Dissecting cross-impact on stock markets: an empirical analysis

Abstract

Abstract. The vast majority of market impact studies assess each product individually, and the interactions between the different order flows are disregarded. This strong approximation may lead to an underestimation of trading costs and possible contagion effects. Transactions in fact mediate a significant part of the correlation between different instruments. In turn, liquidity shares the sectorial structure of market correlations, which can be encoded as a set of eigenvalues and eigenvectors. We introduce a multivariate linear propagator model that successfully describes such a structure, and accounts for a significant fraction of the covariance of stock returns. We dissect the various dynamical mechanisms that contribute to the joint dynamics of assets. We also define two simplified models with substantially less parameters in order to reduce overfitting, and show that they have superior out-of-sample performance.

Statistical Properties of Market Collective Responses

“…Here, we focus on the primary price and liquidity changes, which measure the price and liquidity impact without time lags, respectively. To explore the latent factors of responses, we employ the singular value decomposition [20,[27][28][29]. Furthermore, we discuss the relation between prices and liquidity in view of the overlapping factors.…”

“…The study of cross-impacts, i.e., the response of stock prices to a market order in a different stock, emerged as an obvious challenge [15][16][17]. Large scale data analyses [18][19][20] revealed non-Markovian features in these cross-impacts and in the corresponding trade sign cross-correlators. Consequently, the Efficient Market Hypothesis cannot hold in a strict form [18].…”

A New Attempt to Identify Long-term Precursors for Financial Crisis in the Market Correlation Structures

“…[9,[16][17][18][19][20][21] and identify quasi-stationary markets states in the time evolution of the non-stationary correlation structure. The industrial sectors, which are clearly visible in the correlations [22][23][24][25][26][27][28][29] and covariances [30], and their mutual connections are thereby analyzed in a time resolved fashion. This is accomplished by applying k-means clustering [31][32][33][34][35], a machine learning algorithm, to a set of correlation matrices measured over time in a moving window.…”

Grasping Asymmetric Information in Market Impacts

“…Recently, empirical studies [7,8,14] disclosed that there are also price impacts across stocks. To avoid confusion, the impact in single stocks is named self-impact and the impact between stocks is named cross-impact.…”

“…We empirically analyzed the cross-response and the cross-impact on a physical time scale [7,8,24] with a one-second resolution. In their study, Benzaquen et al used time intervals of five minutes [14]. Both choices of time scales have advantages and limitations.…”