Predicting instability

Unanticipated shocks could lead to instability, which is reflected in statistically significant changes in distributions of random variables. Changes in the conditional moments of stationary variables are predictable. We provide a framework based on a statistic for the <italic>sample generalized variance</italic>, which is useful for interrogating real time data and for predicting statistically significant sudden and large shifts in the conditional variance of a vector of correlated macroeconomic and financial variables. It is a test for a market-wide instability. Central banks can incorporate the framework in the policymaking process.