This paper explores empirically the role of noisy information in cyclical developments and aims at separating fluctuations that are due to genuine changes in fundamentals from those due to temporary animal spirits or expectational errors (noise shocks). Exploiting the fact that the econometrician has a richer data-set in some dimensions than the consumers, we use a novel identification scheme in a structural vector-autoregressive (SVAR) framework. Our results show that noise shocks are more important for business cycle fluctuations than permanent (or technology) shocks. We also show that technology shocks turn negative a few years before recessions, while noise shocks are very positive at the cycle peaks. By contrast, the recovery from recessions is mostly led by technology shocks, noise shocks remaining negative for some time during this business cycle phase.
animal spirits business cycles identification Kalman Filter noise shocks signal-extraction problem structural vector autoregression technology shocks