There is a chance later this week that we will know the decision of the GFXC feedback process on the possible rewording of Principle 17 of the Global Code, so I won’t go into the arguments put forward by Norges Bank Investment Management over last look, aside from observing it didn’t appear to provide any feedback on the process, which is a little strange.
Instead I want to look at the observations around the use of algos and price verification in the paper and then to add to my growing list of foes in the analytical/academic space by pointing out why they are wrong!
Aside from thanking (and yes this is ironic) NBIM for introducing the concept of last look in the algo execution space (please call it something else!), I am unsure its arguments around the provisions of algorithms and liquidity generally work in the FX environment. The overall tone of the report seems to me at least, to support a move to the ‘all-to-all’ equities style market structure, where algo deployment and TCA are pretty easy to come by. The fact that it is more challenging in FX does not necessarily indicate a structural weakness in my view – rather it reflects the multi-dimensional nature of the market.
One sentence caught my eye in particular, where NBIM states, “…there is significant potential for conflicts of interest if the dealer providing the algorithms routes the executions to the same dealer’s trading venue/principal stream.”
To me, this appears to argue against accessing any internalisation flow and I can’t see how this enhances execution quality. I am the first to argue that “true” internalisation rates (something that involves zero signalling, internally or externally) are not what they are proclaimed to be by most players, but there is still valuable liquidity in internalised pools. FX may be the most liquid market in the world but it is also very diverse and matching interests simply do not exist that often – there is always a (sometimes admittedly small) window where the risk is unmatched and held by the dealer. By not accessing this internal pool of liquidity, which can still be significant in size despite my scepticism, the algorithm is missing out on potentially a large amount of the available liquidity –how can that be good?
It seems to me that NBIM is calling for more agency algos – it specifically mentions third party providers being a potential future source of development – something that reinforces the impression that it believes that internalisation is a bad thing.
I fully understand and support the sovereign wealth fund’s argument that FX could do a little more in terms of transparency in certain areas (last look being front and centre) but I also think it has missed the point about how TCA is changing in FX.
Yes, dealers are providing their own TCA reports, which will inevitably leads to suspicions (ungrounded in my view in this day and age) that they are skewed, but independent data and TCA providers are available – it’s one of the bigger growth sectors in FX over the past two years.
So I tend to not only disagree with NBIM over the role of internalisation, I further would argue that it should play a role in providing a superior pool of liquidity to the client and can be independently verified for TCA reports.
In many ways the fact that a major player like the Norwegian sovereign wealth fund is openly commenting on the need for greater transparency and checking serves to highlight just how big a trust problem the FX industry has, but again, I would suggest that is largely in the past. I understand that it is not helped by events relating to activities alleged more than five years ago now, but the FX industry, especially the major dealers, has put its house in order. It does support independent TCA and, in most cases, is a lot more transparent than it was over last look.
I was tempted to observe, reading the paper, that NBIM had used the multiple themes to cover a pointed criticism of the FX industry over its use of last look but I couldn’t see why it would do that. Not only does it explicitly state that last look should not be banned, it also sees some potential benefits for the practice as long as it is more transparent than it currently is.
I stand behind its arguments totally on this issue – the FX industry does need to be more transparent around last look – but I can’t agree with its conclusions on the other themes. On one hand it appears to be arguing for a market structure in which more responsibility is pushed to the client side when it comes to market and execution risk, but on the other it says the risk controls in place around algos are squarely aimed at protecting the dealer, not the client – I would respectfully suggest in this case they are one and the same.
The algo strategies that have been developed are the results of the quant teams’ at the dealers best efforts and they have gone through a robust testing phase. Yes, there is always a danger that an algo could “go rogue” but the dealers are aware that this is, reputationally, a business killer and will make sure their algos are appropriately designed and controlled.
As always, there is a risk of a lower quality strategy being rolled out by smaller, peripheral players that doesn’t have this robust framework going wrong, but such offerings should come with a ‘buyer beware’ sticker. I fail to see how a customer thinking of using an algo, should not do their own due diligence on the risk framework in place at the provider? Responsibility works both ways.
Overall, I feel NBIM has a serious – and in my view, legitimate – beef with the level of transparency around last look, but I feel it has seriously underestimated both the amount of work done at the major dealers in the algo space and the benefits of accessing internalisation programmes in the FX market.
I accept this model does not conform to standard equity market structures, but I would also argue (yet again) that this doesn’t mean it has no value.
In Thursday’s column I said that a 20% move in the value of Bitcoin would lead to a trading shut down on CME. This is, CME tells me, not the case and that the contract can trade at limit up and limit down, but not beyond the limits and trading is not halted. The article has been adjusted and I would like to offer my apologies for the misinterpretation on my part.