SummaryUsing a new dataset for the German market, this article analyses whether modeling time-varying stochastic discount factor parameters in the CAPM of Sharpe (1964), the HCAPM of Jagannathan and Wang (1996) and the CCAPM of Lucas (1978) can help to explain the cross-section of book-to-market, size and industry portfolio returns. In addition to classic financial conditioning variables, we focus on modern consumption-based variables - the consumption surplus ratio of Campbell and Cochrane (1999), the consumption-wealth ratio of Lettau and Ludvigson (2001a) and the labour income to consumption ratio of Santos and Veronesi (2006). Our results show that (a) time-varying parameters can drastically increase the empirical fit of the models and that (b) a CAPM using the labour income to consumption ratio as a conditioning variable proves to be the best model specification.