Empirical Methods: A Brief Introduction to Causal Inference

This chapter provides an intuitive yet rigorous introduction to empirical methods for causal inference, with a focus on evaluating education policies. It is designed for readers without a formal background in econometrics, equipping them with the tools to critically assess causal claims and understand the identification strategies employed in later chapters. The chapter highlights the distinction between correlation and causation, emphasizing the need for well-designed empirical strategies to estimate policy effects accurately. Key methods such as randomized controlled trials (RCTs), difference-in-differences (DiD), instrumental variables (IV), and regression discontinuity designs (RDD) are introduced, with discussions on their assumptions, strengths, and limitations. A central finding is that while RCTs remain the gold standard for causal inference, practical and ethical constraints often necessitate alternative approaches. Furthermore, often the effect of interest is only well defined in the context of a structural economic model and it is different than the reduced-form causal effect RCT and some of the other methods can estimate. The chapter underscores the risks of selection bias and endogeneity and demonstrates how different econometric techniques can mitigate these challenges. It also stresses the importance of external validity, arguing that while controlled experiments provide robust internal validity, their generalizability to broader populations and contexts is often limited. The role of heterogeneity in causal effects on external validity, on the definition of the magnitudes of interest, and on their estimation is discussed. By providing a conceptual framework for causal inference in education economics, this chapter enables readers to critically evaluate empirical findings and assess the robustness of policy recommendations.