The use of aggregate time series for testing conditional heteroscedasticity

Many time series exhibit conditional heteroscedasticity such as stock prices or returns, interest rates or exchange rates. Time series used in empirical analysis are often temporal aggregates. We study the effects of using temporally aggregated time series in testing for heteroscedasticity. The distribution of the test statistics is affected by aggregation which causes a severe power loss that worsens with the order of aggregation. Thus, the tests often fail to detect the heteroscedastic nature of the data which is a misleading outcome and can entail wrong decisions. Our conclusions are illustrated by an empirical application.