New research argues standard migration regressions understate how responsive workers are by ignoring destination choices and correlated shocks across places, biasing estimates toward zero.
Economists often estimate how people move by running a “migration regression” that links changes in a place’s population to a measure of how much local labor demand has risen or fallen. A recurring puzzle is that these regressions frequently produce small, statistically weak estimates—even when the same shocks clearly affect wages, employment, and other labor-market outcomes. In new work, Kirill Borusyak, Rafael Dix-Carneiro, and Brian K. Kovak argue that this pattern should not be read as evidence that workers are immobile or indifferent to local economic conditions.
Their central point is methodological: the standard regression framework can systematically understate how responsive migration actually is because it fails to represent how people choose among multiple possible destinations. In a simple equilibrium model, what matters for a location’s population growth is not just whether it receives a positive or negative shock, but how that shock compares with conditions in the places that would otherwise attract or retain its workers. Migration decisions are inherently relative—workers weigh opportunities across competing locations.
That relative nature creates a key problem for empirical work: local shocks are often correlated across connected labor markets. If a downturn or boom tends to occur simultaneously in a location and in its main alternatives—for example, because regions are exposed to similar industries or macroeconomic forces—then the observed gap between “origin” and “destination” conditions is smaller than the local shock alone suggests. When researchers regress population changes only on the local component, ignoring the correlated conditions elsewhere, the estimated relationship is “attenuated,” meaning the coefficient is biased toward zero.
The authors also identify a second source of attenuation that arises in widely used “shift-share” approaches, where researchers construct local labor-demand shocks from national industry trends weighted by a location’s initial industrial composition. When shocks are defined at the industry level, the conventional shift-share measure may not match a location’s true effective exposure in the equilibrium migration environment the authors describe. In particular, the relevant exposure depends on how a location’s industries and its feasible destinations’ industries jointly shape relative opportunities; the standard measure can mischaracterize that exposure, again pushing estimated migration responses downward.
To quantify how large these distortions can be, Borusyak, Dix-Carneiro, and Kovak run simulation exercises calibrated to detailed administrative data from Brazil and Germany. Across these experiments, they find that attenuation can easily exceed 50%, implying that conventional regressions can miss more than half of the true migration response. They also show that, beyond bias, traditional estimates can be highly imprecise—producing wide confidence intervals that make it difficult to distinguish meaningful mobility responses from statistical noise.
As a remedy, the researchers propose empirical strategies that are explicitly tied to the underlying equilibrium model. These approaches aim to correct the biases created by correlated shocks across destinations and by mismeasured exposure in shift-share designs. By grounding the estimation in theory, they argue, the resulting coefficients recover parameters with a clear structural interpretation and yield sharper, more informative estimates of how strongly people relocate in response to local economic shocks.
The authors acknowledge comments from a wide set of economists across seminars and conferences, note research assistance, and report financial support for Borusyak from the European Research Council under the EU’s Horizon 2020 program (grant 949995). They also state that the views expressed are their own and not necessarily those of the National Bureau of Economic Research.
Why it matters
Migration is a key adjustment mechanism in regional economies, shaping inequality, housing demand, and the local effects of trade and technology shocks. If standard methods systematically understate mobility responses, policymakers and researchers may misjudge how quickly workers can relocate—and which places will bear the lasting costs of downturns. The next step is to see how these theory-based corrections change conclusions in prominent empirical settings that rely on conventional migration regressions or shift-share instruments.