An updated version of this analysis is available. A revised version was published on April 23, 2026, with refined estimates of the strategic-reclassification offset and other corrections. Please read the updated version instead. This page is preserved for reference and is no longer listed on PWBM’s site.

Key Points:

• Under the proposed prevailing wage thresholds, 21 percent of current registrations would fall below the new Level I floor and receive no lottery entries, while only 7 percent would reach Level IV, down from 17 percent.

• Combined with the new wage-weighted lottery, the higher thresholds would raise mean compensation of selected registrants by $20,611 (+18.4%) over the pre-2026 random lottery, nearly double the $11,159 (+10%) gain from wage-weighting alone.

• Raising prevailing wages amplifies the demographic and occupational shifts from wage-weighting: India’s selection share falls 5.2 pp, the doctorate share rises 3.3 pp, and Computer and Mathematical occupations lose 2.8 pp relative to the random lottery.

• Strategic reclassification of occupations could offset 32 percent of the compensation gain, reducing the increase from 18.4 percent to 12.5 percent.

Background

Beginning with the March 2026 lottery, the Department of Homeland Security (DHS) replaced the random H-1B lottery with a wage-level-weighted selection process to allocate the 85,000 cap-subject visas, which are typically oversubscribed by about a factor of five. Registrations receive 1 to 4 lottery entries based on their prevailing wage level: Level I receives 1 entry, Level II receives 2, Level III receives 3, and Level IV receives 4. Our previous analysis projected this rule would increase average compensation of selected registrants by approximately 10 percent relative to the random lottery, with minimal effects on U.S.-born worker wages.¹

The prevailing wage system is administered by the Department of Labor’s (DOL) Office of Foreign Labor Certification (OFLC). While DHS’ new wage-weighted lottery takes the prevailing wage thresholds as given, those thresholds are themselves a policy choice. DOL currently sets the four wage levels at approximately the 17th, 34th, 50th, and 67th percentiles of the occupational wage distribution for each geographic area.² On March 27, 2026, DOL published a Notice of Proposed Rulemaking, “Improving Wage Protections for the Temporary and Permanent Employment of Certain Foreign Nationals in the United States” (hereafter, the 2026 NPRM), that would raise these to approximately the 34th, 52nd, 70th, and 88th percentiles.³ Comments are due by May 26, 2026. If finalized, this would be the first change to the prevailing wage percentile methodology since DOL adopted the current structure in 2005. DOL previously attempted a similar increase through a January 2021 final rule, but that rule was vacated by a federal court before it took effect.

How the 2026 NPRM would change the wage level distribution of H-1B registrations

Under current thresholds, most registrations fall at Levels I (28.6 percent) and II (39.2 percent), with 17.4 percent reaching Level IV. The NPRM’s higher thresholds compress this distribution sharply. We treat registrations whose offered wage falls within 5 percent of the new Level I threshold as Level I, assuming their employers would slightly raise the offer to clear the new floor. Even with that buffer, 20.9 percent of registrations fall below the cutoff entirely (“Too Low”) and would receive no lottery entries. Only 7 percent qualify at Level IV. Figure 1 shows how wage level assignments under current prevailing wages would change under the proposed new scale.⁴

Almost all current Level II registrations drop to Level I, and 73.1 percent of current Level I registrations fall below the new Level I threshold entirely. Even among current Level IV registrations, only 40.5 percent remain at Level IV.

Compensation effects

Table 1 compares the compensation distribution of selected registrants under three scenarios: (1) the pre-2026 random lottery, (2) the wage-weighted lottery with current prevailing wage thresholds, and (3) the wage-weighted lottery with the NPRM thresholds. For scenario (3), registrations with wage offers that are “too low” (20.9 percent of the original pool) are dropped and the remaining pool is resampled back to the original size, assuming enough demand exists to refill it; the composition of the registrant pool changes but the size does not.

The proposed thresholds add $9,452 (+7.7%) to mean compensation above what the current prevailing wage levels produce; this is the incremental impact of the NPRM (row C). Combined with the move to wage-level-weighted lottery beginning in 2026, the total gain over the random lottery reaches $20,611 (+18.4%), nearly double the $11,159 (+10%) the weighted lottery achieves alone (rows B and A).

Wage level distribution of selected registrants

Figures 2 and 3 show the wage level composition of selected registrants in each scenario under the two different prevailing wage scales. Figure 2 classifies all selected registrants using current OFLC thresholds; Figure 3 uses the NPRM’s proposed thresholds. Scenarios (1) and (2) draw from the same registration pool; scenario (3) draws from the reduced pool that excludes the 20.9 percent of registrations that fall below the NPRM’s new Level I floor.

On the current OFLC scale (Figure 2), Level IV’s share among winners rises from 18 percent under the random lottery to 30.9 percent under the current lottery and thresholds to 41.7 percent under NPRM thresholds, while Level I falls from 28.1 percent to 13.3 percent to 6 percent.

The same set of selected winners looks very different when evaluated against the NPRM’s scale (Figure 3). Because the NPRM’s thresholds are higher, a registrant classified as Level II in Figure 2 might fall to Level I or lower in Figure 3. Under the random lottery, 20.6 percent of winners would fall below the proposed Level I (the “Too Low” category) and 47.4 percent at Level I. The wage-weighted rule with current wage thresholds cuts the “Too Low” share to 9.7 percent. With NPRM thresholds, “Too Low” registrants are excluded from the pool entirely, so none appear among winners, but 37.3 percent still land at Level I.

Who gets selected?

The NPRM’s higher thresholds would amplify the demographic and compositional shifts already induced by the move from a random lottery to a wage-weighted lottery rule: older workers, better-educated workers, and a more geographically diverse nationality mix.

The doctorate share rises from 5.7 percent under the random lottery to 7.6 percent under the current weighted lottery to 9.0 percent if the NPRM thresholds were adopted — nearly twice the shift that the weighted lottery alone produces.

India’s share of selections falls from 66.8 percent (random lottery) to 64.1 percent (current prevailing wages) to 61.6 percent (NPRM prevailing wages). Canada sees the largest gain, rising from 2.7 to 4.7 percent.

Computer and Mathematical occupations, which constitute 71 percent of registrations in recent years, lose selection share under both prevailing wage scales. Business and Financial Operations gain the most (+1.98 pp under NPRM thresholds). Educational Instruction and Library occupations also lose share (-0.21 pp), as wages in those fields tend to fall at the low end of the occupational wage distribution.

Strategic reclassification under higher thresholds

Our companion brief found that under current thresholds, 60.7 percent of registrations could achieve a higher wage level by reclassifying into a related occupation (using the 5 nearest O*NET occupations). Under the proposed NPRM thresholds, that share rises to 82.5 percent.⁵ Because the NPRM compresses most registrations into Level I, more of them have room to gain a level by switching to an occupation with lower prevailing wage thresholds.

Table 6 shows the wage-level transitions available to registrants under the proposed NPRM thresholds (5 nearest occupations, first-alternative and maximum-gain strategies). The largest flows are from Level I: 23.4 percent of the pool can reach Level II, 20.8 percent can reach Level III, and 10.6 percent can reach Level IV under the first-alternative strategy. Only 17.5 percent of registrations have no reclassification potential.

Table 7 shows the compensation effects after reclassification. We estimate that 31.8 percent of the expected compensation increase under the weighted lottery with NPRM thresholds is offset by first-alternative reclassification, reducing the gain in average compensation from 18.4 percent to 12.5 percent relative to the random lottery. Under the maximum-gain strategy, 35.8 percent of the gain is offset.

Table A1 in the appendix compares the reclassification offset under current and proposed thresholds. The offset rate falls from 42.2 percent to 31.8 percent, but the absolute dollar offset is larger ($6,552 vs. $4,711). The rate falls because the base compensation gain nearly doubled (+$20,611 vs. +$11,159). The absolute offset grew through two roughly equal channels: more registrations can reclassify (82.5 vs. 60.7 percent of the pool, accounting for 59 percent of the increase in expected extra entries per worker) and each reclassifier gains more lottery weight (1.64 vs. 1.33 average extra entries, accounting for the remaining 41 percent). The intensity increase reflects the NPRM’s compression of registrations into Level I, which creates room for larger upward jumps through reclassification (see our companion brief for the transition matrix under current thresholds). Even after reclassification, the 12.5 percent mean compensation increase exceeds the 10 percent gain under current thresholds without any reclassification.

Figure 5 traces average compensation as the number of related occupations considered for reclassification increases. The “Weighted lottery” dashed line is the upper bound: no reclassification. As registrations reclassify, compensation falls but remains above the random lottery baseline regardless of how many related occupations are considered. Under the maximum-gain strategy with 11 or more related occupations, however, average compensation in the NPRM panel falls below the level the weighted lottery achieves under current thresholds without any reclassification. The first-alternative strategy does not cross that line, suggesting this risk is limited to cases where registrants have wide occupational flexibility and pursue the most aggressive reclassification available.

Figure A1 decomposes the expected wage-level gain by occupation group and source level, showing that Architecture and Engineering contributes the most per worker while Computer and Mathematical occupations have the largest absolute impact due to their 71 percent pool share.

These simulations hold employer behavior fixed: they do not model wage-setting responses to the higher thresholds beyond the 5 percent buffer for registrations that would otherwise fail to qualify. If employers raise offered wages to maintain Level II or III classification under the proposed thresholds, the actual compensation effects could be larger than estimated here.

Appendix

Reclassification offset decomposition

Figure A1 decomposes the expected wage-level gain from reclassification by occupation group and source wage level. For each group and level, the contribution equals the group’s exposure (share of registrations at that level) times the reclassification rate (share that can reclassify) times the average jump in wage levels. Architecture and Engineering contributes the most per worker, driven by a high reclassification rate and large average jump at Level I. Computer and Mathematical occupations rank lower per worker but account for 71 percent of all registrations, so their reclassification behavior has the largest absolute impact on the pool. Educational Instruction and Library contributes the least, a reversal from current thresholds, where it was among the largest contributors. Because the NPRM raises thresholds in related occupations as well, a Level I registrant whose salary once reached Level III in a neighboring occupation may now fall short of Level II, closing most reclassification pathways for lower-wage occupation groups.

Prevailing wage estimation methodology

We estimate a counterfactual prevailing wage schedule under DOL’s 2026 NPRM, focusing on the proposed percentile-based target structure at approximately the 34th, 52nd, 70th, and 88th percentiles. We apply the NPRM to OFLC prevailing wage data (wage year 2022-23, derived from OEWS May 2021, used by the FY 2024 lottery).

Under the legacy OFLC methodology, Level I and Level IV are based on the mean of the lowest one-third and the mean of the upper two-thirds of the OES wage distribution: Level I $= E[W \mid W < Q(1/3)]$ (approximately $p_{17}$), Level IV $= E[W \mid W \geq Q(1/3)]$ (approximately $p_{67}$), with Levels II and III determined by statutory interpolation.² The 2026 NPRM proposes to replace this framework with percentile-based anchors. DOL identified percentile pairs for Levels I and IV that align the average prevailing wage with a “Benchmark Value” — the average of mean wages across LCA-program occupations, weighted by the current composition of employers, occupations, and locations. Starting from initial values of the 35th and 90th percentiles (matching the 2021 final rule), DOL settled on the 34th and 88th percentiles. Levels II and III are then obtained by dividing the dollar gap between Levels I and IV into thirds, as required by the INA. In the NPRM, this yields an overall target structure summarized as approximately the 52nd and 70th percentiles, though in any given SOC-by-area cell the interpolated dollar values need not equal the exact 52nd and 70th quantiles of the fitted wage distribution.

For SOC-by-MSA cells with complete published OEWS percentile data, we fit a log-normal distribution to the five published percentiles (p10, p25, p50, p75, p90), constrained to match the observed OEWS mean wage. Under a log-normal model, the quantile function is

where $\mu$ and $\sigma$ are the location and scale parameters of the log-normal distribution, $\Phi^{-1}$ is the inverse standard normal CDF, and $p$ is the percentile. The mean is $E[W] = \exp(\mu + \sigma^2/2)$. Setting $E[W] = m$ (the observed OEWS mean) pins $\mu = \log(m) - \sigma^2/2$, leaving $\sigma$ as the single free parameter. We choose $\sigma$ to minimize the sum of squared log-residuals:

We validate each fit against published OFLC levels: 79.9 percent of cells pass a 5 percent error threshold; 96.1 percent pass a 10 percent threshold.

For the remaining cells (those without complete OEWS percentile data), we recover $(\mu, \sigma)$ by inverting the published OFLC levels. Under the current methodology, Level I and Level IV are conditional means over the bottom third and upper two-thirds of the wage distribution:

For a log-normal distribution with cutoff $z_c = \Phi^{-1}(1/3)$, these have closed-form expressions in terms of $\mu$ and $\sigma$. Given the published $L_I$ and $L_{IV}$, we solve for $\sigma$ from the ratio $L_I / L_{IV}$ and recover $\mu$ from the implied unconditional mean. Together, the two methods cover 99 percent of all 451,984 OFLC cells.

To illustrate the estimation, consider SOC 15-1211 (Computer Systems Analysts) in MSA 37980 (Philadelphia-Camden-Wilmington). The published OEWS annual wage percentiles are:

We fit a log-normal distribution constrained to match the $105,000 mean, yielding $\sigma = 0.3558$ and a maximum fit error of 4.5 percent across the five percentiles. The predicted current OFLC levels closely track the published values:

Applying the 2026 NPRM’s proposed percentile-based methodology to the same fitted distribution produces the estimated new wage levels:

Under the 2026 NPRM, the Level I threshold for this cell would rise from $69,971 to $85,106, and the Level IV threshold from $122,262 to $149,715. A worker offered $105,000 in this occupation and area would exceed the Level III threshold under current law but fall between the Level I and Level II thresholds under the proposed rule. Figure A2 illustrates this shift.

The full set of estimated 2026 NPRM prevailing wages is available for download (estimated_2026_nprm_pw.zip). The file follows the same structure as the ALC_Export.csv file in OFLC’s published prevailing wage data (OFLC_Wages_2022-23), with estimated NPRM levels replacing the current OFLC levels.

This analysis was produced by PWBM staff under the direction of Alex Arnon.