Industrial organization
Microeconomic theory
Applied econometrics and Computational economics

Industrial Organization
Microeconomics
Applied Econometrics
Computational Methods in Economics

May 1997, (Oral) Industrial Organization, Microeconomic Theory
May 1996, (Written, with Distinction) Microeconomic and Macroeconomic Theory

Technological Change and the Demand for Differentiated Durable Products in the Computer Printer Industry

Professor Steven Berry
Professor Ariel Pakes
Professor Martin Pesendorfer

Summer 2001

M. Phil., Yale University, 2000
M.A., Yale University, 1996
Diplom (M.Sc.) in Robotics (with distinction), Kharkov State Polytechnic University, Kharkov, Ukraine, 1989

Raymond P. Powell Prize for Excellence in Teaching, Yale University, 1998-99
Yale University Dissertation Fellowship, 1999
Yale University Graduate Fellowship, 1995-98
Freedom Support Act Fellowship Program sponsored by USIA, 1994-95
Young Professionals Foundation Award, Kharkov, Ukraine, 1992
Nickolay E. Zhukovsky Fellowship, Kharkov State Polytechnic University, 1986-89

Teaching Assistant, Introductory Macroeconomics, Yale University, Spring 2000
Instructor, Introductory Microeconomics, Yale University, Fall 1998
Teaching Assistant, Intermediate Microeconomics, Yale University, Spring 1998
Teaching Assistant, Introductory Microeconomics, Yale University, Fall 1997
Instructor, Production/Operations Management, Management Information Systems, Kharkov State Polytechnic University, 1993-94
Visiting Lecturer, Kharkov Computer College, Fall 1992

Research Assistant, Professor Steven Berry and Professor Ariel Pakes, 1998-99, for the project "Estimating from the First Order Conditions for Dynamic Controls," Monte Carlo simulations and estimating cost parameters of a dynamic oligopoly.

Research Assistant, Professor Ariel Pakes, Summer 1998, developed public domain Gauss code for the Pakes-McGuire algorithm of Markov perfect industry dynamics.

Research Assistant, Professor Ariel Pakes and Professor Martin Pesendorefer, 1997-98, for the project "An Empirical Analysis of the Market for Military Aircraft," database processing in C and Gauss.

Consultant, The Boslego Corporation, Boston, Summer 1996 and Summer 1997, developed optimal hedging strategies for the international clientele in energy and transportation sectors.

Independent Consultant, SeverGazProm (Northern Natural Gas Industries), Moscow, Russia, 1993-94, design of the pipeline inventory management system.

Research Associate, Kharkov State Polytechnic University, Department of Technical Cybernetics, 1990-93, research projects included optimal government/agency contract design, micro-level computer simulations of inventory flows in flexible manufacturing systems, and design of CAD/CAM software.

Research Engineer, Kharkov State Polytechnic University, Department of Technical Cybernetics, 1989-90, software development.

• "Demand for Differentiated Durable Products: The Case of the U.S.
• Computer Printer Market," mimeo, Yale University, 2000.
• "Estimating Differentiated Durable Products Demand System from the Aggregate Data," mimeo, Yale University, 2000.
• "Markov-perfect Dynamics of a Differentiated Durable Products Oligopoly," work in progress, Yale University.
• "Using Nested Logit Model for Estimating Differentiated Products Demand System with an Application to the U.S. Automobile Industry," mimeo, Yale University, 1998.
• "An Empirical Study of Intertemporal Price Discrimination in Supermarkets," work in progress, Yale University (with M. Pesendorfer).
• "Intertemporal Price Discrimination and Inventory Behavior in Retailing," mimeo, Yale University, 1998.

[In Russian]

• "Mathematical Models of Adaptive Pricing," Kharkov University Press, 1994.
• Adaptive Pricing Strategies: Mathematical Models and Software Support," in Information Technologies in Science, Industry, and Healthcare, ed. by L.G. Raskin, Kharkov, 1993, 243-247 (with V. Zaruba).
• "Efficient Pricing Strategies in Sequential Bargaining," same source, 248-250.
• "Efficient Pricing of Products whose Value Expires with Time," in Business and Science, ed. by L. Kulikovskaya, Kharkov, 1992, 17-19.
• "Production Smoothing Scheduling Algorithms in Flexible Manufacturing," Technicheskaya Kibernetika i ee Prilozheniya (Technical Cybernetics and Applications) 7 (2), 27-30, 1992.
• "SIMAP: Micro-level Computer Simulation of Production Flows in Automated Manufacturing," in Proceeding of the 2^nd East European Conference on Flexible Manufacturing Systems, Sofia, 1991, 246-256 (with V. Nechipurenko, A. Slepchenko, and E. Melzedinov).
• "Mathematical Model of Equipment Replacement in Flexible Manufacturing," in Perspectives of Flexible Manufacturing and Robotics, ed. by V. Chubuk, Kiev, 1990, 51-53.
• "Minimizing Replacement Investment Costs in Flexible Manufacturing Systems," in Technological and Economic Problems of Flexible Manufacturing, Kharkov, 1989, 193-194.
• "Linear-Quadratic Approach to Production Scheduling in Flexible Manufacturing," master's thesis, Kharkov State Polytechnic University, 1989.

My dissertation studies the impact of technological change on the dynamics of consumer demand in the differentiated durable products industry. The dissertation consists of three essays. The first essay develops a model of consumer demand for differentiated durable products and proposes a strategy for identifying model parameters from the aggregate data. The second essay takes the model to data describing the U.S. computer printer market. The third essay (in progress) supplements it with the supply side to allow for an equilibrium analysis of the durable product oligopoly.

Empirical work on estimating differentiated product demand systems is dominated by static discrete choice models with random utility. However, many markets for durable products, such as those for personal computers and accessories, exhibit several features at odds with their assumption of myopic consumer behavior. First, purchase of a durable results in a consumer's exiting the market for a significant period of time. Second, when product quality rapidly improves over time, static models will tend to predict an upward trend in aggregate sales, which fails to account for a variety of patterns commonly seen in the data. Third, the evolution of product quality leads to the possibility of intertemporal demand substitution. By foregoing a current purchase, the consumer retains the option to buy a potentially better product in the future. This trade-off cannot be accounted for in the static framework.

My first essay presents a dynamic model of consumer demand for differentiated durable products that explicitly accounts for consumers' expectations of future product quality and consumers' outflow from the market, arising endogenously from their purchase decisions. In my model, the consumer faces a sequence of static discrete choice problems over a non-stationary choice set. I show that for a subclass of random utility models (which includes multinomial logit and McFadden's (1978) generalized extreme value model) there exists a scalar-valued sufficient statistic that determines the value of the option to postpone the purchase. This allows me to model industry evolution over a vastly reduced state space and to formalize the consumer's decision of when to buy as an optimal stopping problem. A solution to the latter defines the hazard rate of adoptions, while the nested discrete choice model determines the alternative-specific purchase probabilities. Integrating individual decisions over the population distribution generates rich dynamics of aggregate and product level sales. The paper proceeds to present Monte-Carlo evidence that ignoring the time dimension of consumer choice will induce a downward bias in the estimates of consumer preferences over quality, sometimes reversing the sign of the coefficients. I propose a three-stage nested GMM estimator that allows for sequential identification of the model parameters with aggregate data from the relatively short time series.

My second essay estimates the dynamic model of consumer choice described above with data on the U.S. computer printer market. During the 90s, the industry expanded dramatically and was characterized by high rates of product entry and exit and remarkable technological innovation. These trends are indicative of the performance of the computer industry as a whole. Given the annual rate of decline in the quality-adjusted prices of 35%, it appears that intertemporal consumer optimization is an important feature of the market. The data set used in this study is a detailed monthly sampled scan-code level panel on more than 400 printer models. The estimates support the hypothesis of consumer's forward-looking behavior, allowing for better demand forecasts and improved measures of welfare gains from introducing new products.

The third essay (in progress) integrates the above model with the supply side to derive an equilibrium model of the quality-generating process in a durable goods industry. The model builds upon theoretical advances in the sequential bargaining literature and uses the Markov–perfect Nash equilibrium approach to modeling industry dynamics suggested in Pakes and McGuire (1994). Using this framework should solve for the Bayesian equilibrium in a durable goods oligopoly when the number of competitors is small.

Related Papers:
In "Using Nested Logit Model…" I explore a different dimension of consumer demand for durable products, namely, demand implications of consumer heterogeneity. For this study, I use rich consumer-level cross-section on automobile purchases used in Berry, Levinsohn, and Pakes (1998). One computationally simple yet reasonably realistic demand system is the nested logit model of McFadden (1981). Its practical applicability is somewhat hindered by the fact that theory will often offer little guidance on how to assign products to different nests. Clustering products by observable product-specific dummies is problematic since the econometrician does not a priori know whether consumers attach any importance to the attribute in question. The unique feature of the Berry et al. data set is the availability of the second choice information (survey respondents were asked to indicate their second best vehicle). The paper presents a simple approach to incorporating these data into the estimation strategy. I propose using combinatorial optimization methods to define the nests consistent with consumers' revealed preferences. Using this procedure, I obtain sharper estimates of consumer preferences and more realistic substitution patterns than those implied by the conventional nested logit.