The Saturated Bayesian - Break Detection in Panel Data with Short Time Horizons

Abstract: Effectively tackling contemporary challenges posed by climate change and the continued degradation of natural habitats requires swift and decisive actions. Identifying the most effective policies (or a mix thereof) is crucial to inform policy-makers that are often constrained in their choice set. Traditional methods for policy evaluation rely on precise knowledge about the occurrence and timing of interventions. Structural break identification on the other hand has a long tradition in the field of econometrics. Recent approaches cast the search for such breaks in the form of indicator-saturated regressions, identifying step-shifts in relevant time series, but lack a proper framework of uncertainty quantification. We introduce a coherent probabilistic framework for the detection of structural breaks with unknown timing in panel data. The proposed Bayesian setup naturally incorporates the quantification of break uncertainty with little overhead. Simulation studies demonstrate that our approach is competitive to existing approaches in detecting true positives and drastically reduces false positives. We apply our method to replicate studies on the effectiveness of climate policies in the European transport sector and provide novel insights in the dynamics of deforestation in the tropics.

Presented at:

• What Works Climate Solutions (WWCS) 2024
• International Society for Bayesian Analysis (ISBA) World Meeting 2024
• European Seminar on Bayesian Econometrics (ESOBE) 2024

Status: Work in progress

Co-authors: Lukas Vashold, Jesús Crespo Cuaresma

Keywords: Panel Data, Structural Breaks, Bayesian Statistics, Indicator Saturation