Public News Arrival and Cross-Asset Correlation Breakdown: Implications for Algorithmic Trading

Abstract

This study models the role of public news arrivals on asset correlation in a trading environment populated by computerized algorithms. The model is empirically tested with the individual stock futures and its underlying spot markets, which are characterized by the mechanical cost-of-carry relation that is typically exploited by algorithmic trading. In normal circumstances, the return correlation between the stock futures and spot quotes is nearly perfect, because futures market makers peg their quotes to those of the underlying by using computerized algorithms. Our simple model predicts that this near-perfect correlation can occasionally break down with two conditions: one, the futures market is less liquid than the corresponding spot market; and two, the uncertainty surrounding the impact of the news on the underlying stocks is sufficiently large. This breakdown occurs because the futures market makers switch from automating the quote-matching process to manually monitor and update their quotes. By employing the comprehensive RavenPack database with firm-level news releases, we test and confirm our model predictions. In particular, the spot-futures return correlation falls as the news uncertainty rises, and this correlation breakdown is more prominent for small-cap stocks. Furthermore, for actively traded stocks, the impact of the news on the breakdown is more intense. If the overall stock market experiences extreme turbulence, however, this impact is weaker. We discuss the implications for the limits of algorithmic trading.