Events and Meetings of Italian Statistical Society, Advances in Latent Variables - Methods, Models and Applications

Font Size: 
High-frequency modeling for VWAP based trading strategies: a Generalized Autoregressive Score approach
Francesco Calvori, Fabrizio Cipollini, Giampiero M. Gallo

Last modified: 2013-06-16


Market activity is proxied in the literature in a variety of ways (volatility, volumes, number of trades, etc.) as a way to address the latent variable nature of the information flow. In this framework, the Volume Weighted Average Price(VWAP) is a benchmark measure based on volumes and prices at intra-daily intervals frequently used to evaluate a trader’s performance. Under suitable assumptions, splitting a daily order according to ex ante volume predictions is a good strategy to replicate the VWAP. We propose a novel Generalized Autoregressive Score (GAS)based model for predicting volume shares (by intra-daily interval relative to the daily total), inspired by the empirical regularities (intra-daily periodicity patterns, residual serial dependence) of the observed series. Such an approach bypasses the need for modeling volumes directly which may be affected by local trends. An application to 6 stock tickers from the NYSE illustrates the effectiveness of the proposed model.

Full Text: PDF