with Andrey Simonov, Szymon Sacher and Jean-Pierre Dube. Marketing Science (2022)
with Pradeep Chintagunta and Sanjay Dhar
How do households’ joint budget allocations across different store types (grocery vs. discount vs. warehouse vs other stores), category types (food vs. non food) and brand types (private label vs. national brand) change with changes to their income, wealth, employment status, household size and macro factors like the recession? We answer this question using consumer packaged goods (CPG) purchases from more than 1000 categories by a representative panel of US households over fourteen years. We measure within household effects on expenditure shares of joint bivariate (e.g., grocery-food) and trivariate (e.g. grocery-food-private label) baskets along with their effects on univariate baskets. We demonstrate that such an analysis is important for nuanced responses to changes in economic conditions by managers across the different combinations of store formats, product categories and brand types. For example, managers of national brands looking at univariate baskets would conclude that an increase in income is good for their businesses: we find that this would be true for a non-food but not for a food product. We also demonstrate that approximating the joint budget allocation shares or the effects of economic variables on joint budget allocations using separate analyses of store-, category- and brand-type expenditure shares along with an assumption that these decisions are made independently by consumers, can result in incorrect inferences regarding joint baskets.
Consider a display ad for a new pizza brand that announces a discount. Such discount advertising mentions the brand and often shows a picture of the product, thus informing consumers of the brand and some characteristics (the `brand advertising effect’). It also highlights a discount, thus informing consumers of the existence of a discount (the `pure promotion effect’). There could be a third effect, the `discount spotlighting effect’ – a reduction in brand preference from discount advertising that arises from the brand choosing to highlight a discount in its advertising and marketing itself based on a low price. These three effects are typically confounded and, in particular, the `discount spotlighting’ effect has not been studied before. Two identical consumers, one who has seen a brand ad and then found out about the discount separately, and another one who has seen a discount ad, have the same information about the brand and the discount. However, they may have different probabilities of purchase due to the `discount spotlighting’ effect. We demonstrate the existence of this effect using a field experiment on a food delivery app with exogenous variation in ad content (in terms of presence or absence of discount information), and discount percentage. More broadly, we show that discount ads can induce more consumers to engage in search, but yield lower conversion rates to purchase conditional on search for a given consumer, relative to brand ads. Since these two ad types have different strengths, we show that targeted usage of both ad types increases revenue compared to using a single one.
Works in Progress
A New Instrument for Measuring the Effectiveness of TV Advertising: Channel Position Effects
with Jean-Pierre Dube and Andrey Simonov
We use variation in TV viewership that arises from differences in channel positions across different cable systems as an exogenous shifter of advertising viewership. We propose a novel method to measure advertising elasticities for CPG purchases by combining several different datasets on channel position, TV viewership, TV advertising placements and CPG product sales.