A common misconception of FX market liquidity is that it is plentiful, and so market impact does not matter. A dive down into the BIS triennial survey data shows that FX markets are far less liquid than people think, and market impact is a driver of transaction costs at surprisingly low trade volumes.
To illustrate this point, we parsed the EURUSD daily average figures from the BIS and broke these down into flow weighted hourly buckets and further into seconds. The average daily volume for global FX spot and outright transactions in EURUSD was USD720 billion in 2019, but this reduces to just USD18 million per second at the busiest hour of the day (between 3 and 4 pm UK time). This is for the entire global market. When we divide this volume up among the largest FX market makers, we end up with even smaller volume amounts.
Availability of Liquidity: Top 10 FX dealers by Euromoney volume, per second average of USD liquidity in millions, London time:
When a client asks a bank for a price, the bank must either hold the position in inventory (risk warehousing) or cover the risk in the market. Bank policies nowadays tend to require that banks reduce the amount of risk they hold. Accordingly, most customer deals are cleared in the market between 30 seconds and 5 minutes later. Should the trade amount be above the amounts outlined above (the natural absorption rate) then the trade will necessarily be pushed into the market, which will create market impact.
Let’s consider for a moment what market impact is. From a dealer’s perspective, market impact is the principal driver of dealer’s profits. Between order submission and when an order is filled, market impact cost is positively correlated to dealer’s profit. Post trade, market impact is negatively correlated to dealer’s profit.
Fig. 2 The cost profile of a customer trade
In the above example we show trading costs (calculated as the difference between the execution price and the market mid-rate) at different points in time before and after execution at point 0. An order is submitted at 1 minute before execution, is filled by the dealer at 0 and cleared in the market at +1 minute later. Between 5 minutes prior and 1-minute prior the cost to trade is almost unchanged.
Market impact begins when the order is submitted (-1 on the X-axis). At the moment of execution (0 on the X-axis) transaction costs are at their highest and are fully paid by the client. A minute after the trade, the effect of the order has dissipated almost entirely, and the market is roughly where it was before the trade took place.
Market impact is important because it is the base calculation used by the dealer to make a price to the client. The dealer wishes to ensure that the client trade is priced with sufficient market impact ‘insurance’ that he or she can exit the trade, paying for their own market impact on the way out of the trade and still make a profit.
Dealers are very focussed on market impact and will rank counterparts and clients according to the market impact experienced and paid, respectively. When combined with a factor for price volatility and something for credit, one can see how a price is formulated.
As we can in the example above, market impact begins to occur before the trade has actually been executed because it is vital to the dealer’s profitability.
When a quote is asked for on a Request for Quote basis market impact is already impounded in the price. The dealer must include in the spread a charge to cover the risk that prices move in an unfavourable direction before they can clear the trade. People find it more difficult to understand how market impact occurs in streaming prices. How can turning on a stream contain market impact?
All price streams are slightly different. This is the essence of a bespoke OTC market. The first time I trade with a bank, the probability of being a buyer or seller is 50/50. Over time however, this probability changes. The more I trade, the more info the bank has, and the better they can get at assigning a probability to which side I am on.
The evaluation of liquidity providers is often reduced to the simplicity of calculating what the market did next and evaluating the transaction against random points after the transaction. This ‘mark out’ is often supplied by the dealers form of TCA using the same price feed on which the transaction was created. This creates a number of problems.
The first issue is one of circularity, or finance’s analogue of the Heisenberg Uncertainty Principle. Market impact is measured as the difference in price between two moments – a moment that includes the transaction and a hypothetical moment of where the price would have been had I not traded, asked for a price or turned on a stream.
Using the same price feed to measure this will cloak or shroud this part of market impact. Accordingly, we are seeing a number of bank algo providers select liquidity sources for their algos because they know the same liquidity source will then be used to measure their performance in post trade TCA by a third party. This will always tend to flatter the results of the algo provider.
The second issue is one of comparability. If dealers are measured against their own liquidity, it becomes difficult to compare dealers on a like for like basis.
Measuring market impact, evaluating liquidity providers, and ensuring that all costs of transactions can be disclosed requires independent measurement. This is the rationale for comparing execution against an independent, anonymized data feed.
For more information about New Change FX, please contact: email@example.com