Revisiting some Productivity Debates

Researchers interested in estimating productivity can choose from an array of methodologies, each with its strengths and weaknesses. I evaluate five widely used techniques; two nonparametric, index numbers and data envelopment analysis, and three parametric methods, instrumental variables estimation, stochastic frontiers, and semi-parametric estimation. Using two samples of manufacturing firms from developing countries, from Colombia and Zimbabwe, I find that the different methods produce surprisingly similar productivity estimates, even though the input coefficient estimates vary widely. Crucially, on two debates in the literature the different methods also reach very similar conclusions. We find evidence for productivity growth premiums associated with variables that are plausible proxies for knowledge acquisition, as modeled in the endogenous growth literature. Plant-level productivity changes are an important determinant of aggregate productivity growth, especially over shorter time horizons. On a third debate, whether exporters are able to increase their productivity advantage, the parametric methods find that self-selection based on initial productivity level does not explain the entire productivity gap. In contrast, the nonparametric methods do not find a significant productivity advantage for exporters, even including self-selection effects, indicating that technology differences between exporters and nonexporters are likely to be important. ∗University of Toronto, Institute for Policy Analysis, 140 St. George Street, Suite 707, Toronto, ON M5S 3G6, Canada, [email protected]. I would like to thank Jim Tybout, Mark Roberts, and Marcel Fafchamps for providing me with the data and Mel Fuss, Frank Wolak, Marc Melitz, and participants at the NBER Summer Institute and the 2003 summer workshop at the CEU for comments. All remaining errors are my own. Financial support from the Connaught Fund and SSHRC is gratefully acknowledged.