Market structure or traders' behavior? A multi agent model to assess flash crash phenomena and their regulation

Abstract

Analyzing the market reaction to a flash crash caused by an operational shock with different market participants

Acknowledgments

In memory of Russell Todres. We are indebted to him for his editing support. But above all, we are grateful to have known such a radiant personality. The authors thank the anonymous reviewer for his/her insightful comments and suggestions. We are grateful to Labex Refi for access to the Bloomberg data base. The authors thank seminar participants and particularly A.Calamia, R.Douady and D. Ladley at the Labex Refi Scientific Day (2014) as well as participants at the 5th International Conference of the Financial Engineering and Banking Society (2015) and the 32nd Symposium in Money Banking and Finance (2015) for stimulating discussions and comments on earlier versions of this paper. We are grateful to Carol Alexander and Dobrislav Dobrev for fruitful discussion at the International Workshop on Financial Markets and Nonlinear Dynamics (FMND, 2017) held in Paris, which contributed to greatly improving our model and results. We also thank the other participants of this workshop for their comments and suggestions. We are especially grateful to Thomas Lux for useful discussions on the profitability of crash events during the 22nd Annual Workshop on Economic Science with Heterogeneous Interacting Agents held in Milan in June 2017.

Disclosure statement

No potential conflict of interest was reported by the authors.

Notes

1 http://atom.univ-lille1.fr/

2 Some high-frequency trading firms are known to send thousands of orders trying to determine the price institutional or retail investor are offering, then cancel 90% of them a fraction of a second later. Practices such as quote stuffing or momentum ignition are not considered in this paper. We assume that HFTs cancel their orders to trade on short-lived momentum or fundamental information (Foucault et al. 2016).

3 Agents incur adverse selection risk if their liquidity pending in the order book is taken by other, better informed traders. Thus, orders by HFTs are the first to be involved in the transactions initiated by big-volume market order. A deep analysis of the transactions contracted over a 10 second periods after the introduction of an aggressive market order shows that on average, trend followers respectively initiate sell transactions and buy transactions

4 This crash is less remarkable than the flash crash of 6 May 2010, yet it was a day when the U.S. stock market crashed after the Associated Press's Twitter account was hacked and the message about the explosions at the White House was posted. The market recovered when the traders realized the tweet was false.

Additional information

Funding

This work was achieved through the Laboratory of Excellence on Financial Regulation (Labex ReFi) supported by the Center for Research and Higher Education (PRES) of the heSam Université [grant number ANR-10-LABX-0095]. It benefitted from French government support managed by the Agence Nationale de Recherche (ANR) within the ‘Paris-New Worlds’ Investments in the Future project [grant number ANR-11-IDEX-0006-02].