Three Essays on New Product Introduction and Product Failures

Thumbnail Image
Gong, Yiran
This thesis presents an empirical study of new product introduction and product failures. First, I investigate reasons for product failures through a case study of the introduction and subsequent exit of Coca-Cola's Vanilla Coke. Using scanner data, we estimate a structural model of the soft drink industry to infer Coca-Cola’s profit gains from the new brand. Results show that they would not cover associated fixed costs. We then analyze the importance of market variables for explaining the failure and find that Vanilla Coke’s profitability is most sensitive to rival reactions. We also find Coca-Cola did anticipate some rival reaction that made survival harder, but the actual changes were even more intense and contributed to Vanilla Coke's exit. I also study the evaluation of new products’ welfare effects. Structural estimation of such effects largely relies on demand models with a logit error. In markets where consumers have both brand and option preferences, nested logit demand models are used. However, the logit errors lead to welfare overestimation and make welfare estimates sensitive to the number of nests. To address these problems, I develop an empirical framework to estimate a pure characteristics demand model that allows for option choices and eliminates logit errors. I provide an application using scanner data to evaluate four new products introduced to the U.S. shampoo market. Compared with models featuring logit errors, my approach reduces the implausibly large estimates by at least 73%. Finally, I explore supply-side optimization problems when the pure characteristics model is employed on the demand side. Pure characteristics models are well-suited for analyzing problems involving a changing number of products. Examples include pricing new products, product assortment, and merger analysis. However, it is challenging to incorporate a supply side because market share functions can be set-valued and discontinuous. To fill the gap, I utilize a regularization method proposed in the mathematical programming literature to obtain a unique and continuous market share function, which can be estimated feasibly. My simulation results suggest that such a method can provide accurate estimates for firms' marginal costs and new equilibrium prices when the number of products changes.
External DOI