Where the Dividend Stops: Sectors, Samples, and Uncertainty
The Mexican trade-diversion result is strongest when it is kept within its actual boundaries. This chapter locates those boundaries across sectors and municipalities, then treats the robustness checks as answers to competing explanations rather than as a ritual list of statistical tests. The aim is not to undo the export and employment findings. It is to decide which parts of the story travel confidently beyond the measured sample, which parts remain conditional, and why a real near-term job dividend is not yet proof of structural upgrading.
A manufacturing shock does not benefit every sector or place equally
The direct production channel is manufacturing. The exposure measure is built from manufacturing products, and the paper finds significant manufacturing employment growth and the largest sectoral labor-income response in that sector. This is the most direct path from stronger U.S. demand to local jobs: firms connected to the affected export lines have reason to produce more, and production requires additional labor or longer use of existing capacity.
Services receive a positive spillover. A growing manufacturing workforce can raise demand for transport, repair, food, retail, housing, and other local services. Some service firms may also support production directly, while others benefit from workers spending more of their income nearby. The service response is therefore connected to the manufacturing shock, but it is not a second export shock with the same mechanism. The study reports a smaller service-sector response and a sectoral reduction in informality that is driven primarily by services.
Agriculture shows no comparable benefit from the manufacturing tariff exposure. That contrast is informative because it keeps “local” from becoming synonymous with “everything nearby.” A municipality can contain manufacturing, services, and agriculture, but they do not all face the same demand channel. Services may be pulled into an export expansion through local spending and business linkages. Agriculture may be less connected to the particular manufacturing products whose Chinese competitors became more expensive in the United States.
The sector pattern also limits what can be inferred from proximity. Being in a municipality with an exporting factory does not guarantee that every worker or firm receives an opportunity. A spillover depends on the local connections through which additional demand travels. The evidence supports a positive service response and no comparable agricultural response in the reported specifications; it does not provide a complete input-output map of every Mexican municipality or identify each firm that gains or loses.
The paper tests the difference between labor absorption and productive-capacity deepening with Economic Census outcomes. More-exposed municipalities do not show significant changes in firms' employment, value added, or fixed assets in the reported specifications. That does not cancel the employment findings from the labor data. It means that, over the available comparison, the clearest response appears in labor-market adjustment rather than in measured expansion of firm-level productive capacity.
This distinction is central to the development question.