The smallest stocks are not just smaller: global evidence

Using an international Thomson Reuters Datastream database, where size coverage is unusually wide and data errors have been reduced to a low level, we show that some specification decisions, and especially those related to size, may have a significant impact on asset-pricing test results. We also show that, in data with wider coverage with respect to size, the Fama and French factor portfolios need to be adjusted and their number increased. Specifically, (i) standard asset-pricing models leave pricing errors for the 10% smallest stocks, and (ii) two additional risk factors (i.e. one micro-stock factor and one extreme book-to-market factor) are needed to capture this mispricing. This holds both in USA and international data. Further research is needed to measure the separate relevance of the possible economic interpretations and to identify more economic explanations for the additional risks associated with the smallest stocks.