Macro and Micro Dynamics of Productivity: Is the Devil in the Details?∗

semanticscholar(2015)

Cited 1|Views0
No score
Abstract
Does the method of estimating plant-level productivity matter? We attempt to answer this question in the context of key stylized facts and popular estimation methods. Using plant-level manufacturing data for the U.S., we test the robustness of results on five dimensions. First, we find non-trivial differences in estimated factor elasticities, especially for capital, across commonly used methods. These differences yield considerable variation in estimated returns to scale across methods. Second, the variation in elasticities maps into differences in (total factor) productivity dispersion but does not invalidate the general conclusion that productivity differences across establishments within the same industry are large. In addition, the ranking of plants by productivity within industries is also sensitive to method. Third, more productive plants are shown to be more likely to grow and survive, no matter how we estimate productivity. However, outliers in factor elasticities that arise more frequently from some methods non-trivially impact the quantitative marginal effects of productivity on growth and survival. Fourth, all our productivity variants confirm the main conclusions from the structural productivity decomposition literature: reallocation is productivity enhancing, and variation in within-plant productivity seems more important in terms of cyclical fluctuations of aggregate productivity by all methods considered. However, here again there are non-trivial quantitative differences across methods in the contribution of reallocation to aggregate productivity growth. Some methods imply that all or even more than all of aggregate productivity growth is due to reallocation while other methods imply only 25 percent is due to reallocation. Finally, we look at the robustness of productivity dispersion and growth and survival results to imputation and the assumption that elasticities are homogenous within industries. Dispersion is negatively influenced by imputation and the homogeneity assumption. Growth and survival results are also affected but the effect of these factors is more in line with the variation we found in previous exercises. ∗We would like to thank conference participants at the 2013 Comparative Analysis of Enterprise Data in Atlanta and the 2014 Research Data Center Annual Conference for valuable comments. We are grateful to Kirk White for useful discussions and for making his code available to us. Any remaining errors are our own. Any conclusions expressed herein are those of the authors and do not necessarily represent the views of the U.S. Census Bureau. All results have been reviewed to ensure that no confidential information is disclosed. †Corresponding author: zoltan.wolf@census.gov.
More
Translated text
AI Read Science
Must-Reading Tree
Example
Generate MRT to find the research sequence of this paper
Chat Paper
Summary is being generated by the instructions you defined