Understanding the link between micro and aggregate productivity dynamics has been a key theme in the recent empirical and theoretical literature. There has been a rapid increase in the development and availability of longitudinal business databases around the world that permit studying industrial dynamics on many rich dimensions. One key limitation of the typical longitudinal business database (e.g., the LRD for the U.S. manufacturing sector) is that the data do not provide information on plant- or firm-level prices. As such, the typical measures of plant-level output, inputs and productivity, using datasets like the LRD, have been based upon deflating nominal revenue of the plant with a sectoral-level deflator. This procedure is appropriate only if there is price homogeneity within the sector. However, the limited evidence that is starting to emerge indicates that there is substantial within sector price heterogeneity. Given such large heterogeneity, productivity measures which rely on sector-level price deflators capture differences in productivity across plants but also differences in prices across plants within sectors. In the last several years, theoretical and empirical researchers have recognized this limitation and have been developing indirect methods to take into account the presence of within sector price heterogeneity. While considerable progress has been made using these indirect methods, this contribution is directly to assess the importance of using plant-level prices of inputs and outputs for measurement, given the access to a longitudinal business database that includes plant-level prices. This database provides a unique laboratory to explore the role of heterogeneity in plant-level prices in outputs and inputs for the measurement, interpretation and analysis of plant dynamics.
Broader Impact: Output price heterogeneity within sectors may be due to differences in mark-ups, differences in product quality or differences in demand shocks across plants within sectors. Measurement of output price heterogeneity, along with measures of revenue, permits decomposing movements in the traditional "revenue-based productivity" measures into the component associated with movements in physical productivity and the movements due to relative prices. A whole host of research questions can be addressed using such decompositions. First, it is of interest to know whether the main insights from the existing literature on plant dynamics hold up when physical output and relative price movements are measured separately. Second, it is of interest to explore the sources of plant-level price heterogeneity--in particular, does such price heterogeneity reflect differences in mark-ups, product quality or demand shocks. Third, it is of interest to explore the relative role of physical productivity differences and relative price differences in accounting for the observed plant dynamics. For example, a common finding in the literature is that plants with lower measured productivity are more likely to exit. However, an open question, given the available productivity measures until now, is whether exiting plants indeed have lower productivity or whether they simply charge higher prices. Finally, given the absence of plant-level prices in most datasets, it is of interest to explore to what extent the indirect methods developed in the recent literature are successful in controlling for within sector price heterogeneity. The researchers are in a unique position to answer that question given their access to a longitudinal business database that does have direct measures of prices. In terms of broader impact, the findings and analysis in this study are expected to be helpful to those using datasets without plant-level prices on the research strategies that can and should be used when trying to draw inferences using deflated revenue-based measures of output and productivity. Since there are now longitudinal business datasets for more than 50 countries around the world and virtually all of them do not have plant-level prices, the potential broader impact is substantial.
