When a Wage Estimate Becomes a Visa Fee

The subject now is the employer’s decision to use an H-1B slot for a particular worker once the wage comparison is turned into a price. This chapter explains how the Borjas model moves from an adjusted wage gap to payroll savings, willingness to pay, simulated demand, and revenue. The key judgment is that payroll savings are not the same as profit, and a fee prediction is only as strong as the assumptions added between the wage record and the hiring decision.

From a wage gap to payroll savings

The model starts with a specific choice. A firm can hire an H-1B worker at an annual wage, or it can hire a comparable U.S.-born worker at a different annual wage. The H-1B relationship is assumed initially to last for the full six-year visa term. A one-time fee is charged when the visa is awarded. The employer compares the discounted cost of the two alternatives over those years rather than comparing one year’s salaries in isolation.

That comparison needs a term for everything the wage data do not reveal. The model calls it an unobserved cost differential. It can be negative when the H-1B worker brings complementary skills, raises productivity, or reduces turnover costs. It can be positive when sponsorship creates legal expenses, coordination costs, language barriers, or other inefficiencies. The term is not a synonym for exploitation or productivity. It is a container for an unknown difference in the true cost of the two hires.

This hidden term is what keeps the model from making an overly simple inference from a wage gap. Suppose the petition wage is lower than the wage predicted for a comparable native worker. The difference looks like payroll savings, but the H-1B worker may also cost more to recruit, supervise, or retain. Conversely, suppose the H-1B wage is not lower. The worker could still be worth hiring if they bring a productivity advantage or another benefit that the wage comparison cannot see. The observed wage is one component of the employer’s decision, not a complete profit-and-loss statement.

The fact that the worker was hired supplies one useful restriction. During the FY2021–FY2024 period in the data, there was no modeled fee. If a firm hired the worker anyway, the unobserved cost increase could not have been larger than the payroll saving that the firm appeared to receive. In the model’s terms, the hidden differential must lie below the worker-specific savings threshold. This selection rule explains how the simulation can include workers whose observed H-1B wage is higher than the predicted native wage: an unobserved productivity gain or other benefit can make the hire worthwhile.

The author estimates the native wage for each H-1B worker from a regression fitted to the native sample. That regression uses observable skills and fixed effects, and the predicted value serves as the wage the worker would earn if they were native-born under the model’s comparison. The difference between that predicted wage and the petition wage is then treated as the worker-specific payroll saving. It is a model-based counterfactual, not an offer observed from a real alternative employer.