We use a robust regression estimator to analyze the risk premia on siz
e and book-to-market. We find that the risk premium on size that was e
stimated by Fama and French (1992) completely disappears when the 1 pe
rcent most extreme observations are trimmed each month. We also show t
hat the negative average of the monthly size coefficients reported by
Fama and French can be entirely explained by the 16 months with the mo
st extreme coefficients. We argue that further investigation of these
results could lead to an understanding of the economic forces underlyi
ng the size effect, and may also yield important insights into how fir
ms grow.