Small value, as an interaction effect, does not depend on there being a statistically significant main effect for either size or value. In fact, in my home field of experimental social psychology, main effects were considered of minor importance. You published (or not), and got tenure (or not), based on your ability to find an interesting interaction effect. Non-significant main effects were frosting, not poison, adorning the cake, which was a uniquely interpretable interaction effect.
Fear not, smallcap value fans—there is nothing in this thread to threaten you.
The size effect--a brief history
Banz (1981), looking at the bottom quintile of NYSE stocks by capitalization from 1926 – 1980 inclusive, found outperformance by these (very) small stocks. https://citeseerx.ist.psu.edu/document? ... 1c3b5b79ab
Beginning in 1982, Dimensional Fund Advisors launched a mutual fund providing investors an opportunity to invest in the smallest stocks. That fund and its successors (currently Microcap) provide the “small stock” returns from 1982 forward reported in the Stocks, Bonds, Bills and Inflation yearbook.
Banz’ data is typical of historical work: it is completely cost-free, and assumes that investors can always buy or sell at the midpoint of the bid-ask spread, and that if there was no trade on the valuation day, the midpoint of a stale bid-ask spread could be used without much loss of information.
In contrast, the DFA fund beginning in 1982 provides returns that were available in the world, to actual investors, after all costs and fees, reported in keeping with strict government regulations.
A decade later, Fama and French launched the factor tilt revolution, confirming Banz’s data for size, and adding value, momentum, and more. SMB—small minus big, the performance of stocks below median capitalization against those above the median cap—is reported monthly on Ken French’s website, and annual SMB returns from 1927 – 2022 can be downloaded.
*Nisiprius is wont to say that Banz’ work was later found wanting. However, the time sequence matters: regardless of the fate of Banz’ work, if Fama-French still found a size effect a decade later, and if they have not recanted in the decades since, then academic orthodoxy continues to hold that a size effect obtains.
In theory. If able to buy and sell at the midpoint of the bid-ask spread. Last month’s spread if need be.
Note that SMB, as currently constituted, can’t be dismissed as “performance available only from the tiniest stocks, those too small for institutions to own, and too small for anyone to build much of a position without driving up the price.” SMB is all the stocks below the median against all the stocks above the median, bottom half against the top half.
*Because of skewness in the capitalization distribution, above-median stocks account for far more than half of total market capitalization. Thus, Apple alone recently surpassed the entire capitalization of the Russell 2000—which holds 66.67% of stocks in the Russell 3000, i.e., 500 above-median stocks.
The second dismissal of the size effect, with which many BH will be familiar, rests on an arbitrage argument: “The size effect might have been real in Banz’ data. But if it was real before, it would quickly be arbitraged away after publication once knowledge of its existence diffused.” One presumes the same argument would apply to any Fama-French factor after its discovery.
The empirical question remains: could a mutual fund investor have ridden the size effect to superior wealth accumulation?
Mutual fund versus mutual fund
Investors have been able to own the market of large cap stocks ever since 1976 when John Bogle launched the predecessor to today’s Vanguard 500 Index fund. The S&P fund is not strictly cap based, but it does own most of the 500 largest stocks, and most of its holdings rank high up the capitalization ladder. And it has been available for as long as the DFA fund.
Both funds provide after-cost performance on an all-in basis: commissions, transaction costs, fund management costs, client service costs, etc. If small caps outperform in the laboratory under idealized conditions, but this advantage disappears under real world conditions because of liquidity issues, wide bid-ask spreads, difficulty in building positions, yada yada, then that laboratory outperformance will not be found in the DFA fund results.
Conversely, if the DFA fund investor has outperformed the S&P investor over the 41 years available, then that provides an apples-to-apples demonstration that the size effect can be harvested to accumulate greater wealth--by actual investors using real money, as opposed to academics testing beta coefficients for statistical significance.
Here is the chart
Looks like small cap did outperform … 11.35% annualized versus 11.34% annualized. Superiority of +1 bp, what’s not to like?
An investment of $10,000 turned into $822,063 with the DFA microcap fund, but only $818,297 in the S&P 500 index fund. A clear victory amounting to $3,766, or 38% of the starting portfolio value.
Oh wait. I’m told that DFA funds in that era charged a 1% load. So the final portfolio value will be $8,221 lower, corresponding to an annualized return, on the full $10,000, sans load, of 11.33%. Score now for the 500 index fund. Costs matter, as Mr. Bogle liked to say.
Oh well. Forget about the load (maybe your brother-in-law rebated the DFA charge as a credit against his annual advising fee).
Suppose you only allocated 20% to the DFA fund. Factor tilts are fine, you said to yourself, but don’t get carried away. Looks like, on a $10,000 starting investment, you would have made an extra $753 with that 20% tilt to small caps (no rebalancing on this run). Your wealth accumulation is 100.0009202% of what it would have been from the S&P 500 fund alone.
Go, size effect!
Pause for comments, and then I’ll probe these results further.
