Selection Bias in Native Responses: Reconciling the Employment and Wage Impacts of Immigration

Selection Bias in Native Responses: Reconciling the Employment and Wage Impacts of Immigration

A major puzzle in labor economics is why some empirical studies find that immigration has a negative impact on native employment but little to no impact on native wages, while other studies find the reverse. A landmark peer-reviewed study by George J. Borjas and Anthony Edo published in the Journal of Labor Economics (April 2026), titled "Gender, Selection into Employment, and the Wage Impact of Immigration," reconciles this asymmetry by documenting the critical role of selection bias.

Using the French labor market as a natural experiment, Borjas and Edo demonstrate that standard regional wage regressions are heavily contaminated by the non-random self-selection of native workers who choose to exit or remain in the workforce in response to an immigrant supply shock.

Reconciling the Gender Asymmetry in France

Following a 1976 policy shift that granted foreign-born workers the right to family reunification, France experienced a rapid "feminization" of its immigrant labor force, with women accounting for 75.6% of immigrant population growth between 1975 and 2007.

The "raw" regional correlations in France over the 1982–2016 period suggest a stark gender asymmetry:

• For native men (who have highly inelastic labor supply), immigration is strongly negatively correlated with wages but has essentially zero correlation with employment rates.
• For native women (who have highly elastic labor supply), immigration is strongly negatively correlated with employment rates but has a near-zero or weakly positive correlation with wages.

Borjas and Edo show that the apparent "zero wage elasticity" for native women is a compositional illusion. The immigrant supply shock disproportionately pushed low-wage native women out of the labor force. Because these low-wage workers exited, the average wage of the remaining female "survivors" in the region rose mechanically, masking the true downward pressure on wages.¹

Correcting for Selection Bias

By integrating a standard Mincerian labor demand framework with a Heckman selection model, the authors adjust for this compositional shift. Once selection bias is corrected, the wage elasticity for native women becomes negative, statistically significant, and nearly identical to that of native men:

"Adjusting for the selection bias results in a similar wage elasticity for both French men and women (between -0.8 and -1.0)."

The study also quantifies the employment crowd-out effect. The authors estimate a female crowd-out parameter of approximately 0.55, implying that about 6 native women left (or did not enter) the workforce for every 10 new immigrants who entered.

Methodological Implications

The paper argues that the common practice of "tracking" continuously employed workers to bypass selection bias does not actually solve the problem. Continuously employed "survivors" are themselves a self-selected group, and their observed wage trends do not represent the counterfactual wage trends of the broader at-risk population.

This selection bias taints standard spatial correlations and cross-area studies globally. Whenever native workers respond to immigration by changing geographic areas, switching occupations, or leaving the workforce, their sorting is highly unlikely to be random, systematically biasing unadjusted estimates of the wage impact of immigration.