This repository implements a complete causal pipeline for studying how low-friction payment methods (credit/debit/ACH/OBBP) affect daily spending using the 2024 Atlanta Fed Diary of Consumer Payment Choice (DCPC) data.
Our focus is on recovering the causal effect of “friction” on spending using Double Machine Learning (DML), orthogonal DR-loss, and CATE estimation, while adapting the pipeline to the strong constraints of the dataset (only 4 diary days per person, non-random diary assignment, sparse demographic variation).