CMStatistics 2022: Start Registration
View Submission - CFE
Title: Measuring the information content of trades in a time-varying setting Authors:  Francesco Campigli - Scuola Normale Superiore (Italy) [presenting]
Fabrizio Lillo - Scuola Normale Superiore (Italy)
Giacomo Bormetti - University of Bologna (Italy)
Abstract: The estimation of the market impact of trades on prices is important for measuring the information content of trades, for optimal execution, and for transaction cost analysis. Originally, a Structural-VAR (S-VAR) model was proposed to be used, but more recent literature has highlighted some pitfalls of this approach. S-VAR models are misspecified, and the estimates can be contradictory when the permanent impact function has a nonlinear relationship with the trade sign. They are not flexible to parsimoniously exploit the long memory of the order flow variable. The instantaneous impact, which is a measure of market liquidity, is constant, while liquidity is known to be highly fluctuating. A nonlinear modified score-driven version of the original model is proposed. To measure the long-term effect of trade on prices, we compute the asymptotic cumulative impulse response function using Monte Carlo simulations. The analysis indicates that the trade information content varies and it is conditional on past trades and prices. Real-time knowledge of the permanent impact is important for transaction cost analysis. We derive an expression for the permanent impact from the estimated parameters. Simulations and empirical applications suggest that the approach provides reliable estimates of the cost of an optimal execution.