Informed Trading and Portfolio Returns

Abstract

We solve a multi-period model of strategic trading with long-lived information in multiple assets with correlated innovations in fundamental values. Market makers in each asset can only condition their pricing functions on trading in each asset. Using daily non-public data from the New York Stock Exchange, we test the model's predictions on the conditional and unconditional lead-lag relations of institutional order flow and returns within portfolios. We find support for the model prediction of positive autocorrelations in portfolio returns as well as the predictions for how informed order flow positively predicts future returns and future informed order flow. We show that these relations strengthen for portfolios formed from assets within the same industry, which likely have higher correlation of fundamental values. Furthermore, we discuss issues that arise when testing implications of strategic models with imperfect proxies for the underlying strategic behaviour.

The Trading Profits of High Frequency Traders

“…Badrinath, Kale, and Noe, 1995;Boehmer and Wu, 2008;Boulatov, Hendershott, and Livdan, 2011;Hendershott, Livdan, and Schurhoff, 2012). Under the informed trader hypothesis we expect to see HFT profits being small, or even negative, when trading with Fundamental traders.…”

Detecting Informed Trading Activities in the Options Markets

“…Several studies have documented informed trading activities both in stock and option markets; see, e.g., Hasbrouck (1991), Easley and O'Hara (1992), Srinivas (1998), Poteshman (2006), and Boulatov, Hendershott, and Livdan (2012). Keown and Pinkerton (1981) report statistical evidence of (illegal) informed stock trading up to 12 days before the first public announcement of a proposed merger.…”

Dissecting cross-impact on stock markets: an empirical analysis

“…Many partial, competing explanations exist, for a review of recent economics literature on the subject see in Ref. [11]. When choosing their quotes, liquidity providers use correlation models calibrated from real data.…”