A quick review of the relevant literature is necessary at this point. Research evidence suggests that the value premium, whether derived as a function of risk or investor mis-pricing, is found in three distinct, but clearly related areas of the universe of stocks: 1) small-cap stocks [see Kothari, Shanken and Sloan, 1995, Loughran, 1997, among many others72], 2) stocks experiencing some type of distress or poor prior performance [see Fama and French, 1995, Dichev, 1998, Griffin and Lemmon, 2002, and Penman, Richardson and Tuna, 2007], and 3) stocks suffering from illiquidity issues [see Pastor and Stambaugh, 2003 and Dimson, Nagel and Quigley, 2003]. This means that an institutional investor constructing a separate account or mutual fund portfolio specifically designed to capture the value premium would need to alter her stock screens to include small-cap stocks perceived to reflect considerable distress- related risks and trading noise. Moreover, based on problems relating to liquidity, funds would need to hold large numbers of these stocks to minimize any market impact from their purchase. The fund would need to minimize portfolio turnover to again lessen its market footprint.
Fama and French (2007) provide further evidence about optimal portfolio turnover in the quest for the value premium. The authors observe the value premium to be a function of changing fortunes, originating as a function of 3 events: value stocks are either 1) acquired or become defunct during the observation period, or 2) improve operating returns, thus leaving the high BE/ME value category. Growth stocks simultaneously experience diminishing returns, thus leaving the low BE/ME category. The authors observe only a slight performance premium in value stocksvis a visgrowth stocks in portfolios where no migration occurs and a consistent BE/ME characteristic is maintained. Evidence from this research suggests that a fund manager is not likely to capture a large premium by purchasing high BE/ME stocks and utilizing a buy and hold method. According to Fama and French, value stocks that generate superior performance in one period migrate to become stocks that in future periods fail to generate the same relative superior performance. By implication, it would be wise for fund managers to rebalance their portfolios by size and BE/ME characteristics frequently to capture the superior performance of value stocks prior to their ultimate migration (or acquisition). Consistent with this view, Dennis, Perfect, Snow and Wiles (1995) suggest that the optimal rebalancing period for portfolios formed on BE/ME and size for the sample period 1963 to 1988 is 2 years.
However, it is important to make a distinction between rebalancing in the context of the Fama and French portfolio construction method and turnover occurring in actively managed or passive
72Davis, Fama and French (2000) find that large cap stocks produced a larger value premium than small cap stocks between July 1929 to June 1963.
investment portfolios. For portfolio returns constructed as in Fama and French (1993), Lakonishok, Shleifer and Vishny (1994), Banko and Conover (2002) among many others, stocks may or may not be effectively ‘sold’ at each annual rebalancing or portfolio formation period if their relative size and BE/ME characteristics do not change, or the relative size and BE/ME characteristic for the entire sample shifts proportionately to the shift in characteristics of the stock. Therefore, while rebalancing occurs annually using these methods, some stocks may be held in the same Fama and French-style portfolios (size and BE/ME deciles) for many years across many rebalancing periods.
The DFA Small Cap Value Fund investment philosophy and strategy are consistent with research evidence in academic literature on where and how to capture the value premium. The fund begins with a universe of all stocks traded on the NYSE, AMEX and NASDAQ markets and then eliminates certain stocks to capture the rewards within the small-cap investment space. The fund then excludes or screens stocks for cause before reaching its final portfolio. DFA currently defines small-cap for the Small Cap Value Fund as those residing in the last decile of stocks ranked by market capitalization, or stocks ranking smaller than the 1,000th largest US company, whichever results in the largest market capitalization boundary. Detzel (2008) observes that the average annual market cap of the DFA Small Cap Value Fund is only $312 million between January 1994 and December 2005. That size characteristic is somewhat larger than the $269 million average size for the small value mutual fund category and $309 million for the small growth fund category observed by Detzel. The DFA prospectus adds that the fund typically buys stocks in the lowest 8thpercentile of market capitalization and may from time to time buy stocks with larger market caps. Characteristics of stock purchases are only part of the story. A portfolio is clearly influenced by the market cap of stocks that continue to be held long after purchase. Morningstar reports that the average market cap of the DFA Small Cap Value Fund is $673 million at 31 March 2008, nearly double its 12 year average reported in Detzel (2008).
The DFA Small Cap Value Fund undoubtedly operates in the size area shown by research to generate the largest value premium. However, trading rules imposed by the company exclude many of the most illiquid stocks traded on the exchanges. The fund is restricted from buying any of the thousands of OTC stocks not traded on the National Market System (NMS) of NASDAQ. Moreover, the fund eschews NMS stocks with fewer than four market makers, as well as stocks defined as REITs, limited partnerships, closed-end funds, ADRs and foreign stocks, stocks currently in bankruptcy, those stocks valued under $10 million, or stocks priced under $2.00 per share. DFA also tends to wait six
months before purchasing IPOs.73While the above trading restrictions do not violate the concept of investmentpassivity, they do create potential return differences between the DFA small-cap portfolio and the universe of small cap value stocks traded on those three markets. However, since portfolio constituency restrictions of similar definition are also found in academic research (e.g. Fama and French, 1993), such rule restrictions may have little impact on the process of capturing the value premium.
The DFA Small Cap Value Fund officially defines a value stock quite vaguely.74 The company states in their fund prospectus that, “Securities are considered value stocks primarily because a company's shares have a high book value in relation to their market value.” They broaden the definition stating that “cash flow or price-to-earnings ratios may be considered, as well as economic conditions and developments in the issuer's industry.” DFA says that the criteria for “assessing value are subject to change from time to time.”75However, as a practical matter, DFA defines value as the highest 30% of stocks ordered by book-to-market value and defines growth as the bottom 30% prior to portfolio formation, a definition also used in Fama and French (2007). Similarly, DFA trading rules restrict portfolios to stocks exhibiting positive BE/ME characteristics, again similar to most academic research restrictions. Morningstar reports that the average BE/ME of the DFA Small Cap Value Fund observed at 31 March 2008 is 1.10, considerably higher than the 0.70 average for the Morningstar small cap value fund category and 0.50 average for the S&P 500 Index, and virtually identical to the breakpoint value (1.17) for the highest BE/ME decile portfolio at year end December 2007 computed by Fama and French. This suggests that DFA is populating the value portfolio with stocks possessing high BE/ME characteristics consistent with findings of the value premium in academic literature.
Finally, Detzel (2008) observes that the average annual microcap value-oriented mutual fund portfolio turnover is 61% between January 1994 and December 2005, but only 21% for the DFA Small Cap Value Fund. In other words, the DFA fund holds stocks on average about 5 years, a rate apparently at odds with capturing the pattern of migration observed in Fama and French (2007). As discussed earlier, portfolio turnover rates observed in market-based funds, and rebalancing rates observed for
73Source: Keim (1999) and the DFA funds prospectus dated 29 March 2008.
74It is not unusual for fund prospectuses submitted as a legally binding document to the Securities and Exchange Commission to be intentionally vague. The threat of investor lawsuits as well as the dynamics of the competitive marketplace often preclude the presentation of a more specific set of rules and definitions
hypothetical portfolios in academic research, are not necessarily comparable.76 A low turnover rate does, however, allow DFA to tread lightly in the market for illiquid stocks.
The question of whether the DFA Small Cap Value Fund captures the value premium will be evaluated in the next four sections. Section three discusses the choice of return computation used in this essay. Section four defines the investable universe for the DFA fund and provides a return comparison and discussion of the performance tracking error. Section four also tests the question of whether the DFA fund captures the value premium. DFA excess returns are also regressed against the Fama and French 3-factor model to observe the fund’s sensitivities to the size and BE/ME factors. This begins the process of evaluating the drivers of performance and determining why the fund either does or does not capture the value premium. Following the method of Keim (1999), section five decomposes fund returns identifying the return impact from DFA portfolio constituency rules as well as the impact from the company’s unique trading strategies. Section six further tests returns for periodic changes and time varying characteristics while the seventh section tests for seasonality in DFA returns. An appendix to the essay further investigates the drivers of performance using Sharpe’s return based style analysis (RBSA). These tests help to illuminate market-segment influences on DFA returns over the entire sample period, as well as observing any changes in these influences by using a rolling sixty month RBSA computation.