24

This paper sheds a new light on Fama and French (1993) and Carhart (1997) multifactor models estimation focusing on the possibility of errors-in-variables. We use monthly data for the Canadian stock market from July 1960 to December 2004. Fama and French (1997) conclude that “ estimates of the cost of equity for the three-factor model of FF (1993)” are imprecise. Our results show that this imprecision is more severe in presence of measurement errors. We suggest to use Dagenais and Dagenais (1997) higher moments estimator to deal with this problem. This estimator has the advantages of being easy to compute and to require no extraneous information. We show that estimates of the cost of equity on the Canadian stock market obtained with Dagenais and Dagenais estimator sharply differ from biased OLS estimates. This approach revisits performance attribution and abnormal performance (?).