Estimating Treatment Heterogeneity of International Monetary Fund Programs on Child Poverty with Generalized Random Forest

Abstract

<p>A flourishing group of scholars of family sociology study how macroeconomic shockwaves propagate via households dynamics and landing a blow on children’s living conditions; simultaneously, scholars of political economy unravel impacts of such shockwaves on population outcomes. Since these two strands of literature have evolved independently, little is know about the relative importance of societal and family features moderating this impact on children’s material living conditions. In this article, we synthesize insights from these two strands by examining the effect of economic austerity following International Monetary Fund programs—a type of economic shock—on child poverty across a sample representative of about half the world’s population of mainly the Global South. This article addresses the following fundamental sociological questions: to what extent do the pathways of economic austerity propagate through families’ living conditions and societies’ structural and political characteristics. To capture these multiple non-linear heterogeneous relationships between macro and micro traits, we deploy machine learning in the service of policy evaluation. First, our analysis identifies an adverse average treatment effect (ATE) following the implementation of IMF programs on children’s probability of falling into poverty: 0.14, 95% CI 0.03- 0.24. Second, our algorithms identify substantial impact heterogeneity distributed about this ATE. Macro constellation moderate about half of the impact variation on children, and families’ capabilities moderate the other half of this variation. We named this finding the 50-50 impact-moderation rule of thumb. Our algorithm identified family wealth closely followed by governments’ education spending as the critical moderating factors. IMF program affects children residing in the middle of the social stratification more than compared to their peers residing in both the top and bottom of this stratification; for those children residing in societies that have selected into IMF programs and have historically spent most on education, are at a higher risk of falling into poverty. These findings identify the value of combining family sociology and political economy perspectives. Scholars will likely cross-fertilize this research further by testing this 50-50 rule of thumb to other types of economic shocks.</p>

Leveraging Machine Learning to Estimate Effect Modification

“…They optimize the accuracy of individual-level predictions and directly output predictions at the individual level as opposed to the aggregate level of substantive interest. The causal forests, a variant of Generalized Random Forests (GRF), (Athey, Tibshirani, and Wager 2019; Wager 2019) is a prominent example in this category and has been applied in multiple social science settings (Brand et al 2021;Daoud and Johansson 2019;Knittel and Stolper 2019;Tiffin 2019). Other examples include the R-learner (Nie and Wager 2020), the X-learner (Künzel et al 2019), and the Modified Covariate Method (Chen et al 2017;Tian et al 2014).…”

Statistical Modeling: The Three Cultures

“…While average treatment effect analysis focuses on the aggregated effect of an exposure on a population, effect heterogeneity analysis focuses on more granular effects, the group-specific effects disaggregated by subpopulations Shiba, Daoud, Hikichi, et al, 2022;Shiba, Daoud, Kino, et al, 2022). For example, although a famine or an economic crisis is likely to affect an entire country adversely, some combination of socioeconomic factors may protect certain groups better than others (Daoud & Johansson, 2020). In DMC, scholars tend to capture such effect heterogeneity with the help of interaction models.…”

Causal Effects of Landing Parameters on Runway Occupancy Time using Causal Machine Learning Models

“…This model addresses decision-making problems by quantifying the causal effects of treatment or intervention. The framework has enabled numerous applications ranging from quantifying the effectiveness of treatment in the medical settings [29] to policy evaluation (i.e., monetary fund programs on child poverty [30]).…”