A note on business cycle accounting

Chari, Kehoe, and McGrattan (2007) (CKM) show that a large class of dynamic stochastic general equilibrium (DSGE) models with various frictions and shocks is observationally equivalent to a benchmark real business cycle (RBC) model with correlated 'wedges' in the RBC model's first-order conditions. The wedges in the static first-order conditions of the RBC model can be readily computed by evaluating the first-order conditions at the data and then solving for the wedges. In contrast, identification of the 'investment wedge' in the RBC model's dynamic Euler equation requires the researcher to make assumptions about the expectation formation by agents in the RBC model. In particular, CKM assume that expectations are formed as if, from the perspective of the model's agents, wedges followed a vector autoregressive process of order one (VAR(1)). We show that wedges generally do not have a VAR(1) representation, implying that CKM's procedure is based on modelinconsistent expectations. We also provide an alternative, model-consistent approach to modeling expectation formation. On the former issue, we present a necessary and sufficient 'rank condition' under which a detailed economy can be mapped into a benchmark model where wedges follow a VAR(1) process. On the latter issue, we suggest that the information set underlying the expectation formation should not only contain current wedges, but also all predetermined variables.