Review of the FCA paper on UK dark pools and reference prices
The FCA publishes a paper on asymmetries in dark pool reference prices to analyse the role of participant speed and sophistication in driving outcomes in today’s markets.
Following their Occasional Paper #16 on High Frequency Stock Trading earlier in 2016, the FCA’s economists have followed this up with an Occasional Paper on asymmetries in dark pool reference prices (#21). They look at two important aspects of reference prices in dark pools – the prevalence of trades at stale prices and secondly the choice of reference price. To what extent are participants implementing best execution prices when a dark pool references a worse price than the lit market and is this influenced by conflicts of interest within dark pools and participant sophistication?
A ‘dark pool’ is a trading venue with no pre-trade transparency, while ‘lit’ markets allow participants to observe orders submitted by other participants. The main advantage of a dark pool is that the trade intention is not revealed to the entire market. A secondary advantage is that participants can potentially obtain a better price as many dark pools match trades at the mid-price, saving participants half the spread. The main disadvantages of dark pools is uncertainty, as it is impossible to know if there is a willing counterparty, so participants cannot be certain a trade will take place, and that reference prices must be sourced from primary markets.
Firms that regularly conduct detailed risk assessments, that include a comprehensive list of the relevant risks, the key responsibilities for managing and mitigating those risks and the controls employed to achieve their goals were much better at demonstrating that effective controls were in place than those who didn’t.
As its primary market reference price source the FCA use the London Stock Exchange (“LSE”) in their study.
Two forms of latency exist in sourcing data from primary markets; processing latency – where markets are in the same physical location but a delay occurs in hardware and software calculation and dissemination of the market data, and transmission latency – where the markets are in different physical locations and it takes time to transmit the data between one market and the other.
The FCA found that reference prices were sometimes stale in every dark pool they sampled and that this was increasing over time (3.36% in 2014 to 4.05% in June 2015). The increase could be explained by increases in message volumes and volatility over the sample period. This has implications for market infrastructure as volumes grow.
90% of all stale reference prices were 6 milliseconds or less in duration, with the top 5% above 20 milliseconds and the top 1% above 217 milliseconds. All stale prices were long enough for an algorithm to identify and act on, but the top 1% was long enough to be identified by human traders.
The FCA estimated that the cost of stale reference prices is approximately £4.2m per year over all dark venues. This is an insignificant figure given that the average daily order book, by equity value, traded on the LSE is $4.9bn.
Where a trade is executed at a stale reference price one counterparty benefits from this. Latency should affect participants equally, so equal outcomes should be expected across participant types. The FCA found this was not the case as in 96% of the cases High Frequency Traders were on the benefitting side.
MiFID requires that the reference price must be ‘reliable’, which is clearly not happening when the price is affected by latency. MiFID II, which is expected to take effect in January 2018, will require microsecond granularity and a maximum timestamp divergence of 100 milliseconds for venues with less than 1 millisecond gateway to gateway latency. From the FCA’s analysis market data latency regularly exceeds this threshold.
Overall, the FCA study found asymmetric outcomes across market participants where the reference price was stale and when it is inferior to other prices available. This may be the result of market participants’ differing abilities to observe and manage latency and smart order routeing in a fragmented market; the costs involved are more typically borne by those less capable of managing them. But, while the effects are significant in statistical terms, the FCA found that the economic impacts are small.
Dark pools may still offer a valuable service as in most cases they provide price improvement and in all cases they allow investors not to show their hand to the market.
The link to FCA Occasional Paper 21 is here - https://www.fca.org.uk/publication/op16-21.pdf