High-frequency quoting and trading (HFQ) has become a global phenomenon. It’s based on reducing the lag time – known as latency – between order submission and execution or cancellation so that order outcome is reported almost instantaneously. A number of market mishaps, though, have drawn HFQ to the attention of regulators.[1] Whether they have anything to worry about is a question we seek to answer in our recent paper, available here.

The  paper focuses on the introduction of the Arrowhead high-speed-trading platform by the Tokyo Stock Exchange (TSE) in January of 2010. The platform reduced latency from six seconds to two milliseconds and made HFQ possible on the TSE for the first time. By April 2011, HFQ market share soared from non-existent to as much as 36 percent (TSE Annual Report, 2011). The quote-to-trade ratio more than doubled after the launch of Arrowhead. The sudden and exogenous introduction of HFQ on the TSE provides a clean experiment and a fertile research ground to directly assess the impact of HFQ in U.S. markets, where even the gradual increase in HFQ’s market share is confounded by other events.

A number of researchers have investigated the impact of HFQ on market quality measures such as liquidity and cost of trading, but there has been less focus on how HFQ affects systemic risk. Although HFQ can increase volatility, it is not clear whether HFQ affects the severity of losses from the type of episodic illiquidity observed during the Flash Crash of May 2010 in the U.S. (Easley, Lopez de Prado, and O’Hara, 2011) by increasing systemic risks.

In our research, we examine stressful market conditions when systemic risks are most relevant. How did the reduced latency of Arrowhead affect systematic trading risks such as shock-propagation risk, quote-stuffing risk, Limit Order Book (LOB) attrition risk, and tail risk?[2] We also develop systemic risk measures based on correlations, order flow, and Adrian and Brunnermeier’s (2011) CoVaR. ∆CoVaR measures the component of systemic risk that moves with the distress of a particular firm.

Our analysis goes beyond the traditional measures of market quality. That’s important, because HFQ and low latency change the nature of observable data when liquidity suppliers use short-lived order and mixed-equilibrium strategies by injecting endogenous noise into the order flow to hide their information. To quantify the true state of the LOB we compute measures such as the LOB slope and the cost of immediacy (COI), which tend to be more stable than National Best Bid and Offer (NBBO) spreads, and also are more relevant for liquidity demanders with order sizes larger than volume supplied by the best quotes.[3] These LOB measures are particularly relevant in fast-paced markets where orders frequently walk up or down the LOB.[4] The change in the LOB slope (∆SLOPE) measures the resiliency (i.e., the rate at which the LOB refills) of the full LOB while COI comprehensively measures both the spread between bid and ask prices and depth of the LOB (Jain and Jiang, 2014). With our advanced measures, we also study the distinct effects of low latency on high-frequency quoting versus high-frequency trading. Whereas the systemic risks associated with high-frequency trading result from aggressive demand for liquidity, the systemic risks of high-frequency quoting emanate from the cancellation or absence of quotes from liquidity suppliers.

The novel findings relate to the risks of HFQ. We show that HFQ made possible by Arrowhead amplifies systemic risk by increasing shock-propagation risk, quote-stuffing risk, LOB attrition risk, and tail risk. The incidence of extraordinary market-wide volatility in large groups of stocks, such as occurred during the Flash Crash on May 6, 2010, in the U.S., is of significant regulatory interest. Regulatory responses to systemically risky events such as flash crashes include a single stock circuit breaker or limits on the movement up or down of a single stock, but they do not explicitly focus on measures of systemic or correlated risks. We find that Arrowhead increases the exposure to systemic risk even more during tail-risk events, which can potentially lead to a highly destabilizing market situation. One implication of our finding is that low-latency markets may benefit from safety features such as kill switches, circuit breakers, and rigorous software testing, which prevent the proliferation of risks from one stock to another and to the trading system at large. By focusing on the extreme negative market condition we are able to highlight that, whereas prior findings about HFQ’s contribution to market quality discussed at the beginning of this piece are applicable in most normal periods, special attention is required to deal with the adverse impact of HFQ during tail events.

We also investigate whether Arrowhead’s increased HFQ affects market quality measures in Japan similar to HFQ’s impact on market quality reported in studies using U.S. data. These effects include increased trading speed, increased volume and number of trades, and increased LOB liquidity. The similarity of these traditional market-quality effects of HFQ in Japan and the U.S. indicates that our results concerning HFQ’s impact on systematic trading risks are also applicable to other markets.

Increased systemic risk from correlated trading and quoting in a high-speed environment needs special attention and tools. Our study provides a framework for analyzing the systemic microstructure risks of HFQ that can be used in future research to assess the effectiveness of the new regulatory framework, including the single stock circuit breaker or limit up-limit down rules that were a response to market-wide or systemic extraordinary volatility events. In summary, high speed trading significantly increases systemic risks in ways similar to a multi-car accident on a highway. It calls for robust safety measures at the levels of both individual traders and trading systems.

ENDNOTES

[1] In Germany the High Frequency Trading Act entered into force in 2012-2013. (http://regulation.fidessa.com/ataglance/german-high-frequency-trading-act/).  France introduced a high-frequency trading tax in 2012. Italy followed suit with an HFQ tax in September 2013. There is a broader financial transaction tax being considered by eleven Eurozone countries (http://marketsmedia.com/HFQ-makes-last-stand-in-europe-as-ftt-and-mifid-ii-edge-closer/). In the US regulations include the SEC market access rule 15c3-5 (http://www.sec.gov/rules/final/2011/34-64748fr.pdf) and limit up-limit down rule (https://www.nasdaqtrader.com/content/MarketRegulation/LULD_FAQ.pdf). The TSE does not have any such regulations.

[2] Shock propagation risk is the risk of a system wide failure due to failure of a given stock. Quote stuffing risk is the risk of market congestion due to submission of an unwieldy number of orders to the limit order book (LOB). LOB attrition risk is the risk of a quick liquidity dry up. Tail risk captures the market reaction during the extreme negative market conditions.

[3] The LOB slope measure is defined as the weighted average of the change in quantity supplied in the LOB per unit change in the price. COI captures the fact that liquidity demanders incur progressively higher cumulative costs as the available depth at the top of the LOB in fast markets becomes insufficient to fully execute the order. For the COI transaction cost measure, weighted average LOB information for executions at multiple price points resulting from walking up or down the book is used instead of top of the LOB bid-ask spreads. The calculation of both measures is described in detail below.

[4] A Limit Order Book (LOB) is a record of unexecuted limit orders sorted based on price and time priority. A market buy (sell) order with an order size of greater than the volume supplied by the top of the LOB (best quotes), walks up (down) the LOB.

This post comes to us from Pankaj Jain and Thomas McInish, professors at the University of Memphis – Fogelman College of Business and Economics, and Professor Pawan Jain at the University of Wyoming – College of Business. It is based on their recent paper, “Does High-Frequency Trading Increase Systemic Risk?”, available here.