With the full disclosure that the pages and pages of formulae are way beyond me, I have to confess I quite like reading academic papers on the FX market structure – often they state the obvious (but in a clever way), but just as often they get the hamster back on the wheel in my head.
So when, a couple of weeks ago, a paper crossed my virtual desk (I saw it on the Web obviously but academics are not the only ones who can use big language!) entitled Spoofing and Pinging in Foreign Exchange Markets I was obviously intrigued. I have now got around to reading it – it is written by Alexis Stenfors at the University of Portsmouth and Masayuki Susai from Nagasaki University – and find it makes some interesting revelations that prompt a few thoughts on the Mark Johnson case. It also highlights some remarkable statistics around market behaviour.
The aim of the research is to investigate the susceptibility of OTC spot FX markets to spoofing and pinging (sometimes referred to as “flashing”) and uses high frequency data over five days in September 2010 from EBS across its main currency pairs, EUR/USD and USD/JPY, as well as three pairs where EBS was a secondary platform, EUR/SEK. USD/TRY and USD/RUB.
The top level findings are unsurprising to most in the market; the two majors are susceptible to “information-rich” orders, which suggests that spoofing is platform dependent – in other words if a dealer places an order on EBS then it triggers the cancellation of counter-orders and has an impact on other venues – and the three minor pairs are susceptible to “pinging”.
I guess we have to jump straight in here by asking, ‘should that matter to us?’ The paper makes a point this column has raised several times before by noting, “…markets highly populated with algorithmic limit orders could be particularly susceptible to manipulative trading tactics such as spoofing”. In other words, it is easier to spook a machine than an experienced trader.
In some definitions, placing an order on EBS or Matching with the intention of having someone pay you on another venue is spoofing, but I disagree. Clearly it is a tactic – just look at the recent BIS paper on the FX market structure, which observed that while the primary ECNs’ volumes had declined, they were still the main source of price discovery. That is saying what has long happened in FX markets, a bunch of platforms’ liquidity is predicated upon their being a price on a primary venue (I accept the number of primary venues is increasing slightly.
To me this does not enter the realm of spoofing unless the primary venue has last look. I would also point out that he authors note that only orders placed aggressively, i.e. at top of book or inside, actually trigger a cancellation or trade on the other side. This, to me, signals adequately intention to deal and it is vitally important to distinguish this behaviour from that of convicted spoofers in futures markets who often placed their large orders at the fourth or fifth level of depth of book, and cancelled the minute (microsecond) they got near to top of book.
I also think it is important to highlight the nature of market making in FX (and most markets to be fair) and continued existence of high frequency traders in the market. Something that struck me was that in a dataset containing over 2.3 million limit orders, 99% of them were cancelled – that’s a lot of price turnover and reiterates the challenges brought by decimalisation. There are players still who can make a living nicking two or three tenths of a pip thousands of times a day and that commensurately costs the market makers, who in turn, cancel and reprice. That sensible defence mechanism, however, should not be confused with spoofing or pinging – it is updating a price to ensure it is accurate.
That said, however, we enter a grey area around this issue for, as the research finds, the placing of an aggressively priced order on one side of the market (top of book or inside), in 50% of instances, led to the cancellation of an order on the opposite side. Where is the intention to deal now? More pertinently, is this counter price from a true liquidity provider?
I understand that no market maker is going to stand in front of a runaway truck, but the paper also notes that 85% or orders in the two majors on EBS are for the minimum amount of one million units of base currency. I would argue this is not “liquidity providing” because when so many orders are cancelled just because someone signals the desire to trade close to the price level, the legitimate question has to be asked, does this participant really want to deal? I would argue not – they are either sniffing out order patterns – and the paper also notes that split orders lead to increased signalling risk – or, they pricing there to make two-tenths of a pip by hitting a slower player on another venue. Either way, that is not liquidity provision.
The good news for me is that this type of behaviour is exactly what can be highlighted by the data analytics packages provided by NEX Analytics and, more recently Thomson Reuters. Other platforms also provide this data to their customers, so to me the question will be simple – which of them are willing to ask difficult questions of these shadow LPs or recyclers with the ambition to improve hit ratios and market quality on their venues?
Moving on, the pinging research is interesting. I totally get it can be with the intention of sniffing out a larger order, but given how the research shows that the vast majority of orders in the majors are for one million units, there has to be a significant chance that all that will happen is the pinger will meet someone looking to sell one million, not 100 million.
The paper notes how the latter orders in a split parent order have more signalling risk, which I guess is natural as more information has to leak out, but outside of those instances it strikes me that pinging can only really work if the trader gets the name of the counterparty – it’s not about the order itself. It could be one of one, or one of one hundred, as I just pointed out, but the name (with a possibly unhealthy degree of assumption) can provide some interesting information that may indicate there is something more going on.
That said, the research finds that pinging happens when EBS is the secondary venue, which suggests to me it could be someone seeking liquidity who is unwilling to keep the order on the platform for fear of signalling risk through bidding in a pattern across two or more venues. I am not sure, but such are the complexities of the modern foreign exchange market, that it is a credible argument I think.
Other points I would raise having read the paper are that it reaffirms my distaste for primary venues supporting minimum sizes lower than one million units. As the report states, “the vast majority (over 85%) of limit orders submitted in the EUR/USD and USD/JPY markets are for precisely [EUR or USD] 1 million”.
It goes on to observe, “This is consistent with empirical findings of a ‘race to the bottom’ concerning order sizes in other markets increasingly populated with high-frequency traders.”
Clearly the authors (and, not that they care, I agree) believe that lower minimums will dilute the quality of the market. The paper provides a very tidy explanation for why volumes have gone down on the anonymous channels and why the CLOB model has suffered – there are too many players seeking to exploit the transparency available which works against those with genuine, often larger, business to execute.
I also believe this paper highlights why many of the platforms are waking up to the need to better police activity – they have reached the zenith in terms of retail brokers and prop shops, growth from here has to be with “genuine” traders with intention to deal, and those traders are demanding protections.
I want to close out with an observation I flagged at the top of this piece and it is in regards to the Mark Johnson trial.
I found it very interesting to read that out of more than 2.3 million limit orders placed in the five day period, the largest single order was EUR 250 million. Moreover, only 2% of those orders were considered “large” by the researchers, with that label being defined as more than five million units of base currency.
EUR/USD is much bigger than Cable and somehow HSBC was expected to buy GBP 2.25 billion – that’s $3.5 billion remember – in one minute? If the largest single order on a bigger venue, in a bigger market was 250 million in a five minute, how on earth can the prosecution argue that it is possible to buy more than 10 times that in one minute?
I have mentioned before in my columns that I thought a central plank of the Johnson defence case, and now appeal of course, should be the likelihood of a flash event taking place if the order was executed without pre-hedging – this would appear to provide support for that argument. The liquidity to counter such a trade doesn’t exist, and the behaviours highlighted in this paper indicate that the placing or a single, or split, order for a large amount, would have triggered a market response.
The HSBC buying of sterling did provoke a market move of course, but a slower one over the course of the pre-hedging window. To me, this paper, merely reinforces what most market professionals already know, but what we simple have to get through to the authorities – the only way to work a large order is over a longer period of time.