Return reversals and the compass rose: insights from high frequency options data

Abstract

We study the occurrence and visibility of the compass rose pattern in high frequency data from individual equity options contracts. We show that the compass rose pattern in options contracts is more complex than portrayed in prior work with other asset classes. We find that the tick/volatility ratio proposed in prior studies gives inconclusive results on the pattern's visibility. A major contribution arises from linking the compass rose pattern with return reversals, which gives new insights into the pattern's predictability. We show that return reversals are revealed as an element of the compass rose pattern and are particularly evident at higher sampling frequencies. We study the determinants of these reversals and report that return reversals are primarily associated with high transaction frequency and decrease with the presence of additional market makers. Also, the hypothesis that there is a reaction to overnight events which is reflected in prices at the market open is not supported. Reversals are less prevalent for larger firms and when trade sizes are larger.

Keywords:

• compass rose
• high frequency data
• return reversals
• equity options

JEL Classification :

• G12
• C22

Acknowledgements

The authors are very grateful for the comments and suggestions received from two anonymous referees. We also thank David McMillan for his comments as discussant at the conference.

Notes

These patterns refer to simulated scenarios. For a detailed illustration, see Szpiro (1998).

Six (22) firms have a tick size of 0.25 (0.50). No tick size changes are reported during the sample period.

The regression variables are calculated on an hourly basis. In order to show the effect of sampling frequency on the visibility and properties of the compass rose pattern, we also compute returns using 15 and 30 min intervals (following the same procedure).

Designated market makers are assigned by NYSE-Euronext to increase liquidity in specific individual equity options. Out of 28 firms in this sample, 16 are covered by DMMs.

As the returns’ calculation formula is for the nearest-to-mature contracts only, the formulae discussed refer only to different strike prices. If maturity was also allowed to float, the complexity of the formulae would increase.

A similar effect is observed in electricity prices (Batten and Hamada 2008). Lee, Mathur, and Gleason (2005) note that the same feature can emerge using a Monte Carlo simulation.

The release time for the scheduled macroeconomic announcements is 09.30 GMT, and the figure uses hourly data.

Estimation is by OLS. No autocorrelation or heteroskedasticity is detected.