30 Oktober 2013
Dispelling Myths of High Frequency Trading
The purpose of this paper is to dispel misconceptions about the effect of high frequency traders (HFTs) on market making. I will clarify how markets are made on non-HFT exchanges and then explore the efficiencies created for the markets by high frequency traders (HFT). I will demonstrate how HFTs can make tighter and more efficient markets that benefit all market participants through the use of high speed computers. The issues of rebate trading and co-location will also be explored.
Understanding High Frequency trading
When thinking about the term “high frequency trading” a helpful image is that of the specialist. The specialist is the member of the exchange who makes a market in stocks and is responsible for maintaining the quote and the book for stocks. The specialist has always been somewhat mysterious and viewed with skepticism over the years. This stems from a misunderstanding of the risks in managing the specialist book and envy of the monopoly that the specialists enjoy. The HFT performs a similar function of making a market and managing risk, but with computers and at high speeds. Not surprisingly, the HFT is also subject to misunderstanding and envy.
How does the specialist earn a living by “making markets”? Stocks trade in a continuous auction market (think Ebay). There is always a bid price, the price the Specialist is willing to pay to buy the stock, and an offer price, the price at which the Specialist is willing to sell the stock. Here’s a simple example for stock XYZ: the market is $10.00 bid and offered at $10.05; or 10.00/10.05. Customer A is bullish and decides to buy 100 shares, so he/she sends their order to the market and the specialist sells 100 at 10.05. The specialist resets the market to 10.01/10.06 (because of the buy interest) and sometime later Customer B decides to sell 100 shares at 10.01. The specialist buys the shares at 10.01 for a profit of 4 cents on a hundred shares, or $4.00. The specialist will most likely reset the market to 10.00/10.05. The point is that the specialist makes a living by buying on the bid and selling on the offer to “capture” the spread between the two prices.
As you may guess, the reality is not as simple and is far more risky than the example. If customer A wants to buy, it is somewhat likely that Customers B, C, and D may also want to buy at the same time and it would be easy to create a case where the specialist would sell a substantial number of shares before someone else showed up as a seller. So the skill of being a specialist is in determining where the balance of supply and demand, the balance of buyers and sellers, are in the stock and then making a market around that level. This level is typically a moving target during the day. Also, the specialist must manage the risk involved in being long or short the stock. Managing the risk entails not letting the position get too big, either long or short, to protect against adverse price movements ( i.e. having a long position with prices going down or short position with prices going up). If the risk is managed well and the specialist is profitable, the money made from capturing the spread between the bid and ask prices will offset the losses from adverse movement of the position. The specialist will typically have an overnight long or short position which entails additional risk. To compensate for the very real risk involved in making a continuous market, the specialist has historically enjoyed a monopoly on the stocks listed at his post. While this enhanced the specialist’s chances of being successful, it hindered the development of faster more efficient markets.
Enter the high frequency trader, and the HFT computers. The computers are programmed to make a market as described above and the software also computes a new market after each trade and manages the risk of the trade. The reason the HFT is successful is that they have extraordinarily good risk management. This is because the software has perfect discipline; it will always do what it is programmed to do. Completely unemotional about taking a profit or loss, it is significantly better at managing a long or short position than a person. Additionally the software can assess many more factors than a human can; it can have multiple inputs well beyond what a human can observe and so can calibrate a better market. At the end of the day the HFT is almost always flat, meaning that they have no net market impact.* This creates a much more efficient market by tightening the spread between the bid and offer and reducing the impact on the market of a single trade. This benefits all participants of the marketplace.
Exchanges and ECNs offer a rebate for any market participant who posts their market at the exchange and there is a very good reason for this. While specialists are obligated to make markets, the HFT has no such responsibility. In fact, the HFT is wise to hold back and test the waters before jumping in and being the first market maker on the scene. This is because if the HFT posts first with a market that is too high or too low, the HFT runs the risk of multiple losing trades because of the speed at which other HFTs will respond to take advantage of their “bad” market. The HFT may then acquire a significant long or short position that is on the wrong side of the market. To overcome this disadvantage the exchanges offer a rebate to those that post of typically 25 cents per 100 shares, and charge a fee to those who take liquidity of typically 30 cents per hundred shares. The rebates encourage the HFT to take the risk of posting first and also allow them to make money from making a very narrow market.
So here’s what happens when the HFT makes a tighter more efficient market and the simple example (from above) becomes much more efficient and competitive. Instead of the market being 10.00/10.05 the HFT posts 10.02/10.03. Customer A buys 100 shares at 10.03 from the HFT. The HFT resets the quote to 10.03/10.04 and Customer B comes in to sell 100 shares and gets a price of 10.03 the new price from the HFT. Customer A has saved 2 cents per share (less the take fee) by buying at 10.03 instead of the non HFT price of 10.05, and customer B has also saved 2 cents per share(less the take fee) by selling at 10.03 instead of 10.01. The HFT has “scratched” 100 shares (buying and selling 100 shares at the same price), something that the specialist cannot make money doing. But the HFT has received two rebates, one on the sell to Customer A and one on the buy from Customer B. The rebate to the HFT totals 50 cents, and both customers save $1.70 on this sample trade. The rebate was the incentive for the HFT to make a much more efficient market by making a tighter bid/ask for the customers of 10.027/10.033, which in the days of manual trading would have not been profitable. The HFT makes money by repeating this process thousands and thousands of times in the course of the trading day and effectively managing the risk from such a narrow market.
In addition to the risk of declaring the market first, the HFT also runs the risk of having the posted quote become stale due to rapidly changing market conditions and being traded against with a quote that is not current. The principal reason the HFT’s need to co-locate is to compete with each other, not to compete with the customers. If you are a slow HFT, other HFTs will take advantage of your markets. To keep the quotes as current as possible the HFT will position their computers in the same facility as the exchange to cut down on the travel time it takes to refresh their quotes at the exchange. Co-location is available to anyone who is willing to take on the expense and management of this opportunity. It becomes fairly expensive when you factor in the servers, power, telecommunications and staff that it takes to operate in this fashion. HFTs will occasionally trade with each other, but it is the competition between them for customer orders that keeps the markets narrow and efficient.
High frequency trading is an automated version of the specialist model. The HFTs have brought significant efficiencies to the market place by breaking wide open the specialist monopoly and competing with each other for customer orders. HFTs offer customers tighter and better pricing (less cost to trade) and much faster access to the markets for quicker fill times. The HFTs speed and efficiency coupled with rebates make it possible for HFTs to earn a living by making a market that is effectively one half cent wide and available to all buyers and sellers.
The Next Generation of HFT Trading
In the specialist model a good floor broker would ask the question of the specialist “What is the market?” and get a picture from the specialist of what size might be available up or down a few ticks or where there might be size available to trade. The lack of this information in the HFT model has made it more difficult to execute large orders. Traders have resorted to parsing out their orders into the small lot market. There is a new ATS, called PDQ which restores this functionality in the high speed electronic marketplace, causing the HFTs to compete for orders. A full discussion of PDQ, the next stage in the evolution of electronic markets, exceeds the scope of this paper. To learn more go to http://www.PDQATS.com.
*HFTs should not be confused with other forms of electronic trading. Traders can now use computers to automatically execute trades in a similar fashion to the old manual days, such as rebalance portfolios, trend follow, invest and speculate. Unlike HFTs, these programs are not involved in making markets and providing liquidity and they do have market impact.