A Structural Model of Mental Accounting

We construct and estimate a formal economic model of mental accounting where utility maximizing consumers facing rationality constraints infrequently update their desired expenditure budgets for different categories of consumption. We use latent Bayesian inference to estimate the model on a unique dataset of consumer-level weekly consumption expenditure in order to infer mental accounting decision variables. Our formulation reconciles the variability of observed high-frequency consumption expenditure patterns. We find that consumers optimally re-evaluate only about 25% to 50% of their expenditure budgets each period, providing empirical evidence that consumers regulate their expenditure behavior with mental accounting heuristics. Compared to two separate models where consumers face no cognition frictions or engage in full, sticky mental accounting, the model with both consumer- and time-dependent heterogeneous rationality constraints fits the data best. Counterfactually, we find that implementing policies which encourage consumers to stick with weekly expenditure budgets leads to, on average, higher savings rates and welfare improvements.