The Department of Labor’s (DOL) new rule changes how it calculates the mandatory minimum wage—called the prevailing wage—for employers of H-1B and permanent foreign workers. DOL adopts a fundamentally flawed methodology as its basis to inflate the prevailing wage. But a bigger issue is that DOL itself failed to understand how much its methodological changes would artificially raise the required wages. DOL estimated the wage effects of its rule using completely erroneous assumptions, and so it understates to the public the wage increases by, in many cases, as much as 26 percent.

The prevailing wage is supposed to approximate the wages of similarly skilled U.S. workers. DOL currently uses the Bureau of Labor Statistics’ (BLS) Occupational Employment Statistics (OES) survey to create a prevailing wage for four skill levels within each occupation in every area of the country. The creation of the skill levels—which is the focus of this rule—is contentious because the OES doesn’t directly record skills. Instead, BLS creates these skill levels mathematically based on the reasonable assumption that higher wages within an occupational category within a specific area generally reflect higher skills.

Table 1 compares the new and old prevailing wage methodologies. Previously, DOL had assumed that the bottom third of the wage distribution represented entry level wages, while the rest (the top two thirds) were not entry level. After averaging the wages in the bottom third and top two thirds to create the Level 1 and Level 4 wages, it placed the two other wage levels equally distant between them. The new rule, however, ignores all wages below the 40^th percentile—averaging the 40th to 50th percentiles to determine the starting wage—and rather than averaging all other wages to produce Level 4 as before, it only averages the top decile of wages (the 90^th to 100^th percentile).

The big error that DOL makes is that it assumes that averaging the top decile will equal the 95^th percentile. Since the top decile (the top 10 percent of wage earners) includes some extreme outliers and a very small sample size, those outliers skew the level 4 wage far higher than the 95^th percentile. Because the Level 2 and Level 3 wages are dependent on the Level 4 wage calculation, it also mistakes where these wage levels will fall on the wage distribution. This becomes incredibly important when DOL then tries to estimate the wage impact of the new rules:

To estimate the wage impacts of new percentiles contained in this [rule], the Department used publicly available BLS OES data that reports the 10th, 25th, 50th, 75th, and 90th percentile wages by SOC code and metropolitan or non‐​metropolitan area. In order to estimate wages for the new IFR levels of 45th, 62nd, 78th, and 95th percentiles, the Department linearly interpolated between relevant percentiles for reported wages at each SOC [job] code and geographic area combination. [Endnote: For example, if OES reports a wage of $30 per hour at the 25th percentile and $40 per hour at the 50th percentile then the 45th percentile is interpolated as $30+($40-$30)*((45–25)/(50–25)) = $38 per hour.] For the 95th percentile, the Department used OES wages reported for the 90th percentile at each SOC code and geographic area combination. (p. 126)

In other words, DOL did not even try to estimate what the actual 95^th percentile would be. For reasons that are only explicable as laziness or hastiness, it did not ask BLS to calculate the new prevailing wages until after it issued the rule. Instead, it just settled on using the 90^th percentile as a stand‐​in for the level 4 wage to estimate the rule’s effects because it thought that it would be similar. It then tried to calculate the level 3 and level 2 wages assuming that they would be around the 78^th and 62^nd percentiles, respectively. But this was wrong.

Table 2 is an example of DOL’s assumed wage levels for San Francisco architects and New York City management analysts under the new rule and what the actual wage levels now are. For wage levels 2–4, DOL’s estimate is wildly inaccurate—underestimating the actual amount by 15–18 percent for level 2, 23–25 percent for Level 3, or 24–26 percent for level 4. These huge differences appear in every single occupational category. Notice that in the examples below, the actual level 2 wage is higher than what DOL assumed would approximate the level 3 wage, and the actual level 3 wage is higher than what DOL assumed would approximate the level 4 wage.

DOL repeatedly refers to the wage levels as “percentiles,” misstating its own methodology which is based on averaging the 5th decile and the top decile and then imputing the middle two wage levels equally distant from the other levels. Because the level 4 wage is so inflated, the other two middle wage levels are also widely off to the point that what DOL refers to as the “78^th percentile” (level 3) is actually above the 90^th percentile in the two examples above.

Because DOL decided not to ask BLS for the actual OES data that it would use to create the new wage levels, it missed how egregious this mistake was. Moreover, because DOL refused to let the Office of Management and Budget review the rule, this massive error slipped through unnoticed. Finally, because DOL made the rule take effect almost immediately and refused to accept public comments on the rule, it has no time to correct it before the rule begins to effect tens of thousands of American businesses. DOL should immediately rescind the rule and start the process over.

Post was updated to clarify that the methodology of the entry level wage is the average of the 5th decile, not the 45th percentile exactly.