Household Scanner Panel Data Research Articles

Outside good utility and substitution patterns in direct utility models

This paper investigates the role of the outside good utility function on admissible substitution patterns in multiple discrete/continuous demand models. We first present a set of novel results that characterize the functional form of quantity price effects within this class of models. The results highlight the relative inflexibility of many standard outside good utility functions. We then propose a new outside good utility function that admits more flexible marginal utility curves. Our empirical analysis uses household scanner panel data from the potato chip category, where we find empirical support for non-standard rates of satiation for the outside good. We then show how the restrictive substitution patterns induced by standard utility specifications may distort price elasticities and the evaluation of loyalty coupon targeting programs.

A State Duration Model for Brand Choice and Inter-Purchase Time

A new approach for analyzing state duration data in brand-choice studies is explored. This approach not only incorporates the correlation among repeated purchases for a subject, it also models the purchase timing and the brand decision jointly. The former is accomplished by applying transition model approaches from longitudinal studies while the latter is done by conditioning on the brand choice variable. Then mixed multinomial logit models and Cox proportional hazards models are employed to model the marginal densities of the brand choice and the conditional densities of the interpurchase time given the brand choice. We illustrate the approach using a Nielsen household scanner panel data set

Going Online for Groceries: Drivers of Category-Level Share of Wallet Expansion

Some grocery product categories may be more successful than others in terms of stimulating consumers to increase their share of wallet (SoW) when they start buying through the online channel of a grocery chain. This study explores the circumstances in which online and multichannel marketing mix instruments determine the extent of category-level SoW expansion. To do so, the authors use U.K. household scanner panel data, covering online and offline purchases by 3,311 households in 59 categories of four multichannel retail chains. The results indicate that the effectiveness of online and multichannel marketing mix instruments for stimulating expansion is moderated by category characteristics, such that a selective approach to making decisions about the online price, online assortment breadth, online/offline assortment integration, and online national brand proliferation, tuned to account for category differences, can increase category-level SoW for the online-visited chain.

Capturing Flexible Price Elasticities in Direct Utility Models

This paper investigates the role of the outside good utility function on admissible substitution patterns in direct utility models of discrete/continuous demand. We first present a set of novel results that characterize the functional form of price effects within this class of models. The results highlight the relative inflexibility of many standard outside good utility functions. We then propose a new outside good utility function that admits more flexible marginal utility curves. Our empirical analysis uses household scanner panel data from the potato chip category, where we find empirical support for non-standard rates of satiation for the outside good. We then show how the restrictive substitution patterns induced by standard utility specifications may distort price elasticities and optimal pricing decisions.

Characteristics of Demand Structure and Preferences for Wild and Farmed Seafood in Germany: An Application of QUAIDS Modeling with Correction for Sample Selection

AbstractThe production of farmed fish is growing rapidly and presents a sustainable and possibly low-cost alternative to wild fish. Thus, we may expect retail prices of farmed to be lower than prices of wild fish and demand to be less elastic. Otherwise, marketing of farmed fish may generate some extra value that justifies higher prices and may exhibit more elastic demand. To test these hypotheses, we employ monthly household scanner panel data for Germany from 2006 to 2010 for six frozen seafood products that include farmed and wild fish. A QUAIDS model is estimated by a consistent two-step procedure to account for censoring of the dependent variable. We find consumers to be price sensitive, particularly with regard to the high-value seafood species salmon and shrimp. This price elastic market implies that the German seafood industry still has the potential for growing revenues if production increases.

Consumer Response to Package Downsizing: Evidence from the Chicago Ice Cream Market

It is common among producers of consumer packaged goods to reduce the volume of product per package such that the new size replaces the old one. This tactic is commonly referred to as package downsizing. In this article, we investigate the extent to which consumers have different sensitivities to package price and package size in order to shed light on the managerial implications of package downsizing. To do so, we estimate a random utility model of demand to measure consumer response to price and package size using household scanner panel data on bulk ice cream purchases in Chicago. The estimation framework involves modeling household heterogeneity, addressing price endogeneity and accounting for unbalanced choice alternatives. Our main finding is that consumers are less responsive to package size than to price; the demand elasticity with respect to package size is approximately one-fourth the magnitude of the demand elasticity with respect to price. This result implies that marketing managers can use downsizing as a hidden price increase in order to pass through increases in production costs, that is, cost of raw materials, and maintain, or increase, their profits.

Consumption-Based Cross-Brand Learning: Are Private Labels Really Private?

While some researchers view private labels (PLs) as a key tool for retailer differentiation, others imply that consumers do not distinguish between PLs of different chains. This debate raises the question: Do a retail chain's PL investments subsidize rival PLs? In addition, how does this affect their choice share vis-à-vis national brands? The authors investigate whether consumers generalize knowledge from product experience across PLs of different retail chains and whether such cross-brand learning depends on the PL brands' link with the chain name or on their quality differences. The proposed brand choice model captures cross-brand learning through quality perception spillovers (consumers adjust beliefs about PL quality on the basis of consumption experience) and familiarity spillovers (uncertainty about a PL diminishes with rival PL consumption). Household scanner panel data on dish soap and breakfast cereals yield clear evidence of cross-retailer learning among standard PLs, regardless of their name or quality differences. The results also reveal that familiarity spillovers dominate quality-level spillovers, implying that the presence of cross-learning benefits PLs and enhances their market position relative to national brands. The authors conclude with a discussion of managerial implications.

Distinguishing between state dependence and unobserved heterogeneity in a discrete choice model

This paper proposes a methodology to distinguish state dependence from unobserved heterogeneity in the estimation context of a discrete choice model using household scanner panel data. With traditional maximum-likelihood approaches, the individual effects become correlated with the covariates. In contrast, the proposed estimation method formalizes the correlation structure and thereby enables its consideration during the estimation process. In the resulting model, state dependence becomes a less significant variable, and the proposed model yields better measures of predictive performance than a conventional model. The proposed approach is illustrated through an empirical application in marketing.

The Effect of Media Advertising on Brand Consideration and Choice

The nature of the effect of media advertising on brand choice is investigated in two product categories in analyses that combine household scanner panel data with media exposure information. Alternative model specifications are tested in which advertising is assumed to directly affect brand utility, model error variance, and brand consideration. We find strong support for advertising effects on choice through an indirect route of consideration set formation that does not directly affect brand utility. Implications for media buying and advertising effects are explored.

The Effect of Media Advertising on Brand Consideration and Choice

The nature of the effect of media advertising on brand choice is investigated in two product categories in analyses that combine household scanner panel data with media exposure information. Alternative model specifications are tested in which advertising is assumed to directly affect brand utility, model error variance, and brand consideration. We find strong support for advertising effects on choice through an indirect route of consideration set formation, and not directly affecting brand utility. Implications for media buying and advertising effects are explored.

Behavioural loyalty towards store brands

This paper applies a consumer brand choice model to measure store brand (SB) loyalty. The aim of this paper is to examine whether SB loyalty is different across categories, and we focus on risk perception as an explanatory variable. The model is estimated using ACNielsen Spanish household scanner panel data on two laundry detergent categories over a 2-year period for more than 1107 households. Loyalty, price, socio-demographics and shopping behaviour variables are included. The discrete-choice model formulation is the logit model.

Beyond the Endogeneity Bias: The Effect of Unmeasured Brand Characteristics on Household-Level Brand Choice Models

We investigate the role of potential weekly brand-specific characteristics that influence consumer choices, but are unobserved or unmeasurable by the researcher. We use an empirical approach, based on the estimation methods used for standard random coefficients logit models, to account for the presence of such unobserved attributes. Using household scanner panel data, we find evidence that ignoring such time-varying latent (to the researcher) characteristics can lead to two types of problems. First, consistent with previous literature, we find that these unobserved characteristics may lead to biased estimates of the mean price response parameters. This argument is based on a form of price endogeneity. If marketing managers set prices based on consumer willingness to pay, then the observed prices will likely be correlated with the latent (to the researcher) brand characteristics. We resolve this problem by using an instrumental variables procedure. Our findings suggest that simply ignoring these attributes may also lead to larger estimates of the variance in the heterogeneity distribution of preferences and price sensitivities across households. This could overstate the benefits from marketing activities such as household-level targeting. We resolve the problem by using weekly brand intercepts, embedded in a random coefficients brand choice model, to control for weekly brand-specific characteristics, while accounting for household heterogeneity. Overall, our results extend the finding on the endogeneity bias from the mean of the heterogeneity distribution (i.e., the price effect) to include the variance of that distribution.

Is 75% of the Sales Promotion Bump Due to Brand Switching? No, Only 33% Is

Several researchers have decomposed sales promotion elasticities based on household scanner-panel data. A key result is that the majority of the sales promotion elasticity, approximately 74% on average, is attributed to secondary demand effects (brand switching) and the remainder is attributed to primary demand effects (timing acceleration and quantity increases). The authors demonstrate that this result does not imply that if a brand gains 100 units in sales during a promotion, the other brands in the category lose 74 units. The authors offer a complementary decomposition measure based on unit sales. The measure shows the ratio of the current cross-brand unit sales loss to the current own-brand unit sales gain during promotion; the authors report empirical results for this measure. They also derive analytical expressions that transform the elasticity decomposition into a decomposition of unit sales effects. These expressions show the nature of the difference between the two decompositions. To gain insight into the magnitude of the difference, the authors apply these expressions to previously reported elasticity decomposition results and find that approximately 33% of the unit sales increase is attributable to losses incurred by other brands in the same category.

Estimating an SKU-level Brand Choice Model Combining Household Panel Data and Store Data

The extant literature using household scanner data to estimate consumer choice models has identified two key sources of bias in estimated mean responses to marketing variables. Omitted heterogeneity may bias mean responses towards zero. At the same time, omitted time-varying characteristics of alternatives that influence consumer choices may also bias mean responses towards zero if these characteristics are correlated with observed factors such as price - the endogeneity bias. Both these issues have been well recognized, and methods have been proposed to address them using household scanner panel data. However, when estimating a choice model with these data at the SKU or the UPC level, one may not observe choices for each item in each of the time periods under consideration. Without such information, one cannot control for item and time period specific unmeasured characteristics, as there is no information on alternatives during those periods in which they are not purchased by any of the panelists. In general, when a product category has many alternatives, each with fairly small shares, the household sample may not contain sufficient choices for each alternative, negatively impacting the ability to control for endogeneity with household data. In contrast, as aggregate store-level data are the true aggregation of purchases by all households visiting the store, they contain the time-period specific item level information required to account for endogeneity as long as each item has some sales in each time period. Given the relative merits of household data to estimate the distribution of heterogeneity and store-level data to address the endogeneity problem, we propose an integrated estimation procedure that uses the information in both sources. Our approach provides consistent estimates of the mean responses to marketing variables and the heterogeneity distribution and also controls for potential endogeneity due to correlation between unmeasured item-level characteristics and prices.

Die Bewährung von Ankerpreismodellen bei der Erklärung der Markenwahl

We evalute reference price models with regard to their ability to explain brand choices of individual households. Reference price models are of the adaptive expectations and extrapolative expectations types. Brand choice is analyzed by means of multinomial logit (MNL) models. We specify the deterministic utility component of MNL-Models as both conventional linear function and nonlinear function. Nonlinear utility is approximated by an appropriate neural network, a feedforward multilayer perceptron with sigmoid hidden units. Reference price models of the extrapolative expectation type formed by lagged prices and a time trend are superior to those of the adaptive expectation type for household scanner panel data. Improvements of posterior probabilities of choice models due to the inclusion of reference prices, losses and gains are greater if nonlinear utility choice models are used.

A Flexible Class of Purchase Incidence Models

Purchase incidence models estimated on household scanner panel data typically assume the household's decision interval to be one week. However, it is well known in the econometrics literature that discrete-time models are highly sensitive to the assumed time interval of decision-making. In this study we investigate the consequences of endogenizing the household's decision interval, instead of restricting it to be one week. We characterize the household's random utility maximization problem, and therefore its purchase likelihood function, as a function of the household's decision interval. Such a flexible purchase incidence model is then used to explicitly estimate households' decision intervals in addition to their response to marketing activity and their baseline hazard functions. The proposed model of purchase incidence not only nests traditionally used choice models (such as the binary logit model) and hazard models (such as the discrete hazard model), but also allows for a gamut of more flexible parametric specifications. We estimate the proposed model across four category-level scanner panel datasets and find that the traditional assumption of restricting the household's decision interval to be one week may be too restrictive. We find that households are not only quite heterogeneous in their decision intervals but often have decision intervals longer than a week. From a managerial perspective, we show that estimated price elasticities are systematically understated if one does not allow for the effects of decision intervals. We demonstrate, using a fourth product category, that the results obtained from the category-level analyses generalize to the context of a full model of purchase incidence and brand choice.

A Guadagni-Little Likelihood Can Have Multiple Maxima

Despite many advances in marketing models, the Guadagni–Little (1983) model is still in widespread use by both practitioners and academics. For many new marketing models, the Guadagni–Little model serves as a benchmark. The key variable that allows the Guadagni–Little model to accurately fit data is the loyalty variable, which is an exponential smoothing of past purchases. In this paper, I show that inclusion of this variable in the logit model may result in a likelihood function that can have multiple maxima. I am able to demonstrate this using simulated data and actual household scanner panel data. In addition, I document a systematic relationship between the loyalty coefficient and the loyalty smoothing parameter. Insight for this systematic relationship and the multiple maxima is obtained by recognizing a trade-off between capturing household heterogeneity and state dependence. Finally, in the Guadagni–Little model extreme parameter values capture two different idealized forms of consumer behavior. However, reported studies rarely find these extreme parameter values. I show that procedures commonly used to initialize loyalty biases against these extreme parameter values. This bias offers some explanation for the observed empirical regularity in Guadagni–Little parameter estimates and suggests that researchers should be cautious concluding these parameters capture regularity in consumer behavior.

The Estimation of Pre- and Postpromotion Dips with Store-Level Scanner Data

One of the mysteries of store-level scanner data modeling is the lack of a dip in sales in the weeks following a promotion. Researchers expect to find a postpromotion dip because analyses of household scanner panel data indicate that consumers tend to accelerate their purchases in response to a promotion—that is, they buy earlier and/or purchase larger quantities than they would in the absence of a promotion. Thus, there should also be a pronounced dip in store-level sales in the weeks following a promotion. However, researchers rarely find such dips at either the category or the brand level. Several arguments have been proposed to account for the lack of a postpromotion dip in store-level sales data and to explain why dips may be hidden. Because dips are difficult to detect by traditional models (and by a visual inspection of the data), the authors propose models that can account for a multitude of factors that together cause complex pre- and postpromotion dips. The authors use three alternative distributed lead and lag structures: an Almon model, an unrestricted dynamic effects model, and an exponential decay model. In each model, the authors include four types of price discounts: without any support, with display-only support, with feature-only support, and with feature and display support. The models are calibrated on store-level scanner data for two product categories: tuna and toilet tissue. The authors estimate the dip to be between 4 and 25% of the current sales effect, which is consistent with household-level studies.

Measuring the effects of new brand introduction on inter-brand strategic interaction

This paper empirically investigates the effects of a new brand introduction on the market structure of a product market, i.e., the nature of competition among existing brands. The model is based on the discrete choice multinomial logit model in which the household-specific brand intercepts are decomposed into brand locations in attribute space and households' importance weights for these attributes. The formulation also incorporates the effects of marketing variables on brand choice behavior. The relative proximity of brands in the attribute space provides a measure of the intensity of competitive rivalry among brands. A new brand introduction results in an additional location in this space. Three consequences of the introduction are investigated. (i) Impact on brand locations of extant brands. (ii) Change in the importance weights assigned by households to the attributes. (iii) Effects on households' sensitivities to marketing activities. As some or all of the above could change, these three effects result in eight combinations that need to be studied. The general model is one in which all three components are allowed to change and this nests the remaining seven specifications. The specification that is most consistent with the data can, therefore, be isolated via a statistical test of nested hypotheses. Identifying the consequences of the introduction for market structure has implications for managerial action. An empirical application of the model to household scanner panel data using liquid laundry detergents is provided.

A Relationship between Market Share Elasticities and Brand Switching Probabilities

The authors derive a theoretical relationship between the aggregate market share elasticity matrix and the aggregate brand switching matrix on the basis of a logit model of heterogeneous consumers choosing among competing brands in a product class. Aggregate cross-elasticities are shown to be proportional (through a single scaling constant) to their corresponding aggregate row-conditional brand switching probabilities. Aggregate own-elasticities are shown to be proportional (through the negative of the same scaling constant) to one minus their corresponding aggregate row-conditional repeat purchase probabilities. An empirical analysis conducted on household scanner panel data in the liquid laundry detergent category shows that the theoretical correspondence holds as a very good approximation. An illustrative use of the relationship in estimating aggregate (store-level) models of market share indicates that the relationship helps improve predictive validity in a holdout period.