Does trading activity contain information to predict stock returns? Evidence from Euronext Paris

Abstract

This article aims to examine the causal and dynamic relationship between trading activity and stock returns, using detailed intraday data from Euronext Paris. We distinguish between two measures of trading activity: the raw volume metric (the nondirectional volume) and the directional volume. In line with the existing literature, we find a unidirectional causality running from stock returns to nondirectional volume. Furthermore, we highlight a strong bidirectional relation between stock returns and directional volume. This result is interesting and has several implications. First, it provides evidence that the directional volume is more informative than the nondirectional volume. Second, it shows that the directional volume helps predict stock returns. Third, it provides an empirical test for the Mixture Distribution Hypothesis (MDH) and the sequential arrival hypothesis, which posit that the information content of the trading activity affects future returns.

Keywords:

• directional volume
• stock index returns
• VAR model
• Granger causality
• market microstructure

JEL Classification::

• G14
• G12

Notes

¹ Lisbon, Porto and the London International Financial Futures and Options Exchange (LIFFE) joined the group later.

² We have also computed the directional volume as the daily number of buyer-initiated minus seller-initiated trades scaled by the total number of trades and we have obtained similar results.

³ We can cite the spread decomposition model of Lin et al. (1995) or the price decomposition methodology of Hasbrouck (1991).

⁴ See Odders-White (2000, p. 262).

⁵ We also tested stationarity with the Phillips–Perron (1988) test, and find similar results rejecting the nonstationarity hypothesis.

⁶ The number of lags was chosen on the basis of both the Akaike Information Criterion (AIC) and the Schwarz criterion.