James P. O’Shaughnessy’s What works on Wall Street is one of my favorite books on investing. The thing that I like most about the book is O’Shaughnessy use of data to slaughter several sacred value investing cows, one of which I mentioned yesterday (see The Small Cap Paradox: A problem with LSV’s Contrarian Investment, Extrapolation, and Risk in practice).
Another sacred cow put to the sword in the book is the use of five-year earnings-per-share growth to improve the returns from a price-to-earnings screen. O’Shaughnessy describes the issue in this way:
Some analysts believe that a one-year change in earnings is meaningless, and we would be better off focusing on five-year growth rates. This, they argue, is enough time to separate the one-trick pony from the true thoroughbred.
So what does the data say?
Unfortunately, five years of big earnings gains doesn’t help us pick thoroughbreds either. Starting on December 31, 1954 (we need five years of data to compute the compound five-year earnings growth rate), $10,000 invested in the 50 stocks from the All Stocks universe with the highest five-year compound earnings-per-share growth rates grew to $1,287,685 by the end of 2003, a compound return of 10.42 percent (Table 12-1). A $10,000 investment in the All Stocks universe on December 31, 1954 was worth $3,519,152 on December 31, 2003, a return of 12.71 percent a year.
O’Shaughnessy interprets the data thus:
Much like the 50 stocks with the highest one-year earnings gains, investors get dazzled by high five-year earnings growth rates and bid prices to unsustainable levels. When the future earnings are lower than expected, investors punish their former darlings and prices swoon.
The evidence shows that it is a mistake to get overly excited by big earnings gains.
Five-year growth rates are clearly mean reverting, and I love to see an intuitive strategy beaten by a little reversion to the mean.