This thesis includes three essays which examine the implications of incorporating parameter heterogeneity, consideration set heterogeneity, and decision rule heterogeneity, respectively, in brand choice models. In the first essay, we identify the conditions under which unaccounted for price response heterogeneity results in a spurious sticker shock effect. We show, using an analytical derivation, a simulation study and an empirical application to scanner panel data, that estimates of the sticker shock effect may be biased if households that are price sensitive in their brand choice decision are also more likely to respond to category marketing activity in their purchase timing decision. The empirical results, from two product categories, show that the sticker shock coefficient from a Hierarchical Bayes model (which continuously accounts for price response heterogeneity) is statistically insignificant, providing no evidence of the existence of a sticker shock effect. In contrast, the corresponding coefficient from the standard model, which ignores this heterogeneity, is highly significant and supports the existence of a sticker shock effect. A posterior analysis of household parameters confirms the hypothesized relationship between price sensitivity in brand choice and responsiveness to promotional activity in purchase incidence, and is consistent with our explanation of the underlying cause of the bias in the standard model. The second essay develops a new consideration set model that can be estimated with scanner panel data. In contrast to many previous approaches, which require enumeration of all possible consideration sets, we directly model uncertainty about including a brand in the consideration set. The resulting inclusion probabilities for brands reflect a "fuzzy" consideration set in the sense that a brand belongs to the consideration set only probabilistically. The proposed fuzzy set model outperforms several previous consideration set models in two product categories (yogurt and ketchup). We then apply the fuzzy set approach to examine the role of the consideration set in moderating the impact of advertising on price sensitivity. In contrast to the experimental findings of Mitra and Lynch (1995), we find no positive relationship between consideration set size and price sensitivity. Further empirical test may be necessary to confirm the hypothesized relationship. In the third essay, we investigate the role of decision rule heterogeneity in brand choice behavior. We develop a flexible model, which allows for the uncertainty in decision rules used by the consumer. Specifically, we develop a Hierarchical Bayes model of reference price effects that accommodates both the sticker shock and reference-dependent formulations. In addition, we also incorporate the possibility that consumers may mix the two decision rules probabilistically. Therefore, the proposed model allows for three different decision hierarchies which incorporate sticker shock, reference-dependent and mixed rules respectively. The empirical results show that consumers differ not only in their preference and response but also in their decision rules. On average, half the sample households appear to show loss aversion, i.e., follow a reference-dependent decision rule, while the remaining households do not seem to respond to reference prices. The proposed model provides a richer description of consumer choice processes than the comparison models that allow for only one model structure and ignore model uncertainty.

This thesis includes three essays which examine the implications of incorporating parameter heterogeneity, consideration set heterogeneity, and decision rule heterogeneity, respectively, in brand choice models. In the first essay, we identify the conditions under which unaccounted for price response heterogeneity results in a spurious sticker shock effect. We show, using an analytical derivation, a simulation study and an empirical application to scanner panel data, that estimates of the sticker shock effect may be biased if households that are price sensitive in their brand choice decision are also more likely to respond to category marketing activity in their purchase timing decision. The empirical results, from two product categories, show that the sticker shock coefficient from a Hierarchical Bayes model (which continuously accounts for price response heterogeneity) is statistically insignificant, providing no evidence of the existence of a sticker shock effect. In contrast, the corresponding coefficient from the standard model, which ignores this heterogeneity, is highly significant and supports the existence of a sticker shock effect. A posterior analysis of household parameters confirms the hypothesized relationship between price sensitivity in brand choice and responsiveness to promotional activity in purchase incidence, and is consistent with our explanation of the underlying cause of the bias in the standard model. The second essay develops a new consideration set model that can be estimated with scanner panel data. In contrast to many previous approaches, which require enumeration of all possible consideration sets, we directly model uncertainty about including a brand in the consideration set. The resulting inclusion probabilities for brands reflect a "fuzzy" consideration set in the sense that a brand belongs to the consideration set only probabilistically. The proposed fuzzy set model outperforms several previous consideration set models in two product categories (yogurt and ketchup). We then apply the fuzzy set approach to examine the role of the consideration set in moderating the impact of advertising on price sensitivity. In contrast to the experimental findings of Mitra and Lynch (1995), we find no positive relationship between consideration set size and price sensitivity. Further empirical test may be necessary to confirm the hypothesized relationship. In the third essay, we investigate the role of decision rule heterogeneity in brand choice behavior. We develop a flexible model, which allows for the uncertainty in decision rules used by the consumer. Specifically, we develop a Hierarchical Bayes model of reference price effects that accommodates both the sticker shock and reference-dependent formulations. In addition, we also incorporate the possibility that consumers may mix the two decision rules probabilistically. Therefore, the proposed model allows for three different decision hierarchies which incorporate sticker shock, reference-dependent and mixed rules respectively. The empirical results show that consumers differ not only in their preference and response but also in their decision rules. On average, half the sample households appear to show loss aversion, i.e., follow a reference-dependent decision rule, while the remaining households do not seem to respond to reference prices. The proposed model provides a richer description of consumer choice processes than the comparison models that allow for only one model structure and ignore model uncertainty. / Business, Sauder School of / Graduate

Discrete choice models and the assortment optimization problem are the fundamental aspects of the broader field of revenue management, which now spans a broad array of industries such as airlines, hotels and online advertising. The main focus here is to first study the consumer preferences and their substitution behavior when they are faced with multiple options, explain those observed behaviors with mathematical models and then identify an optimal set of options to offer to maximize revenues. This dissertation enriches the choice models and assortment optimization fields by studying the setting when such options are available in multitude, either to the sellers or to the consumers to choose from. The first half of this dissertation focuses on the situation when sellers have access to a vast array of features to be chosen for products they want to offer. The second half of the dissertation focuses on the situation when customers are faced with a lot of options to choose from. This dissertation formulates concrete mathematical discrete choice models to tackle those situations, then studies the assortment optimization problem of maximizing the expected revenue resulting from these newly introduced choice models, and finally also designs efficient algorithms to solve them. Chapter 1 explores discrete choice models which capture consumer behavior and choices when faced with a set of different alternatives, and the resulting assortment optimization problem along with the different existing algorithms for solving them as well as the existing challenges therein. Chapter 2 models and solves the problem when the sellers have access to a vast array of inventory of products. Chapter 3 models dynamic preferences of consumers and the choice overload phenomenon when the customers are faced with a lot of options, and solves the ensuing optimization problem. Chapter 4 showcases the applicability and effectiveness of such models and approaches on high dimensional data from a field experiment on Flipkart, the largest e-commerce firm in India.

This dissertation is an empirical study of demand and supply in differentiated products markets using supermarket scanner data on two particular product categories - canned tuna and hot-breakfast cereals. First, I study the impact of retailers' price promotions on consumer demand and retailer profits in the canned-tuna product category. Since canned tuna is storable, I examine whether consumers stock up during sales. The results suggest that only a limited amount of stockpiling exists in this product category. Since inventory is not very important, consumer demand is thus modeled by a static demand model with a random-coefficients-nested-logit specification, which is estimated by the Markov Chain Monte Carlo method. The unit-sales decomposition results show that on average 36% of the demand response to price promotions comes from brand-switching, so market expansion effects due to consumers switching from the outside good and to higher quantities usually dominate the brand-switching effect. Using the demand estimates, I compute optimal retail prices assuming that stores are local monopolists and choose prices to maximize static category-level profits. I find that regular prices at "high-low" stores are typically at or slightly below the optimal prices, but that regular prices at "every-day-low-price" stores are substantially below the optimal prices. These results suggest that retail price levels and price promotions are more likely related to local market conditions such as retail competition. In addition, I study the effects of store-brand (SB) entry on the demand elasticities of incumbent national brands (NB), consumers' substitution patterns for national and store brands, and the implications for consumer welfare in the hot-breakfast-cereals product category. A random-coefficients model of consumer demand is estimated by the generalized-method-of-moments approach. The empirical findings are: (1) After the entry of SB's, demand becomes more elastic for non-imitated NB's, and either more elastic or shows no change for imitated NB's; (2) in general, substitution patterns for NB's and SB's are asymmetric, i.e., when the prices of their favorite products increase, most NB buyers tend to substitute to other NB products, but SB buyers will substitute to the corresponding imitated NB's; (3) the increase in consumer surplus due to SB entry is trivial for an individual consumer, but the aggregate benefit could be quite substantial.