Abstracts of papers by Ed Hopkins

"Learning, Matching and Aggregation,Games and Economic Behavior, January, 1999.

 

Fictitious play and reinforcement learning are examined in the context of a large population where agents are repeatedly randomly matched. We show that the aggregation of this learning behaviour can be qualitatively different from learning at the level of the individual. This aggregate dynamic belongs to the same class of simply defined dynamic as do several formulations of evolutionary dynamics. We obtain sufficient conditions for convergence and divergence which are valid for the whole class of dynamics. These results are therefore robust to most specifications of adaptive behaviour.

 

 

"A Note on Best Response Dynamics," Games and Economic Behavior, October, 1999.

 

We investigate the relationship between the continuous time best response dynamic, its perturbed version and evolutionary dynamics in relation to mixed strategy equilibria. We find that as the level of noise approaches zero, the perturbed best response dynamic has the same qualitative properties as a broad class of evolutionary dynamics. That is, stability properties of equilibria are robust across learning dynamics of quite different origins and motivations.   

"The Stability of Price Dispersion under Seller and Consumer Learning", with Robert M. Seymour.  International Economic Review, November, 2002.

In many markets it is possible to find rival sellers charging different prices for the same good. Earlier research has attempted to explain this phenomenon by demonstrating the existence of dispersed price equilibria when consumers must make use of costly search to discover prices. We ask whether such equilibria can be learnt when sellers adjust prices adaptively in response to current market conditions. With consumer behaviour fixed, convergence to a dispersed price equilibrium is possible in some cases. However, once consumer learning is introduced, the monopoly outcome first found by Diamond (1971) is the only stable equilibrium.

"Two Competing Models of How People Learn in Games".  Econometrica, November, 2002.

Reinforcement learning and stochastic fictitious play are apparent rivals as models of human learning. They embody quite different assumptions about the processing of information and optimisation. This paper compares their properties and finds that they are far more similar than were thought. In particular, the expected motion of stochastic fictitious play and reinforcement learning with experimentation can both be written as a perturbed form of the evolutionary replicator dynamics. Therefore they will in many cases have the same asymptotic behaviour. In particular, they have identical local stability properties at mixed equilibria. The main identifiable difference between the two models is speed: stochastic fictitious play gives rise to faster learning.

 “Adaptive Learning Models of Consumer Behavior.  Journal of Economic Behavior and Organization, November, 2007.

This paper applies recent advances in the theory of learning to the analysis of consumer behavior in a dynamic duopoly. Nash equilibrium play is characterized when consumers learn adaptively about the relative quality of the two products.  A contrast is made between belief-based and reinforcement/familiarity-based learning.  In the latter case, consumers can become locked into the habit of purchasing inferior goods. Such lock-in permits the existence of multiple history-dependent asymmetric steady states in which one firm dominates. In contrast, belief-based learning rules must lead asymptotically to correct beliefs about the relative quality of the two brands and so in this case there is a unique steady state. However, if consumers' initial estimate of a firm's quality is high (low), a firm has an incentive to charge above (below) the myopic duopoly price in order to slow (speed up) learning.

Learning in Perturbed Asymmetric Games”, with Josef Hofbauer.  Games and Economic Behavior, July, 2005.

We investigate the stability of mixed strategy equilibria in 2 person (bimatrix) games under perturbed best response dynamics. A mixed equilibrium is asymptotically stable under all such dynamics if and only if the game is linearly equivalent to a zero sum game.  In this case, the mixed equilibrium is also globally asymptotically stable. Global convergence to the set of perturbed equilibria is shown also for (rescaled) partnership games (also known as games of identical interest). Some applications of these results to stochastic learning models are given.

Learning, Information and Sorting in Market Entry Games: Theory and Evidence”, with John Duffy. Games and Economic Behavior, April, 2005.


Previous data from experiments on market entry games, N-player games where each player faces a choice between entering a market and staying out, appear inconsistent with either mixed or pure Nash equilibria. Here we show that, in this class of game, learning theory predicts sorting, that is, in the long run, agents play a pure strategy equilibrium with some agents permanently in the market, and some permanently out. We conduct experiments with a larger number of repetitions than in previous work in order to test this prediction. We find that when subjects are given minimal information, only after close to 100 periods do subjects begin to approach equilibrium. In contrast, with full information, subjects learn to play a pure strategy equilibrium relatively quickly. However, the information which permits rapid convergence, revelation of the individual play of all opponents, is not predicted to have any effect by existing models of learning.

“Running to Keep in the Same Place: Consumer Choice as a Game of Status”, with Tatiana Kornienko.

 American Economic Review, September 2004.

 If individuals care about their status, defined as their rank in the distribution of consumption of one ``positional'' good, then the consumer's problem is strategic as her utility depends on the consumption choices of others. In the symmetric Nash equilibrium, each individual spends an inefficiently high amount on the status good. Using techniques from auction theory, we analyze the effects of exogenous changes in the distribution of income. In a richer society, almost all individuals spend more on conspicuous consumption, and individual utility is lower at each income level. In a more equal society, the poor are worse off.  

Attainability of Boundary Points under Reinforcement Learning” with Martin Posch. Games and Economic Behavior, October, 2005.

This paper investigates the properties of the most common form of reinforcement learning (the ``basic model'' of Erev and Roth, American Economic Review, 88, 848-881, 1998). Stochastic approximation theory has been used to analyse the local stability of fixed points under this learning process. However, as we show, when such points are on the boundary of the state space, for example, pure strategy equilibria, standard results from the theory of stochastic approximation do not apply. We offer what we believe to be the correct treatment of boundary points, and provide a new and more general result: this model of learning converges with zero probability to fixed points which are unstable under the Maynard Smith or adjusted version of the evolutionary replicator dynamics. For two player games these are the fixed points that are linearly unstable under the standard replicator dynamics.

“Inequality and Growth in the Presence of Competition for Status”, with Tatiana Kornienko. Economics Letters, October, 2006.

We investigate a simple endogenous growth model where agents care about their social status. As greater equality tends to provide greater incentives to differentiate oneself, redistribution may increase wasteful competitive consumption and lead to lower growth.

“Cross and Double Cross: Comparative Statics in First Price Auctions”, with Tatiana Kornienko. The B.E. Journal of Theoretical Economics, 2007.

This paper analyses comparative statics for first price auctions, winner pays and all pay. In all-pay auctions, bidders with low values will respond to a stochastically higher distribution of types by playing less aggressively. In winner pays, a similar change results in all types playing more aggressively. Furthermore, we show that a decrease in dispersion of values, in the sense of a refinement of second order stochastic dominance, although also associated with an increase in competitiveness, may in addition result in less aggressive play by bidders with high values in both auction forms. We also find similar considerations in an oligopoly game with incomplete information: stochastically lower costs can lead to higher prices.

“Status, Affluence, and Inequality: Rank-Based Comparisons in Games of Status”, with Tatiana Kornienko.

This paper considers the effects of changes in the income distribution in an economy where agents' utility depends both on consumption and on their rank in the distribution of consumption of a positional good. We introduce a new methodology to compare the behavior of agents that occupy the same rank in the two different income distributions but typically have different levels of incomes. We analyze the effect of changes in the distribution of income on equilibrium choices and welfare of every member of society. If an income transformation raises incomes at the lower end of the income distribution, the poor will typically be better off. But because such an income transformation also increases the degree of social competition, the middle class will typically be worse off - even if they have higher incomes as well. We find a sufficient condition to make all better off is an increase in income accompanied by an increase in income dispersion. Our new techniques highlight the importance of density of social space as we demonstrate that one can have an increase both in income and relative position but still be worse off.

Learning in Games with Unstable Equilibria”, with Josef Hofbauer and Michel Benaim.

We investigate games whose Nash equilibria are mixed and are unstable under fictitious play-like learning processes. We show that when players learn using weighted stochastic fictitious play and so place greater weight on more recent experience that the time average of play often converges in these ``unstable'' games, even while mixed strategies and beliefs continue to cycle. This time average is related to the best response cycle first identified by Shapley (1964). For many games, the time average is close enough to Nash equilibrium to create the appearance of convergence to equilibrium. We discuss how these theoretical results may help to explain data from recent experimental studies of price dispersion.

“Job Market Signalling of Relative Position, or Becker Married to Spence”

We consider a matching model of the labour market where workers that differ in quality send signals to firms that are also vertically differentiated. Signals allow assortative matching in which the highest quality workers send the highest signals and are hired by the best firms. Matching is consider both when wages are fixed (non-transferable utility) and when they are fully flexible (utility is transferable). In both cases payoffs are determined by relative position - the best worker gets the best job. The standard signalling model which communicates the signaller's absolute type is a special case of the current model of signalling relative position. Furthermore, in the relative model, equilibrium strategies and payoffs depend on the distributions of types of workers and the distribution of firms. This is in contrast with separating equilibria of the standard model which do not respond to changes in supply or demand. With sticky wages, despite incomplete information, equilibrium investment in education by low ability workers can be inefficiently low and this distortion can become worse in a more competitive environment. In contrast, with flexible wages, greater competition improves efficiency.

 Which Inequality?  The Inequality of Endowments Versus the Inequality of Rewards”, with Tatiana Kornienko.

The economic effects of inequality are investigated in a tournament model of social and economic competition. Contestants who are differentiated in ability, choose a level of output. The participant with the highest output gets the highest reward (the best match), the second placed competitor gets the second-highest reward and so on. In this model, equilibrium strategies and payoffs depend on the distributions of types of the contestants and the distribution of rewards. Thus, inequality affects individuals through conventional economic variables such as consumption and leisure. Moreover, changes in the distributions of the resources and of the rewards tend to have opposite effects on equilibrium strategies and payoffs. An increase in the inequality of competitors' resources makes many (possibly all) competitors better off, while an increase in the dispersion of the rewards makes many (possibly all) competitors worse off. Typically, more dispersed rewards (resources) induce lower (higher) effort by the low ability.

 

“A Simple Test of Learning Theory”, with Jim Engle-Warnick

 

We report experiments designed to test the theoretical possibility, first discovered by Shapley (1964), that in some games learning fails to converge to any equilibrium, either in terms of marginal frequencies or of average play. Subjects played repeatedly in fixed pairings one of two 3x3 games, each having a unique Nash equilibrium in mixed strategies. The equilibrium of one game is predicted to be stable under learning, the other unstable, provided payoffs are sufficiently high. We ran each game in high and low payoff treatments. We find that, in all treatments, average play is close to equilibrium even though there are strong cycles present in the data.