Heckman correction

A two-step or full-likelihood econometric method that models sample selection and adds the implied inverse Mills ratio to correct outcome estimates under joint-normality assumptions.

Version

v2 · 2026-09-19 · History

Domain-specific #

4843

Domain group

Social Sciences

Origin domain

Economics & Finance

Subdomains

Econometrics, Sample Selection Models → Economics & Finance

Core Idea

The Heckman correction estimates an outcome relationship when observation of that outcome is nonrandom and correlated with unobserved outcome determinants. A first-stage probit estimates selection probability; the conditional expectation of the outcome error yields an inverse-Mills control term included in the outcome regression, or both equations are estimated jointly. The abstraction is therefore identified by a declared carrier, a transformation or constraint over that carrier, and an invariant that tells an analyst whether the named structure is genuinely present.

How would you explain it like I'm…

Who-Showed-Up Fixer

Imagine trying to learn how much people earn, but you can only ask the people who have jobs. The people without jobs might be different in ways you can't see. That would make your answer wrong. The Heckman correction is a trick that first figures out who ends up being seen, and then uses that to fix the answer.

The Two-Step Missing-People Fix

Sometimes we only get to see results for some people, and the reason they're seen is linked to hidden things that also affect the result. For example, we only see wages for people who have jobs, and the people who chose to work might be different in ways we can't see. The Heckman correction works in two steps. First it guesses each person's chance of being seen. Then it uses that guess to add a fix-up number when studying the result, so the answer is less biased.

Two-Step Selection Bias Correction

The Heckman correction deals with selection bias: the outcome is observed only for a nonrandom group, and the selection is correlated with unobserved factors that affect the outcome. Step one is a probit model that predicts the probability each case is observed. From it, you compute the inverse Mills ratio, a term capturing the expected hidden error given that a case was selected. Step two adds this term as an extra variable in the outcome regression. Alternatively, both equations can be estimated together. Unlike simply dropping missing cases, this method explicitly models why data are missing, but it relies on stated assumptions about the error distributions and how selection works.

 

The Heckman correction estimates an outcome equation when the outcome is observed only under a selection rule whose unobservables correlate with the outcome error. The model has two equations. A selection equation determines observability, and an outcome equation is observed only when selection occurs. Under the standard joint-normal error assumption, a first-stage probit estimates the selection probability. The conditional expectation of the outcome error given selection then equals a coefficient times the inverse Mills ratio. Including that ratio as a control function in the outcome regression removes the selection bias. Alternatively, both equations can be estimated jointly by maximum likelihood. Validity depends on stating the selection and outcome equations, the functional-form assumptions, and the identification strategy, which is typically an exclusion restriction affecting selection but not the outcome. Without these, identification rests only on nonlinearity. It is a parametric control-function method, distinct from generic missing-data imputation.

Scope of Application

Heckman correction belongs to econometrics and is useful where the analyst can specify a latent selection equation, observed selection indicator, outcome equation observed conditionally, covariates and exclusion restrictions, correlated disturbances, probit estimate, inverse Mills ratio and corrected coefficients, then evaluate selection and outcome equations, functional-form assumptions and identification strategy are stated and the outcome is observed only under the modeled selection rule. The scope is broad within that domain but bounded by the need for selection and outcome equations, functional-form assumptions and identification strategy are stated and the outcome is observed only under the modeled selection rule. The entry records a descriptive analytical identity; practical use requires the governing domain's evidence, standards, and safety obligations.

Clarity

The abstraction clarifies a crowded vocabulary by making selection and outcome equations, functional-form assumptions and identification strategy are stated and the outcome is observed only under the modeled selection rule the center of the account. A claim should name the carrier, the governing operation or relation, the applicable assumptions, and the recognition test. A bare label is insufficient because the name Heckman correction can be used for a formal identity, an implementation, or a neighboring result unless carrier and convention are stated.

Manages Complexity

Without the abstraction, an analyst must reason directly over many local details: the carrier roles, admissibility assumptions, competing conventions, derived invariants, boundary cases, and proof or validation obligations specific to Heckman correction. Heckman correction compresses them into the roles in the structural signature. That compression permits comparison across instances without erasing the variables that determine validity. It also exposes which details may be varied safely and which are constitutive.

Abstract Reasoning

1. Identify the carrier. State what the elements, states, objects, or observations are: a latent selection equation, observed selection indicator, outcome equation observed conditionally, covariates and exclusion restrictions, correlated disturbances, probit estimate, inverse Mills ratio and corrected coefficients. Reject examples whose alleged carrier belongs to a different problem.
2. Lock the constitutive rule. Express selection and outcome equations, functional-form assumptions and identification strategy are stated and the outcome is observed only under the modeled selection rule independently of one notation or implementation.

Knowledge Transfer

Knowledge transfers strongly among subfields of econometrics because they reuse a latent selection equation, observed selection indicator, outcome equation observed conditionally, covariates and exclusion restrictions, correlated disturbances, probit estimate, inverse Mills ratio and corrected coefficients, A first-stage probit estimates selection probability; the conditional expectation of the outcome error yields an inverse-Mills control term included in the outcome regression, or both equations are estimated jointly., and type the carrier, state every parameter and convention in the definition, test that selection and outcome equations, functional-form assumptions and identification strategy are stated and the outcome is observed only under the modeled selection rule, compare the nearest accepted identity, and report counterexamples, uncertainty, and limiting cases.

Relationships to Other Abstractions

Current abstraction Heckman correction Domain-specific

Parents (1) — outgoing typed relations

• Heckman correction is a kind of Bias Prime

  The method addresses bias induced by nonrandom sample selection.

Hierarchy path (1) — routes to 1 parentless root

• Heckman correction → is a kind of Bias

Neighborhood in Abstraction Space

Heckman correction sits in a moderately populated region (44^th percentile for distinctiveness): it has near-neighbors but no dense thicket of look-alikes.

Family — Dispersion Measures & Multilevel Regression (11 abstractions)

Nearest neighbors

• Regression analysis — 0.88
• Polytomous choice — 0.88
• Thurstonian model — 0.86
• Heteroskedasticity-consistent standard errors — 0.86
• Pareto index — 0.86

Computed from structural-signature embeddings · 2026-10-25