This chapter details the two-stage semi-parametric Data Envelopment Analysis (DEA) approach, designed to identify the determinants of production efficiency. It begins by outlining the shortcomings of traditional second-stage methods, such as Tobit regression, which produce inconsistent estimates due to the unique statistical properties of DEA scores, including serial correlation and endogeneity. The core of the chapter is a comprehensive explanation of the Simar-Wilson (2007) method, which employs truncated regression and advanced bootstrap algorithms to correct for estimation bias and facilitate valid statistical inference. Furthermore, the chapter delves into the scholarly debate between the Simar-Wilson model and the alternative OLS-based framework proposed by Banker and Natarajan (2008), emphasizing the crucial role of underlying assumptions. Finally, it presents a practical extension by integrating the bootstrap-truncated methodology with a Difference-in-Differences (DID) model, creating a robust tool for assessing the causal impact of policy interventions on efficiency.