Kydland and Prescott (1982) and Prescott (1986) established that productivity shocks could account for most post-World War II business cycle volatility. Business cycle volatility was roughly constant up through the period studied by Kydland and Prescott, but has changed substantially since then. Kim and Nelson (1999), McConnell and Perez-Quiros (2000) and Stock and Watson (2002) all identify a large and statistically significant permanent decline in U.S. GDP volatility beginning in the first quarter of 1984.
This paper examines this decreased volatility through the lens of neoclassical business cycle theory. We focus our analysis on changes in the variance of the Hodrick-Prescott cyclical component of real GDP, its components, labor input, and total factor productivity (TFP). All of these variances are about 30-50 percent smaller in the post-1983 period compared to the 1955-83 period.
Within the neoclassical framework, changes in cyclical volatility are the result of either changes in the volatility of the exogenous shocks that are fed into the model, and/or changes in the structure of the model that maps the exogenous shocks into the endogenous variables. We focus our analysis on changes in the exogenous shock volatility.
We first evaluate the ipact of changes in the volatility of TFP shocks. We find that the volatility of this shock declines about 50 percent after 1983. We find that this volatility change reduces the volatility of output and its components, and labor input also by 50 percent in the Hansen (1985) model. This finding suggests that lower productivity shock volatility can be a significant factor underlying lower cyclical volatility. Some economists will question this finding, however, because they argue that TFP shocks are not productivity shocks per se, but rather the endogenous consequence of other shocks operating through unmeasured capital and labor utilization.
This "mis-measurement" view of TFP would suggest that the change in TFP volatility is due to the change in the volatility of some other shock, combined with unmeasured changes in factor utilization. We therefore pursue this possibility using the model of Burnside et al. (1996) that features both variable capital and labor utilization. We follow Chari, Kehoe, and McGrattan (2002, 2006) and Cole and Ohanian (2002) who focus on three other shocks for understanding fluctuations in the growth model: a shock to the household's static first order condition, a shock to the household's dynamic first order condition, and an additive shock to the resource constraint, such as government spending shocks.
Download
Why Have Business Cycle Fluctuations Become Less Volatile?
