William J. Bernstein
Value Stocks—Hidden Risk or Free Lunch?
Few notions are as well embedded in financial thought as the relationship between return and risk. At ground zero are treasury bills. Historically not returning much more than inflation, they are considered to be a "riskless asset." Perhaps not absolutely riskless, but dire indeed would be the circumstances under which the Treasury of the United States defaults on its short term debt.
Stocks have a much higher return than t bills, but also expose the owner to the possibility of severe capital loss. Consider for a moment what would happen if tomorrow the US Treasury offered a 1 year security with a yield of 11.22%—the annualized return of stocks for the past 73 years. The price of that security would immediately be bid up high enough to reduce the yield to its current 4.6% return. The free lunch would not last a nanosecond.
However, once we start looking at different categories of stocks things get a bit sticky. It turns out that certain kinds of stocks have higher returns than others, namely value stocks. Fama and French sort stock returns by size and price and book-to-market (BtM) ratios, and obtain the following returns and risk metrics:
I've also graphed out the cumulative return of $1.00 invested in each style category on July 1, 1963:
Note the large differences between the returns of value and growth stocks. Classic theory predicts that the risks of value companies should also be higher. Yet the data in the above table suggests otherwise; by any measure (standard deviation, 1973-4 bear market return, worst 12 month rolling return) value stocks actually look less risky.
For a while, this flummoxed even Fama and French. They wrote vaguely of "hidden risk"—it was there, you just couldn't measure it with crude statistics like standard deviation or raw calendar losses. Others pointed out a simple bond market analogy. It's a well know fact that the standard deviation and raw losses sustained by treasuries are a bit larger than those of high grade corporate debt of similar duration. By that measure, at least, the corporate paper is less risky than the treasury securities. And yet, "everybody knows" that treasuries are safer than corporates. In other words, there are other dimensions to risk besides simple numerical concepts.
Fama and French finally settled on the "sick company" theory of risk. Value companies are "sick" and because they are less likely to survive, must offer higher returns to offset this risk. (Some spoke in hushed tones of a massive financial intensive care unit, presumably somewhere in lower Manhattan, containing vast numbers of companies maintained on the fiscal equivalent of ventilators and potent antibiotic and heart medicines.)
I highly recommend a lucid and entertaining piece by Fama's son (think Paul Samuelson meets Jerry Seinfeld) on this conundrum. A few passages express the risk/return equation beautifully:
. . . .there is nothing special about book-to-market. It does not describe risk. However, sorting stocks by BtM also seems to sort them by their true underlying source of risk—the level of their distress. The key to book/market lies in the denominator, market price. High BtM stocks are lower-priced stocks. This is usually because the stock is a poor earner, because it is riskier. Riskier means higher returns.
Suppose Microsoft and Apple Computer each go to the bank for a loan. Which company will have to pay the higher interest rate? Apple will—its future is uncertain ant the bank will need to be paid to take the extra risk. Apple therefore pays a higher cost for its capital. The stock market works the same way. The market expects a higher return for Apple stock than for Microsoft stock. This induces investors to purchase Apple even though Microsoft seems to have better earnings prospects (it seems safer). Put differently, if the two companies had the same expected return, no one would buy Apple.
Fama fils then goes on to attack standard deviation as a measure of risk, applying the label mean variance preferenced to investors who are concerned mainly about SD. (This gets my vote for finance buzzword-of-the-year). He goes on:
If the only risk you fear is fluctuation of returns, you should use a mean-variance optimizer, and the optimizer will tell you to overweight value heavily. This is a perfectly legitimate approach. However very few investors care only about standard deviation. If you care only about SD, you don't care about tracking drift. You don't mind if the market is going strong for several months and your portfolio is flat, or negative. You don't care if your portfolio is dominated by bank stocks and has no technology stocks.
I'm not entirely convinced. Of course most investors don't care about mere "fluctuation of returns;" they care about loss of capital. And, as the above table shows, SD is a superb proxy for this. Most mutual fund shareholders (and particularly index fund shareholders) could care less if several percent of the companies their fund owns wind up on the wrong side of the daisies each year as long as the whole portfolio does well. The key point being that the risk of owning sick value companies is for the most part nonsystematic—it is easily eliminated by owning a diversified portfolio of sick companies. And, as any efficient market student knows, you are not rewarded for bearing nonsystematic risk. (Or, in the words of Paul Samuelson, you are not rewarded merely for going to Las Vegas.)
Well, perhaps not entirely. Maybe 1973-4 just wasn't bad enough to push enough sick companies over the brink. Fama and French also have data going back to 1927, and it shows that during the Great Depression value investing was indeed riskier than growth investing. From January 1929 to July 1932 large value stocks lost 85.6% versus "only" 80.0% for large growth stocks. For small stocks, though, the situation was reversed, with small growth losing 98.1% and small value stocks losing "only" 90.0%.
The tracking error issue is also not entirely convincing. Eugene Fama Jr. works in the world of big-money pension plans. These large funds, together with the mutual fund industry, are indeed exquisitely sensitive to tracking error. Underperform your benchmark by more than a year or two and you're toast. In fact, Robert Haugen, in The New Finance, identifies this as the source of the value premium. Since growth stocks track the S&P more closely than value stocks, they are overowned, and have lower returns. There is only one thing wrong with this argument. As Fama pere points out, the portion of institutionally managed money has increased by an order of magnitude over the past several decades, and yet the value premium seems not to have changed.
Further, I doubt that tracking error is that big of an issue with small investors. Nobody is going to fire you or me if our portfolios underperform the S&P for a few years. And frankly, having seen the long-term returns of technology stocks and IPOs, I don't want to be near the things.
But perhaps the most persuasive argument against the risk premium story for value stocks comes from the current internet stock mania. These stocks trade at astronomical multiples of book value, and most will disappear without a trace of earnings. Yet the risk premium theory predicts that the low expected returns of these stocks (and boy, have they ever got that one right) is due to their safety. Sure.
Perhaps in an era when individual investors held undiversified portfolios consisting of a few issues the "sick company" theory made sense. But the shareholders of the Vanguard and DFA value funds do not seem overly troubled by the fact that they own the equivalent of a financial hospital.
What I find most worrisome is the possibility that investors are now much more willing to own sick value companies than they have been in the past. For if they are, then the value premium may be lower than it has been, or even disappear entirely. I don't pretend to have the answer to that one. I was somewhat shocked to find that as of 3/31/99 over 3 times as much was invested in Vanguard's growth as in their value index funds, so perhaps not all of the value premium has been arbitraged from the financial landscape yet.
Perhaps the real risk of value investing is not that many of these companies will die, but that the value premium itself has disappeared, and that one will be rewarded just as well for owning big, safe companies as small or sick companies. And that alone may be a risk worthy of a premium.
copyright (c) 1999, William J. Bernstein