.
the fixer wrote:Factor analysis certainly can help tell you if or how an index or fund has captured the market, small cap, and value premiums, but it leaves out many real world considerations important in day to day investing. Here are a few that can tilt the advantage in favor of certain DFA allocations such as the Vector portfolios that I mentioned above (relative to factor matched ETF or index fund strategies) assuming reasonable assumptions:
Here are some views on each, including the extent the listed factors are excluded from a factor load analysis.
Turnover: (i) internal costs (a) brokerage commissions: my earlier look at this was that the brokerage commissions of the DFA vector fund were 50 percent larger than the iShares S&P600 index. Both had a turnover of 14% last year. The reason IMO: it takes more transactions to get the 14% turnover with the US Vector fund (3,300+stocks), than with an index like the S&P600 with about 600 stocks, (b) bid-ask spread: DFAs patient trading could reduce bid-ask spreads relative to the less patient trading of index type funds. However, Bridgeway also uses patient trading (and has a primary focus on reducing trade costs), the S&P600 screening approach reduced less liquid stocks thereby likely reducing bid-ask spread costs, and these spreads are fairly narrow when getting up to mid-cap stocks. (ii) tax costs – the ETF structure benefits from the internal creation and redemptions process which should add to tax efficiency. The low turnover (10% or so) of the open-ended vector type fund should add to tax efficiency, but its future expected tax efficiency is unclear IMO. The DFA US large value portfolio, for example, seems to have roughly similar turnover, yet its 10-year tax-cost has been 1.3 percent of returns.
Internal rebalancing: Much of the stock migration (turnover) seems to occur from value to netural (SV to SN) [re: the FF migration paper]. From 1926-2005, on average 17% of SV stocks moved annually from SV to SN and SG (also about 17% of market cap), while 3% of stocks moved from SV to LV (about 8% of market cap). As the MSCI and S&P value funds cover much of the SV-SN spectrum (at least more so than the DFA component value funds), they already include some migration benefit. Coupled with the ETF structure, and reduced brokerage commissions, I don’t think internal rebalancing trade costs and tax impacts will be dramatically different from a Vector type open ended fund approach (and I think there is some double counting between turnover and internal rebalancing).
Securities lending: DFA has a larger program than Vanguard, iShares or Bridgeway. However, securities lending is not a risk-free exercise (see Swensen’s book), so maybe larger but no free-lunch IMO.
Utilities and REITS: Several/many concentrated sector portfolios don’t seem to perform as their factor exposure would suggest i.e. low R^2 (as in the FF Industry Cost of Equity paper) – perhaps a reflection of idiosyncratic industry risk. Some blue chips also have low beta, so I don't think REITs are unique in this respect. Excluding REITs may raise portfolio beta (in the FF3F regressions) to above 1 percent, as with many of the DFA funds. But any resulting added return is from higher risk (not alpha).
Liquidity: I think Bridgeway does as good a job as any in navigating illiquidity in the micro-cap space. And if you believe the size premium can be captured by underweigthing CRSP1 relative to all other deciles then this is much less of an issue and don’t see much of a DFA vector liquidity benefit.
Momentum: Here is
an earlier analysis, using very similar analysis time periods, which suggests that DFA funds have historically had a larger negative momentum load than many/most of the retail index funds (likely from more frequent rebalancing/reconstitution relative to annual rebalancing). Will the future performance be different? Not sure.
Seasonal factor premiums and index reconstitution frequency: Estimates of annual average factor loads include any seasonal effects. i.e. the annual value premium = 4%, a fund value load of 0.4 indicates the fund captured 1.6% of this premium (4*0.4). The only way to capture more of this value premium is to have a higher annual value load (e.g. 0.45). i.e. you can’t capture more of the annual premium, by having more value exposure in Dec-July and less in Aug-Nov. while maintaining a 0.4 estimated average value load, the higher exposure when premiums are high will show up in a higher value load (that’s how the OLS regressions work – as I understand them).
Adding up: Following the above, in my view the utilities/REITs, momentum, and reconstitution frequency show up in factor exposure estimates. Momentum has some negative alpha effects. So we are left with turnover & internal rebalancing effects, securities lending and liquidity effects which, following the above, I don’t think will result in the 1 to 2 to 3%(!) alpha. Just my take.
Now I agree on some of the other reasons often listed to hold DFA funds: e.g. simplicity, ability to take more even factor exposure across US:EAFE:EM, if want large/'extreme' value loads, and a fund company with product lines (and ongoing research) linked to a common (Fama-French) framework (all of which I think are credible reasons), but not the 1 to 2 (to 3)% in alpha above factor exposure. Again just my take.
Robert
.