Handbook of Empirical Corporate Finance SET

5.4 Robustness issues

The matched-firm technique discussed above uses firm characteristics (size and B/M) to adjust for priced risks, while the factor regression approach uses a set of pre-specified portfolios as proxies for pervasive risks. Either approach suffers from potential “bad model” problems in terms of representing the true asset pricing model. Since tests for abnormal returns are always a joint test of the risk factors assumed to generate expected return, it is therefore useful to provide information on the sensitivity of abnormal return estimates to alternative model specifications. Moreover, factor regressions may suffer from non-stationarity in the estimated parameters that may be predictable using publicly available information. ...