Instrumental Variables and the Search for Identification

Causal Question / Estimand

The causal effect of an endogenous regressor on an outcome — and, with heterogeneous effects, the LATE for compliers — recovered by an instrument that is as-good-as- randomly assigned and affects the outcome only through the treatment.

Identification Strategy

A conceptual history and guide to IV. Traces IV from its origins in identifying simultaneous supply-and-demand systems to its modern use exploiting natural experiments — sources of variation (institutional rules, weather, birth timing) that are plausibly independent of the outcome’s unobservables. The defining move is to find an instrument satisfying independence and the Exclusion-Restriction rather than imposing a full structural model. Stresses that the quality of the instrument (its exogeneity story), not statistical machinery, is what makes an IV estimate credible.

Key Assumptions

Instrument independence (Randomization-like as-good-as-random assignment), Exclusion-Restriction, a first stage / relevance (see Weak-Instruments), and — for a causal LATE interpretation — Monotonicity. SUTVA.

Threats to Validity

Instruments that fail exclusion (affect the outcome through other channels); weak first stages that bias estimates toward OLS and distort inference; instruments whose exogeneity is asserted rather than argued.

Setting / Data

n/a — methodological/expository, with classic illustrations (supply–demand, quarter-of-birth and schooling, etc.).

Key Claims

  • IV’s modern value is in exploiting credible natural experiments, not in mechanically solving simultaneity.
  • The exclusion restriction is the binding, untestable assumption; a good instrument needs a substantive exogeneity argument.
  • Weak instruments are a real hazard; report and scrutinize the first stage.

Connections

Citation

Angrist, J. D., & Krueger, A. B. (2001). Instrumental Variables and the Search for Identification: From Supply and Demand to Natural Experiments. Journal of Economic Perspectives, 15(4), 69–85.