Assignment to Treatment Group on the Basis of a Covariate

Causal Question / Estimand

The average effect $\tau$ on outcome $Y$ of Treatment 1 versus Treatment 2 over the relevant population, when assignment is governed entirely by an observed covariate $X$. See Causal-Estimand.

Identification Strategy

When the probability of treatment is a known function of $X$ alone — assignment is ignorable given $X$ — the effect is identified by estimating the conditional expectations $E[Y\mid X]$ in each group and averaging their difference over the distribution of $X$. This is the canonical statement of selection-on-observables (Ignorability) identification, and the conceptual seed of propensity-score methods (assignment as a function of covariates).

Key Assumptions

Ignorability (assignment depends only on the recorded $X$), Overlap (“substantial overlap in the distribution of $X$” across groups, or strong prior information), Potential-Outcomes.

Threats to Validity

Sensitivity to the model for the conditional expectations: if the regressions are non-parallel or nonlinear, simple covariance adjustment is biased — Rubin urges searching for nonparallelism/nonlinearity. Coarse blocking degrades estimates. Identification fails if any assignment-relevant variable beyond $X$ is omitted.

Setting / Data

n/a — methodological. Running example: assigning children to a compensatory reading program by a reading pretest score $X$.

Key Claims

• If assignment is based solely on $X$, conditioning on $X$ removes confounding — no need to worry about unreliability of $X$ or unmeasured background variables.
• The correct estimator averages the difference in estimated $E[Y\mid X]$ over the population; gain scores and reliability-adjusted scores are generally wrong.
• Under parallel linear regressions, this reduces to the simple covariance-adjusted estimator.
• Overlap in $X$ (or strong priors) is required for the techniques to be useful.

Connections

• Builds on: Rubin1974-EstimatingCausalEffects
• See also: Rubin1978-BayesianCausalEffects, Holland1986-StatisticsAndCausalInference, Foundations

Citation

Rubin, D. B. (1977). Assignment to Treatment Group on the Basis of a Covariate. Journal of Educational Statistics, 2(1), 1–26.