Article révisé par les pairs
Résumé : The most common definition of efficient stock markets states that in such markets prices fully reflect all available information. This obviously implies that in order for a market to be deemed (semi-strongly) efficient, it should not be possible to earn abnormal returns using trading strategies based on publicly available accounting data. However, especially since the late 1980s, a number of such trading strategies have been shown to be lucrative. In the underlying paper the profitability of two investment strategies that were originally developed on the basis of U.S. data is investigated for the Belgian stock market on the basis of accounting data for the period 1990-1998. The two replicated studies are Ou and Penman (1989) and Holthausen and Larcker (1992). With respect to On and Penman's results, we are unable to draw an unequivocal conclusion. On the one hand, we find an indication that some earnings-based variables may be predicted on the basis of financial statement data. The trading strategies building on these forecasting abilities do not, however, systematically earn abnormal returns. This is due to the fact that our proxies for unexpected earnings appear to lack value relevance. We also find that Holthausen and Lacker's s results cannot be considered valid for the Belgian stock market, simply because the LOGIT models estimated on the basis of accounting data do not produce sufficiently accurate forecasts of the signs of future abnormal returns.