Definition
Structural econometrics estimates the parameters of an explicit economic model — preferences, technology, constraints, and the optimizing behavior they imply — so that the estimated parameters are interpretable and can be used for counterfactual and policy analysis. It is usually contrasted with the experimentalist (or "atheoretic"/reduced-form) approach, which puts the identification of causal effects from natural experiments and instrumental variables front and center and deliberately minimizes reliance on economic theory. Keane (2010) argues the contrast is a false dichotomy: both rely heavily on a priori assumptions, differing mainly in whether those assumptions are made explicit.
Key Ideas
- Explicit vs. implicit assumptions. The structural approach writes its assumptions into a model; the experimentalist approach also makes strong assumptions (instrument exogeneity, exclusion restrictions, external validity) but leaves them implicit. The difference is transparency, not quantity.
- The experimentalist program. Identify causal effects from natural experiments — exogenous variation from "events or situations" — using instrumental variables, without specifying a behavioral model (Angrist-Krueger 1999). Prized for credible identification of a specific causal contrast.
- What IV actually estimates. Absent a model, an instrument identifies a local average treatment effect (LATE) — the effect for the subpopulation whose behavior the instrument shifts (the compliers) — not a structural parameter. Different instruments identify different LATEs.
- The counterfactual/extrapolation gap. Policy questions typically ask about interventions not yet tried, or effects outside the range the natural experiment covers. Extrapolating a LATE to such settings is itself an (implicit) assumption; a structural model makes the extrapolation explicit and disciplined by theory.
- Interpretability as the goal. Making assumptions explicit is what renders estimates interpretable and cumulative — "a prerequisite for scientific progress" — even though explicit assumptions can, of course, be wrong.
Why It Matters
- Frames the credibility-revolution debate. The structural-vs-experimentalist argument is one of the defining methodological disputes of modern applied microeconometrics; Keane's "explicit vs. implicit" reframing clarifies what is actually at stake.
- Guides method choice by question. For estimating a single well-identified causal effect in a studied population, natural-experiment/IV methods are compelling; for counterfactual policy analysis and extrapolation, structural models are often unavoidable.
- Connects to identification broadly. The same theme — that identification always rests on assumptions that should be stated and, ideally, have their uncertainty carried into the answer — underlies the Bayesian treatment of SVAR identification and the interpretation of external instruments.
Open Questions
- Transparency ≠ correctness. Explicit structural assumptions can be wrong in ways that are hard to check; the credibility critique is that this is a real cost, not just an implicit-assumptions problem.
- How much structure is enough? The spectrum from fully specified dynamic models to lightly parameterized "semi-structural" designs is contested, and the right point depends on the question and the data.
- Combining approaches. Using natural-experiment variation to identify or validate parts of a structural model (and carrying identification uncertainty through) is an active reconciliation of the two camps.
Related