1 Panel-Data Estimates of the Production Function and the Revenue Function: What Difference Does it Make?#

The lack of individual firm information on output prices is a major problem in the econometrics of production. In particular, one would expect that it may account for a significant share of the large discrepancies found in productivity panel-data studies between the cross-sectional and time-series estimates of capital and scale elasticities (Klette and Griliches, 1996). For this study, we were able to obtain individual information on output price indices—as well as on capacity utilization rates—for a balanced panel of about 450 French manufacturing firms over four years and an unbalanced panel of 675 Spanish manufacturing firms over nine years. However, whether we rely on OLS or IV panel-data estimation methods, estimating the revenue function (using a nominal output measure) or the production function proper (using a real output measure) makes little or very little difference on our results. The biases due to other sources of specification errors are probably more important. Needless to say, the availability of individual output price information remains essential to investigate and model firm behavior thoroughly. Key words: production function; revenue function; panel data; estimation bias; specification errors; price dispersion; capacity utilization. JEL classification: C23; D24. The first author was inspired to write this paper following long and friendly discussions with Zvi Griliches and Tor Klette. The authors are particularly grateful to Rozenn Desplatz, Alberto Lopez and Benoît Mulkay for their contributions in helping in the construction and exploratory investigation of the samples. They are also thankful to Bruno Crépon and Patrick Sevestre for remarks and suggestions. An account of preliminary results on the French sample has been published in French in: Mairesse et Desplatz, 2003: "Econométrie de la production sur données de panel et dispersion des prix de production: quel