Nimalendran Mahendrarajah co-authored this article.
In a recent research article, we examined "Informational Linkages Between Dark and Lit Trading Venues" (Forthcoming in the Journal of Financial Markets; view the working paper version here). We found that the principal source of information leakage was from smaller trades initiated by algorithms that seemed to be spread across dark and lit platforms. We concluded this was a result of liquidity seeking algorithms parsing out orders over time and across venues.
The big block crosses did not show similar leakage. This makes sense because no one would be using an algorithm while leaving a sitting block order, for fear of leakage and getting gamed.
[HFT Debate: Two Experts Square Off On Market Structure ]
The upshot we took away from this was that execution traders can choose between gambling on an uncertain big block cross with minimal price impact or an algo suite parsing orders for a more certain fill at what would likely be a worse price. Perhaps a mixture of the two could also be used, but it was unlikely both would be used at the same time.
These different trading strategies suggests that the size of the order is strongly correlating with how truly "dark" the order was. In some senses, the small cross venues seem more like kiddy pools (shallow and without scope for large liquidity), and these were the pools that were more likely to be leaking.
Our paper documented a variety of other facts which were consistent with this story, but the differences in these types of trades remain one of the main findings. As such, there are concerns when we see articles regarding dark trading that do not make this distinction clear (or refer to dark pools without qualification), as the markets for the two types of dark trades have very different impacts on stock prices and cost of trading.