Government pay surveys are totally missing the real wage gap in tech because they ignore the stock options people get.
March 26, 2026
Original Paper
<div> Vetter's Paradox in Real Data:&nbsp; </div> <div> Senior Wage Premiums, Employment Reallocation, and the Measurement Blindness of Government Wage Surveys </div>
SSRN · 6347239
The Takeaway
While official Bureau of Labor Statistics data suggests top software engineers earn about 3 times more than entry-level staff, the actual ratio is often over 10 to 1 when stock options are included. Because government surveys primarily track base salaries, policymakers are making decisions based on data that misses up to 70% of the total compensation at the top of the market.
From the abstract
The Acemoglu-Restrepo framework predicts monotonic automation displacement. Deobhakta (2026) shows theoretically that when within-task competence heterogeneity (κ) is substantial, displacement follows a spike-cliff trajectory. This paper tests for early-stage spike dynamics using three data sources: BLS Occupational Employment and Wage Statistics (2019-2024), CPS microdata (2018-2025), and Levels.fyi total compensation reports. In government wage data, the evidence is suggestive but limited: mea