Skip to Main content Skip to Navigation

Empirical study, modelling and applications of multiple limits trades in limit order books

Abstract : This thesis aims at studying particular events occurring in the limit order books - the ’tradesthrough’. In the first chapter, we define trades-through as those who consume the liquidity available on several limits of the limit order book, without waiting for the best limit to be filled with new incoming limit orders. We study their empirical properties and present statistics about their liquidity, their volume, their arrival time distribution, their clustering and the spread relaxation that follows their arrival. Their market-impact is higher than the one of the other trades, even with a comparable trading volume : trades-through have a higher informational content. We present two applications linked to the lead-lag between assets/markets : to find which asset moves first, and also to measure the trades-through intensity signal in a simple trading strategy based on lead-lag. The next chapter goes into more detail about the trades-through arrival time clustering. We model their arrival time with self-excited stochastic processes (Hawkes processes). A statistical study of the calibration obtained with models based on exponential-decay kernels for the temporal impact ensures a satisfactory modelling with two independent processes, one for the bid and one for the ask. The model class under scrutiny for the calibration is well-adapted as no inhibitory effects are measured after trades-through arrival. We use those results to compute an intensity indicator based on trades-through arrival, and thus we enhance a simple trading strategy that relies on them. Finally, a non-parametric calibration of the empirical decay kernel for the temporal impact of trades-through indicates a decrease faster than exponential, and closer to a power-law. The last chapter recalls a general statistical method robust to market microstructure noise to find jumps in prices/returns time series. We generalize the empirical results already known in the literature to new financial indices and we adapt this statistical jump detection method to intraday trajectories in order to obtain the intraday proportion of detected jumps. Extreme values and biggest intraday variations of this jump proportion occurs at very specific hours of the day (14:30, 15:00 and 16:30, Paris time reference), already linked with trades-through. Using trades-through, we explain the main characteristics of the intraday proportion of detected jumps with the test using a modification in the relative importance of each jump component in the assets trajectories (the continuous moves component and the pure-jumps component).
Document type :
Complete list of metadatas
Contributor : Abes Star :  Contact
Submitted on : Tuesday, November 5, 2013 - 9:07:32 AM
Last modification on : Wednesday, December 16, 2020 - 10:12:03 AM
Long-term archiving on: : Thursday, February 6, 2014 - 4:35:07 AM


Version validated by the jury (STAR)


  • HAL Id : tel-00879857, version 1



Fabrizio Pomponio. Empirical study, modelling and applications of multiple limits trades in limit order books. Other. Ecole Centrale Paris, 2012. English. ⟨NNT : 2012ECAP0050⟩. ⟨tel-00879857⟩



Record views


Files downloads