Abstract: This paper proposes a new method for estimating average treatment effects of a binary policy in staggered adoption settings. The method combines modern causal inference methods for policy evaluation with theoretical insights from the large panel data literature. Using theory and data-driven simulations, we show that this estimator improves over the available alternatives in settings with persistent errors. We recommend that applied researchers use the new method for policy evaluation in settings with many units and time periods.
Abstract: We propose a new estimator, Sequential Synthetic Difference in Difference (Sequential SDiD), for estimating treatment effects in staggered treatment rollout settings. Our method builds on the original SDiD estimator by sequentially applying it to aggregated data. We demonstrate the estimator’s asymptotic equivalence to an infeasible OLS benchmark, ensuring asymptotic normality and efficiency. This approach offers a robust alternative to traditional Difference-in-Differences methods for staggered adoption designs, and we demonstrate its advantages using empirical applications and data-driven simulations.
Abstract: This paper assesses a pilot project in the Canary Islands designed to reduce the digital divide and enhance employability among disadvantaged individuals aged 45-64 with low educational attainment. Participants were randomly assigned to one of three groups: those receiving tablets with internet access, tablets with accompanying digital skills training, or serving as controls. The group that received both tablets and training demonstrated notable improvements in digital skills and job search capabilities, though no significant impact on employment status was observed. The findings indicate that offering both tablet access and digital skills training can help tackle socio-economic challenges.