Working Paper: NBER ID: w31608
Authors: Ben Gilbert; Hannah Gagarin; Ben Hoen
Abstract: We use restricted-access, geocoded data on the near-universe of workers in 23 U.S. states in order to quantify the impact of wind energy development on local earnings and employment, by race, ethnicity, sex, and educational attainment. We find the largest relative impacts for workers without a high school education, or workers with a college education, in addition to other systematic differences across sub-populations. We compare these results to estimates using county aggregates of the worker-level data, such as can be obtained using publicly available data. We find that (a) county-level estimates are dramatically dampened relative to geocoded worker-level estimates, and (b) the degree of bias differs by sub-population such that qualitative comparisons of impacts are not consistent using restricted-access data versus county-level data for most sub-populations. We discuss implications for achieving equity goals within energy transition policies.
Keywords: energy transitions; employment impacts; wage impacts; wind energy; environmental justice
JEL Codes: Q4; Q42; Q43; R11; R12
Edges that are evidenced by causal inference methods are in orange, and the rest are in light blue.
| Cause | Effect |
|---|---|
| Utility-scale wind capacity within 20 miles of a worker's residence (R29) | Increase in earnings (J31) |
| Utility-scale wind capacity within 20 miles of a worker's residence (R29) | Increase in employment (J23) |
| Utility-scale wind capacity within 20 miles of a worker's residence (R29) | Largest proportional marginal impact for black workers (J79) |
| Utility-scale wind capacity within 20 miles of a worker's residence (R29) | Larger gains for men than women (J79) |
| Utility-scale wind capacity within 20 miles of a worker's residence (R29) | Largest proportional impacts for workers without a high school diploma (F66) |
| County-level aggregates (R10) | Attenuated estimates compared to worker-level data (J39) |
| Worker-level data (J29) | Statistically significant impacts (C29) |
| County-level data (R10) | Effectively zero impacts (F69) |