Mexico | Cash vs. Card Consumption Patterns: A Machine Learning Approach

We offer a novel methodology combining high frequency card transaction data from BBVA and point-of-sale data from cash operations registered at convenience stores from Frogtek by Clarity AI to study changes in consumption patterns relative to variations in income, including changes in items consumed and payment channel.

Key points

• Key points:
• The analysis of household consumption patterns is relevant for social welfare, policy design, and economic analysis. Recent advances in Big Data and data science techniques allow us to measure consumption trends in an extraordinarily granular way.
• Traditional empirical analyses of consumer behavior based on household consumer surveys provide an incomplete picture, which is why in this analysis a new method is used.
• BBVA and Clarity AI have joined forces to contribute knowledge to society by showing the analysis of how individuals allocate their card and cash purchases using a variety of econometric and Machine Learning Models.
• Shapley values are used to gain explainability in the Machine Learning models applied, finding Random Forest to be the champion model, which achieves R2 scores above 0.92.
• The results show that the most relevant variables to increase the expenditure by card relative to cash are the changes in income, living in an urban center, and the financial deepening effects.

Authors

Geographies

Topics

Tags