Some issues with the factor enthusiasts' arguments:
1) To the extent that factor (out)performance in the past may have been behavioural, market structure has likely changed enough to cast serious doubt on forward performance.
The equity markets have changed DRAMATICALLY in the last few decades, from very little institutional investment, and little use of mutual funds, as late as the 1970s, to the modern era rife with mutual funds in general, and index funds in particular, taking a LOT of the casual investors out of the price-setting equation entirely (they're in index funds), and transferring a lot of what active decision-making remains into the hands of professionals. Yes of course there are still individual, poorly informed investors out there making bad decisions, but their likely impact now is MUCH smaller, and there is a lot more savvy money taking the opposite side of their bets.
Limits to arbitrage generally implies that it can be hard to short certain overvalue securities. But it's easy to be long undervalued securities. The amount of money chasing smart beta/factors/etc is likely much larger now than it was in 2000. How much have DFA's AUM grown in that time? How about all the other smart beta funds? If the only behavioural advantage of DFA and the like is avoiding a relatively small % of overvalued, hard-to-arb securities, I'm skeptical of the size of the advantage that would provide vs. total market (more or less) investing.
2) To the extent that it's RISK (or supposed to be):
Many portfolios that focus on only a subset of the market will be riskier than the market as a whole. That's why we prefer diversified portfolios rather than single stock or small baskets of stocks. The question isn't, "Does your portfolio's SD go up if you only invest in ~3% of the market (by market cap)", but rather "Does your portfolio's SD go DOWN if you omit certain slices of the market, or overweight others?"
By analogy, if we consider the BIG market of investment opportunities - stocks and bonds mainly, if you lower your stock allocation, relative to the global average allocation to stocks, your expected risk will generally fall in a significant and observable way. And furthermore, that's widely understood by investors, and their advisers. It's why those with less risk tolerance are advised to be 40/60 or 20/80 or whatever, and those with more risk tolerance are pushed the other way.
But how many folks out there are being advised to overweight large growth within their portfolio, to lower risk?
Or, if we take a more Sharpe-ian view of things, if the default recommendation is to hold the global market portfolio of all investment assets, with deviations from that per the particular circumstances of various investors (including, but not limited to tax situation and risk tolerance), what group is being advised to overweight large growth? To underweight small-value, within their equity allocation?
3) The particulars of determination of value:
F-F's research used price-to-book as primary sorting criteria. The implication is that book value has strong meaning, and a decent relationship to a company's true economic value. And during the primary period for F-Fs original research, 1926-199x, that might have been true. But the economy then was more industrial and more physical. Companies created profits in relation to fixed investments - factories, railroads, etc. If a widget factory sat on the books at $5M, but the stock market valued it at $2.5M, there's a good chance the market undervalued it.
But markets have changed. Economic leadership is much more about intellectual property and non-tangible assets (and liabilities) nowadays. Book value may translate to minimal, if any liquidation value if a company goes under or is sold for a pittance. Furthermore, companies buying back stock at various rates (sometimes leveraging up to do so), creating, if I understand things correctly, further serious distortions in book value per share as a useful measure.
Yes, you can supplement price to book with price to earnings or other measures. But if you do so, you're introducing new problems, and moving the goalposts some versus the original claim (roughly, that low price-to-book companies are, over time, a superior investment).
4) The argument equating market factor (beta) with size, value and other factors is weak:
In response to skepticism about factors, the argument is sometimes made that beta is just another factor like the others, has been negative for long stretches, and yet even factor skeptics will generally accept beta as expected to be positive, and these skeptics do generally have equity market exposure.
But there's a key difference, IMO.
Forward estimates of the equity risk premium (ERP, or Beta), are at least moderately observable. Forward estimates of size and value premiums are much less so.
* You can readily observe interest rates, both nominal and real, over various time frames. While one should be mindful of reinvestment risk and default risk of even safe instruments, I think it's fair to say that the forward return of bonds (assuming you match duration to the timeframe of interest to you), is fairly clear (EDIT - had said "quite high" by mistake), barring extreme outlier situations in the future.
* You can readily observe price to earnings (or better E/P) ratios of stocks, and, with slightly more difficulty, you can at least get some idea of the volatility of earnings over time. Making fairly plausible assumptions about growth of E based on long historical sequences, or just common sense and intuition, you can estimate the earnings yield of stocks. This gives you some basis to estimate the long term return of stocks, from current levels. Of course, volatility in valuations (i.e. P/E of market going way up or down) could (and has, historically), cause substantial variation between realized and expected returns. And different folks can reasonably disagree about expected changes in earnings levels of stocks. But, at the least, it's not hard to start with present values of E/P and do some hypothesizing about expected returns, from that.
* But for small, value, and the other factors, this is much less the case. While you can dig around in French's or others' data and try to do some historical analysis and draw some conclusions, I think you're on much shakier ground here than with equities (ERP/beta) as a whole. And current and historical relationships between value and growth, small and large sectors of the market are less available, and probably less uniform (i.e. definitions of "value" seem to vary widely) than the market.
So, I can look at current equity valuations and hazard an estimate that future ERP, over a long enough period of time, should be positive and reasonably substantial. It would be rather harder for me, or others, to hazard such an estimate about small and/or value.
(Long post already, so I'll continue in another post)
Last edited by psteinx
on Wed Jul 17, 2019 3:12 pm, edited 1 time in total.