We examine the source of permanent shocks to the variance of a set of emerging and developed markets. By using the ICSS algorithm, Bai-Perron (2003)’s test for structural breaks in mean level, and wavelets, and analyzing weekly data for 18 countries over the January 1996 – April 2006 period, we find significant numbers of variance breaks. In very few instances, the three methods provide the same dates of variance shifts. Therefore, they can be used as complementary tools. In particular, the ICSS algorithm and wavelets tend to detect more variance shifts than Bai-Perron’s. We conjecture that this is an outcome of how volatility is gauged and of the design of Bai-Perron’s algorithm. We confirm earlier findings that the majority of variance breaks appear to be associated with local rather than international events.