| A consumer's mood, opinions, and general disposition can influence her demand for a product. While an empirical specification that models the correlation between attitudes and purchasing decisions will allow for a more accurate prediction of demand, there is disagreement among researchers over the role of attitudinal responses in estimation. This paper examines the relationship between the consumer's responses to questions about attitudes and demand for goods in order to demonstrate the most appropriate use of attitudinal data in empirical work. A theoretical model of consumer behavior identifies several latent factors that simultaneously influence the individual's purchasing decisions and responses to attitudinal questions. This correlation, driven by unobserved heterogeneity such as the individual's mood, opinions, discount factor, expectations over future outcomes, and unreported experiences, causes a simultaneity bias in specifications that include attitudinal responses as explanatory variables. Instead, I jointly estimate purchases and attitudinal responses using a random effects model to accurately capture the relationship between both observed outcomes. An econometric proof, a Monte Carlo experiment, and a data application show that, compared to commonly-used specifications, this jointly-estimated model improves the accuracy and efficiency of the estimated response parameters of the covariates that explain consumer demand. |
