Low Frequency Movements and SVAR Analyses

This paper studies the consequences of using a deterministic steady state in Vector Autoregressive (VAR) models, when the data may display structural breaks, transitional dynamics or low-frequency fluctuations. We document substantial upward biases in the estimated coefficients, with distortions further amplified by the identification scheme. Allowing the steady state to be stochastic, however, reduces these distortions. To address this issue, we propose a spike-and-slab prior to differentiate between two alternative long-run specifications. Finally, we apply our empirical framework to revisit two well-known debates in macro: (i) the dynamics of hours in response to technology shocks; (ii) the habit formation hypothesis and the humpshaped response of consumption to business cycle shocks.