The method and the man — 7 500 companies, 27 years, ten deciles
Chapter 1 of 14 · 10 min
David Dreman is the value investor who built his thesis on data instead of anecdotes. Between 1970 and 1996 he tracked on the order of 7 500 American companies, divided into deciles on four valuation measures and rebalanced mechanically every year. The setup is the model for AK1A's own factor tests: unbiased, long, and with the dead companies kept in the material.
Dreman began as a journalist and practitioner, not as an academic. After burning himself on deeply underpriced growth stocks in the late 1960s he wrote Psychology and the Stock Market (1977), founded Dreman Value Management in the same decade and went on to write a contrarian column in Forbes for more than thirty years. The main works are The New Contrarian Investment Strategy (1982) and Contrarian Investment Strategies: The Next Generation (1998), where the large body of data is found, followed by the psychologically deepened edition from 2012.
The thesis is the same through all the books: the market's mistakes are not random — they are systematic and go in one direction, and therefore they can be harvested.
The method is pure mechanics. The entire universe — on the order of 7 500 companies over 27 years, 1970–1996 — is sorted every year on a valuation measure, for example P/E, and divided into ten equally large deciles. Decile 1 is the ten percent cheapest companies, decile 10 the ten percent most expensive. You buy the decile portfolio, hold for exactly one year and repeat the procedure.
The database is Compustat with research files, which means that bankrupt companies and delisted companies remain in the material — survivorship bias is thus handled, in contrast to most backward-looking winner lists. The comparison group is the S&P 500 with dividends.
The result in rough terms: the cheapest decile on low P/E grew from 10 000 dollars to over 1,2 million dollars during the period, while the comparison groups — the most expensive decile and the market — landed on the order of 240 000 dollars or below. Measured as an annual interest rate, that means the low deciles ran at around 15–20 percent per year depending on measure and period, against roughly 11–14 percent for the market and the most expensive deciles.
The difference sounds modest per year, but after 27 years of compounding the interest rate on the interest rate, that difference is the difference between becoming a millionaire and becoming a quarter-millionaire. And crucially: the pattern repeated on all four measures Dreman tested.
The most important thing about the evidence is not the top decile but the staircase. If low P/E had won only in a single group it could have been a lucky streak, but Dreman shows a monotone pattern: the return rises essentially step by step from the most expensive decile to the cheapest. A relationship that holds across ten steps, four different measures and 27 years is not chance — it is a structure.
It is exactly the bar AK1A lays on all its own factors and that O'Shaughnessy showed with sixty years of data in What Works on Wall Street: gradation separates factor from fluke.