SPREAD REVISITED.
By Yossi Brandes, Managing Director at ITG Analytics.
How informative are spreads for market participants who engage in FX trading? Can proper spread monitoring and analysis further enhance the investment and trading process? If so, how exactly? Can this relatively “old”, sometimes overlooked metric, provide insights into important factors like the likelihood of execution and the cost of liquidity in this OTC market? Can spread analysis support real trading decisions (for example, trading now vs later)?
We find that a relatively simple but enhanced spread metric, size adjusted spread, can help answer these questions. This simple metric can help measure the empirical cost of an immediate execution (“instantaneous execution”) and, once calculated over time, can reveal interesting facts about the quality of liquidity for various currency pairs. Size adjusted spread can lay the foundations for a framework that can simulate the trade-off between an immediate execution and execution over a certain time horizon.
The electronification and fragmentation of the FX market coupled with a best execution theme and a regulatory push have created a need for more transparency.
In addition to that, execution arrangements for all asset classes are being overhauled and trading decisions are under the microscope in all investment desks and trading rooms. The FX market is no exception and there is a clear need for more colour on the market, including decision making tools that can help with investment and trading decisions.
Advancements on the electronic side and the slow but steady move from voice to electronic trading complement this shifting environment very well. FX quotes and trades can now be stored and analysed in a relatively quick manner. Spreads and other spread-type measures can be easily computed and examined.
We have the tools to do so. In what follows, we discuss our data as well as some empirical stylised facts related to our spread metric.
ITG has created a unique FX market database that processes feeds from multiple banks, ECNs and other data providers. On a daily basis around 100 million records, including tradable quotes, indicative quotes and trades, are being processed and stored. That in return, allowed us to reconstruct a synthetic order book for the FX market at every point in time.
Moreover, it allowed us to calculate the spread at the top of the book as well as the spread that a trader will have to pay by “climbing up the order book” or in other words; the cost of an immediate execution in this synthetic order book given time of day and deal size as additional parameters. Size adjusted spread is a metric designed to do just that.
The calculation starting point is the mid quote at the point in time where the trade has to be immediately executed. The actual calculation is a weighted average of the liquidity at each level and the distance from the mid quote of the different quote levels of the order book that need to be touched in order to complete the execution.
An illustration of the order book and size adjusted spread metric can be found in Figure 1
In Table 1, an immediate buy of 7 million EUR/USD at 8:05 will have an empirical cost of 0.207 pips while an immediate sell of 15.75 million EUR/USD will have an empirical cost of 0.193 pips. The implications of this metric are instant. Automation coupled with simulation can allow for multiple applications that can answer multiple questions; for example, at what time of the day is the size adjusted spread typically at its lowest level?
Generalising on size adjusted spreads and running it across different dimensions, for instance size, time of day and currency pairs, indeed allows for interesting observations on both the quality of execution as well as the possible spread cost of an immediate execution in different times of the day.
In Figure 2, one can observe low levels of sizeadjusted spread in the morning times. Figure 2 illustrates the distribution of size-adjusted spread and hence can provide the range of possible results that one can expect at different times of the day (given probabilities). While historical averages are smooth, actual real time values can vary significantly throughout the day in a fraction of a second. Monitoring these fluctuations, with historical levels as references, could help improve trade timing and reduce costs. Assuming enough computer power, the foundation and one of the building blocks for modelling cost for various currency pairs seem to be just lying there. Hence spread revisited.
Displaying the size adjusted spread over time begs for further research on the trade-off between an immediate execution and an execution over a certain time horizon. That in return requires a measure of volatility that will help assess the risk that the market will move against the residuals. The trade-off between immediate execution and a delayed one can be quantified using size adjusted spread on one hand and an FX volatility metric on the other. Figure 3 below, depicts one version of intra-day volatility metric that can be used to measure the delay costs.
Summary
Changes to the OTC markets on the technology and on the regulatory front will continue to drive significant changes in the FX market structure as well as in the way investment and trading decisions are being made. One safe bet in this environment is that data will be more readily available and its quality will improve. Creating and reconstructing an order book in an OTC market has the potential to shape future trading and investment decisions.
How? Research around the dynamics of the order book will provide more transparency into liquidity and will continue to improve the execution quality. Simple but enhanced spread metrics that make use of an order book, like size adjusted spread, provide the foundations not only for cost modelling but also for the introduction of cost effective algorithms that can balance the trade-off between an immediate execution and an execution over a specific time horizon.
Spread, like Bob Dylan’s Highway 61, electrified and revisited.
[divider_line]© BestExecution 2015
[divider_to_top]
