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 opposite. A groundbreaking study by George Borjas (2026), published in the Journal of Labor Economics, resolves this apparent contradiction by demonstrating that native labor supply adjustments and self-selection into employment systematically bias standard empirical estimates.
When an immigration-induced labor supply shock occurs, some native-born workers—particularly women and lower-wage workers—exit the labor force (the crowd-out effect). Because the workers who exit are disproportionately from the lower end of the wage distribution, the average wage of the remaining employed natives in repeated cross-sections is artificially inflated. This positive selection bias masks the true negative wage impact of immigration.
The Methodological Framework: Correcting for Selection and Crowd-Out
To reveal the true underlying labor demand elasticity, Borjas (2026) models the wage impact of immigration by incorporating two factors that standard literature typically ignores or assumes to be fixed:
- Sample Selection (The Inverse Mills Ratio): Adjusting for the self-selection of native workers in and out of the labor force using a Heckman-style correction.
- Native Labor Supply (Log N): Explicitly controlling for the size of the native labor force rather than assuming it is constant.
Using regional data from France over the 1982–2016 period and instrumenting the immigrant share with a classic shift-share instrument, Borjas compares standard OLS and IV estimates with selection-corrected models.
Key Empirical Findings
- Native Women (The Impact of Selection): In the simplest IV model that does not correct for selection or control for native labor supply, the estimated impact of immigration on the wages of native women is close to zero and statistically insignificant (-0.08). However, when correcting for sample selection (using the inverse Mills ratio), the coefficient becomes significantly negative at -0.42. When the model is expanded to control for both sample selection and the size of the native labor force (revealing the crowd-out effect), the estimated wage elasticity falls to -0.91. In other words, an immigration-induced 10% increase in the labor force is predicted to lower the wages of native women by about 9%.
- Native Men: For native men, the selection bias is less severe but still present. The estimated wage elasticity is negative and significant across all specifications, lying between -0.50 and -0.80 (with the fully corrected specification at -0.70).
- Convergence of Elasticities: Once both selection bias and native labor supply adjustments are corrected, the estimated wage elasticities for native men (-0.70) and native women (-0.91) are large, negative, and statistically indistinguishable from one another.
Economic Implications
This research has profound implications for how economists interpret natural experiments and regional studies. It shows that the "no-effect" finding of many local-area studies—such as those tracking the The Mariel Boatlift Debate: Card's No-Effect Finding vs. Borjas's Reanalysis or general regional inflows—may be an artifact of native flight and selection. When low-skilled immigrants enter a local market, the exit of low-wage natives prevents the local average wage from falling, creating the illusion of wage resilience.
By demonstrating that the labor supply curve is indeed downward-sloping once behavioral responses are accounted for, Borjas (2026) provides a rigorous reconciliation of the wage and employment impacts of immigration, showing they are two sides of the same coin.
Representative Quotes
From George Borjas (2026)
"Equation (11) shows that immigration has three distinct effects on the wage change observed in market k. The first is the direct short-run effect of the shock Dmk, captured by the (negative) wage elasticity h. The percent change in the number of native workers generates its own attenuating effect, as that supply response moves the labor market back up the labor demand curve... The correction of the biases introduced by the crowd-out effect and the self-selection of workers results in a wage elasticity that has roughly the same value for the two groups. In fact, the difference between the -0.7 elasticity for men and the -0.9 elasticity for women reported in column 8 is not statistically significant." — George Borjas, Journal of Labor Economics (2026)
"The OLS coefficient of the immigrant share becomes significantly negative, with a value of -0.44 (0.07). The change in the impact of immigration between columns 1 and 2 is predicted by our theoretical framework if the women who exit the labor force in the postmigration period have relatively low wages." — George Borjas, Journal of Labor Economics (2026)