Currently, there is an important debate about the relative merits of models with discrete and continuous representations of consumer heterogeneity. In a recent JMR study, Andrews, Ansari, and Currim (2002; hereafter AAC) compared metric conjoint analysis models with discrete and continuous representations of heterogeneity and found no differences between the two models with respect to parameter recovery and prediction of ratings for holdout profiles. Models with continuous representations of heterogeneity fit the data better than models with discrete representations of heterogeneity. The goal of the current study is to compare the relative performance of logit choice models with discrete versus continuous representations of heterogeneity in terms of the accuracy of household-level parameters, fit, and forecasting accuracy. To accomplish this goal, the authors conduct an extensive simulation experiment with logit models in a scanner data context, using an experimental design based on AAC and other recent simulation studies. One of the main findings is that models with continuous and discrete representations of heterogeneity recover household-level parameter estimates and predict holdout choices about equally well except when the number of purchases per household is small, in which case the models with continuous representations perform very poorly. As in the AAC study, models with continuous representations of heterogeneity fit the data better.

Empirical studies in Marketing have typically characterized a household’s purchase incidence decision, i.e. the household’s decision of whether or not to buy a product on a given shopping visit, as being independent of the household’s purchase incidence decisions in other product categories. These decisions, however, tend to be related both because product categories serve as complements (e.g. bacon and eggs) or substitutes (e.g. colas and orange juices) in addressing the household’s consumption needs, and because product categories vie with each other in attracting the household’s limited shopping budget. Existing empirical studies have either ignored such inter-relationships altogether or have accounted for them in a limited way by modeling household purchases in pairs of complementary product categories. Given the recent availability of IRI market basket data, which tracks purchases of panelists in several product categories over time, and the new computational Bayesian methods developed in Albert and Chib (1993) and Chib and Greenberg (1998), estimating high-dimensional multi-category models is now possible. This paper exploits these developments to fit an appropriate panel data multivariate probit model to household-level contemporaneous purchases in twelve product categories, with the descriptive goal of isolating

The importance of pricing decisions for firms has fueled an extensive stream of research on price elasticities. In an influential meta-analytical study, Tellis (1988) summarized price elasticity research findings until 1986. Empirical generalizations on price elasticity, however, require modifications because of: a) changes in market characteristics, i.e. characteristics of brands, product categories, and economic conditions, and b) changes in the research methodology to assess price elasticities. Therefore, we present a meta-analysis of price elasticity with new empirical generalizations on its determinants. Across a set 1851 price elasticities based on 81 studies, the average price elasticity is −2.62. One of the most salient findings is that over the past four decades, sales elasticities have significantly increased in magnitude, whereas share and choice elasticities have remained fairly constant. Across all determinants studied, we find that accommodating price endogeneity has the strongest (magnitude-increasing) impact on price elasticities. A striking null result is that accounting for heterogeneity does not affect elasticities significantly. We also present an analysis that explains the difference between our findings and Tellis (1988), and indicate which new price elasticity studies are most desirable

We develop a statistical model of browsing behavior by predicting the number of web pages, in a particular category, that are viewed by a user in a single web session. The purpose of this analysis is to better understand web browsing behavior, and to help predict which sessions are likely to result in retail visits. A single record in our database consists of the number of web pages viewed by a user during a single session from each of the following categories: portals, services, entertainment, retail, auctions, adult, and others. This dataset can be characterized as multivariate count data, where many of the counts are zero. We consider the use of Poisson and discretized tobit models, and contrast both univariate and multivariate versions of these models. Additionally, as our dataset is characterized by a great deal of heterogeneity in usage across users and also a good deal of persistence in viewership, we propose a new multivariate tobit model with a mixture process whose multiple states are governed by an unobserved (hidden) Markov chain. We find that users move between sessions that are characterized by browsing behavior that is focused in specific categories

St. Louis, and Wharton for their helpful suggestions. The authors are especially grateful to John Lynch for his help in developing a behavioral underpinning for their model. David Bell, Pete Fader, and Bruce Hardie generously provided the data. The authors also thank three anonymous for their many suggestions. Wagner Kamakura provided numerous detailed and helpful suggestions.

It is well known that in package goods categories a temporary price cut of a brand leads to increase in the sales of that brand in the current period. Under the assumption of stable consumption rate, previous literature has identified that the sources for the increase in sales are brand switching, purchase acceleration, and increase in quantity. However, there are very few studies that have formally modeled the impact of price promotions on consumption when consumption rate in a category is not constant. In this paper we offer a methodology to decompose the effects of price promotions into brand switching, stockpiling and change in consumption and explicitly allow for consumer heterogeneity in brand preferences and consumption needs. A dynamic structural model of a household that decides when, what, how much to buy as well as how much to consume to maximize its expected utility over an infinite horizon is developed. By making simplifying behavior assumptions we enable to reduce the dimensionality of the problem. We estimate the proposed model using scanner panel data of 1000 households on canned tuna purchases for 12 product alternatives over a two-year period. The results from the model shed insights on the decomposition of the price elasticity into its components that could help managers make inferences about for which brands and categories

In usual practice, brand-choice models are specified and estimated from purchase data, ignoring those observations where category incidence does not occur, i.e. the no-purchase observations. This practice can be problematic if there are unobservable factors that affect both the no-purchase and brand-choice decisions. Under such a correlation, it is important to simultaneously model the no-purchase and brand-choice decisions. This paper is an effort in this direction. We propose a model suitable for scanner panel data in which the nopurchase decision depends on the price, feature and display of each brand in the category and the household’s stock of inventory. This no-purchase model is linked to the brand-choice outcome through marketing-mix covariates and, importantly, through unobservables that affect both outcomes. Model parameters are assumed to be heterogeneous across households with the coefficients in the no-purchase model correlated with those in the brand-choice model in a completely general way. This model formulation is more general than what is possible from either a nested logit model or a translog utility model and from models in which the no-purchase outcome is modeled as an additional outcome with the deterministic component of its utility set equal to zero. Estimation of the proposed model is by Bayesian Markov Chain Monte Carlo methods (Chib and Greenberg 1995, 1998) based on the approach of Albert and Chib (1993). The estimation methods are applied to scanner panel data on the

Working Draft. Please do not quote or distribute.

Discrete choice methods model a decision-maker’s choice among a set of mutually exclusive and collectively exhaustive alternatives. They are used in a variety of disciplines (transportation, economics, psychology, public policy, etc.) in order to inform policy and marketing decisions and to better understand and test hypotheses of behavior. This dissertation is concerned with the enhancement of discrete choice methods. The workhorses of discrete choice are the multinomial and nested logit models. These models rely on simplistic assumptions, and there has been much debate regarding their validity. Behavioral researchers have emphasized the importance of amorphous influences on behavior such as context, knowledge, and attitudes. Cognitive scientists have uncovered anomalies that appear to violate the microeconomic underpinnings that are the basis of discrete choice analysis. To address these criticisms, researchers have for some time been working on enhancing discrete choice models. While there have been numerous advances, typically these extensions are examined and applied in isolation. In this dissertation, we present, empirically demonstrate, and test a generalized methodological framework that integrates the extensions of discrete choice.