It is widely acknowledged that the recent generation of DSGE models failed to incorporate many of the liquidity and financial accelerator mechanisms revealed in the global financial crisis that began in 2007. This paper complements the papers presented at the 2009 BIS annual conference focused on the role of banks and other financial institutions by analysing the role of household decisions and their interplay with credit conditions and asset prices in the light of empirical evidence. In DSGE models without financial frictions, asset prices are merely a proxy for income growth expectations and play no separate role. On UK aggregate consumption evidence, section 2 of the paper shows this is strongly contradicted by the data, for all possible discount rates and both for a perfect foresight and an empirical rational expectations approach to measuring income expectations. However, an Ando-Modigliani consumption function generalised to include a role for liquidity, uncertainty, time varying credit conditions, wealth and housing collateral effects, as well as income expectations, explains the data well. Section 3 reports new evidence on the striking rejection on aggregate data of the consumption Euler equation central to all DSGE models. Section 4 shows that UK micro evidence is consistent with the generalised Ando-Modigliani model. Section 5 discusses the limitations of recent DSGE models with financial frictions and housing. Section 6 discusses some business cycle implications of amplification mechanisms and non-linearities operating via households and residential construction. It reconsiders econometric methodology appropriate for designing better evidence-based central bank policy models.