Xavier Porterfield
Xavier Porterfield

Why you need independent data to differentiate between execution processes

By Xavier Porterfield CFA, Head of Research at New Change FX

All transaction costs can be more accurately defined as search costs.  In over the counter (OTC) markets, such as Foreign Exchange, the fair price is hidden.  To discover the best available price participants must deploy resources (search) to achieve a good deal. The FX market is de-centralised, or fragmented. This means the market clearing price cannot be inferred from individual market quotes. At what point can market participants determine whether they have identified the best available price and call the search off?  This ambiguity creates opacity, producing a positive incentive for market-making, but also a negative incentive, from the customers’ perspective, of discriminatory pricing which treats customers unequally.  Typically, it’s the smaller participants who get the worst deal. 

The emergence of independent mid-rates has made price opacity optional.  To paraphrase Gibson, a fair and efficient FX market is already here, it’s just not evenly distributed yet.  Let’s establish what makes a mid-rate independent and then examine how independent mid-rates are being employed in a number of different market solutions to distribute the benefits of price transparency beyond market insiders.

How the Fair Price is hidden

In OTC markets dealers quote or stream the (two-way) prices at which they will buy and sell a given currency pair for a given amount.  The mid-rate is the inferred mid-point between the bid and the offer price.  It is also the theoretical equilibrium rate at which two equal and opposite transactions with the same dealer would clear.  A consolidated mid-rate is therefore the natural clearing rate at which all market orders would clear.  Changes to the mid-rate represent changes to the equilibrium level of the market that matches supply and demand. 

A consolidated view of the FX market is difficult to achieve.  Unlike lit markets such as equities, there is no central counterparty that publishes a consolidated market tape. The view of the market is fragmented.  Prices on different venues may differ, and no one venue can claim to be the undisputed primary venue for Foreign Exchange.  The top venue perhaps sees as much as 10% of average daily volumes.  That is less than the average daily volume of the some of the largest market makers.  Each FX venue aggregates pricing from panels of bank and non-bank market makers but the prices on the venues differ from one another.  The mid-rate on one venue may be higher than on another.  

These differences reflect biases that arise from the positioning of market makers, the costs and privileges of venue participants and other factors.  These differences in prices are called skew.  Pricing from one venue will never be able to identify whether better pricing is available on another venue.  

Single source pricing, whether it is aggregated or not, can never be independent.  And without an independent mid-rate, it is not possible to identify skew. 

Inability to identify skew is not the only problem that arises from not having access to an independent mid-rate.  Two-way pricing reflects a cost of risk transfer in that market, but the mid-rate from a single counter party or platform as we have just discussed offers a very restricted view of the market.  This is further evidenced when we examine the cost of risk comparing independent mid-rates that are constructed from multiple platforms and venues, to the cost of risk calculated from a single venue sourced mid-rate. 

To illustrate, we collected aggregated data from a single ECN and compared it to the NCFX Mid-rate. The only difference between the data sets is that the NCFX data set is aggregated from multiple ECNs and venues, while the single source ECN data is created from aggregating data from a single venue. To compare the data sets we calculated realized volatilities from each data set.

The first thing to note is that just as there are absolute differences between independently aggregated consolidated mid-rates and single venue/provider mid-rates, the realized volatilities are also different. This means the behaviour of single source venues relative to broader consolidated market is not the same.  These differences can translate into significant mispricing of risk. 

Article Image
Fig 1. NCFX calculated historic volatility over 1 hour between 2-3 PM 30th June 2021
Article Image
Fig 2. USD Cost of Volatilities calculated using NCFX independent mid-rates and Single Venue Mid-rates
Article Image
Fig 3. The % differences between NCFX calculated USD cost of volatility using NCFX independent mid-rates and single source mid-rates

In Fig. 1 we show realized volatilities for four currency pairs over a one-hour period at one of the busiest times of day, from 2pm to 3pm local time in the London trading session.  To help comparability the volatilities have all been annualized. 

The differences are obvious, but it is difficult to translate annualised volatility into monetary exposure over the given period. In the case above, the historic volatility was calculated over 1 hour.  What does this translate to in USD terms? 

To better understand what volatility is showing us we translate period volatility into USD by using NCFX’s USD cost of volatility method.  We multiply the period volatility (not the annualised number) by the base amount and convert into USD. 

For USD 10 million of exposure, over the period of 1 hour the respective USD cost of Volatilities are given in Fig 2.

When volatility is low, differences between single source and independent mid-rates are still noticeably different. In USD terms this gives us the following differences.

These mispricing can be both positive and negative. The quantum of mispricing is given by the percentage differences (Right Hand Scale). 15% to 25% mispricing of risk for 3 of the top 4 most actively traded currency pairs. 
However, when markets become very active, particularly around well-known events such as month end Fixing, these differences become more pronounced.

Calculating the USD Cost of Volatility over the 1 hour leading up to the 4 pm fix we find the USD Cost of Volatility for AUD calculated using the NCFX Mid and the Single ECN mid jumps to $15366 and $20,711 respectively.

A summary of the differences in the calculation of risk in the hour leading up to the end of Quarter 4PM Fix is given in Fig 4.

We can see that the mispricing of risk has become strongly positive.  This makes intuitive sense. In periods of high activity, the single source view magnifies the impact of increased trading.  Greater disparity of pricing is not captured by the single source venue. Its sample size becomes less representative (and sampling error increases). 

Article Image
Fig 4. NCFX calculated mispricing of risk in percentage terms at the quarter end 4 pm Fix
Article Image
Fig 5. Unit Cost of Volatility calculated using NCFX Mid or Single ECN mid-rates

Sampling error is strongly correlated with volatility as we might expect. In a short ten-minute burst of volatility in AUDUSD on 3rd August, annualised vol reached 30% calculated using NCFX independent mid-rates, while the single source venue showed annualised volatility of 39%. An hour later when the market had stabilised AUDUSD annualised volatility had fallen to 6.6% (Independent) and 6.94% (single source)

One of the most useful tools we use at NCFX to understand costs is the Unit Cost of Volatility metric. This contextualises transaction costs by dividing cost by the USD cost of volatility to which the trade was exposed to. For instance, a trade for 10 million euros in EURUSD that takes 2 minutes to complete is exposed to 2 minutes’ worth of volatility. By calculating costs as a ratio to risk we can compare transaction costs across different currency pairs, different trade amounts, different market conditions. 

The lower the ratio of cost to USD cost of volatility, the lower the normalised cost. By calculating volatility using a myopic, single ECN sourced rate, not only is the price of risk miscalculated, but normalised costs appear lower than they actually are. The choice of mid-rate hides costs.

To illustrate, look what happens to the unit cost of volatility for AUDUSD calculated using USD cost of volatility with the NCFX mid and the Single Source ECN mid. Keeping trade costs constant at $1000 we find the Unit Cost of Volatility is almost 30% lower using single ECN mid-rate calculated volatility. 

It might be tempting for a broker to prefer costs be measured against a single ECN sourced mid-rate, but President Lincoln once said “you can fool all of the people some of the time” but this isn’t a long-term strategy. 

Article Image

SIREN – An alternative to the 4 pm Fix

We have seen that broker and venue neutral independent mid-rates play a key role in correctly identifying transaction costs and calculating risk exposure.  The error term rises substantially around key market events, such as the 4 pm Fix. The Fix has well known flaws that accentuate market volatility.  It takes market prices from one or two ECNs and equally weights observations during a 5-minute window taking no view on market impact or market risk. 

A new alternative to the FIX, the SIREN benchmark proposes a methodology based on an optimisation framework that seeks to minimise market impact for the market risk exposure over a 20-minute time frame.  The siren benchmark is calculated polling the underlying NCFX mid-rates every second through the window and smoothing participation exponentially into the close, specifically aiming to minimise market impact.  The difference in outcome for users of SIREN benchmarks are compelling. 

As the chart above shows, GBPUSD showed evidence of strong pre-hedging going into the 4pm fix on 27th July. This accentuated price impact resulted in buyers at the 4pm fix paying substantially more ($1395 per million more) for the 4pm fix than buyers of the SIREN 4 pm benchmark. 

Reducing Market Impact

Market impact increases costs and those costs increase with trade size.  Prices from the market contain biases and skew which lead to price uncertainty. How much information from a large order will leak into the market and how much can be saved by reducing orders before they go to market?

Siege FX peer to peer platform went live this year. It brings together buyside firms to match interests at the regulated, broker and venue neutral NCFX mid-rate. Back-testing netting results with external market transaction cost analysis suggested that potential savings are in excess of $100 per million matched on average and almost double that for less liquid pairs. Siege participants (peers) include top global asset firms, pension plans and corporates. 

Article Image
Fig 6. How Siege Pre-Trade Netting works

Independent TCA

The market prices participants receive from dealers contain biases and skew.  It is difficult to de-construct dealer bids and offers to determine the market clearing rate.  Independent mid-rates, rates that are broker and venue neutral provide complete post trade transparency. 

As we have seen single ECN venue mid-rates hide the bias and skew of the platform and misstate the true cost of risk. It follows therefore that transaction costs must be calculated against data that is venue neutral.  Otherwise, the TCA is about as useful as setting foxes to guard the hen house.

The publication of independent mid-rates and the dissemination of independent mid-rate data through various market access channels acts to redress the information deficit that market participants face. Independent mid-rates exist to police the market to make sure that clients are treated fairly. And that is good for business.