Subgame-perfect equilibria of finite- and infinite-horizon games

We show that subgame-perfect equilibria of infinite-horizon games arise as limits, as the horizon grows long and epsilon small, of subgame-perfect epsilon-equilibria of games which are truncated after a finite horizon. A number of applications show that this result provides a useful technique for analyzing the existence and uniqueness of infinite-horizon equilibria. We extend our result to the sequential equilibrium concept.     

Learning in experimental games

Summary Experimental games typically involve subjects playing the same game a number of times. In the absence of perfect rationality by all players, the subjects may use the behavior of their opponents in early rounds to learn about the extent of irrationality in the population they face. This makes the problem of finding the Bayes-Nash equilibrium of the experimental game much more complicated than finding the game-theoretic solution to the ideal game without irrationality. We propose and implement a computationally intensive algorithm for finding the equilibria of complicated games with irrationality via the minimization of an appropriate multi-variate function. We propose two hypotheses about how agents learn when playing experimental games. The first posits that they tend to learn about each opponent as they play it repeatedly, but do not learn about the population parameters through their observations of random opponents (myopic learning). The second posits that both types of learning take place (sequential learning). We introduce a computationally intensive sequential procedure to decide on the informational value of conducting additional experiments. With the help of that procedure, we decided after 12 experiments that our original model of irrationality was unsatisfactory for the purpose of discriminating between our two hypotheses. We changed our models, allowing for two different types of irrationality, reanalyzed the old data, and conducted 7 more experiments. The new model successfully discriminated between our two hypotheses about learning. After only 7 more experiments, our approximately optimal stopping rule led us to stop sampling and accept the model where both types of learning occur.We acknowledge the financial support from NSF grant #SES9011828 to the California Institute of Technology. We also acknowledge the able research assistance of Mark Fey, Lynell Jackson and Jeffrey Prisbrey in setting up the experiments, recruiting subjects and running the experiments. We acknowledge the help of the Jet Propulsion Laboratory and its staff members for giving us access to their Cray XMP/18, and subsequently their Cray YMP2E/116.     

Tax schemes in a class of differential games

This paper evaluates tax schemes in a class of differential games. The results indicate that there are many tax schemes that support efficient resource usage, but each may fail to implement the targeted resource because of the multiplicity of equilibria. Since all of the equilibria are subgame perfect, it is difficult to predict which specific one arises. Care must then be taken in using a tax scheme as a remedy for the “tragedy of the commons.” The advantages of other policy instruments (including command-and-control regulation and a tradable permit system) are also discussed. I am indebted to Gerhard Sorger, Koji Shimomura, and two anonymous referees for their helpful comments. The detailed suggestions of one of the referees in particular have markedly improved the paper. Any remaining errors are mine. Partial financial support from MEXT KAKENHI(11730017, 18078004) is also gratefully acknowledged.     

Sample sizes in experimental games

Limited use has been made of power analyses in experimental economics. Very often, the outcome of the control group is associated with a random variable of interest that shows little variance, in which case, there is often not much to learn from the control group. In such cases, control groups of lesser size are more desirable for they give the same message with fewer resources. I demonstrate that the central limit theorem cannot be blindly relied upon in experimental economics. I propose a general solution for a class of problems that interest experimental economists both in the field and the lab. I show that even when the distribution of the outcome variable is not known or assumed, one can (non-parametrically) arrive at a satisficing sample size that has sufficient power for testing the null hypothesis of an assumed mean for the control group.     

Equilibrium selection in experimental cheap talk games

In the past, many refinements have been proposed to select equilibria in cheap talk games. Usually, these refinements were motivated by a discussion of how rational agents would reason in some particular cheap talk games. In this paper, we propose a new refinement and stability measure that is intended to predict actual behavior in a wide range of cheap talk games. According to our Average Credible Deviation Criterion (ACDC), the stability of an equilibrium is determined by the frequency and size of credible deviations. ACDC organizes the results from several cheap talk experiments in which behavior converges to equilibrium, even in cases where other criteria do not make a prediction.     

Competition and coordination in experimental minority games

This work presents experimental results on a coordination game in which agents must repeatedly choose between two sides, and a positive fixed payoff is assigned only to agents who pick the minoritarian side. We conduct laboratory experiments in which stationary groups of five players play the game for 100 periods, and manipulate two treatment variables: the amount of information about other players’ past choices and the salience of information regarding the game history (i.e., the length of the string of past outcomes that players can see on the screen while choosing). Our main findings can be summarized as follows: aggregate efficiency in the game is in most cases significantly higher than the level corresponding to the symmetric mixed strategy Nash equilibrium. In addition, providing players with information about individual choices in the group does not improve aggregate efficiency with respect to when such information is absent. Displaying information about more rounds than just the previous one, on the other hand, seems to have a positive effect on aggregate efficiency. At the individual level, we find a stronger statistical relation between players’ current choices and their own past choices than between players’ choices and previous aggregate outcomes. In addition, the depth of the relation between present and past choices seems to be affected by the prompt availability of information about the game history. Finally, we detect evidence of a mutual co-adaptation between players’ choices over time that is partly responsible for the high level of efficiency observed.      

Expressed preferences and behavior in experimental games

Participants in experimental games typically can only choose actions, without making comments about other participants' future actions. In sequential two-person games, we allow first movers to express a preference between responder choices. We find that responder behavior differs substantially according to whether first movers express a hope for favorable or unfavorable treatment. Responders largely ignore first movers' expressed preferences for favorable responses, however, when the first movers misbehave. As in earlier experiments without preference expression, subjects assign a high positive weight to another person's payoffs when ahead and misbehavior elicits a strong negative response. Logit regressions estimate the weight placed on another (non-misbehaving) person's payoffs to be positive, even when one is behind. There is suggestive evidence that positive reciprocity is enhanced when a preference for favorable treatment is expressed.     

Level-k analysis of experimental centipede games

As one of the best-known examples of the paradox of backward induction, centipede games have prompted a host of studies with various approaches and explanations (McKelvey and Palfrey, 1992, Fey et al., 1996, Nagel and Tang, 1998, Rapoport et al., 2003, Palacios-Huerta and Volij, 2009). Focusing on initial plays observed in experiments, this paper attempts to offer another explanation based on thorough study of level-k models as applied to these games. Borrowing ideas from the cognitive hierarchy model (Camerer et al., 2004), the authors constructed a group of models based on levels of rationality, and also tested for various assumptions on the play of the most naïve player type in these models. It was found that level-k models generally perform better than the agent quantal response equilibrium (AQRE) model and its variant with altruistic player types for increasing-pie centipede games, while the AQRE model with altruistic player types performs better in constant-pie games.     

An experimental investigation of Colonel Blotto games

This article examines behavior in the two-player, constant-sum Colonel Blotto game with asymmetric resources in which players maximize the expected number of battlefields won. The experimental results support the main qualitative predictions of the theory. In the auction treatment, where winning a battlefield is deterministic, disadvantaged players use a “guerilla warfare” strategy that stochastically allocates zero resources to a subset of battlefields. Advantaged players employ a “stochastic complete coverage” strategy, allocating random, but positive, resource levels across the battlefields. In the lottery treatment, where winning a battlefield is probabilistic, both players divide their resources equally across all battlefields. However, we also find interesting behavioral deviations from the theory and discuss their implications.     

Investigating the failure to best respond in experimental games

Experimental Economics - We examine strategic sophistication using eight two-person 3?×?3 one-shot games. To facilitate strategic thinking, we design a ‘structured’...     

Cognitive ability and human behavior in experimental ultimatum games

Recent researches have shed light on the effect of cognitive ability on economic decision-making. By measuring cognitive ability applying Raven's progressive matrix test, we obtain two significant results that this effect affects decision-making in two types of experimental ultimatum games. First, the higher the cognitive ability, the larger the amount a sender offers when the offer is smaller than or equal to the half split. Second, the higher the responders’ cognitive ability, the smaller the offer they accept, when they accept it or not with the strategy method. This study not only finds new factors that affect decision-making in experimental ultimatum games, but also provides more evidences that cognitive ability influences economic decision-making.     

Strategic saving decisions in the infinite-horizon model

We study a mechanism that prevents the long-run distribution of wealth from becoming degenerate in the Ramsey–Cass–Koopmans model when households have different time-preference rates. This mechanism is based on the observation that price-taking behavior is no longer justified when all wealth is owned by a single household. Formalizing this observation, we obtain a model with a unique stationary equilibrium in which, depending on the parameter constellation, any number of households can own positive stocks of capital. We characterize this equilibrium and show for example that an increase in the dispersion of the time-preference rates across households unambiguously increases aggregate output. Whereas the main results are derived for a rather general class of production functions, we devote a separate section to the special case of the Cobb–Douglas technology for which the equilibrium conditions are particularly simple. The research reported in this paper forms part of the project “Economic Growth with Strategic Saving Decisions” supported by the Austrian Science Fund (FWF) under project number P17886. Comments from Robert Becker, Edward Green, Takashi Kamihigashi, David Levine, Fabrizio Zilibotti, anonymous referees, and participants at various conferences and seminars are gratefully acknowledged.     

Efficiency in one-sector,discrete-time,infinite-horizon models

In recent years, a considerable number of papers have been devoted to the problem of characterizing efficient programs in infinite-horizon models. In the present paper on one-sector models, it is shown that (1) the study of production or consumption efficiency can be reduced to the study of a single model, and (2) certain measures of curvature are crucial for the efficiency of programs. Criteria for efficiency of programs using these curvature measures are given and it is indicated how the existing results in the field can be obtained using the present criteria.     

Punishment strategies in repeated games: Evidence from experimental markets

An experiment is designed to provide a snapshot of the strategies used by players in a repeated price competition game with a random continuation rule. One hundred pairs of subjects played the game over the Internet, with subjects having a few days to make their decisions in each round. Occasionally subjects are asked to enter one-period-ahead pricing strategies instead of prices. According to the elicited strategies, between 90% and 95% of subjects punish less harshly (in their initial response to a deviation) than implied by the grim trigger strategy, and do so in a way that depends on the size of the other subjectʼs deviation. Future earnings are highest for subjects adopting the tit-for-tat strategy, even after controlling for a subjectʼs past earnings. Punishment strategies are generally softer and more graduated than implied by a grim trigger strategy, and do better as a result.     

The benefits of voluntary leadership in experimental public goods games

We study the effects of voluntary leadership in experimental public goods games when each group member can volunteer to contribute before the other members. We find that voluntary leadership increases contributions significantly, compared to a treatment where leadership is enforced exogenously.     

Coordination and learning in dynamic global games: experimental evidence

Coordination problems are ubiquitous in social and economic life. Political mass demonstrations, the decision whether to join a speculative currency attack, investment in a risky venture, and capital flight from a particular country are all characterized by coordination problems. Furthermore, all these events have a dynamic nature which has been largely omitted from previous experimental studies. Here I use a two-stage variant of a dynamic global game to study experimentally how the arrival of information in a dynamic setting affects the relative aggressiveness of speculators. In the first stage, subjects exhibit excess aggressiveness, which appears to be driven by beliefs about others’ actions rather than an intrinsic taste for attacking. However, following a failed first-stage attack, subjects learn to be less aggressive in the second stage. On the other hand, the arrival of new, more precise information after a failed attack leads to an increase in subjects’ aggressiveness. Beliefs, again, play a crucial role in explaining how the arrival of information affects attacking behavior.     

Valuation in infinite-horizon sequential markets with portfolio constraints

Summary. We develop a theory of valuation of assets in sequential markets over an infinite horizon and discuss implications of this theory for equilibrium under various portfolio constraints. We characterize a class of constraints under which sublinear valuation and a modified present value rule hold on the set of non-negative payoff streams in the absence of feasible arbitrage. We provide an example in which valuation is non-linear and the standard present value rule fails in incomplete markets. We show that linearity and countable additivity of valuation hold when markets are complete. We present a transversality constraint under which valuation is linear and countably additive on the set of all payoff streams regardless of whether markets are complete or incomplete. Received: March 9, 2000; revised version: February 13, 2001