Labor Productivity: Structural Change and Cyclical Dynamics

A longstanding puzzle of empirical economics is that average labor productivity declines during recessions and increases during booms. This paper provides a framework to assess the empirical importance of competing hypotheses for explaining the observed procyclicality. For each competing hypothesis we derive the implications for cyclical productivity conditional on expectations of future demand and supply conditions. The novelty of the paper is that we exploit the tremendous heterogeneity in long-run structural changes across individual plants to identify the short-run sources of procyclical productivity. 'University of Maryland and NBER; **Free University of Amsterdam (on leave from Federal Reserve Board); ***University of Maryland and NBER, CES Fellow, Census Bureau. The views expressed herein are solely the authors'. We wish to thank the staff and visiting researchers a t the Center for Economic Studies, U.S. Bureau of the Census, for providing a stimulating research environment. We would also like to thank participants of the NBER Summer '93 workshop, and the Econometric Society meetings, January 1995, for valuable comments and suggestions. Table 1: 1987 Summary Statistics: A11 Mfg. vs. LRD Sample All Mfg. LRD Sample Sample as % of All Establishments 368897 8669 2.34 Output (Billions) 2329.6 1100.7 47.25 Employment (000) 17406.5 6332.8 36.38 Avg. Estab. Size 47 73 1 Plant Distribution By SIC 20 2 1 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 16573 142 0.86 * Excludes computers, SIC 3573. groupings using information from the Current Population Survey. In later analysis, we also look at variants of employment and productivity, with output measured by double-deflated value added, and employment measured by total number of employees. The central point of Chart 1 is that the basic aggregate properties of output, employment, and productivity fluctuations for our sample are the same as those exhibited by the published series. Specifically, for our sample and the published series, output and total hours are procyclical, output is more cyclically volatile than total hours and, accordingly, average labor productivity is procyclical. There are some quantitative differences in the nature of the cyclicality between our sample and the published series that we explore below. 3.2 The Contribution of Simple Composition Effects We now turn to the last hypothesis for procyclical productivity, that of composition effects. As stated earlier, we are not able to discuss the effect of plant turnover on productivity movements directly. As seen in the bottom panel of chart 1, however, our sample appears to have more cyclical productivity than the aggregate for all manufacturing. This would agree with our prior belief that less productive plants are more likely to exit in a trough, and reduce the procyclicality of the aggregate. For our sample of continuing plants, we can decompose annual productivity changes into a 'within' plant component, and two mix-shift terms: an employment share effect, and a cross term. Ant Xi 4t-l,iAnt,i + E, &t,int-l,i + E, A&,iAnt,i -nt-1 nt-1 nt-1 nt-l , (.l)