Monday 29 October 2012

The Problem with the Foresight Report on Computer Trading

It turns out that Paul Wilmott shares my views on the BIS Foresight  Future of Computer  Trading in the Financial Markets.  I have made the point that the Report is running the risk of appearing as what Roger Pielke Jr describes as "stealth advocacy", scientific research presenting itself as neutral technical analysis while it is in fact advocating a specific course of action.

My main issue with the Report is captured in its observation that (p 140, ss8.2)
In financial markets, the ideal is to discover the ‘fundamental’ price of assets
This statement is economically controversial, I do not think Keynes would agree with it.  Intuitively, if you have ever bought a house, did that house have a 'fundamental' price? Rather than develop these more philosophical points I shall focus on the mathematical objections to the statement after highlighting how this assumption drives the Report's overall conclusions.

If there exists a fundamental price, then the markets are attempting to solve an epistemological problem, and wish to identify the true price in a mass of noisy data.  This means that financial stability is defined as (p 19 n1)
the lack of extreme movements in asset prices over short time periods
and is closely related to volatility (p 19 n2)
variability of an asset’s price over time
which is a bad thing, so the report looks for evidence of HFT increasing volatility (which it does not find, but the FT cites research that suggests there is a link between increased volatility and HFT).  

The definition does mean, however, that the pathological behaviour described in Figs 4.2-4.6 of the Report is not market instability, they are localised transient effects that are quickly corrected.  The analogy is that it is OK for an aeroplane to go out of control, providing it does not crash, I do not think the aeronautical industry would allow itself to be run on this basis.

The problem is with the Report's treatment of High Frequency Trading.  The Report recognises that this aspect is controversial, particularly with pension managers and conventional fund managers.  Because it takes an approach that will see HFT as beneficial, because it increases information flows that help resolve the epistemological problem,  the concerns of those not involved in HFT, like the pension funds,  are counter-balanced/nullified. 

My concern is that in taking the epistemological approach to markets, it is inevitable that HFT is beneficial.  Just as if you model  credit default as some form of contagious agent, dynamical systems analysis will point to a solution involving a few, large, well defended financial institutions (i.e. how you deal with mad cow disease), while modelling the banking system as a communications network, loans are 'packets' that move around the network, a distributed system like the internet involving many small institutions with lots of connections, is best.  The result, by and large, depends on how you approach the problem.  The Bank of England seem to be moving to the Internet model of banking, rather than the Contagion model. 

The mathematical objection to the approach the Report takes is rooted in the Fundamental Theorem of Asset Prices, which indicates that a 'true' price of an asset only exists in an idealised situation, the second statement of the Theorem is explicit in stating that in actual markets, a unique, 'fundamental'  price is impossible to identify, there is a financial Heisenberg uncertainty principle going on.  The problem of a price is ontological, not epistemological.  I suspect this is one way of summarising the points Wilmott made to the enquiry, because it is such a dominant theme in contemporary mathematical finance.

The line the Report takes is explainable in terms of the background of the bulk of the contributors to the evidence based, they are in the main from dynamical systems (rooted in ergodic systems) and computer science (data and rules to manipulate data).  There are only a few who contributed to the evidence base whom I recognise would be familiar with the FTAP (Cont, Shied, Avellaneda, Mitra, Jaimungal, Cvitanic out of over 200, 2 of whom declare a commercial interest.  There were 11 alone at the Complex Systems workshop).

However, the Report, in taking a very particular approach  narrows the scope of discussion and leads the reader of the Report to a conclusion that is heavily dependent on the assumption that markets solve an epistemological problem.  The fact that there is such a bias towards market insiders on the High Level Stakeholder Group, there is no representation from pension funds but the ISDA, a lobby group for derivatives traders that gave us the infamous Potts opinion, are there,  lays the report open to the accusation that it is stealth advocacy.  Given that the public have been angered by the apparent privatisation of profits and socialisation of losses by banks in the past, appearing to side with the proprietary traders over pension fund managers, seems a very short-sighted approach to take.
I think these observations are compounded by the fact that the Report could have been clearer in distinguishing the various impacts computers are having on markets, rather than a focus on addressing HFT within this context described above.  The Report's Executive Summary opens with
A key message: despite commonly held negative perceptions, the available evidence indicates that high  frequency trading (HFT) and algorithmic trading (AT) may have several beneficial effects on markets.  However, HFT/AT may cause instabilities in financial markets in specific circumstances. This Project has shown that carefully chosen regulatory measures can help to address concerns in the shorter term. 
This distinction is lost in discussing the benefits of Computer Based Trading (the combination of the two) as improved liquidity, reduction in transaction costs and improved market efficiency.  

The risks are observed collapses in liquidity and instability.  When discussing instability the observation is made that HFT does not appear to increase volatility, but there are issues about stability, which the report describes in terms of non-linearities, incomplete information and what the report calls 'normalisation of deviance' but sociologists describe as 'counter-performativity' (markets follow a model and then discover the model is wrong).  

HFT is highlighted in regard to market abuse, which the report argues there is no evidence.  However the distinction between AT, usually conducted for agency trading, and HFT, usually conducted by proprietary traders, in their contribution to the risks is not really developed.  Can we explore this issue?

The report defines 'liquidity' (p 19, n3) as 

the ability to buy or sell an asset without greatly affecting its price. The more liquid the market, the smaller the price impact of sales or purchases
AT has made a significant contribution to this, particularly through the work developing out of Chriss and Almgren's pioneering research in optimally executing large trades.  The report's definition of liquidity is somewhat biased , a different definition associates liquidity with the 'depth' of the market, the ability to actually buy or sell an asset in the market.  HFT will not increase depth, but it may present a mirage of improved liquidity as assets are churned by proprietary traders.  This is associated with the Flash Crash, p 56. 

Liquidity is related to what the Report describes as  (p 19, n4)

price discovery ... the market process whereby new information is impounded into asset prices.
Clearly if trades are executed quickly, price discovery improves.  However there are issues with mechanistic approaches to news processing (Derwent Capital Markets' closure).

The Report notes that there has been a reduction in transaction costs since the introduction of CBT, but this cannot be ascribed to HFT and the FT cites 2012 research that suggests the reduction in costs occurred before HFT emerged.

Overall I think there is a good argument that the Report is far from being "pure science" and lacks the balance of "honest brokerage".


  1. How many leading academics and experts does it take to change a light bulb?

    The report of Foresight: The Future of Computer Trading in Financial Markets suggests that 150 is not the right number. What is a light bulb? Why does it need to be changed? What does it need to be changed into? Maybe it's the light switch that needs to be changed? Is the lack of light really a problem? Research shows that blundering around in another room looking for something else does not reveal a problem with the bulb. Why can't these pesky humans put up with being kept in the dark anyway?

    Given the time, money and sheer number of international academics involved in this Foresight Project, its conclusions and recommendations are banal and embarrassingly thin.

    More at

  2. Fascinating blog, and I agree with much of what you've stated.

    I would like to give a trading perspective, although I'm familiar with the math. Research such as Almgren and Chriss, is academically nice, but it has little relevance to real trading, since they are really optimising over the costs, and not optimising over the price. Of course price is included in their model, so they might think that they are optimising over the price, but as a by-product of the model chosen, you lose the reality. There's a trader saying related to liquidity...'you can have as much as you want, if you're wrong'. There is a price objective for trading, not a transaction cost objective. Yes there's a total cost of the two, but price is going to win that battle every time.

    If one has a much better price to trade at than the price objective, then the movement in price related to liquidity and large size ordes may not be a problem at all. Then there are other issues, such as the fact that not everyone trades market orders. A limit order placed, will not move the market at all. It either executes at that price or better or it doesn't. Then there are stop-limit orders which specify the range you're willing to trade at, and so on. Given the need to execute before a fixed time, a trader will have an objective target, and anything better than that can be executed. This will often be done by limit order, and this results in zero cost (aside from exchange fees etc) in terms of price movement against you, and typically results in slippage for you, i.e. price improvement. Does the research take into account any of this? Now as the time to liquidate becomes very close, one may have to use market orders which may move the market, but that shouldn't be the starting point. That's the desperation point, which is a complicated function of both price and time left.

    Another point you raise is about fundamental price. The answer to what the fundamental price is, is very simple, it's the current price. Nothing else matters. Something is only fundamentally worth what someone is willing to pay for it, whether that seems reasonable to others who are not trading it or not. All else is nonsense.

  3. On financial is a market, and its price needs to change, and it must change in ways that are difficult to discern. It can't be any other way. Therefore, by its nature it must in some sense be unstable. A market which barely moves is a dead market, not a free one. So stability is not something that needs to be sought after. If you want that, then fix the price and face the consequences. Yet people are looking for market stability. It's baffling. Of course they are perhaps scared by the magnitude of some moves. Well again from a trading perspective, the magnitude of moves and the effect of those moves to your wealth, is related by how leveraged you are. Therefore is the problem really with the size of the movements, or is the problem with the account holder, whether that be an individual trader, an entire bank or hedge fund, or the government? I would suggest the latter. People with no idea on the gamble are leveraged up to the eyeballs.

    Final point about HFT/AT. CBT has reduced the costs of trading, just as the internet reduces costs of many businesses, but HFT has increased costs. And I don't see a benefit. It increases costs to the exchanges (some of which can't cope), it increases costs to those competing on the HFT war, it increases the costs associated with data, quote stuffing etc. and hence the cost of storing, transmitting and using the data, and because of a combination of these, it increases costs to others too. HFT doesn't really provide liquidity, it provides fake liquidity. You may be interested in taking a look at some of the research done by nanex ( in relation to HFT.

    I don't believe HFT provides any price discovery (perhaps we have a different concept of price discovery), nor do I believe there is any convincing evidence for that, although that one is harder to debate. I also don't believe they cause flash crashes and so on, we're quite capable of doing that ourselves. When talking about reduced volatility, well what does that really mean in a case such as this? We have markets which go through cycles of being volatile and less volatile. Even on a daily basis, volatility might reduce at lunch time for example. The statistical measurement of something like this, given the timeframe HFT has been active/dominant, leads to results that can't be relied upon. And I do always find it funny how people like to equate the price with "new information" being "impounded into asset prices", the only new info is the price itself.


Note: only a member of this blog may post a comment.