Industrial organization
Econometrics

Microeconomic Theory
Industrial Organization
Econometrics

May 1998 (Oral) Industrial Organization and Microeconomics
May, 1997 (Written, with Distinction) Microeconomic and Macroeconomic Theory

Essays on Collusion: Dynamic Models for Dynamic Markets

Professor David Pearce
Professor Ariel Pakes
Professor Steven Berry

May 2001

M.Phil. 1998, Yale University
M.A. 1997, Yale University
B.Ec. (First Class Honours) 1993, University of Adelaide

Yale University Fellowship, Fall 1996 – Spring 2000
The John Lorenzo Scholarship Prize, University of Adelaide, 1993
Professor Tew's Prize for First-Year Economics, University of Adelaide, 1990
The Economics Society Prize for Economics 1, University of Adelaide, 1990

Head Teaching Assistant, Financial Theory, Yale University, Fall 2000
Teaching Assistant, Intermediate Macroeconomics, Yale University, Spring 2000
Teaching Assistant, Graduate Microeconomics, Yale University, Fall 1998

Research Assistant to Professors Ariel Pakes and Martin Pesendorfer, Fall 1998–Summer 1999
Research Assistant to Professor Ariel Pakes, Summer 1998
Reserve Bank of Australia, Sydney, January 1994–June 1996

"A Model of Collusion Timing," manuscript, Yale University, 2000.

"Examining Models of Collusion: The Market for Lysine," manuscript, Yale University, 2000.

"An Empirical Note on the Influence of the US Stock Market on Australian Economic Activity" (with Bill Russell), Australian Economic Papers, 39(3), September 2000.

"The Determination of Voter-Turnout Across States," manuscript, Yale University, 1998.

"Towards an Understanding of Australia’s Co-Movement with Foreign Business Cycles" (with Bill Russell), Reserve Bank of Australia Research Discussion Paper 9607, 1996.

Economic theory admits a spectrum of predictions of firm behaviour under collusion, depending on the specific modeling assumptions employed. A recent high profile case of collusion occurred in the market for lysine. My dissertation begins by examining whether existing models of collusion can explain observed behaviour in the lysine market. Estimates of demand and cost parameters are derived, and suggest that none of these models can satisfactorily match actual firm behaviour. I conjecture that their lack of descriptive power stems from the common assumption that firms operate in a stable environment. The literature is extended to allow firm decisions about collusion, entry, exit, and investment within an evolving environment. Verifying my conjecture, I find that these factors can improve the descriptive power of theoretical models.

The first paper of my dissertation uses events and empirical observations in the lysine market to assess the explanatory power of alternative models of collusion. The lysine market provides an ideal setting in which to conduct my investigation for several reasons. First, the existence of collusion is confirmed by the confessions of the leading firms. Second, price, quantity, and cost data is available over a sample period encompassing two price wars and a period of successful collusion. Detailed information is also available about the operation of the cartel. Finally, being a chemical compound, lysine is a homogeneous product. This makes analysis considerably simpler. I combine cost data for a major firm in the market with estimates of market demand to calculate the markup over marginal costs of that firm. This observed markup is then compared to the predicted markup of different classes of collusion models. I find that static models such as Cournot or joint-profit maximisation provide little guidance for firm behaviour. Models of repeated games with imperfect information are roughly descriptive of some of the major events in the lysine market. However, to better understand the determinants of firm behaviour under collusion, I propose that we need to relax the assumption, common to much of the collusion literature, that firms operate in a stable environment.

In particular, most theoretical models of collusion ignore the role of entry. Motivated by this limitation, in my second paper, I develop a dynamic model of collusion. The model provides one rationale for the breakdown of collusion upon entry, and a means for examining the timing of the decision to recommence collusion. The typical assumption that firms are identical and operate in an unchanging environment is relaxed. I endogenise entry, exit, investment, and collusion decisions. The driving force behind the results is the nature of the collusive agreement adopted by the participating firms. Motivated by observations in the lysine market, this rule specifies that for the life of the agreement the market share allocated to each firm is given by the market share of that firm at the time of the agreement.

In the model, there are three competitive regimes yielding different profit opportunities: a non-cooperative, a collusive, and a punishment regime. Firms engage in the following sequence of activities each period. First they make a collusion decision. If we are currently in the non-cooperative regime, firms decide whether they wish to collude. If we are in the collusive regime, firms decide whether they wish to abandon collusion. This triggers the punishment regime. Following the collusion decision, incumbent firms decide whether to exit. Firms then make output decisions which determine profits. Finally, potential entrants decide whether to enter, and incumbent firms make investment decisions which stochastically determine future capacity and thereby future profit opportunities. I simulate the model using parameter values based on demand and cost estimates for the lysine market.

The model is particularly successful in explaining the observation in the lysine market that entrants tend to wait until they have captured a decent market share before agreeing to collude. In addition, the framework I use allows us to contrast an environment with and without collusive possibilities and draw conclusions concerning market structure and consumer and producer welfare. In particular, allowing collusion generates an environment in which a greater number of firms can be supported. Producer surplus is increased and consumer surplus reduced relative to a model that does not allow collusion.

Two further projects on the topic of collusion are in progress. The first is a model of reputation formation in collusion. The uncertainty that a rival is adhering to a specified collusive agreement is at the heart of most price war models. However, no role is allowed for reputation. I posit a model of collusion where firms develop and maintain a reputation for cooperative behaviour. In an environment of demand uncertainty and unobservable actions, firms update the reputations of their rivals each period in Bayesian fashion after they observe profits.

There has been a recent explosion in the number of international cartel prosecutions. The largest case involves a constellation of vitamins markets. An intricate collusive scheme was set up involving many firms and markets. The second paper in progress compares the recent experience of collusion in the lysine market, the citric acid market, and several of the vitamins markets. US import quantities and implicit import prices are obtained for the vitamins markets. These data are combined with data and information in the markets for lysine and citric acid to reveal an interesting and diverse pattern of collusion across markets.