The missing intercept problem

Benjamin Moll and Oliver Hanney explain a critical flaw in how microeconomic estimates get scaled to the macro level.

Cross-sectional research designs (DiD, RDD, shift-share) identify relative effects — how outcomes change in high-exposure places compared to low-exposure places. Researchers then routinely multiply that coefficient by a national change and claim an aggregate result. The canonical example: Autor, Dorn and Hanson’s claim that “import competition explains one-quarter of the aggregate decline in US manufacturing employment.”

The problem: aggregate spillovers (trade, labour mobility, demand linkages) get silently absorbed into the regression intercept because there’s no cross-sectional variation in the aggregate. The estimate captures β, not β + γ. The line shifts, not just the point along it.

Rimsha Arif’s LinkedIn commentary adds a nice echo — she recognizes the pattern from her own work building district-level elasticities and projecting national outcomes from them, and flags that the headline numbers that travel furthest are often built exactly this way.

Moll’s bottom line: if you care about macro effects, combine causal micro estimates with a general equilibrium model instead. The bar for the facts that travel furthest should be as high as the bar for the underlying identification.

The “missing intercept” problem with going from micro to macro