Richard Roll (1977) criticized the use of proxies for the market portfolio when testing the Capital Asset Pricing Model (CAPM). He showed that in order to test the model the econometrician needs to have access to the entire set of risky assets. In this paper we develop an Estimated Beta Pricing Model (EBPM) that allows us to test a linear relation similar to the one of the CAPM. The advantage of EBPM is that it holds under less restrictive assumption and can be tested using a subsample of risky assets. This model is essentially a one factor model where the common source of risk is captured by the excess return of the portfolio most correlated with the stochastic discount factor with respect to its zero-beta portfolio. When we compare the EBPM to the Fama-French Three-Factor model we find that both models capture the same source of common risk. In fact, we show that the three variables used as factors in the Fama-French model are correlated with the same source of common risk with the consequence of reducing the dispersion in betas when all are used in the same regression.