Source Themes

This paper proposes decomposing macroeconomic fluctuations into three components based on their contributions to fluctuations of different frequencies. A first component explains business cycle fluctuations but has no long-run effect; a second component also influences business cycles but may have a long-run effect; a third component has only long-run effects. The decomposition has several desirable properties. The two business cycle components jointly explain 99% of business cycle fluctuations and deliver comovements consistent with a structural interpretation as demand and supply shocks, respectively. The presence of long-run, non-business cycle fluctuations biases full-information estimation of business cycle models, however, distorting conclusions about the sources of business cycles. Estimating a model to match only the business cycles shocks resolves this issue and gives more plausible parameter estimates.

This paper investigates whether the documented shortcomings of medium-scale DSGE models, as outlined by Beaudry et al. (2020), stem from parameter misspecification or a fundamental absence of endogenous propagation mechanisms. Beaudry et al. (2020) identify that the spectral density of “hours worked” displays two distinct peaks, one at business cycle frequencies and another at lower frequencies. They argue that medium-scale DSGE models fail to accurately capture both the position and magnitude of the spectral density peak at business cycle frequencies, attributing this limitation to insufficient endogenous mechanisms and the restrictive nature of AR(1) shocks. In contrast, we argue, both analytically and through model simulations, that these models can reproduce the spectral density peak at business cycle frequencies when AR(1) shocks are employed, provided there is additional endogenous propagation in labor markets. Importantly, this propagation mechanism has significant implications for counterfactual policy analysis.