Trading networks

We model and measure patterns of order execution in order driven markets. We model these patterns as trading networks - graphs consisting of traders (nodes) and pairwise buy-sell connections among them (edges) that occured within a period of time. We con- struct a time series of over 12,000 trading networks using audit trail, transaction-level data for the September 2009 E-mini S&P 500 futures contract. We find that network metrics that quantify patterns of order execution are highly contemporaneously correlated with returns, volatility, volume, duration, and market liquidity. We also find that net- work metrics strongly Granger-cause intertrade duration and trading volume, but are not Granger-caused by them. We conclude that network metrics can be used to characterize the interaction between latent order submission strategies and transaction prices, trading volume, intertrade duration, and liquidity.