Venue Information Share
I came across a recent paper by Bj?rn Hagstr?mer and Albert J. Menkveld , "Trades, Quotes, and Information Shares" in which the authors quantify how much information trades and quotes on different London venues (lit and dark) contribute to price formation. I love the premise but find the methodology too mathematical. It does not lend itself easily for a non-academic audience.
Plus, it would be nice to have some numbers for German-listed stocks. So, here is my much simpler methodology.
Methodology
Given the BBO time series from Xetra, Cboe (lit), and Aquis (lit), calculate the time series of the EBBO microprice defined as
where mp denotes the microprice, bp and ap are the European best bid and ask respectively, and bq and aq are the quantities on the EBBO across all venues. Note that we cap the deviation from the mid price to half a tick size (ts). This is a necessary modification of the microprice for securities with multi-tick spreads.
For each update of the microprice also store the venue which caused it, the relative change from the previous value (in bps), and the T+10 s microprice.
We then decompose changes into signal and noise. Signal is any change which moves the microprice towards its T+10 s value. Noise is any change in the opposite direction or any overshoot.
Finally, aggregate the signal and noise contributions across securities and dates.
Properties
This KPI has a number of desirable properties:
Results
The result is shown in Fig. 1 below for DAX40 constituents on Xetra, Cboe (lit) and Aquis (lit) for April 2024.
The orange markers denote the net signal (signal - noise). The annotations are the net information share and the information per BBO update on this venue (Xetra = 1).
Obervations
References
Hagstr?mer, Bj?rn and Menkveld, Albert J., Trades, Quotes, and Information Shares (February 13, 2023). Available at SSRN: https://ssrn.com/abstract=4356262 or https://dx.doi.org/10.2139/ssrn.4356262.
Stefan Schlamp could this be a side effect of Xetra’s physical proximity to Eurex/DAX execution and therefore the market leading information is really coming out of Eurex? Do these dynamics hold for FTSE?
Professor of Finance
9 个月Stefan Schlamp, thanks for your interest in our work. Indeed, the strength of your approach is that it is straightforward to calculate and it has intuitive appeal. But, if one is interested in identifying how much information arrives at all venues simultaneously, confidence intervals, convergence beyond T+10 s, impulse-response functions, etc., then one needs an econometric approach. Bj?rn and I follow in the footsteps of Joel Hasbrouck's seminal work on dissecting price discovery. Again, thanks alerting us to your analysis. It would be nice to compare what you find to what our approach would yield.
Great article! If you add the latency, it may amplify your result
Professor of Finance at Stockholm Business School
9 个月Stefan Schlamp, this is great! The results make a lot of sense and I like that you base it on the microprice rather than the midpoint. Our approach differs in the way we separate signal and noise, but in this case I think the results would be similar. Ping Albert J. Menkveld