Bence Toth

Increasing market efficiency: Evolution of cross-correlations of stock returns

We analyse the temporal changes in the cross-correlations of returns on the New York Stock Exchange. We show that lead–lag relationships between daily returns of stocks vanished in less than 20 years. We have found that even for high-frequency data the asymmetry of time-dependent cross-correlation functions has a decreasing tendency, the position of their peaks is shifted towards the origin while these peaks become sharper and higher, resulting in a diminution of the Epps effect. All these findings indicate that the market becomes increasingly efficient.

Why Is Equity Order Flow so Persistent

Order flow in equity markets is remarkably persistent in the sense that order signs (to buy or sell) are positively autocorrelated out to time lags of tens of thousands of orders, corresponding to many days. Two possible explanations are herding, corresponding to positive correlation in the behavior of different investors, or order splitting, corresponding to positive autocorrelation in the behavior of single investors. We investigate this using order flow data from the London Stock Exchange for which we have membership identifiers. By formulating models for herding and order splitting, as well as models for brokerage choice, we are able to overcome the distortion introduced by brokerage. On timescales of less than a few hours the persistence of order flow is overwhelmingly due to splitting rather than herding. We also study the properties of brokerage order flow and show that it is remarkably consistent both cross-sectionally and longitudinally.

How Does the Market React to Your Order Flow

We present an empirical study of the intertwined behaviour of members in a financial market. Exploiting a database where the broker that initiates an order book event can be identified, we decompose the correlation and response functions into contributions coming from different market participants and study how their behaviour is interconnected. We find evidence for the following. (1) Brokers are very heterogeneous in liquidity provision—some appear to be primarily liquidity providers while others are primarily liquidity takers. (2) The behaviour of brokers is strongly conditioned on the actions of other brokers. In contrast, brokers are only weakly influenced by the impact of their own previous orders. (3) The total impact of market orders is the result of a subtle compensation between the same broker pushing the price in one direction and the liquidity provision of other brokers pushing it in the opposite direction. These results enforce the picture of market dynamics being the result of the competition between heterogeneous participants, interacting to form a complex market ecology.

The Epps effect revisited

We analyse the dependence of stock return cross-correlations on the data sampling frequency, known as the Epps effect: for high-resolution data the cross-correlations are significantly smaller than their asymptotic value as observed for daily data. The former description implies that a changing trading frequency should alter the characteristic time of the phenomenon. This is not true for empirical data: the Epps curves do not scale with market activity. The latter result indicates that the time scale of the phenomenon is related to the reaction time of market participants (this we denote as the human time scale), independent of market activity. In this paper we give a new description of the Epps effect through the decomposition of cross-correlations. After testing our method on a model of generated random walk price changes we justify our analytical results by fitting the Epps curves of real-world data.

Studies of the limit order book around large price changes

AbstractWe study the dynamics of the limit order book of liquid stocks after experiencing large intra-day price changes. In the data we find large variations in several microscopical measures, e.g., the volatility the bid-ask spread, the bid-ask imbalance, the number of queuing limit orders, the activity (number and volume) of limit orders placed and canceled, etc. The relaxation of the quantities is generally very slow that can be described by a power law of exponent ≈ 0.4. We introduce a numerical model in order to understand the empirical results better. We find that with a zero intelligence deposition model of the order flow the empirical results can be reproduced qualitatively. This suggests that the slow relaxations might not be results of agents’ strategic behaviour. Studying the difference between the exponents found empirically and numerically helps us to better identify the role of strategic behaviour in the phenomena.

Anomalous impact in reaction-diffusion financial models.

We generalize the reaction-diffusion model A+B→0 in order to study the impact of an excess of A (or B) at the reaction front. We provide an exact solution of the model, which shows that the linear response breaks down: the average displacement of the reaction front grows as the square root of the imbalance. We argue that this model provides a highly simplified but generic framework to understand the square-root impact of large orders in financial markets.

On the origin of the Epps effect

The Epps effect, the decrease of correlations between stock returns for short time windows, was traced back to the trading asynchronicity and to the occasional lead-lag relation between the prices. We study pairs of stocks where the latter is negligible and confirm the importance of asynchronicity but point out that alone these aspects are insufficient to give account for the whole effect.

Accurate estimator of correlations between asynchronous signals

The estimation of the correlation between time series is often hampered by the asynchronicity of the signals. Cumulating data within a time window suppresses this source of noise but weakens the statistics. We present a method to estimate correlations without applying long time windows. We decompose the correlations of data cumulated over a long window using decay of lagged correlations as calculated from short window data. This increases the accuracy of the estimated correlation significantly and decreases the necessary effort of calculations both in real and computer experiments.

Linear Models for the Impact of Order Flow on Prices I. Propagators: Transient vs. History Dependent Impact

Market impact is a key concept in the study of financial markets and several models have been proposed in the literature so far. The Transient Impact Model (TIM) posits that the price at high frequency time scales is a linear combination of the signs of the past executed market orders, weighted by a so-called propagator function. An alternative description -- the History Dependent Impact Model (HDIM) -- assumes that the deviation between the realised order sign and its expected level impacts the price linearly and permanently. The two models, however, should be extended since prices are a priori influenced not only by the past order flow, but also by the past realisation of returns themselves. In this paper, we propose a two-event framework, where price-changing and non price-changing events are considered separately. Two-event propagator models provide a remarkable improvement of the description of the market impact, especially for large tick stocks, where the events of price changes are very rare and very informative. Specifically the extended approach captures the excess anti-correlation between past returns and subsequent order flow which is missing in one-event models. Our results document the superior performances of the HDIMs even though only in minor relative terms compared to TIMs. This is somewhat surprising, because HDIMs are well grounded theoretically, while TIMs are, strictly speaking, inconsistent.

Modeling the Epps effect of cross-correlations in asset prices

We review the decomposition method of stock return cross-correlations, presented previously for studying the dependence of the correlation coefficient on the resolution of data (Epps effect). Through a toy model of random walk/Brownian motion and memoryless renewal process (i.e. Poisson point process) of observation times we show that in case of analytical treatability, by decomposing the correlations we get the exact result for the frequency dependence. We also demonstrate that our approach produces reasonable fitting of the dependence of correlations on the data resolution in case of empirical data. Our results indicate that the Epps phenomenon is a product of the finite time decay of lagged correlations of high resolution data, which does not scale with activity. The characteristic time is due to a human time scale, the time needed to react to news.