Checkpoint: From an Economic Story to a Testable Claim

The relationship between minimum wages and robot adoption now has two necessary pieces. A higher wage floor can change the relative cost of routine work, and the Census study observes one kind of adoption when a manufacturing firm imports an industrial robot. The judgment this checkpoint supports is whether those pieces form a testable claim: what is directly seen, what is inferred through the economic mechanism, and what remains outside the evidence.

One claim, with three different layers

The direct observation is concrete. A firm in the study is recorded as adopting a robot when it imports an identified industrial-robot product in a given year. The business records locate that firm, describe its industry and employment, and connect it to the state minimum wage that applies to its location. The data can therefore support a comparison of adoption events across firms, states, and years.

The economic interpretation is one step removed. A higher minimum wage raises the effective cost of some labor, especially where production depends on routine work near the wage floor. That changed cost can make a robot-enabled process more attractive relative to the existing arrangement. Whether the firm adopts still depends on technical fit, demand, capital, integration, financing, and capability. The mechanism makes adoption plausible; it does not appear as a field in the business record.

The unmeasured outcomes are farther away still. The import event does not reveal which task changes, how many workers are affected, whether output expands, whether productivity rises, or whether workers move into new occupations. It does not resolve whether the firm would have adopted without the wage change. Those are not minor details of the same variable. They are different questions requiring different observations.

Keeping the layers separate prevents two opposite mistakes. One mistake is to treat the import record as too narrow to tell us anything about the policy relationship. It is a useful, repeatable outcome that can be linked to local policy and firm characteristics. The other mistake is to treat it as a complete measure of automation and its consequences. A recorded adoption is evidence that a particular event occurred under a defined measurement rule, not a summary of the firm's entire production system.

flowchart LR A[Minimum-wage policy] --> B[Relative cost of routine task] B --> C[Firm choice under constraints] C --> D[Import-based adoption event] D --> E[Measured association] B -. mechanism inferred .-> C D -. does not observe .-> F[Worker and productivity outcomes] E -. cannot alone establish .-> G[Universal causal effect]

The arrows deliberately use different meanings. The policy-to-cost link and the firm decision describe the economic channel. The import event is the measured outcome. The association is an empirical result built from comparisons. The worker and productivity outcomes sit beyond the endpoint of this particular measure.

Why evidence must widen and narrow

A policy relationship can look different at different scales.