Working Paper: NBER ID: w31588
Authors: David Card; Jesse Rothstein; Moises Yi
Abstract: We revisit the estimation of industry wage differentials using linked employer-employee data from the U.S. LEHD program. Building on recent advances in the measurement of employer wage premiums, we define the industry wage effect as the employment-weighted average workplace premium in that industry. We show that cross-sectional estimates of industry differentials overstate the pay premiums due to unmeasured worker heterogeneity. Conversely, estimates based on industry movers understate the true premiums, due to unmeasured heterogeneity in pay premiums within industries. Industry movers who switch to higher-premium industries tend to leave firms in the origin sector that pay above-average premiums and move to firms in the destination sector with below-average premiums (and vice versa), attenuating the measured industry effects. Our preferred estimates reveal substantial heterogeneity in narrowly-defined industry premiums, with a standard deviation of 12%. On average, workers in higher-paying industries have higher observed and unobserved skills, widening between-industry wage inequality. There are also small but systematic differences in industry premiums across cities, with a wider distribution of pay premiums and more worker sorting in cities with more high-premium firms and high-skilled workers.
Keywords: wage differentials; firm-based approach; industry premiums; labor market; worker sorting
JEL Codes: J31; J62
Edges that are evidenced by causal inference methods are in orange, and the rest are in light blue.
| Cause | Effect |
|---|---|
| systematic skill sorting across industries (L00) | actual wage premiums (J31) |
| unobserved heterogeneity in wage premiums across different firms within the same industry (J31) | bias in estimated industry effects (C51) |
| the hierarchy term (gap between the pay premium at the actual workplace and the average premium in the associated industry) (J31) | changes in industry premiums (L16) |
| ground-up approach (B50) | standard deviation of industry premiums (G22) |
| local labor market characteristics (J29) | industry premiums (G22) |
| higher values of local skill sorting (J24) | employment in high-premium industries (J39) |
| estimated industry effects from a richly specified cross-sectional model (C21) | magnitude of the premiums from an AKM approach (C58) |
| industry premiums from a model with person and industry effects (C29) | effects from a ground-up AKM approach (C90) |