Abstract

Uplift modeling corresponds to an area of machine learning focused on capturing causal relationships on various observational and experimental data. Currently it has several applications, particularly in the marketing area, focused on customer segmentation and the establishment of advertising campaigns. This research proposes an novel economic uplift approach, using branching causal algorithms to estimate the individual treatment effect on real GDP growth and changes in expansionary economic policy on a quarterly basis from an OECD dataset. The developed framework reveals positive causal effects on economic growth driven by expansionary policies, generalized for all countries under study. In addition, lagged causal effects on these policies, exerted by the economic cycle, are captured. The results obtained not only show a performance similar to that of the literature, but also conform to the theoretical and actual macroeconomic behavior.