The Clock and the Map: Where the Dispute Lives

The H-1B wage estimate depends on two coordinates that are easy to hide inside a table: the time at which a wage is measured and the place assigned to a worker. This chapter examines how the Borjas analysis and the Economic Innovation Group’s reconstruction align those coordinates, and why the resulting disagreement is about the object being compared rather than a minor arithmetic adjustment. The practical judgment is to treat a reported percentage as conditional on a clock and a map.

The wage clock is not the fiscal-year label

A fiscal year is an administrative label, not a complete description of when a worker was paid. The H-1B lottery usually occurs in March before the start of the fiscal year. In the example used by the NBER paper, the FY2024 lottery was held in March 2023, and winning firms filed I-129 petitions soon afterward. The petition reported the wage that would be paid once employment began, but the data did not record the actual employment start date.

That creates a gap between the petition and the paycheck. Employment could not begin before the later of the effective H-1B status date and October 1, 2023. Even with no delay, the paper argues, at most one quarter of the offered salary could have been paid in calendar 2023. Delays make the share smaller: using evidence about consular processing, the paper estimates that at most about 15 percent of the offered wage was likely paid in that calendar year for the relevant group.

The ACS has a different clock. It samples households throughout the year, and reported earnings refer to the previous twelve months. A person interviewed in 2024 may therefore report earnings received partly in 2023 and partly in 2024. Putting an I-129 offer next to an ACS annual-earnings figure without deciding how these windows relate can mix a wage offer, a start date, and a trailing income period.

Borjas’s revised NBER paper addresses the problem by using the date on which H-1B status becomes effective to place the wage in real terms. It deflates the H-1B wage using the average CPI over the twelve-month period after that effective date. It uses a related adjustment for ACS earnings, taking the mean CPI for the two years that make up the survey’s trailing earnings window. The goal is to express both sides in constant dollars at the time earnings were actually received, with reported dollar values in 2025 terms.

EIG treats the same calendar differently. It argues that the FY2021–FY2024 petition observations correspond to wages from calendar years 2020 through 2023, while the native ACS baseline covers 2023 only. In its account, inflation over that period was 15.1 percent, and wages in software occupations rose 18.7 percent. Comparing earlier H-1B wages with later native wages therefore makes the H-1B side look artificially low. EIG reports that either comparing the 2023 H-1B observations—those labeled FY2024—with 2023 ACS data, or converting all years to real 2023 dollars, roughly halves the estimated gap.

The disagreement is not merely about whether inflation exists. Both sides say that wages should be expressed in comparable real terms.