NONLINEARITIES AND NONSTATIONARITIES IN STOCK RETURNS

This article addresses the question of whether recent findings of nonl inearities in high-frequency financial time series have been contamina ted by possible shifts in the distribution of the data. It applies a r ecursive version of the Brouk-Dechert-Scheinkman statistic to daily da ta on two stock-market indexes between January 1980 and December 1990. It is shown that October 1987 is highly influential in the characteri zation of the stock-market dynamics and appears to correspond to a shi ft in the distribution of stock returns. Sampling experiments show tha t simple linear processes with shifts in variance can replicate the be havior of the tests, but autoregressive conditional hereroscedastic fi lters are unable to do so.