Abstract Predicting share-of-wallet and size-of-wallet (i.e., category expenditure) of customers requires a firm to have, in addition to its own sales, an estimate of customer spending at competing firms. Given data on competitive spending from a sample of customers, this study considers the problem of predicting consumer expenditures at competing firms when data is unavailable. The proposed methodology, designed for multi-category firms, is a simultaneous equations Tobit model with latent classes which can handle three complicating factors: (i) heterogeneity in spending patterns; (ii) interrelationship of expenditures across firms and categories, called simultaneity; and (iii) data censoring, which occurs when consumers have zero expenditure in a category. The model is estimated on credit card data using Bayesian estimation. Two segments are revealed. One comprises 76% of consumers and is characterized by habitual spending patterns. The other 24% of consumers spend based on income allocation. Segments show different interrelationship of expenditures across firms and categories. Thus, this paper contributes a methodological tool for more accurate prediction of size-of-wallet and share-of-wallet and better targeting of customers when expenditures at competing firms are unavailable. This is accomplished by considering consumer heterogeneity and asymmetric complementary and substitute interactions between expenditures.
© 2001-2024 Fundación Dialnet · Todos los derechos reservados