Censoring and its impact on multivariate testing of the Capital Asset Pricing Model

The primary objective of this paper is to assess the affect of data 'censoring' on asset pricing tests. This is achieved by modifying tests to incorporate a 'selectivity bias' correction factor in a Gibbons (Journal of Financial Economics, 10, pp. 3-27, 1982) multivariate framework. The sample comprises daily Australian stock returns for 524 companies over the five-year period 1995 to 1999. First, it is found that the use of a 'selectivity bias' correction factor is generally justified in stocks with a degree of censoring at about 50% or above. This represents approximately 52% of the sample. Second, despite the first finding no evidence is found supporting the need for such a correction in asset pricing tests - the degree of support for the CAPM is not materially affected.