HFT: Highly Fragilizing Trading

High Frequency Trading is a vastly complex topic, far too vast to discuss in detail in a couple hundred words.  However, it is also an important topic, so I will try to do it justice with the following brief synopsis.

First, about the whole "rigged"[1] business:  Other definitions of rigged aside, I don't think the presence of HFT trading has made the stock markets fraudulent.  If you submit an order to an exchange, and that exchange has the liquidity to fill that order at the price you entered, then the transaction will process as expected. If the misleading 60 Minutes piece[2] has confused you, see my previous post for a more accurate analogy.

However, when you realize that all of the HFT activity happens in a matter of nanoseconds[3], it does sounds REALLY fast.  But fast is good, right?  The typical argument is based on a typical Econ 101-esque line of thinking[4] and focuses on deterministic producer and consumer surpluses along with dead-weight loss. This technique is completely inadequate for this domain because it is entirely deterministic.  In reality, there is a GREAT deal of complexity, uncertainty, and non-linearity build into the trading systems.  As a result, the system itself is highly non-deterministic.  The Econ 101-type argument will focus on surpluses and losses which have a single value at worst, at best the argument will be hand-waved into an averages-based argument; e.g. "on average," "on expectation," or "in aggregate."  One small issue remains...in highly complex, non-linear systems, the expectation is not well defined mathematically. The best approach in such domains is not to try to reason about deterministic value or averages, but rather to look at the situation from the standpoint of fragility[5].

When viewed from this frame of reference, several things become clear.  First, HFT speeds up markets[6].  The faster a market moves, the more frequently bubbles form and the more frequently they burst by way of crashes.  Second, HFT consolidates liquidity provision into fewer sources.  Gone are the days of boutique market-making shops on Wall Street.  Instead, fewer larger institutions provide liquidity.  In extreme domains, the consolidation of the industry leads to more fragile systems because there exists less diversity amongst the herd.  The smaller and less diverse the herd, the more likely it is that a single unanticipated event can sink the entire industry.  Finally, by way of HFT competition, the liquidity provision system gets more and more complex by the day.  As systems get more complex, they become more highly non-linear, and in this case more fragile.  Small changes in circumstances can lead to big changes in outcomes.  Flash Crashes are a good illustrator of complexity fragilizing the market place. As seen in flash crashes, the feedback loops present in the algorithmic HFT system can be very damning.  Worse yet, they are entirely unpredictable and highly uncontrollable.

This fragilization of the capital markets by activities such as those HFT firms engage in is the real issue.  Exposure to limited upside (potential reduction in spread width) paired with unquantifiable potential losses is the very definition of fragility. By shifting to algorithmic HFT-based liquidity providers, we have exposed one of the central pillars of our economy to a great deal of unquantifiable, systematic risk.  Let's be clear: given today's highly-connected, complex global economy, this it is not risk for one person, or one HFT firm, or for one bank, or for one sector, or even for one economy, but rather a system-wide risk that affects us all.  Worse, so far as the aim of the capital markets is to facilitate the efficient provision of capital HFT makes no clear additions. As Mark Cuban pointed out on his interview with CNBC[7], this is the much more troubling issue with HFT. The market is not rigged, but the market is broken.  It hasn't fallen apart yet, but if we don't do something to fix the broken pillar now the next shock might just cause the whole thing to come crumbling down.

1. Definition via Google search:  "rig" verb. past tense: rigged; past participle: rigged.
1. manage or conduct (something) fraudulently so as to produce a result or situation that is advantageous to a particular person.
2. cause an artificial rise or fall in prices in (a market, esp. the stock market) with a view to personal profit.
2. "Is the U.S. stock market rigged?" http://www.cbsnews.com/news/is-the-us-stock-market-rigged/
3. The light from your laptop screen, by comparison, takes a couple nanoseconds to get from your screen to your eyes.  The technology to accomplish these sort of latency-based arbitrages is astounding, and more and more of it is not software related, but rather being implemented by hardware itself by way of Field Programmable Gate Arrays (FPGAs).  To see how far some of the optimizations go, see http://dspace.mit.edu/openaccess-disseminate/1721.1/62859
4. It goes something like this: these HFT titans use considerable capital and brainpower to narrow spreads in equities, lowering the cost for everyone to enter into and exit out of investment positions.  Markets are more liquid with HFT than they would be without!  Couple issues with that. First, as the 60 Minutes piece (and presumably the Flash Boys) demonstrates: there is a cost for that liquidity provision!  Pension funds spend hundreds of millions of dollars in the form of increased position entry/exit costs if they choose to subject themselves to poorly routed orders.  The alternative is not cheap either:  they can develop or buy order routing systems specifically designed to minimize their entry/exit costs.  Either way is more costly than if only those pesky HFT folks would just disappear...right?  Well...maybe.  Maybe not.  Presumably, being insanely fast gives them an advantage over your average joe investor, and as such they can narrow markets more than the average investor would.  As a result, when they leave, markets would be wider, and markets would be "less liquid"and that would be more expensive for investors than the cost of being latency arbitraged. Well....maybe.  Which effect is bigger is a matter of guesswork.  Also, thankfully and wonderfully, it doesn't matter!
5. This idea is entirely Nassim Taleb's.  The markets lie squarely in 'Extremistan' (to borrow a term of his) and in such domains, the Law of Large Numbers no longer saves you. See Antifragile: Things that Gain from Disorder and Silent Risk
6. The speed here is not necessarily how fast the matching engine runs, or how quickly new quotes are registered and published, but rather the rate of trade.  If anyone has data on time between trades and/or daily volume figures looking back pre-electronic and pre-high frequency trading, I would be very interested to see. 
7. http://www.cnbc.com/id/101539820