Home Address:
  703 Whitney Avenue, 3rd Suite
  New Haven, CT 06511

Telephone: (203) 606-5767

Industrial Organization
Applied Econometrics
Applied Microeconomics

Industrial Organization
Econometrics
Microeconomics

May 2002 (Oral) Industrial Organization and Econometrics (with distinction)
May 2001 (Written) Microeconomic Theory and Macroeconomic Theory (with distinction)

Entry and Competition in the Retail and Service Industries

Professor Steven Berry (co-chair)
Professor Penny Goldberg (co-chair)
Professor Hanming Fang
Professor Philip Haile

May 2006

M.Phil. (2003), Department of Economics, Yale University
M.A. (1999), Department of Economics, Tufts University
B.A. (1997), with distinction, Department of Economics, Fudan University, China

Horowitz Foundation Fellowship, 2005
Yale University Dissertation Fellowship, 2004
Cowles Foundation Prize, Yale University, 2004
Carl Arvid Anderson Prize Fellowship, Yale University, 2003
John Perry Miller Fund, Yale University, 2003
Fan Family Fellowship, Yale University, 2000–2004
Champion in Ningxia Province, National Math Olympiad for High School Students, China, 1993

Teaching Assistant, Industrial Organization, Yale University, 2005
Teaching Assistant, Graduate Econometrics II, Yale University, 2003
Teaching Assistant, Intermediate Microeconomics, Yale University, 2002
Teaching Assistant, Microeconomics, Tufts University, 1999

Research Assistant, Professor Penny Goldberg, Yale University, 2002
Research Assistant, Professor George Hall, Yale University, 2001
Research Assistant, Professor Ioannides, Tufts University, 1998–1999
Quantitative Analyst, Property & Portfolio Research, Inc., Boston, MA, 1999-2000

International Industrial Organization Conference, 2005
Chinese Economic Association in North American Annual Conference, 2004
Northeast Universities Development Consortium Conference, 2003

Programming:  MATLAB, SAS, STATA, RATS
Large Data Set Experience:  PSID, AHS

Professor Steven Berry
Department of Economics
Yale University
PO Box 208268
New Haven, CT 06520-8268
Phone: (203) 432-3571 and 432-3556
Fax: (203) 432-6249
Email: steven.berry@yale.edu

Professor Hanming Fang
Department of Economics
Yale University
PO Box 208264
New Haven, CT 06520-8264
Phone: (203) 432-3547
Fax: (203) 432-6323
Email: hanming.fang@yale.edu

My thesis studies entry and competition in the retail and service industries. The empirical Industrial Organization literature has a long tradition of using observed firm entry and exit decisions to make inferences about profit functions and to analyze the nature of competition among firms in local markets. My thesis builds on this tradition and seeks to contribute in two ways. First, it extends the existing methodology by relaxing three sets of commonly used assumptions. Second, it applies these extensions to analyze policy issues in the retail and service industries.

Most entry models assume that entry decisions in different markets are independent. While this assumption simplifies the estimation, it is clearly problematic in the case of retail chains that operate multiple stores in several markets, where scale economies in the distribution system can lead to substantial cost savings. The first chapter formulates and estimates a model that explicitly captures this effect and allows chains to make joint entry decisions in a large number of markets. The model is applied to the discount retail industry to quantify the impact of chain stores on small retailers.

The second chapter relaxes the distributional assumption for the error term in entry models. This is motivated by the observation that the maximum likelihood estimator (MLE) is inconsistent if the assumed distribution is wrong. Klein and Spady (1993) proposed a consistent and efficient semi-parametric estimator that does not require the error term’s distribution to be specified. Their approach is designed for single-index models. I extend their method to an entry model with multiple indices and apply this extension to the dry cleaning industry.

The presence of multiple equilibria is an unfortunate feature of many entry models. The common approaches are either to select an arbitrary equilibrium, or to impose additional assumptions that guarantee uniqueness. The third chapter (joint with Donald Andrews and Steven Berry) proposes as an alternative a set inference method. The method is used to analyze chain stores’ location choices.

Chapter 1: What Happens When Wal-Mart Comes to Town: An Empirical Analysis of the Discount Retail Industry (Job Market Paper)

In the past few decades the retail industry has experienced substantial growth in multi-store retailers, especially chains with a hundred or more stores. At the same time, there is a widely reported public outcry over the impact of these chain stores on small retailers and local communities. In this chapter I identify the effects of chain stores on small retailers and quantify the size of the scale economies within a chain.

I develop and estimate a model that has two key features. First, it allows for fully flexible competition patterns among all the players. Second, it incorporates the scale economies that arise from operating multiple stores in nearby regions; this is one of the primary distinctions between chain stores and single-unit firms. The scale economies cause a chain’s profit to be cross-sectionally dependent. As a result, its profit maximization and location choices lead to a complicated problem with a large number of discrete choice variables. To solve this problem, I exploit the unique features of the scale economies and propose a bound approach that reduces the number of necessary computational evaluations from 10⁶⁰⁰ to a manageable number. I also take advantage of the supermodular property of the profit functions in analyzing the interaction of the scale economies and the competition effects.

To estimate the model I collected a large data set that covers the entire discount retail industry from 1988 to 1997. The model is estimated by simulated method of moments. The results indicate that Wal-Mart’s expansion from the late 80s to the late 90s explains about fifty to seventy percent of the net change in the number of small discount retailers. Failure to address the endogeneity of the firms’ entry decisions results in underestimating this impact by about sixty percent. In addition, I find that direct subsidies to either chains or small firms in this industry are not likely to be cost effective in increasing the number of firms or the level of employment. The Wal-Mart stores that received subsidies in the last decade are on average more profitable than the unsubsidized ones. Last, scale economies were important in Wal-Mart’s early expansion period in the late 80s, but their impact diminished greatly in the late 90s.

Chapter 2: Semi-parametric Estimation of the Distribution of Fixed Costs in Entry Models

In many entry models, profit is a reduced form function of market size variables and an unobserved entry cost (i.e., the error term). Most of the time, the distribution of the error term is selected for reasons of computational simplicity or tractability, and the parameters are estimated by MLE. The MLE estimator, however, is inconsistent if the chosen distribution is wrong. The approach of Klein and Spady (1993), which was originally applied to binary response models and later extended in Klein and Sherman (2002) to ordered response models, provides a consistent semi-parametric estimator without specifying the error term’s distribution.

The Klein–Spady method applies to single-index models in which the conditional mean of the dependent variable depends on the explanatory variables through one index. I adapt their method to an entry model with multiple indices, and transform the choice probabilities in a way that reduces the multiple indices to a single index. Applying the method to the dry cleaning industry in six hundred and eighty-four small towns, I find evidence that the distribution of the entry cost is not symmetric, and that the asymmetry drives the differences between the parametric estimates (which assumes the error term to be normally distributed) and the semi-parametric estimates. Therefore, the distributional assumption does seem to be important. However, I cannot reject the null hypothesis that the parametric and the semi-parametric estimates are the same. The paper proposes this method as a robustness check in situations where the commonly used distributions might be problematic.

Chapter 3: Confidence Regions for Parameters in Discrete Games with Multiple Equilibria, with an Application to Discount Store Locations (joint with Donald Andrews and Steven Berry)

The third chapter addresses the difficulties in estimation due to the existence of multiple equilibria that plagues many discrete choice games, including entry models. Without a proper equilibrium selection rule, the presence of multiple equilibria makes it problematic to apply the traditional MLE method, which requires a one-to-one mapping between regions of the unobservables (model predicted probabilities) and the observed equilibrium outcomes.

A common approach is to seek additional assumptions that guarantee the uniqueness of certain types of the equilibria, but this involves a loss of information and becomes increasingly difficult as the model gets more realistic. This paper proposes a set inference method as an alternative: we exploit the equilibrium necessary conditions to place a system of inequality restrictions on the parameters, and pursue set estimation and inference directly without attempting to solve the equilibrium selection problem. We focus on the distribution of the inequality constraints and how the variance of these constraints affects the inference of the identified set of parameters. Our approach differs from, for example, that of Chernozhukov, Hong and Tamer (2004), who concentrate on the distribution of a minimized criterion function.

First, we provide a full set of sufficient conditions for the consistent estimation of the identified set. Then, we propose several approaches to construct asymptotically valid confidence intervals for real-valued functions of the parameters. Last, we discuss two Monte Carlo examples and an application to chain stores’ location choices in the discount retail industry.