Estimating Macroeconomic News and Surprise Shocks
研究发现常用TFP最大份额估计器在DSGE模型数据中易有偏，提出一种替代的新闻最大份额估计器，能减少偏差和脉冲响应估计的均方根误差，应用于美国数据后发现新闻冲击对TFP扩散更慢、对实际活动影响更小。
ABSTRACT A common VAR approach is to identify responses to TFP news shocks by maximizing the variance share of TFP over a long horizon. We find that these TFP max share estimators tend to be biased in large samples when applied to data generated from DSGE models with shock processes that match TFP moments in the data, especially in the presence of TFP measurement error. We propose an alternative max share news estimator that reduces this bias and the RMSE of the impulse response estimates, even when there is sizable measurement error in the news variable. When applying this estimator to U.S. data, we find that news shocks are slower to diffuse to TFP and have a smaller effect on real activity than implied by the TFP max share estimator.