One of the difficulties in testing the CAPM, as highlighted in the Roll's critique, is how to select a suitable market proxy. The literature has proposed a number of alternatives to the use of Stock Exchange Indices as a proxy for the Market Portfolio. All of them looked at a broader measure of assets which are not sold at Stock Exchanges. This paper follows Hou (2002) by testing the CAPM through a hypothetical aggregate portfolio that generates GDP as its dividend. The goal of the paper is to evaluate if the proxy proposed by Hou (2002) can be used as a good proxy for the market portfolio in Brazil. In addition, we compare the performance of the synthetic return proposed by Hou (2002) to the Ibovespa return, a very popular proxy for the market portfolio in empirical studies in Brazil. Results, especially the ones related to the Black version of the CAPM, show that the synthetic return measure is not mean-variance efficient for the full sample (1991-2002). In spite of not being able to meet all the requirements imposed by the CAPM model in all sub-samples, the Ibovespa return has a better performance compared to the synthetic return. In fact, it is mean-variance efficient for the full sample and for the majority of sub-samples.
CAPM; GDP; synthetic portfolio
