Learning to be unpredictable : an experimental study.
This study tests experimentally whether the ability of subjects to play a noncooperative game's mixed-strategy equilibrium (to make their play unpredictable) is affected by how much information subjects have about the structure of the game. Subjects played the mixed-strategy equilibrium when they had all the information about other players' payoffs and actions, but not otherwise. Previous research has shown that players of a game can play a mixed-strategy equilibrium if they observe the actions of all players and use sophisticated Bayesian learning to infer the likely payoffs to other ...
The U.S. economy in 1989 and 1990: walking a fine line
Delayed financial disclosure: Mexico's recent experience
This article documents a delay in the public release of Mexican international reserve data in the months before Mexico's debt crisis at the end of 1994. The article establishes that in that year investors did not know the level of Mexican reserves before October; yet this lack of information did not seem to reduce investor confidence in the Mexican economy. The article does not establish whether the delay in releasing reserve data was due to logistical problems or to a government strategy. The possibility that the delay was strategic is evaluated by developing an economic model that captures ...
Bad news from a forecasting model of the U.S. economy
This paper describes and analyzes the 1990-92 economic forecasts of a Bayesian vector autoregression model developed by researchers at the Minneapolis Fed. The model's 1990 forecast was pretty bad - too optimistic about both inflation and economic growth, especially growth in consumption and housing. An analysis of the model's errors, however, turns up no reason to think the model is unsound. Based on data available on November 30, 1990, the model predicts weak economic conditions for the next two years: a likely recession in 1991 and moderate inflation and weak overall growth in 1991-92. The ...
Statistical inference in the multinomial multiperiod probit model
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. ...
Why no crunch from the crash?
Vector autoregressions and reality
The statistical significance of variance decompositions and impulse response functions for unrestricted vector autoregressions is questionable. Most previous studies are suspect because they have not provided confidence intervals for variance decompositions and impulse response functions. Here two methods of computing such intervals are developed, one using a normal approximation, the other using bootstrapped resampling. An example from Sims? work illustrates the importance of computing these confidence intervals. In the example, the 95 percent confidence intervals for variance decompositions ...
Old ideas at work in the new economy
A fine time for monetary policy?
Recent research in evaluating the effects of monetary policy is potentially tainted by the problem of time aggregation: that is, effects may be incorrectly estimated using quarterly data if the effects of policy occur rapidly. This study evaluates whether time aggregation is a serious problem in a simple vector autoregression. It shows time aggregation has little impact on evaluating the effect of monetary policy in a simple vector autoregression including total reserves, nonborrowed reserves, and the federal funds rate. This finding suggests that time aggregation is unlikely to be important ...