Experiments with Longitudinal Elements

Chapter 9 — Experiments with Longitudinal Elements Many economic questions involve outcomes that unfold over time—requiring experimental designs that account for treatment duration, timing, and repeated exposure. This chapter develops the potential outcomes framework for longitudinal settings, covering staggered experimental designs, leveraging pre-treatment outcomes to increase statistical power, and panel data estimation strategies. It also addresses the challenge of outcomes that can only be observed long after treatment ends, introducing statistical surrogates as a solution when direct measurement is infeasible.

1. SUTVA, perfect compliance, and observability assumptions must hold for longitudinal designs to recover valid causal effects.
2. Outcomes can depend on treatment duration and timing, requiring careful design of repeated exposure experiments.
3. Pre-treatment outcomes can be leveraged to improve precision and statistical power.
4. When key outcomes are measured long after treatment ends, statistical surrogates can be employed, given the assumptions of comparability and surrogacy are met.