Risk-off episodes—periods of increased global risk aversion—are characterized by sharp increases in credit spreads, high volatility in equity markets, and appreciation of reserve currencies. We identify changes in global risk sentiment using data on the excess bond premium (EBP) introduced in Gilchrist and Zakrajsek (2012). First, we show that changes in the EBP predict popular proxies for global risk such as the VIX index or the BBB corporate spread. Second, employing structural vector auto regressions (SVARs), we find that identified innovations in the EBP have statistically significant and sizable effects on the global economy: a 1.5 percentage point increase in the EBP—as in the lead-up to the dot-com bubble burst, or between the summer of 2007 and the near-failure of Bear Stearns in March 2008—takes 1 percentage point off world inflation and 2 percent off the level of world GDP over the year following the global risk shock. We propose a mechanism based on the differentiated demand for safe assets to rationalize the empirical findings from the SVAR within a large class of open-economy DSGE models.