The District Is Not an Island

The Vietnamese robot-adoption study treats migration and neighboring districts as part of the labor-market outcome, not as background geography. This chapter follows two movements that an employment total can hide: more-skilled workers move toward exposed districts while some low-skilled workers move away, and demand generated by automation reaches places that did not receive the same direct exposure. Understanding these flows clarifies both the denominator behind local employment rates and the extent to which automation's gains or losses travel across borders.

A district is a useful unit for measuring local adjustment, but it is not a sealed container. Firms recruit from elsewhere, workers commute or move, suppliers serve more than one place, and rising incomes create demand beyond the factory gate. If the analysis ignores those connections, it can mistake labor-supply changes for hiring, treat displacement as a local absence, or attribute neighboring growth entirely to unrelated development.

Migration changes what a local employment gain means

The study finds that robot-exposed districts expand in total and working-age population and receive more labor-force participants. An additional robot per thousand workers is associated with roughly a 6–10% increase in labor-force participants. In-migration has an elasticity of about 14%, while out-migration is positive at approximately half that magnitude but imprecisely estimated. The exact figures are local estimates, but their relationship is important: the districts gaining jobs are also gaining people who can fill them.

This helps explain the unchanged employment rate. A district can record more employed people without raising the share of its working-age population that is employed if new residents arrive at roughly the same time as new jobs. The employment count answers “How many people are working here?” The employment rate asks “What share of the relevant population is working?” When population is endogenous to the technology shock, the two measures describe different aspects of the same adjustment.

In-migration also changes the composition of labor supply. The study finds that inflows into robot-adopting districts are driven primarily by young-to-middle-aged and more-skilled workers, with broadly similar patterns across genders. These are the workers most likely to complement expanding production or to compete successfully for the formal opportunities created around it. Their arrival can help a firm scale up, but it can also mean that the measured employment gain belongs partly to people who were not in the district when the automation episode began.

The direction of out-migration is different. Significant out-migration is concentrated among low-skilled workers. This is consistent with a worker whose routine formal job has been displaced leaving the district rather than appearing as locally unemployed or informal. It does not prove that every departing low-skilled worker was displaced by a robot: people move for many reasons, and the study estimates group-level responses rather than individual histories. It does mean that a local measure of low-skilled harm can be incomplete if the most affected workers exit the area being measured.