio of each of the other nine portfolios to D10.
The graph shows distinct phases during which the size premium behaved in different ways. Small stocks sharply underperformed during the 1929 - 1932 crash, and then outperformed from 1932 until around 1946. The size premium then remained close to zero for about twenty years. It showed significant gains from 1965-1969, most of which it gave back from 1969-1974. 1975 started another bull run, which would peak in 1983, but from 1983 to 1999 it went negative again, except for a brief bounce in the early 90s. Since 1999 it has generally been positive, although its growth seems to have slowed over the last few years. I leave any macroeconomic interpretation of these results to an author who knows more about that realm than I do.
Tying The Deciles To Today's Investment Options
To be useful, this information needs to be translated to some investable assets. Using the November 2013 ME breakpoints from French's site, index constituent lists from S&P and Russell, and market capitalization data from finviz.com, I estimated the decile distribution for some popular indices. Of course market capitalizations change over time so these are by no means carved in stone.
(click to enlarge)
|ETFs||Expense ratios||CAGR*||Excess annual returns*||Annual std. dev.*||Sharpe ratio*|
|S&P 500||IVV;SPY;VOO||0.05% - 0.11%||9.71%||7.53%||18.37%||0.41|
|Russell 1000||IWB;ONEK;VONE||0.10% - 0.15%||9.93%||7.83%||18.87%||0.41|
|Russell 3000||IWV;THRK;VTHR||0.10% - 0.20%||10.12%||8.15%||19.59%||0.41|
|S&P 500 Equal Weight||RSP||0.40%||10.43%||8.48%||19.94%||0.42|
|Russell 1000 Equal Weight||EWRI||0.42%||11.07%||9.37%||21.51%||0.43|
|Russell Midcap Equal Weight||EWRM||0.42%||11.44%||9.87%||22.35%||0.44|
|S&P 400 MidCap||IJH; MDY; IVOO||0.15% - 0.25%||11.73%||10.29%||23.13%||0.45|
|S&P 600 SmallCap||IJR; SLY; VIOO||0.17% - 0.24%||12.11%||11.40%||26.62%||0.43|
|Russell 2000||IWM;TWOK;VTWO||0.12% - 0.25%||12.12%||11.49%||26.98%||0.43|
|Russell 2000 Equal Weight||EWRS||0.43%||12.28%||12.31%||30.07%||0.41|
|*Projected based on estimated decile distribution (above) & historical (1927 - 2013) decile performance|
The table shows projected returns, volatility and Share ratios for each index fund based on the weighted average of the 85-year historical data for the decile portfolios that make up each index. (These should probably not be interpreted as short-term market forecasts). As noted above, the historical size premium has behaved quite differently during different time periods, so (as always) caution is advised. Also, equal-weight portfolios may differ from market-weight portfolios for other reasons besides their size exposure. The table is only intended to give a very general description of risk-reward characteristics of different indices from a broad historical perspective. The decile portfolios are strictly hypothetical and include no fees, so I listed the fees here.
- Due to the enormous market caps of the firms in D10, even a very broad index that is market-cap weighted - like the Russell 3000 - that doesn't specifically exclude these firms is going to end up heavily weighted in D10. Investors who seek meaningful exposure to the other, more aggressive deciles will need to choose either an index that excludes very large-cap D10 firms, or an equal-weight index.
- Although D1 historically has offered the highest CAGR, it is expensive and difficult to gain exposure to these stocks using a passive approach. Due to the often illiquid nature of these stocks, tracking error may also be an issue for funds operating in this space.
- The Russell Midcap Index and S&P 400 MidCap Index have similar names but offer exposure to slightly different spaces, with the Russell Midcap focusing more on D9 & D8, and the S&P 400 focusing more on D7 and D6.
Readers familiar with the Fama-French work may note that I've failed to mention the value, momentum and quality premiums, any of which may be at least partially intertwined with the size premium. Those are topics for other articles. I will say here that if the size and other premiums are related, an index that chooses which part of the size space to live in but ignores other factors (and all those discussed above do just that) will also indirectly pick up any part of the other premiums that happen to occupy that part of the size space.