By TIMOTHY COGLEY AND JAMES M. NASON *
The time-series literature reports two stylized facts about output dynamics in the United States: GNP growth is positively autocorrelated, and GNP appears to have an important trend-reverting component. This paper investigates whether current real-business-cycle (RBC) models are consistent with these stylized facts. Many RBC models have weak internal propagation mechanisms and must rely on external sources of dynamics to replicate both facts. Models that incorporate labor adjustment costs are partially successful. They endogenously generate positive autocorrelation in output growth, but they need implausibly large transitory shocks to match the trend-reverting component in output. (JEL E32, C52)
There is an extensive empirical literature on the time-series properties of aggregate output. For example, prominent univariate studies include papers by Charles R. Nelson and Charles I. Plosser (1982), Mark W. Watson (1986), John Y. Campbell and N. Gregory Mankiw (1987), John H. Cochrane (1988), and James D. Hamilton (1989). Multivariate analyses include papers by Olivier Jean Blanchard and Danny Quah (1989), Robert G. King et al. (1991), and Cochrane (1994). This literature documents two stylized facts about output dynamics in the United States. First, GNP growth is positively autocorrelated over short horizons and has weak and possibly insignificant negative autocorrelation over longer horizons (e.g., Nelson and Plosser, 1982; Cochrane, 1988). Second, GNP appears to have an important trend-reverting component that has a hump-shaped impulse-response function (e.g., Blanchard and Quah, 1989; Cochrane, 1994).
This paper links the empirical literature on output dynamics with the theoretical literature on real-business-cycle (RBC) models. In particular, it considers whether various RBC models are consistent with these stylized facts. Our approach is similar in spirit to the “test of the Adelmans” (e.g., Irma Adelman and Frank L. Adelman, 1959; King and Plosser, 1994; Scott P. Simkins, 1994), except that we concentrate on a different set of stylized facts. 1,2 We ask how often an econometrician armed with the techniques used in the time-series literature would observe the same kind of stylized facts in data generated by RBC models. To
* Economic Research Department, Federal Reserve Bank of San Francisco, 101 Market St., San Francisco, CA 94105, and Department of Economics, University of British Columbia, 1873 East Mall, Vancouver, BC V6T 1Z1, respectively. We are grateful to two referees for their comments; we have also benefited from discussion and correspondence with Jack Beebe, John Cochrane, Roger Craine, Allan Gregory, Thomas Sargent, Anthony Smith, Gregor Smith, Richard Startz, and George Tauchen. Desiree Schaan provided exceptional research assistance. Much of this work was done while the authors were visiting the Haas School of Business at UC-Berkeley, whose hospitality is gratefully acknowledged. Opinions expressed in this paper do not necessarily represent the views of the Federal Reserve Bank of San Francisco or the Federal Reserve System.
1 These authors consider whether various models can replicate Burns-Mitchell stylized facts.
2 There is also an extensive literature on testing for unit roots. All the models that we study replicate the univariate persistence found in U. S. GNP. This follows directly from the specification of technology shocks. In models where technology shocks are difference-stationary, output has a unit root. In models where technology shocks are trend-stationary, output has a near unit root that conventional tests cannot distinguish from unity.