Friday, September 4, 2009

HALT: Imposing Limits on High Frequency Trading

In my first post on high frequency trading, I ended with a somewhat tongue-in-cheek proposal for a High-Frequency Arms Limitation Treaty, or HALT. The more I observe of the concern for this strategy, the more pervasive it becomes, and the more apparent abuses that come to the fore, the more this proposal moves from the realm of satire to the real.
As I mentioned in that post, I do not think the market benefits from moving trading speeds faster and faster in the millisecond range. But what the need for speed does do is burn through untold hundreds of millions of dollars for all the competitors to keep up with one another. And just as the complexity of derivatives can lead to the obfuscation of non-economic or manipulative operations, so can operations that blur past the screen before anyone can observe what is happening.

Here are some questions regulators should ask -- maybe they already are asking them -- to see if HALT makes sense:

What is the economic benefit of trading with twenty millisecond latency versus thirty millisecond latency?

What is the economic cost and what are the barriers to entry erected by pushing the latency envelope?

What speed of trading leads the marginal costs of obfuscation to dominate the marginal benefit for the end investor? By obfuscation, I mean the creation of a cloud around the trading activity that prevents the regulators from being able to assure the investors are being protected and the market is operating transparently and fairly.
What level of trades per second becomes problematic in terms of obfuscation versus practical and economic value? The focus with high frequency trading is the latency, but focus also should be given to the number of trades done per second. It is not just a matter of speed of trading, it is the cloud of noise that comes from what can be hundreds of different trades on one security all flowing from one trading firm into the market in a very short time period.
The answer to these question might be limits on the latency of trades and the number of trades per second allowed in any one security by private and proprietary trading firms.


  1. Happy to see someone bringing some real life relevance to this conversation.

    To extend it a bit, if the capital markets are there ultimately to allow users of capital to access it, latency and trades per second have no relevance. The topic is more the extent the looters are allowed a share of the process.

    To think this is getting the attention it is...

  2. On the other hand, the speed of response to a bid or offer has always been a part of the market making game.

    Another thought. "Wash Sales" are already illegal, since long ago.. how can a firm be certain to not afoul of the wash sale rules while executing a hundred trades a second ?

  3. The problem with high frequency trading will occur when the markets gap. High frequency trading is not deigned to handle gaps and during market dislocations gaps do occur.

  4. It would seem pretty difficult to regulate latency in a continuous market. Turning to frequency instead addresses a problem, but does not prevent the latency arbitrage arms race since one would still practice this at longer periods between trading episodes in a single security. Simpler than regulation or taxation is to move to a discrete-time trading model, as I've advocated here.

  5. The idea in the last post and its link to have discrete trading is an interesting one, and one that hearkens back to the 'old days' of Walrasian markets. These markets, used as the basis for most economic analysis up until a decade or two ago, assume people enter in orders, and then the person running the market calls out prices until there is a balance of trades.

    Walrasian markets were thought of in once-a-day sort of time frames. Today, even once a second or once every ten seconds would do the trick.