What׳s news in News? A cautionary note on using a variance decomposition to assess the quantitative importance of news shocks

This paper points out a conceptual difficulty in using a variance decomposition to assess the quantitative importance of news shocks. A variance decomposition will attribute to news shocks movements in endogenous variables driven both by news about future exogenous fundamentals that has yet to materialize (what I call “pure news”) as well as movements driven by realized changes in fundamentals that were anticipated in the past (what I call “realized news”). I present a stylized model in which news about yet unrealized changes in fundamentals is irrelevant for output dynamics, but in which a variance decomposition may nevertheless attribute a large share of the variance of output to news shocks. I then revisit the quantitative importance of news in the model of Schmitt-Grohe and Uribe (2012). In their model news shocks account for 40 percent of the variance of output growth, but this is mostly driven by realized news.