HFT and fallacy of spreads
Gust post by Nanex.
Occasionally a new study comes out claiming HFT narrows spreads. There are a number of questions you need to ask about these claims, such as how recent is the data set analyzed, what is the resolution (1 second, minute or daily), how are spreads measured, and what subset of symbols were used in the study? For example, this article from The Economist uses the following graphic supplied by a pro-HFT camp which purports to show tighter bid/ask spreads. Never mind the obscure explanation (or lack of) for how this one line is calculated, you simply need to look at the dates to realize someone doesn’t know their history very well. HFT trading was born with Reg NMS in early 2007 (left circle), and shortly thereafter, spreads actually spike higher. Furthermore, spreads are right back to where they started (right circle), but we are stuck with all the negative HFT baggage. Remember, this image was supplied by the pro-HFT camp!
The chart below on the left illustrates one aspect of just how difficult measuring quote spreads can be. It shows the number of stocks with increasing (red) and decreasing (black) spreads for each second for about 1 minute on February 27, 2012 around the time that market traffic exploded with active ETFs such as SPY, IWM, QQQ, etc. experiencing locked or crossed markets.
Note how the red line spikes suddenly at 12:50:41, indicating almost 1,600 stocks had increasing spreads. The very next second, the red line drops, and the black line spikes indicating approximately 1,100 had decreasing spreads. This repeats again just 8 seconds later. Note how every red spike is immediately followed by a black spike, and in most cases, the red spike is higher and narrower than the black spike. This indicates the highly volatile nature of quote spreads: where the bid/ask spread of 1,000 or more stocks stocks can suddenly increase within a second.
The chart on the right is a randomly selected chart with a similar surge in market activity on a day before Reg NMS (and HFT). Note the scale difference: many fewer stocks had volatile spreads. The chart on the right is typical of spread behavior before HFT.
See also Quote Spread Disintegration.
|February 27, 2012||January 25, 2007|