Alexander Barzykin : Dealing With Uncertainty of Execution in Delocalized High-Frequency Liquidity Landscape

In the high-frequency world of electronic markets what you see is not necessarily what you can get. This is particularly true in foreign exchange where the specifics of market operation naturally implies uncertainty of execution. While the uncertainty of a passive fill is easily understandable and often related to the nature of spread, the uncertainty of aggressive crossing is sometimes overlooked. There has been a growing interest in exploring the effect of market mechanics on price formation and the choices that market participants have to make when exercising their liquidity requirements. The role of last look when trading in and/or streaming into the aggregator has recently been emphasized. Last look creates uncertainty of execution for a liquidity consumer who is already happy to cross spread but by no means is it the only source. Fragmentation and geographical delocalisation of FX introduces latency as an inherent and significant source of uncertainty. Multiple order book events can take place while the information travels from one liquidity centre to the other. Additional latency components arise due to throttling of certain datafeeds. The liquidity consumer sees the order book snapshot which is already old and may have changed, and will change even more during the time until the order reaches the liquidity pool for execution. As a result, the liquidity consumer is faced with a probabilistic framework when optimising allocation of market orders. The optimisation procedure has to be online in order to deal with ever changing market conditions, take into account the specifics and geographical location of different liquidity pools as well as the impact of allocation strategy on the forthcoming order flow. After presenting relevant stylised facts, the formulation of this optimization problem from practitioner’s viewpoint will be discussed with outline of potential implementation.

Pierre Collin-Dufresne: Liquidity, Volume, and Volatility

We examine the relation between liquidity, volume, and volatility using a comprehensive sample of U.S. stocks in the post-decimalization period. For large stocks, effective spread and volume are positively related in the time series even after controlling for volatility, contrary to most theoretical predictions. This relation is mostly driven by the systematic component of volume. In contrast, for small stocks the evidence matches the predictions of standard adverse selection models. We show that the volatility of order imbalances can reconcile our puzzling finding with standard intuition. Order imbalance volatility is strongly associated with spreads both in the time series and cross-section and makes the relation between spread, volume, and volatility close to what is predicted by invariance theories. We develop a continuous-time inventory model to explain our empirical findings. Evidence from intraday patterns and other liquidity measures (price impact and depth) support our interpretation. Joint work with Vincent Bogousslavsky.

Available online at https://papers.ssrn.com/sol3/papers.cfm?abstract_id=3336171

Albina Danilova: Risk Aversion of Market Makers and Asymmetric Information

In this talk I will explore differences and similarities of informed trading in the markets with different types of liquidity providers. In particular, I will present two types of markets: the one populated by perfectly competitive liquidity providers, who take the asset supply as given, and the one populated by market makers, who compete for total demand by posting price schedules. A particular focus will be on the impact of risk aversion of market makers/liquidity providers on the resulting equilibrium price process.

I will demonstrate that, although the resulting equilibrium price is different, the markets are quite similar qualitatively: the demand is mean reverting and the price process exhibits price reversal in both markets. Furthermore, in both markets an increase in risk aversion leads to lower market depth, less efficient prices, stronger price reversal and slower convergence to fundamental value. The endogenous value of private information, however, is non-monotonic in risk aversion and differs in two markets. Joint work with Umut Cetin

Available online at https://projecteuclid.org/euclid.aoap/1472745450

Olivie Guéant : Neural Networks for Optimal Control in Finance: Three Different Stories

Neural networks are now ubiquitous in science. They seem indeed to be « the » solution to a lot of problems from supervised learning, to pattern recognition, to optimal control, etc. In this talk, we present 3 research projects carried out through the Research Initiative HSBC-College de France that involved neural networks. They address questions related to execution and hedging in equity and equity derivatives, including the case of corporate derivatives for which neural networks have a lot to offer.

Nikolaus Hautsch : Limits to Arbitrage in Markets with Stochastic Settlement Latency

Distributed ledger technologies replace trusted intermediaries with time-consuming consensus protocols to record the transfer of ownership. This settlement latency exposes arbitrageurs to price risk and imposes limits to arbitrage. We derive theoretical arbitrage boundaries that increase with expected latency, latency uncertainty, volatility and risk aversion. Using Bitcoin orderbook and network data, we calibrate arbitrage boundaries of on average 90 basis points, covering 81% of the observed cross-market price differences. Consistent with our theory, periods with high latency risk exhibit large price differences, while asset flows chase arbitrage opportunities. Decentralized settlement thus introduces a non-trivial friction that affects market efficiency. Joint work with Christoph Scheuch and Stefan Voigt

Available online at https://arxiv.org/abs/1812.00595

Sergey Nadtochiy: Explaining the Nonlinear Price Impact

I will present a simple model of market microstructure that explains the nonlinear price impact. In this model, the local relationship between the order flow and the fundamental price (i.e. the local price impact) is linear, which makes the model dynamically consistent. Nevertheless, the expected impact on midprice from a large sequence of co-directional trades is nonlinear and asymptotically concave. In addition to theoretical results, I will present empirical evidence that supports the main conclusions of the model.

Roel Oomen: The Practice of Electronic Trading in OTC Markets

In this talk I will review some recent work on the topic of electronic execution in OTC markets, go over some case studies and practical illustrations, and make a few suggestions for future research.

Emiliano Pagnotta: Becker Meets Kyle: Inside Insider Trading

How do illegal insiders trade on private information? Do they internalize legal risk? Using hand-collected data on insiders prosecuted by the SEC, we find that, consistent with Kyle (1985), insiders manage trade size and timing according to market conditions and the value of information. Gender, age, and profession play a lesser role. Various shocks to penalties and likelihood of prosecution show that insiders internalize legal risk by moderating aggressiveness, providing support to regulators’ deterrence ability. Consistent with Becker (1968), following positive shocks to expected penalties, insiders concentrate on fewer signals of higher value. Thus, enforcement actions could hamper price informativeness. Joint work with Marcin T. Kacperczyk

Available online at: https://papers.ssrn.com/sol3/papers.cfm?abstract_id=3142006

Mikko Pakkanen: Modelling State-Dependent Endogeneity in High-Frequency Limit Order Book Wata

Hawkes processes have been successfully applied to the modelling of high-frequency financial data during the past two decades. They are able to capture the endogeneous dependence of future order flow on the past order flow through their self- and cross-exciting behaviour. Empirical data suggest, however, that future order flow is also influenced by the limit order book (LOB) state, for example through spread or volume imbalance measures, which is not captured by models built on ordinary Hawkes processes. In this talk, I introduce a novel state-dependent extension of the Hawkes process, where the Hawkes point process is coupled to a jump process that describes the LOB state. The state process in the model modulates the excitation behaviour of the Hawkes process, while state transitions occur as prompted by events in the Hawkes process, giving rise to two-way feedback. Despite its complex feedback structure, the new model is in practice not harder to estimate than an ordinary multidimensional Hawkes process. Finally, I demonstrate empirically, by applying the model to Nasdaq LOB data, that high-frequency market endogeneity is indeed highly state-dependent, being more pronounced when the LOB is “out of equilibrium”. Joint work with Maxime Morariu-Patrichi.

Available online at: https://arxiv.org/abs/1809.08060

Mathieu Rosenbaum: Some Recent Advances in Market Design

In this talk, we first consider an exchange wishing to establish a suitable make-take fees policy in order to attract transactions on its venues. We derive the equations defining the optimal contract to be set between the market maker and the exchange. We solve them explicitly when the market maker can access a lit pool only. When a dark pool is available, we use deep reinforcement learning algorithms to approximate efficiently the optimal controls and optimal incentives to be provided by the exchange.

We are then interested in comparing auction markets and limit order book markets. We build a methodology enabling us to derive the optimal auction frequency for a given asset when the criterion is the quality the price formation process. We compute the optimal duration of the auctions for 77 stocks traded on Euronext and compare the efficiency of price formation process under this optimal value to the case of a continuous limit order book. Continuous limit order books are found to be usually sub-optimal. However, in terms of our metric, they only moderately impair the price formation process. Joint work with Bastien Baldacci, Paul Jusselin, Iuliia Manziuk and Thibaut Mastrolia.

Pamela Saliba: From Glosten-Milgrom to the Whole Limit Order Book and Applications to Financial Regulation

We build an agent-based model for the order book with three types of market participants: informed trader, noise trader and competitive market makers. Using a Glosten-Milgrom like approach, we are able to deduce the whole limit order book (bid-ask spread and volume available at each price) from the interactions between the different agents. More precisely, we obtain a link between efficient price dynamic, proportion of trades due to the noise trader, traded volume, bid-ask spread and equilibrium limit order book state. With this model, we provide a relevant tool for regulators and market platforms. We show for example that it allows us to forecast consequences of a tick size change on the microstructure of an asset. It also enables us to value quantitatively the queue position of a limit order in the book. Joint work with Weibing Huang and Mathieu Rosenbaum

Available online at: https://papers.ssrn.com/sol3/papers.cfm?abstract_id=3343779

Yajun Wang: Trading in Crowded Markets

We study crowded markets using a symmetric model of trading among strategic informed traders. We model market crowdedness by assuming that traders may have incorrect beliefs about the number of traders following similar strategies; this distorts traders' inference, strategies, and market prices. If traders underestimate the degree of crowdedness, then markets are more liquid, both permanent and temporary market depths tend to be higher, traders take larger positions and trade more aggressively on short-run profit opportunities. In contrast, as soon as traders start overestimating the degree of crowdedness, they believe markets to be less liquid and trade more cautiously on both their private information and supplying liquidity to others. Fears of crowded markets may also lead to ``illusion of liquidity'' when the actual endogenous market depth is even lower than what traders believe it to be. Crowded markets are fragile, because flash crashes, triggered whenever some traders liquidate large positions at fire-sale rates, tend to be more pronounced and price recoveries tend to be slower. Joint work with Stephan Gorban, and Anna A. Obizhaeva

Available online at https://papers.ssrn.com/sol3/papers.cfm?abstract_id=3152743

Stefan Zohren: Deep Learning Applied to Limit Order Book Data

After a brief introduction to deep learning architectures and their usage in finance, a deep learning model for limit order book data is proposed. Through usage of both convolutional as well as recurrent network layers, the model is able to automatically extract features from sequences of limit order books which can be used to predict market moves within short time intervals. Interestingly, those features are universal in the sense that the model can be used for different assets which were not part of the original training data. This is reminiscent of transfer learning in image recognition. We conclude by showing several extensions of this work.