🔎 Why Profiling Compliers Matters

Instrumental-variable (IV) estimation is essential for researchers handling randomized control trials with noncompliance or exploiting partially exogenous treatment variation in observational studies. The potential outcomes framework underpins identification of the local average treatment effect (LATE) and partitions the sample into compliers, always-takers, and never-takers. Despite this, applied work has paid little attention to the characteristics of those groups—information that is crucial for understanding which subpopulation the IV estimate targets and for assessing the external validity of the estimated LATE.

📚 What This Letter Does

• Discusses the assumptions required to profile compliers and noncompliers, noting these assumptions are weaker than those needed to identify the LATE when the instrument is randomly assigned.
• Introduces a simple, general method to characterize compliers, always-takers, and never-takers in terms of their observed covariates.
• Supplies easy-to-use software implementing the estimator in R and Stata.

🧭 How the Profiling Method Operates

• Relies on the potential outcomes framework and observed covariates to stratify the sample into compliance types.
• Requires assumptions that are explicitly weaker than full LATE-identifying assumptions under random assignment of the instrument, making profiling feasible in a broader set of applied settings.
• Produces descriptive profiles of each compliance group that clarify which subpopulation the IV estimates apply to.

🛠️ Software and Practical Use

• Ready-to-run code in R and Stata implements the proposed estimator for profiling compliers and noncompliers.
• Designed for straightforward integration into standard IV workflows so profiling can accompany any IV analysis.

📌 Why It Matters

Profiling compliers and noncompliers is a necessary first step for interpreting IV estimates: it makes explicit the subpopulation targeted by LATE estimates and aids assessment of external validity. The proposed method and accompanying software aim to make such profiling a standard practice in applied IV work.