Use Occam’s Razor to Fix Reg NMS’s Market Data Revenue Allocation

In the movie, Contact, Jodie Foster’s character helped make Occam’s Razor famous.  As she explained, it states that all things being equal, the simplest answer is usually the right one…  Unfortunately, however, when the SEC wrote Regulation NMS, they did not follow that route and instead introduced complexity in several of the specific rules.   The section which changed the formula for apportioning market data revenue, in particular, is guilty of unnecessary complexity and fails to achieve its stated objective.    That formula, which allocates the revenue that derives from the professional users of the consolidated NBBO and trade feeds (commonly referred to as the SIP) was undertaken with a clear goal:  to increase the incentive for displayed orders.

The adopting release for Reg NMS describes the intention of the changes as providing incentives for participants to display quotes in order to improve price discovery.  Unfortunately, the rule, as written, does not achieve that goal.  To better understand this claim, consider the history.

During the period before Reg NMS was passed, market data was allocated based on the number of individual executions reported by each exchange.  This was flawed for two reasons:  First, it encouraged tape “shredding”, which was reporting a single trade in many small pieces (As an example, a 5000 share trade might have been reported to the SIP as 50 individual 100 share trades).  Second, and more important, the rule did not provide any incentives to post displayed quotes, which meant that it did not promote price discovery.  To correct this, the decision was made, in classic government fashion, to create an overly complex compromise solution.

The resulting rule within Reg NMS was a formula with 50% of the revenue to be allocated based on trades and 50% of the revenue based on quotes, but there are flaws in both implementations.  The trades component, to remove incentives for tape shredding, contained two changes.  First, the new rule introduced the concept of a “qualified trade” (of $5000) before a trade would count towards the new revenue allocation.  This, they reasoned, would reduce the amount of tape shredding, at least to a size of $5000 per individual trade.  Second, they changed the allocation methodology to give half of the credit based on number of trades and half of the credit to the dollar volume of the trades.   Interestingly, the SEC calculated that this would decrease the incentive for tape shredding by 87.5%[1], so they were aware that there was still an incentive.  In addition, this change did not fix the problem that trades which resulted from matched “dark” quotes counted the same as those that occurred when one side of the trade was a displayed quote.

The quote based calculation in the rule is also flawed, although, to be fair, the flaw is directly related to advances in technology.  The rule provides “credit” to quotes of at least 1 second duration, whether or not those quotes were executed.   With today’s technology, this rule makes no sense, but in those days, when only the fastest trading systems operated sub-second, and the majority of trading was still done by human traders reacting to quote changes, this was defensible.  Today, however, with the majority of quote interactions being driven by Smart Order Routers (SORs) or algorithms and with modern trading systems operating at speeds per decision that are sub millisecond, it is hopelessly out of date.   In addition, the rule is biased in favor of the exchanges that contribute least to price discovery, since it rewards those exchanges that are least likely to trade at a particular price level   Market participants know which exchanges are most likely to get executed first or last when they decide where to post orders.  Those that want to get filled, will judge the order queues and each exchange’s fill probability and place their orders accordingly. (Those are the orders that the designers of Reg NMS wanted to promote.)  On the other hand, if gaining market data revenue is the goal, participants might place orders on the exchange least likely to be executed, since those orders are more likely to last the requisite one second.

A far better solution would be to consider Occam’s Razor.  Since the goal of the rule is to provide incentive to display quotes for the purpose of promoting price discovery, there is a simple solution.  Instead of the current formula, allocate market data revenue based entirely on trades executed against displayed orders.   While this would not have been possible before the 2015 SIP upgrade, it is possible today.  (That upgrade required exchanges to provide their own quote and trade time-stamps to the SIP).  The logic would be to compare every reported trade on each exchange to that exchanges pre-existing best bid and offer.  Any trade printed at a price that matched that exchange’s bid or offer from immediately before the trade would qualify for credit.  Since each exchanges time-stamp for trade reporting and quoting should be synchronized, it would be relatively transparent and simple to calculate each exchanges allocation per stock.   The nature of how to aggregate all of the symbols to calculate the final revenue shares could be decided separately, either by trade weighting, equal weighting, market cap weighting, or some hybrid weighting scheme.  It could, however, be completely transparent if the rules were known in advance.

Note that this solution would provide no revenue to trades executed against hidden orders whether on exchange or reported on either TRF, but that is by design.  If the SEC wanted to incentivize dark trading, then the rule could be adjusted.  The key to this proposal is simplicity and transparency as well as aligning it to the goals of the rule.  It is also worth noting that this commentary should not be considered as support for the current SIP architecture or even for the concept of a single SIP.  I think that when there is a holistic review of Reg NMS that both a competing consolidator model or a multi SIP solution (with at least one SIP instance per main data-center should be considered.

The point is that this solution would allocate revenues to those markets that most contribute to price discovery, which is best determined by the quotes which are actually executed.  It would eliminate the (silly) notion that a quote only matters if it exists for over 1 second without an execution and, by doing so, reduce the incentive to display quotes without an intent to trade.  As a side note, this solution also reduces the issues raised over paying quote fees to exchanges with speed-bumps, since those can suppress execution rates.  Simplifying the rule in this manner would be consistent with Occam’s Razor, would be relatively easy to implement, and would be a major improvement to today’s system.   As a result, it is the type of targeted de-regulation that could be undertaken relatively swiftly.

[1] Page 266 of Regulation NMS adopting release (

One thought on “Use Occam’s Razor to Fix Reg NMS’s Market Data Revenue Allocation

  1. Indeed, this was the proposed model by us at the COBA Project for a European consolidated tape. I guess we did not do a very good job marketing our proposal if you had to source inspiration from Jodie Foster on market structure. Our model was slightly more granular in that data resulting from certain types of displayed events such as auctions and other price-forming events were awarded higher value relative to others through the ‘MMT model’ who’s primary prpose was to normalize market data into a common classification for consolidation. Of course, also as you suggest, this was all based on a competing consolidator model (which was confusing to some as being counter-intuitive to enabling consolidation) but more importantly, it allowed for open participation to non-exchanges to encourage innovation.

    Tape-shredding is illegal by the way so why must there be incentives to not conduct it? Incentives should be focused on rewarding value. Competing in price discovery is expensive yet, there is no cost to importing prices from elsewhere. I continue to believe that until the relative information value associated with market events can be recognized, there will always be dysfunctional markets.

    One could go further by suggesting that the information value harvested from these events which inform the entire market is more valuable than the activities which underpin them which serve only two parties. The commercial models used to harvest this value however, are of course ‘challenged’ to say the least…


Leave a Reply

Fill in your details below or click an icon to log in: Logo

You are commenting using your account. Log Out /  Change )

Google+ photo

You are commenting using your Google+ account. Log Out /  Change )

Twitter picture

You are commenting using your Twitter account. Log Out /  Change )

Facebook photo

You are commenting using your Facebook account. Log Out /  Change )


Connecting to %s