The High Frequency Trading Arms Race is a Problem But Taxing isn’t the Solution

Posted on August 11, 2012 by


Discrete time markets are better than taxing.

related: why 24/7 trading isn’t happening

About High Frequency Trading (HFT)

HFT accounts for as much as 40% of the volume on the major exchanges, and it sometimes drives prices in unexplainable directions.  You’re probably aware of some of the ways High Frequency Trading (HFT) has impacted our markets: There’s the flash crash in May 2010, and more recently the Knight Capital debacle. Read more here.

For those who are not familiar with the term, HFT refers to trading methods used by hedge funds and investment banks in which their computers are co-located on the exchange floors to conduct rapid trading. Holding times range from seconds to minutes. HFT algorithms benefits from co-location by being able to observe open orders to infer upcoming price movements earlier than those who are located remotely.

Network delays, including the speed of light, mean that traditional traders are at a relative disadvantage to those who have the fastest algorithms and machines on the exchange floor.

On the positive side, HFT provides broad liquidity to the markets, meaning that there is much more likely to be someone on the other side of a trade when you want to buy or sell a thinly traded equity. On the negative side we have rapid market movements that don’t seem to make sense.

The Flash Crash of May 6, 2010.

How HFT works in continuous markets:

Each major exchange publishes an order book that shows for each bid and ask price how many shares are available. These prices reflect limit orders entered by traders. There is always a gap between the lowest ask price and the highest bid price. This gap is called the spread. There are no transactions until someone “crosses the spread” with a market order or a limit order.

As an example, assume an order book where the lowest ask price is $101.50 with 50 shares available, and 100 more shares were available at $101.51. With this order book, if someone submitted a market order to buy 100 shares it would be executed with 50 shares at $101.50, and 50 more at $101.51 (for an average of $101.505). On the other hand, a limit order for 100 shares at $101.50 would only be filled with 50 shares.

As you might expect, large market orders can cause significant price changes are they “blow through” the shares available at each bid or ask level as they consume those shares.  This was apparently the initial cause of the flash crash: a “fat fingered” oversized market order.

All these trades execute in continuous time as the orders arrive. The HFT machines on the exchange can observe the trades and the order books as they change.  HFT algorithms are able to make predictions about price moves by observing the balance of buy (bid) and sell (ask) orders on the books.

HFT algorithms do more than just observe though, they actively probe the market. They take advantage of their ability to rapidly publish then cancel orders in order to conduct “phishing” expeditions to see how much others are willing to pay. 90% of all HFT orders are canceled.

Because this sort of trading can be so lucrative, hedge funds are willing to spend a lot of money to gain this advantage.  They’re competing with each other of course, so we’ve seen an “HFT Arms Race” emerge as firms deploy faster and more expensive hardware and algorithms.

The at home day trader has no chance. In fact the at home day trader is the poor sap who’s losing out to the HFT robots.

One solution: If it moves, tax it.

One proposal on the table is a small tax on every transaction. From Daniel Dicker’s article at huff post:

The idea is simple: Charge a miniscule extra fee on every trade executed at every registered exchange. Such a simple charge would make much algorithmic ‘churning’ of volume uneconomic, while bringing institutional and retail trade back to the major exchanges and away from the ECN “dark pools.” You’d begin to re-level the playing field for retail and private investors, generate confidence by returning fairness to the stock markets, greatly reinvigorate employment in the financial markets and gain a nice source of unexpected revenue too.

It’s not just Dicker’s idea. This is also on the table, more formally, in Europe as well (WSJ article).

This approach would likely have the intended effect of slowing the markets down. But there are a number of arguments against such a tax, including some significantly likely unintended consequences.

First of all a transaction tax will make all trading more expensive, even for those who it is not intended to punish.We would also see larger spreads and less volume which would inhibit those liquid markets which enable investors to get to the positions they desire quickly.  If you need to sell NOW you might not find someone willing to buy.

Additionally, and perhaps more importantly, trading may move from taxed markets to others that are less expensive. Here’s what happened in Sweden when they created a transaction tax (from wikipedia):

During the first week of the tax, the volume of bond trading fell by 85%, even though the tax rate on five-year bonds was only 0.003%. The volume of futures trading fell by 98% and the options trading market disappeared. 60% of the trading volume of the eleven most actively traded Swedish share classes moved to the UK after the announcement in 1986 that the tax rate would double. 30% of all Swedish equity trading moved offshore. By 1990, more than 50% of all Swedish trading had moved to London. Foreign investors reacted to the tax by moving their trading offshore while domestic investors reacted by reducing the number of their equity trades.

Supporters of a transaction tax argue that Sweden’s is not a good example because their tax was significantly higher than those that are currently proposed.

Another idea: Instead of taxing the markets, change their mechanics

Michael Wellman proposed this alternative a few years ago on his blog:

An alternative would be a discrete-time market mechanism (technically, acall market), where orders are received continuously but clear only at periodic intervals. The interval could be quite short–say, one second–or aggregate over longer times–five or ten seconds, or a minute. Orders accumulate over the interval, with no information about the order book available to any trading party. At the end of the period, the market clears at a uniform price, traders are notified, and the clearing price becomes public knowledge. Unmatched orders may expire or be retained at the discretion of the submitting traders.

I happen to prefer the longer interval, up to a minute, because this will go further to level the field for traders located more distantly, or with less sophisticated computing. The approach has a number of advantages, including

…the call market totally eliminates any advantage of HFT systems. It does not eliminate the opportunities for algorithmic trading in general–just those that come from sub-second response time. No party has privileged information about order flow, and no party benefits by getting a shorter wire to the “trading floor”.

The exchanges already have the technology to do this. They use these auction methods to set prices at the open and close of each trading day. They’re called the opening cross and the closing cross at NASDAQ. You can read the details here. Wellman’s idea could be implemented with a cross executed once each minute.

There are lots of reasons to like this approach.  Another one is that it would be easy for exchanges to apply limits that would prevent, dramatic moves. As an example, the volume at each auction could be limited. This would allow prices to fluctuate with demand, but also slow them down. In the case of a major event, the price would begin to move immediately, and it would eventually converge to equilibrium, but a flash crash could be averted.