Essays in spatial econometrics and causal inference

This thesis explores the intersection of spatial econometrics and causal inference, two fundamental fields in economics that remain relatively disconnected in the literature. The primary objective is to contribute theoretically to both disciplines while bridging the methodological gap between them. The research is structured into three independent yet thematically related chapters. The first chapter, A Survey of Spatial Analysis Methods for Causal Inference, provides a comprehensive literature review of emerging methods that integrate spatial effects into causal inference. Given the increasing interest in causal analysis across empirical economics and the social sciences, the omission of spatial dependence and heterogeneity in conventional methods can lead to biased estimates. This chapter categorizes and examines alternative approaches that address these spatial challenges, presenting their theoretical underpinnings and empirical applications while identifying directions for future research. The second chapter, Spatial Synthetic Difference-in-Differences, introduces a novel estimator that extends the Synthetic Difference-in-Differences (SyDiD) method proposed by Arkhangelsky et al. (2021) to account for spatial spillover effects. By incorporating the Spatial Difference-in-Differences framework of Delgado and Florax (2015) into SyDiD, the estimator decomposes the Average Treatment Effect (ATE) into direct and indirect components, allowing for a more nuanced understanding of treatment effects in spatial settings. The proposed approach is benchmarked against existing estimators, demonstrating advantages in estimating indirect treatment effects while preserving the desirable properties of SyDiD for direct effect estimation. The third chapter, New Developments on the Spatial Econometric Model Specification Search, investigates the model selection process in spatial econometrics through Monte Carlo simulations. Despite the extensive availability of spatial econometric models, the identification of the most appropriate specification remains an open question. This chapter systematically evaluates two widely used strategies—Specific-to-General (STGE) and General-to-Specific (GETS)—assessing their performance under different data-generating processes. By leveraging recent computational advancements, the study contributes to the ongoing debate on optimal model selection, providing insights into the comparative effectiveness of these strategies and suggesting avenues for further research. While the chapters are conceptually linked, they are designed as standalone contributions that can be read independently. The thesis also emphasizes transparency and reproducibility, with all empirical analyses conducted in Python and the corresponding code made available for future research. Through these contributions, the thesis aims to advance methodological developments in spatial econometrics and causal inference, fostering a more integrated framework for addressing spatial dependencies in empirical economic analysis.