Instrument Assignment Based on LATE Results with Defiers

Abstract

While most previous studies have focused on treatment assignment, this thesis emphasises instrument assignment to estimate Local Average Treatment Effects (LATE). We introduce a refined methodology to address limitations in the traditional assumptions set by Imbens and Angrist (1994), specifically by relaxing the monotonicity assumption to account for defiers. This work extends the contributions of Liao (2022) by integrating continuous control variables, namely KWW(Knowledge of the World Work) and IQ scores, into the LATE framework. We empirically validate our approach using college proximity as an instrumental variable. This study examines the impact of college attendance on wage outcomes, offering new insights into how treatment effects vary among individuals with different cognitive and job-related skills. The results conclude that individuals with lower KWW and IQ scores experience the highest wage gains from college attendance, highlighting the potential of targeted educational interventions. By accommodating defiers, this approach improves the robustness and applicability of LATE estimation across diverse empirical settings.