Information, Cash Or In-Kind? Evidence From A Large Field Experiment To Encourage Debit Card Use

We conduct a large field experiment in Mexico to understand the relative effectiveness of cash and in-kind incentives, and informational interventions, designed to encourage debit card use. Larger treatment effects are found with increased monetary value of the incentive whether it is cash or in-kind, and the longer the incentive is active for. However, higher value incentives do not necessarily come with better cost-benefit ratios. The best performing informational treatment performs equally as well as the 300 MXN cashback for amount of transactions and is not statistically different from all except the largest cash incentive for number of transactions. The exact wording of the informational message matters as we find no effects for alternative informational treatments. The modest average treatment effects mask significant heterogeneity in treatment effects, suggesting targeted marketing campaigns will be more cost effective. Using machine learning methods for causal inference, we find that the top quartile exhibits higher positive treatment effects, while the other quartiles have treatment effects that are either not statistically different from zero or negative. Finally, for customers in the top quartile of predicted treatment effects we find no evidence of either an increase or a decrease in total spending.