An empirical test of the effect of the return interval on conditional volatility

Autoregressive Conditional Heteroscedasticity (ARCH) effects have been hypothesized to be caused by variations in the rate of information flow. Further, Nelson (1990, 1992) argues that ARCH effects and persistence in conditional variances should vary across sampling frequencies. However, these claims have not been subject to empirical tests on stock market data when the return interval is measured using intraday data. This paper presents such a test. The results support Nelson's claim that mis-specification in the conditional mean has only a small influence on the estimated conditional variance. Thus, mis-specified ARCH class models can be a consistent filter if high frequency data are employed. At the micro-structure level, the results are consistent with the view that ARCH effects are generated by variability in the rate of information arrival.