This Week In Finance

Technical Trading Rules and the EMH: New Evidence

A perennial question among academics and practitioners is that of capital market efficiency. Specifically, to what extent are future security price changes predictable via the use of publicly available price and volume data (technical analysis) and/or news and financial statement information (fundamental analysis)? Until fairly recently, the bulk of the published empirical research indicated strongly that domestic capital markets are at least weak-form efficient (i.e., publicly-available price and volume information are impounded rapidly in current market values).

However, in 1992 an article by three professors, Brock, Lakonishok, and LeBaron was published in The Journal of Finance, which seemed to indicate that at least some technical trading rules have predictive power. The authors tested 26 technical trading rules using daily values of the Dow Jones Industrial Average over the period 1897 - 1986 and found that a small number of the trading rules tested outperformed a benchmark portfolio.

As is often the case, the results raised as many questions as they answered. A serious problem with the running of a series of tests using a single data set is that the results can be contaminated by data-snooping. Put another way, it is theoretically possible to develop a successful predictive "trading rule" using a set of random numbers (which are, by definition, not predictable) if one tries enough combinations, and restricts his attention to only that data set.

The October, 1999 issue of The Journal of Finance contains an interesting update of the previously-cited article. In "Data-Snooping and Technical Trading Rules" by Sullivan, Timmerman , and White, the authors reexamine the Brock, et al. study by testing 7,846 (!) technical trading rules using the DJIA. However, they test the rules determined to have been successful over the 1897 - 1986 period during the subsequent period ending in December, 1996. What do they find?

Well, if you're a technician, the news is not good. After performing numerous statistical tests, the authors conclude that "the superior performance of the best technical trading rule [in the 1897 - 1986 period] is note repeated in the out-of-sample experiment covering the 10-year period 1987-1996. In this sample the results are completely reversed and the best-performing trading rule is not even statistically significant at standard critical levels. . . there is no evidence that any trading rule outperforms over the sample period." (STW, page 1683)

Does this end the debate? Of course not! On the other hand, it does suggest that we might be back to square one in the efficient markets debate: those who believe technical trading rules outperform buy-and-hold are still searching for quantifiable evidence.