Learning, Capital-Embodied Technology and Aggregate Fluctuations

Recent evidence suggests that agents’ expectations may have played a role in several cycli¬cal episodes such as the U.S. "new economy" boom in the late 1990s, the real estate boom in Japan in the 1980s and the real estate boom in the U.S. which ended in 2008. One chal¬lenge in the expectations driven view of fluctuations has been to develop simple one sector models that can give rise to such fluctuations without a compromise on other dimensions. In this paper we propose a simple generalization of the Greenwood et al. (1988) one sec¬tor model and show it can generate fluctuations that arise as a result of agents difficulty to forecast productivity embodied in new capital. The two key assumptions in the model are: (1) the vintage view of capital productivity, whereby each successive vintage has (po¬tentially) different productivity and (2) agents’ imperfect information and learning about this productivity. The model is consistent with second and third moments from U.S. data. Simulations of the model suggest that, (a) noise amplifies fluctuations and (b) pure noise can trigger recessions that mimic in magnitude, duration and depth the typical post WW II U.S. recession.