Research Database

We report on experiments that tested the predictions of competing theories of learning in games. Experimental subjects played a version of the three-person matching-pennies game. The unique mixed-strategy Nash equilibrium of this game is locally unstable under naive Bayesian learning. Sophisticated Bayesian learning predicts that expectations will converge to Nash equilibrium if players observe the entire history of play. Neither theory requires payoffs to be common knowledge. We develop maximum-likelihood tests for the independence conditions implied by the mixed-strategy Nash equilibrium. We find that perfect monitoring was sufficient and complete payoff information was unnecessary for average play to be consistent with the equilibrium (as is predicted by sophisticated Bayesian learning). When subjects had imperfect monitoring and incomplete payoff information, average play was inconsistent with the equilibrium.

Why are methods of production used in an area when more “efficient” methods are available? This paper explores a “resistance to technology” explanation. In particular, the paper attempts to understand why some industries, like the construction industry, have had continued success in blocking new methods, while others have met failure, like the dairy industry's recent attempt to block bST. We develop a model which shows that how easily goods move between areas determines in part the extent of resistance to new methods in an area.

Dynamic general equilibrium models that include explicit household production sectors provide a useful framework within which to analyze a variety of macroeconomic issues. However, some implications of these models depend critically on parameters, including the elasticity of substitution between market and home consumption goods, about which there is little information in the literature. Using the PSID, we estimate these parameters for single males, single females, and married couples. At least for single females and married couples, the results indicate a high enough substitution elasticity that including home production will make a significant difference in applied general equilibrium theory.

Statistical inference in multinomial multiperiod probit models has been hindered in the past by the high dimensional numerical integrations necessary to form the likelihood functions, posterior distributions, or moment conditions in these models. We describe three alternative approaches to inference that circumvent the integration problem: Bayesian inference using Gibbs sampling and data augmentation to compute posterior moments, simulated maximum likelihood (SML) estimation using the GHK recursive probability simulator, and method of simulated moment (MSM) estimation using the GHK simulator. We perform a set of Monte-Carlo experiments to compare the performance of these approaches. Although all the methods perform reasonably well, some important differences emerge. The root mean square errors (RMSEs) of the SML parameter estimates around the data generating values exceed those of the MSM estimates by 21 percent on average, while the RMSEs of the MSM estimates exceed those of the posterior parameter means obtained via agreement via Gibbs sampling by 18 percent on average. While MSM produces a good agreement between empirical RMSEs and asymptotic standard errors, the RMSEs of the SML estimates exceed the asymptotic standard errors by 28 percent on average. Also, the SML estimates of serial correlation parameters exhibit significant downward bias.

This paper catalogues formulas that are useful for estimating dynamic linear economic models. We describe algorithms for computing equilibria of an economic model and for recursively computing a Gaussian likelihood function and its gradient with respect to parameters. We display an application to Rosen, Murphy, and Scheinkman's (1994) model of cattle cycles.

This paper develops a simple model of sovereign debt in which defaulting nations are excluded from capital markets and regain access by making partial repayments. This is consistent with the historical evidence that defaulting countries return to international loan markets soon after a settlement, but after varying periods of exclusion.

We consider a dynamic, stochastic equilibrium business cycle model which is augmented to reflect seasonal shifts in preferences, technology, and government purchases. Our estimated parameterization implies implausibly large seasonal variation in the state of technology: rising at an annual rate of 24% in the fourth quarter and falling at an annual rate of 28% in the first quarter. Furthermore, our findings indicate that variation in the state of technology of this magnitude is required if the model is to explain the main features of the seasonal cycle.

This study examines common stock prices around ex-dividend dates. Such price data usually contain a mixture of observations—some with and some without arbitrageurs and/or dividend capturers active. Our theory predicts that such mixing will result in some nonlinear relation between percentage price drop and dividend yield—not the commonly assumed linear relation. This prediction and another important prediction of theory are supported empirically. In a variety of tests, marginal price drop is not significantly different from the dividend amount. Thus, over the last several decades, one-for-one marginal price drop has been an excellent (average) rule of thumb.

We integrate search theory into an equilibrium framework in a new way and argue that the result is a simple but powerful tool for understanding many issues related to bilateral matching. We assume for much of what we do that utility is less than perfectly transferable. This turns out to generate multiple equilibria that do not arise in a standard model, with transferable utility, unless one adds increasing returns. We also provide simple conditions for uniqueness that apply to models with or without transferable utility or increasing returns. Examples, applications, and extensions are discussed.

This research compares several approaches to inference in the multinomial probit model, based on Monte-Carlo results for a seven choice model. The experiment compares the simulated maximum likelihood estimator using the GHK recursive probability simulator, the method of simulated moments estimator using the GHK recursive simulator and kernel-smoothed frequency simulators, and posterior means using a Gibbs sampling-data augmentation algorithm. Each estimator is applied in nine different models, which have from 1 to 40 free parameters. The performance of all estimators is found to be satisfactory. However, the results indicate that the method of simulated moments estimator with the kernel-smoothed frequency simulator does not perform quite as well as the other three methods. Among those three, the Gibbs sampling-data augmentation algorithm appears to have a slight overall edge, with the relative performance of MSM and SML based on the GHK simulator difficult to determine.