We study the dynamics of good-by-good real exchange rates using a micro-panel of 270 goods prices drawn from major cities in 71 countries and 245 goods prices drawn from 13 major U.S. cities. We Þnd half-lives of deviations from the Law-of-One-Price for the average good is about 1 year; somewhat lower for U.S. cities and somewhat higher for cities in the OECD with LDC cities in between. This speed of adjustment is well below the concensus range of estimates of 3 to 5 years for purchasing power parity deviations yet consistent with plausible 'price-stickiness.' We further construct price indices using our micro data and Þnd that aggregation bias combined with small sample bias accounts for a large part of the difference between micro and macro estimates for the OECD.

We examine whether the news shocks, as explored in Beaudry and Portier (2004), can be a major source of aggregate fluctuations. For this purpose, we extend a standard dynamic stochastic general equilibrium model of Christiano, Eichenbaum, and Evans (2005) and Wouters (2003, 2007) by allowing news shocks on the total factor productivity (TFP), and estimate the model using Bayesian methods. Estimation results on the U.S. and Japanese economies suggest that (i) news shocks play a relatively more important role in the United States than in Japan, (ii) a news shock with a longer forecast horizon has larger effects on nominal variables, and (iii) the overall effect of the TFP on hours worked becomes ambiguous in the presence of news shocks.JEL codes: E30, E40, E50

This paper considers the bootstrap for the GMM estimator of overidentified linear models when autocorrelation structures of moment functions are unknown. When moment functions are uncorrelated after finite lags, Hall and Horowitz, [1996. Bootstrap critical values for tests based on generalized method of moments estimators. Econometrica 64, 891-916] showed that errors in the rejection probabilities of the bootstrap tests are oðT À1 Þ. However, this rate cannot be obtained with the HAC covariance matrix estimator since it converges at a nonparametric rate. By incorporating the HAC covariance matrix estimator in the Edgeworth expansion of the distribution, we show that the bootstrap provides asymptotic refinements when the characteristic exponent of the kernel function is greater than two. r

Introduction 2. Model 3. Theoretical results 4. Discussions 5. Simulation results 6. Application to financial data 7. Conclusion Appendix References Tables 1-5