The renaissance of field experimentation in evaluating interventions

Area of Science:

• Social Sciences
• Economics
• Political Science

Background:

• Field experiments have a four-decade history with notable successes and failures.
• Their application is growing for answering causal questions in real-world settings.
• Both randomized and nonrandomized experimental designs are increasingly utilized.

Purpose of the Study:

• To review the history and current renaissance of field experimentation.
• To discuss technical developments enabling accurate effect estimation in field studies.
• To cover empirical evidence on the validity of nonrandomized experiments.

Main Methods:

• Review of historical use and technical advancements in field experiments.
• Focus on four key designs: randomized experiments, regression discontinuity, interrupted time series, and nonrandomized comparison groups.
• Examination of methods for handling partial treatment, attrition, nested designs, and propensity score analysis.

Main Results:

• Technical developments have overcome previous limitations in accurate effect estimation.
• New analytic methods improve analysis for regression discontinuity and interrupted time series designs.
• Empirical evidence suggests nonrandomized experiments can approximate randomized experiment results under certain conditions.

Conclusions:

• Field experiments are a powerful tool for causal inference, enhanced by recent methodological progress.
• Advances in statistical analysis and design address challenges inherent in real-world research.
• The use of sophisticated nonrandomized designs offers a viable alternative to randomized trials in many contexts.