This paper proposes the estimation of small-scale dynamic stochastic general equilibrium (DSGE) monetary models under the quasi-rational expectations (QRE) hypothesis. The QRE-DSGE model is based on the idea that the determinate reduced form solution associated with the structural model, if it exists, must have the same lag structure as the ‘best fitting’ vector autoregressive (VAR) model for the observed time series. After discussing solution properties and the local identifiability of the model, a likelihood-based iterative algorithm for estimating the structural parameters and testing the data adequacy of the system is proposed. A Monte Carlo experiment shows that, even controlling for the omitted dynamics bias, the over-rejection of the nonlinear cross-equation restrictions when asymptotic critical values are used and variables are highly persistent is a relevant issue in finite samples. An application based on euro area data illustrates the advantages of using error-correcting formulations of the QRE-DSGE model when the inflation rate and the short-term interest rate are approximated as difference stationary processes. A parametric bootstrap version of the likelihood-ratio test for the implied cross-equation restrictions does not reject the estimated QRE-DSGE model.
