Measuring Financial Wellbeing with Self-Reported and Bank-Record Data

Abstract

This study develops multi-item scales of the financial wellbeing of customers of a major Australian bank using self-reported survey data that are matched with the customers' financial records. Using Item Response Theory (IRT) models, the study develops: First a Reported Financial Wellbeing Scale that is formed from responses to 10 questions about people's experiences and perceptions of financial outcomes, and second an Observed Financial Wellbeing Scale that is formed from five financial-record measures of customers' account balances, net spending, and payment problems. The IRT models show that each scale reliably differentiates between a wide range of outcomes and that the components within each scale have similar power to discriminate. We validate the scales by estimating Least Absolute Shrinkage and Selection Operator machine-learning models of how they correlate with other measurable characteristics. Savings habits, spending habits, credit card behavior, household income, education, difficulties with housing payments, and the use of and access to social or government support are each associated with both types of financial wellbeing.