Simulation of Execution Costs in the Global Institutional Spot FX Market

Abstract

We discuss execution costs in the global spot FX market where only limit orders are permitted. In this work we expand our approach for simulations of maker loss (Schmidt [2008]) to large orders that might notably affect price dynamics if they were submitted in real life. We discern three types of orders: (1) Maker orders: bid (offer) orders submitted at best bid (offer) price until they are completely filled. (2) Auto - match orders: bid (offer) orders submitted at best offer (bid) price but their unfilled remainders are submitted as maker orders. (3) Taker orders: buy (sell) orders submitted at sufficiently high (low) price to be immediately filled. We used the EBS/ICAP market data for EUR/USD in February 2009 for calibrating our simulation model and estimating maker loss. Besides sequential runs upon historical market data, we used also several resampling techniques: bootstrap, Monte Carlo Markov Chain, and random entry. We found that the latter method is preferable for our simulations. Our main conclusion in the early work for one-million orders (the minimal order size in the EBS market) holds also for multi-million orders: expected maker loss is lower than the taker loss.