Stats › Intraday and market microstructure › Algorithmic trading
Algorithmic trading narrowed large cap spreads by half and raised the market maker's take
| Measure | Largest quintile | Second | Third |
|---|---|---|---|
| Quoted spread | -0.53, t = -3.23 | -0.42, t = -2.21 | -0.43, t = -1.44, not significant |
| Effective spread | -0.18, t = -2.67 | -0.32, t = -2.23 | -0.35, t = -1.56, not significant |
| Five minute realised spread | +0.35, t = 3.52 | +0.76, t = 3.97 | +1.03, t = 2.06 |
| Five minute adverse selection | -0.53, t = -3.56 | -1.07, t = -4.08 | -1.39, t = -2.06 |
Instrumental variable coefficients in basis points per unit of algorithmic trading, where a unit is 100 dollars of volume per electronic message, with the sign flipped so that positive means more messages per dollar traded. 1,082 US stocks, daily, 2 December 2002 to 31 July 2003, identified by the staggered rollout of automatic quoting from 29 January to 27 May 2003, with stock and day fixed effects. First stage F statistics run from 5.88 to 7.32, which is weak instrument territory and worth flagging.
Using a staggered exchange upgrade as an instrument, algorithmic trading narrows the largest quintile's quoted spread by 2.41 basis points on a mean of 5.19. The realised spread rises at the same time, so the intermediary captures more per trade.
Members only
The rest of this page is for members
Below this point there are 5 sections, 1 table and 4 named sources, roughly 700 words of it. Every figure carries the source it came from and the date the data is from.
You can keep browsing every statistic in the library for free. The intro and the summary are always open.