Yesterday I took the stocks in the top 2% of my ranking system (about 60 stocks) and measured their excess return (rolling backtest, 3-month holding period). Then I took only the stocks in that top 2% that had a TTM net income of less than 0 and compared them to those with an income greater than 0. The negative earners outperformed the whole by 9% and the positive earners underperformed the whole by 5%–a 14% difference. When I tested further, I found the following:
Net Income over $50M: excess return of 1%
Net Income between $10M and $50M: excess return of 20%
Net Income between $0 and $10M: excess return of 34%
Net Income between -$20M and $0: excess return of 41%
Net Income less than -$20M: excess return of 33%
Given these results, it looked like I should add to my ranking system the simple factor NetIncBXorTTM with lower values better. So I did so, and it improved my overall results. And almost all of the companies I would invest in would be “losers.”
(A few words about my ranking system: the most heavily weighted factors are operating income growth, price to sales, earnings growth, volume (lower better), accrual ratio (cash flow minus net income divided by total assets, lower better), and CurFYEPSMean/Price; there are a dozen other factors too.)
This goes completely against the grain of everything I’ve ever learned about finance. Not only that, but as I understand it from Marc’s excellent course, one should only use ranking factors that make intuitive sense with the DDM model, and as far as I know there’s nothing in Graham and Dodd that recommends investing more in companies that lose money than in companies that make money. Is there any academic research on this that anyone knows of?
Conceptually, I initially approached using P123 as a way to rank companies according to their most important fundamentals. I felt that using screening alone, without ranking, was flawed, as the process necessarily ignored certain very important fundamental factors, and you could get a very good result by, say, buying only stocks with a p/s less than 0.02, while ignoring all the other things that might make a stock rise or fall in price. I have always viewed ranking as a way of evaluating a stock; a system that ranks stocks with terrible earnings higher than those with strong earnings goes completely against that, and directly conflicts with factors that I use like p/e and return on capital. (Of course, I could simply screen out all companies that earn more than $50M, or earn more than $10 M . . .)
Intuitively, I understand why the stocks ranked highly by my system might do better if they were not yet making a profit than if they were. And the idea of contrarian investing, of betting on the underdog, is an old and good one. But I’m still very uneasy with the idea of ranking stocks higher if they’re earning less. Perhaps if I could find a ranking ratio or formula that favors stocks with low earnings I’d be happier . . .