Reproduce “Learning Treatment Effects during Resource Allocation via Priority-Queue Randomization” on OpenResearch
Turn your favorite coding agents into research agents that review literature, develop hypotheses, and run experiments.
Universal app for Apple silicon and Intel. Requires macOS 11 or later.
About this paper
JungHo Lee, Johnna Sundberg, Pim Welle, Bryan Wilder
This work introduces a queue-randomizing design framework for evaluating public service programs that allocate scarce resources via priority queues. It enables rigorous causal inference, including identification of Local Average Treatment Effects with endogenous arrivals, while simultaneously optimizing experimental designs to balance statistical efficiency with programmatic prioritization goals.
Key insights
- Queue randomization enables causal identification: ATE is identifiable with exogenous arrivals, while Local Average Treatment Effects (LATEs) are identified via instrumental variables with endogenous arrivals.
- Despite complex dependencies from dynamic queuing, standard estimators achieve asymptotic variances matching iid semiparametric efficiency bounds, simplifying design objectives.
- The asymptotic treatment propensity is an affine function of queue assignment probabilities, allowing for tractable multi-objective optimization to balance statistical efficiency and prioritization utility.
Summary by alphaXiv.
Tip: change arxiv.org to autoarxiv.org in any arXiv abstract link to open the paper here.