Abstract

This presentation will share the findings of an exploratory study that used machine learning to evaluate the effectiveness of labour market programs from a Gender-Based Analysis Plus (GBA Plus) perspective. Our study uses the Modified Causal Forests (MCF) algorithm on rich administrative integrated datasets, enabling a granular examination of program impacts across intersecting identity factors (e.g., gender and visible minority status).

Our study reveals that machine learning can help assess the distribution of program impacts across participants and show how they may differ according to different sociodemographic characteristics. Results suggest that the program participants, on average, improved their labour market outcomes, and that some subgroups benefit relatively more than others for some specific interventions.

Our study underscores the potential of AI in evaluation practice while emphasizing the importance of high-quality data. By integrating new computational methods with inclusive evaluation methodologies, we can move the needle towards a more nuanced understanding of “what works for whom”, and better inform policy and program from a GBA Plus perspective.