Regress and Progress! An Econometric Characterization of the Short-Run Relationship between Productivity and Labor Input in Brazil

Abstract

What do technology shocks do? This is a hard question to answer. Real-Business-Cycle (RBC) models have provided a better understanding of the effects of technology shocks over business-cycle frequencies. Still, some problems remain. This paper addresses the empirical adequacy of first-generation RBC models through the use of structural vector autoregressions (SVAR) models. We employ an identification condition that imposes few a priori restrictions upon the data and is consistent with a broad class of macroeco- nomic models. Based on those conditions, we are able to obtain conditional correlation coefficients and impulse response functions that may be confronted with the theoretical implications of RBC models. We also report evidence related to short-run increasing returns to labor (SRIRL) in the Brazilian industry. Our results cast doubt on some RBC models’ main predictions. In particular, the estimated conditional correlations between labor input and productivity measures are negative for technology shocks and positive for non-technology shocks, with the labor input displaying a negative response to technol- ogy shocks over business-cycle horizons. These results are robust to several specification issues, such as sample instability and the consideration of higher-order systems during estimation.