Industrial organization
Applied microeconomics
Marketing

Industrial Organization
Microeconomics
Econometrics and Computational Economics

1997 (Oral) Macroeconomics, International Economics
1996 (Written) Microeconomic and Macroeconomic Theory (with distinction)

Essays on the Economics and Marketing of New Goods

Professor Steven Berry
Professor John Rust
Professor Christopher Timmins

Summer 2001

M.Phil., Economics, Yale University, 1998
M.A., Economics, Yale University, 1997
Mag., University of Vienna, 1995

Cowles Foundation Prize Fellowship, 2000
Carl Anderson Prize Fellowship, 1998–1999
University Fellowship, Yale University, 1996–1999
Fellowship, Ministry of Science, Austria, 1995–1997

College Tutor, Yale, 2000
Teaching Assistant, Game Theory, Yale, 1999
Teaching Assistant, Principles of Macroeconomics, Yale, 1997–1998
Teaching Assistant, Intermediate Microeconomics, Vienna, 1992 – 1995

Research Assistant, Prof. Steven Berry, 1999–2000
Assisted Prof. Berry, who is a government expert witness in the forthcoming U.S. v. American Airlines predatory pricing antitrust case. The work involved extensive econometric analysis and proofreading.

Summer Intern, The World Bank, 1997
Participated in a research project on industrial development in several Central and Eastern European transition economies. Research on the effects of privatization, and the effects of foreign direct investment and import competition on firm conduct and market performance

• "Estimating a Model of Product Introduction," manuscript, Yale University, 2000.
• "A Comparison of Discrete and Parametric Approximation Methods for Continuous-State Dynamic Programming Problems," (with Hugo Benítez-Silva, George Hall, Giorgio Pauletto, and John Rust), manuscript, 2000.
• "Sticky Prices and Adjustment Hazards in U.S. Manufacturing Industries," manuscript, Yale University, 1998.
• "Estimating Product Demand Systems with Intertemporal Linkages," in progress, 2000.

The introduction of new products has received much attention in the theoretical industrial organization literature and in marketing. Existing empirical research, however, has shed little light on the reasons why new products are introduced. Product diffusion models, for example, have failed to provide convincing structural foundations for the empirical patterns typically observed over the product life-cycle. New products often exhibit distinct patterns from established products. In the ready-to-eat cereal industry, the sales and advertising levels of newly introduced products peak initially and subsequently decline. Most of these products exit from the industry, and only a few survive and become established products. In contrast, mature products are typically characterized by stable market shares and advertising spending. The theoretical IO literature and marketing research suggest several possible explanations for these empirical patterns, such as market experimentation and learning, entry deterrence, and variety-seeking. In my dissertation, I develop and estimate a dynamic model of product introduction that explains the patterns observed in the data and incorporates the aforementioned reasons for product introductions. Under certain restrictions the model is able to distinguish between the different hypotheses. The model also has normative implications because it explicitly shows how managerial decisions should depend on the current state of the product life-cycle, and may provide forecasting power superior to that of reduced-form diffusion models.

My model provides a realistic description of product introduction in the U.S. ready-to-eat breakfast cereal industry, which can also be applied to other industries (with changes made to certain institutional details). It models the behavior of a firm that has just introduced a new product. Managers are modeled as rational, forward-looking decision makers who set prices, advertising levels, and decide each period whether the product should exit or stay in the market. These managers are initially uncertain about the level of sales and the profitability of the product. Uncertainty about product quality is modeled in the form of a belief. The managers incorporate additional information on the unknown product quality from realized sales and update their belief according to Bayes’ rule. The explicit treatment of learning through time allows for market experimentation, i.e., currently unprofitable products may not exit if the value of staying in the market to learn about the true product quality is sufficiently large. The model also allows for inter-temporal dependencies in demand. One such dependency is due to advertising carryover, where current advertising flows add to a goodwill stock, which affects product demand in the future.

The model is solved using numerical dynamic programming techniques. The parameters are estimated using a nested maximum likelihood algorithm from a rich panel data set of established, entering, and exiting breakfast cereals. The structural assumptions made allow me to infer the history of unobserved beliefs and goodwill levels from the observed data. I can thus distinguish between the contribution of each unobservable to the market share and advertising dynamics, and I can test for learning, even though beliefs are unobserved. The model also allows me to infer the value managers assign to the product at each point in time. Knowing this value is useful, because it allows me to distinguish between a scenario where the firm pursued an aggressive deterrence strategy and introduced a product with negative ex-ante value, and a case where the firm expected the product to be profitable but made losses ex-post. The estimation results show that the effect of advertising on demand is persistent and much larger than the estimated effect in previous studies, which have not allowed for long-lived advertising effects. The persistence and size of this effect implies that a firm which attempts to enter the cereal market needs to invest large amounts in advertising in order to compete with the incumbent brands.

Work in progress: Another part of my dissertation is concerned with the estimation of a demand system for a frequently purchased product, where current marketing choices have long-run effects on demand. I propose a structural approach which allows for the conduct of policy experiments, and hence, the evaluation of long-run effects from alternative marketing strategies. The demand system incorporates purchase reinforcement, where the current consumption experience of a consumer changes her preferences over different brands, and advertising carryover, where current advertising influences the consumer’s purchase decision in the future. Also, past and present advertising exposure and consumption experiences may alter the price sensitivity of a household. The empirical analysis will be performed on household-level data.