High-Frequency Trading In The Bund Futures Market
Goethe University Frankfurt
In this work, I study the impact of high-frequency trading (HFT) on price discovery and volatility in the Bund futures market. Using a new dataset based on microseconds, the focus of the study is on the reaction of high-frequency traders (HFTs) to major macroeconomic news events. I show that through their fast and strong reaction to news, HFTs contribute more to price discovery compared to Non-HFTs, but also add a higher share to noise than to permanent volatility. Moreover, I find evidence that HFTs tend to supply less liquidity after an unexpected rise in market volatility and prior to upcoming macroeconomic news events. These findings suggest that in times of high market stress, HFT behavior may exacerbate intraday price volatility and amplify the risk of market disruptions in fixed income markets.
High-Frequency Trading In The Bund Futures Market – Introduction
High-frequency trading (HFT) has become a dominant tool in many liquid financial markets, such as equity and futures markets. Being considered a subcategory of Algorithmic trading (AT), high-frequency traders (HFTs) typically invest in technological infrastructure that allows them to process a variety of trading signals and send orders to marketplaces in a fraction of a second. In 2012, HFT accounted for about 40% of the activity on European and over 50% of US equity markets (see, e.g., TABB Group (2012)).
Their speed advantage enables HFTs to react to new information faster than other traders (see, e.g., Foucault, Hombert, and Rosu (2015), Zhang (2013)). However, a common point of criticism is that HFTs overreact to important news announcements and thereby generate excessive short-term volatility. Such episodes of extreme volatility can even induce so-called “ash events”, characterized by a rapid and strong fall or rise in asset prices followed by a reversal in a matter of minutes or seconds. Especially since the US equity ash crash on 6 May 2010, HFT has attracted the attention of academics, regulators and policy makers around the globe. Recently, similar ash events could also be observed in fixed income markets. In particular, the Flash rally on 15 October 2014 in US treasury markets and the Bund tantrum in the German government bond market in April/ May 2015 have initiated discussions about the resiliency of fixed income markets and the impact of new market participants like HFTs (BIS (2016a)). Moreover, since the implementation of unconventional policy measures, such as the broad-based asset purchase programs in Europe or the US, many observers fear an increased vulnerability of fixed income markets. Hence, the question whether HFTs’ trading strategies pose an additional risk to market functioning is especially relevant nowadays. Against this background, different regulatory institutions and central banks have investigated the evolution and potential causes of events like the US Flash rally ((Joint Staff Report (2015)) and the Bund tantrum (BIS (2016a), BIS (2016b)) in detail. The debate has mainly focused on the impact HFT has on liquidity, price discovery and excessive volatility. To the extent that HFT affects these features of market quality, it also has implications for financial stability (Benos and Sagade (2013)). Therefore, policy makers and regulators, especially in the US but lately also in Europe, have been discussing whether stricter regulation should apply to the HFT practice. However, the market impact HFT and AT activities have during those events remains difficult to establish and also depends on the strategies followed by the respective HFT firm (Joint Staff Report (2015)).
Based on a very recent and unique dataset from Deutsche Borse on a microsecond frequency, I study the role of HFTs in the price discovery process in the Bund futures market. Since macroeconomic news announcements usually have a strong impact on bond prices (see, e.g., Altavilla, Giannone, and Modugno (2014)), I investigate whether HFT aids or hinders the market’s incorporation of such news into asset prices. If HFTs trade on information faster than other traders, they should contribute to price discovery by accelerating the speed at which new information is impounded into prices. However, a common point of critique is that the fast reaction by HFTs to news may also be associated with prices overshooting, leading to noise that cannot be related to information about fundamentals (Benos and Sagade (2013)). Using the methodology by Hasbrouck (1991b) to decompose the price in its transitory and permanent components allows me to study this question in detail. Another recent point of criticism is that HFTs supply liquidity when volatility is low but withdraw from markets when volatility rises (ESMA (2011), ASIC (2012)). This might pose an additional risk for financial stability because if a market shock induces HFTs to suspend their liquidity provision, the shock may be amplified. Hence, in the final part of this work, I study the behavior of liquidity-providing HFTs considering two different types of volatility. I distinguish between phases of unexpected increases in market volatility due to higher risk perception by market participants and phases of anticipated volatility that follow from macroeconomic news releases in order to find out whether HFT behavior depends on the nature of volatility.
This paper represents the first study on the impact of HFT on price discovery and volatility in European bond markets. The few existing fixed income studies are limited to US Treasury markets (Jiang, Lo, and Valente (2013), Liu, Lo, Nguyen, and Valente (2014)) but suffer from less granular data. The present study is based on a new Bund futures dataset from 2013 to 2014 which is in contrast to existing empirical work that mostly rely on US stock market data (e.g., Gao and Mizrach (2013), Zhang (2013) Brogaard, Hendershott, and Riordan (2013)) with some exceptions for FX data (Chaboud, Chiquoine, Hjalmarsson, and Vega (2012)) that are already several years old. Although there might still be some useful conclusions to be drawn on the effect of HFT using older data, HFT activity has developed strongly in the last years. Particularly the dimensions of speed and latency have changed substantially over the last decade, supporting the need for data timeliness. A unique HFT flag assigned by Deutsche Borse makes it possible to distinguish the trading behavior of HFTs from that of other market participants, which is not available in most other studies. The existing literature often relies on proxies for AT or HFT based on certain data characteristics like reaction times typical for HFT (e.g., Jiang et al. (2013), Hendershott, Jones, and Menkveld (2011), Zhang (2010)). Moreover, the data are based on microseconds, which enables a more granular and deep analysis than in other empirical HFT studies that are at most based on milliseconds (e.g., Brogaard et al. (2013), Gao and Mizrach (2013), Zhang (2013)). Given the improvement in trading speed over the last years, the extremely high frequency of the Bund futures data is of particular value and helps to understand the reactions of HFTs to major macroeconomic news on a tick-by-tick basis. Since macroeconomic news has been proven to be a crucial driver of bond price movements on a lower data frequency (see, e.g., Altavilla et al. (2014), Jones, Lamont, and Lumsdaine (1998)), the analysis of the HFT price impact around important news releases reveals new insights about HFT strategies in these highly volatile market episodes. Also, to my knowledge the contribution of HFT trading to the noise- and information-related determinants of the variance has not been studied around the publication of macroeconomic news before. Moreover, investigating the liquidity supply by HFTs in two different volatility environments, i.e. after macroeconomic news releases and in times of higher risk aversion, allows new insights into HFT market making behavior. While most existing studies rely on trading volume data to measure liquidity provision (Brogaard et al. (2013), Chaboud et al. (2012)), this study focuses on order deletion activity, which is often associated with HFT behavior (e.g., ESMA (2014)). The major advantage with this approach is that order deletions can be directly related to volatility, whereas any causal link between trading volume and volatility is hard to establish as the two variables cannot be considered separately from each other.
See full PDF below.