Academic journal article Journal of Money, Credit & Banking

The Response of Hours to a Technology Shock: A Two-Step Structural VAR Approach

Academic journal article Journal of Money, Credit & Banking

The Response of Hours to a Technology Shock: A Two-Step Structural VAR Approach

Article excerpt

THE RESPONSE OF hours to a technology shock is the subject of many controversies in quantitative macroeconomics. The contributions of Gall (1999), Basu, Fernald, and Kimball (2006), and Francis and Ramey (2005) show that the short-un response of hours worked to a technology shock is significantly negative in the U.S. economy. Gall (1999) and Francis and Ramey (2005) obtain this result using a structural vector autoregression (SVAR) of labor productivity growth and hours in first difference (DSVAR) with long-run restrictions (see Blanchard and Quah 1989). Basu, Fernald, and Kimball (2006) use a direct measure of aggregate technology change, controlling for imperfect competition, varying utilization of factors and aggregation effects, and find that hours fall significantly on impact after a technology improvement.

Moreover, Gali (1999, 2004) shows that the level of hours significantly decreases in the short run in all G7 countries and the euro area as a whole, with the exception of Japan. These results are in contradiction with Christiano, Eichenbaum, and Vigfusson (2004). Using a SVAR with a level specification of hours (LSVAR), they find a positive and hump-shaped response of hours after a technology shock. Moreover, they show that the LSVAR specification encompasses the DSVAR specification.

The specification of hours in level or in difference appears to be the core issue of the controversy. Galf (1999), Gall and Rabanal (2004), and Christiano, Eichenbaum and Vigfusson (2004) perform various unit root tests, but it becomes hard to obtain a clearcut evidence in favor of level or difference specification. Furthermore, recent contributions proceeding with simulation experiments point out that the specification of hours in SVARs using long-run restrictions can alter significantly the estimated effect of a technology shock on hours. For example, Chari, Kehoe, and McGrattan (2008) simulate a business cycle model estimated by maximum likelihood on U.S. data with multiple shocks. They show that the DSVAR specification leads to a negative response of hours under a real business cycle (RBC) model in which hours respond positively. As pointed out by Christiano, Eichenbaum, and Vigfusson (2004), the DSVAR specification may induce strong distortions if hours worked are stationary in level.

In this paper, we use a simple method that allows us to consistently estimate technology shocks and thus the responses of hours to a technology improvement. In contrast to existing LSVAR and DSVAR specifications, we choose to exclude hours worked series from SVARs to identify technology shocks. I The proposed approach consists in the following two steps. In a first step, a SVAR model with a long-run restriction that includes well-chosen covariance stationary variables allows to properly identify the technology shock series. Among these variables, the consumption-to-output ratio seems to be a promising candidate. Two reasons motivate this choice. First, as argued by Cochrane (1994), this ratio may help to better predict the permanent and transitory components of output. Indeed, using a simple permanent income argument, permanent (technology) shocks can be separated from other (nonpermanent and nontechnology) shocks because these latter do not modify the consumption. The joint observation of output growth and consumption-to-output ratio allows then the econometrician to properly identify permanent and transitory shocks. Second, both from the simulations of a DSGE model and the actual data, we obtain that the consumption-to-output ratio displays less persistence than hours. When a SVAR model with long-run restrictions includes variables characterized by a highly persistent process (typically hours with the level specification), the identification of the responses of hours to technology shocks can be seriously disturbed. Gospodinov (Forthcoming) using a near-unit root process for the hours and Christiano, Eichenbaum, and Vigfusson (2004) using a unit root process have shown that the LSVAR specification for such highly persistent processes leads to an inconsistent estimator of the technology shocks. …

Search by... Author
Show... All Results Primary Sources Peer-reviewed

Oops!

An unknown error has occurred. Please click the button below to reload the page. If the problem persists, please try again in a little while.