Discussion

Abstract

John Haltiwanger opened the discussion by bringing attention to the dataused in thepaper. The analysis useddata fromtheNational Establishment Time Series (NETS). The authors argued that the NETS is a reliable source because it aligns on several dimensions with another data set, the County Business Patterns (CBP). However, Haltiwanger argued that NETS overstates employment because of imputations. Further, NETS sales data are not reliable, as documented by Barnatchez, Crane, and Decker (“An Assessment of the National Establishment Times Series [NETS] Database,” Finance and Economics Discussion Series 2017 [2017]: 110). The authors recognized that there are imputations in NETS. However, they emphasized that their findings are validated by several robustness checks. In addition, they noted that similar results hold when using a different data set, the Longitudinal Business Database (LBD). The next comments were related to the suitability of the HerfindahlHirschman Index (HHI) as ameasure ofmarket power at different levels of geographic aggregation. Jeffrey Campbell pointed out that the HHI is not a reliable measure of market power for large cities, where there are many firms in themarket and there is substantial variation across neighborhoods. He encouraged the authors to repeat their analysis for small towns. The authors responded that they replicated their results for different measures of concentration and that their findings are robust regardless of the measure considered. Erik Hurst questioned why the authors decided to start the analysis at the smallest area of aggregation, the ZIP-code level, as opposed to a larger area of aggregation, such as the county level. Seconding Campbell’s remark, the authors responded that themeasurement ofmarket power ismore reliable for smaller geographical