Market Data Feed Fight, a SIP Alternative and … Uber

23 Mar 2016 by Maystreet

The question of market data fees is central to the entire high-frequency trading discussion and to exchanges’ profitability. But the exchanges’ move to continue to raise the costs of data feeds could backfire, as alternatives could emerge that would make higher-quality data available to a much wider audience and undercut exchange volume.

In a recent commentary on market data fees, TABB Group CEO Larry Tabb touched on several points that deserve additional thought, including how fewer market participants reduce diversity and thereby harm volume; market alternatives for SIP; and an open-ended question of how innovation may solve the current market-data conundrum (“Fight Over Market Data Fees Is Going to Get Ugly“).

Impact of reduced diversity

Facing significant increases in feed costs, many small firms have chosen not to migrate to newer, more expensive versions of existing feeds. Fewer participants mean less diversity of strategies, which can increase volatility.

Here’s an example that doesn’t require low-latency data but demonstrates that participant diversity matters: the quant crash of 2007. Many quant funds were working with similar ideas for models and one of the funds needed to liquidate. Shortly thereafter, a much larger portion of the market had the same type of position that the liquidating firm did even if they didn’t start with that position at the start of the market turmoil. Why’s that? Because as the failing firm started to sell its position, those same positions looked attractive to similar, non-liquidating strategies, and as time passed, the market as a whole had purchased some or the entire portfolio being liquidated. A common reaction for many firms that newly entered the position was to exit them even more quickly. At that point in time, it’s easy to see how, under certain conditions, everyone thinking similarly causes them to rush for the exists at the same time, but possibly even faster because there are more of them trying to get out of the way.

The conclusion: Creating regulations that knowingly accept a reduction in participant diversity is bad policy for all investors. Ensuring participant diversity does fit within the SEC’s mandate of ensuring orderly markets.

Still want to take a SIP?

Another point Tabb made is that allowing market versions of SIP (Securities Information Process) would be highly beneficial. I agree, as it would reduce the systematic risk that exists now in the SIP, improve service and offer a much wider range of functionality that reduces the technology burden for many firms.

For example, here is a recent plot of how much faster and less variable proprietary feeds are:

(For an interactive version, showing 3-second; and 1-, 3- and 10-minute data resolutions covering BATS, BATSY, DirectEdgeA, DirectEdgeX and NASDAQOMX, see http://maystreet.com/app/sipdiff.)

It shows the difference in receipt time of trades/quotes from the SIP feeds (CTS, CQS, UQDF, UTDF) and the proprietary depth-of-book feeds in microseconds. The difference between 0 and the baseline is the average of ~800us, a difference that could likely be nearly eliminated and variation significantly reduced, given the ability for market-based solutions.

Execution quality

This raises a question: How can brokers actually provide best-ex if they have data this bad?

Here’s a wild idea to fix market data fees, an idea I’ve heard in a few places and thought worth repeating here: What if a number of important market participants were willing to contribute, in low-latency real time, their order and trade information to some third-party consolidator as they send those same orders to an exchange? The orders would be anonymized and built into a new feed. The consolidator could package and distribute a feed that’s not a 100% match of the feed – but could in principle be faster than exchange feeds and provide a give-to-get business model. In other words, firms that contribute can use the feed for some minimal charge.

Solving a conundrum?

The question of market data fees is a complex one, central to the entire HFT discussion. As Tabb suggested, the exchanges would likely be well-served to propose some revenue-neutral alternatives allowing for wider use of existing feeds by smaller firms – instead of what may be a legal decision that goes far worse for them. What’s more, an improved SIP would make higher quality data available to a much wider audience, especially brokers, who in effect have difficulty answering what best execution is because they’re using SIP data.

Finally, if things don’t change soon, some firm may just come along and build something that causes exchange volumes to change in a way that’s anything but revenue-neutral.

If you don’t think it’s remotely possible, just ask a cab driver what he thinks about Uber.

Sources: