Brokers Routing to their own ATSs Face Scrutiny

A FINRA working paper has raised concerns about broker order routing to affiliated alternative trading systems (ATSs), shedding light on order routing decisions and potential conflicts that can impact execution quality.

The major finding is that institutional orders routed by brokers that send a relatively high percentage of institutional orders through affiliated ATSs receive lower fill rates and higher execution costs, according to a working paper by FINRA’s Office of the Chief Economist, published in late January.

Though this is not the case with all brokers that own ATSs, FINRA was able to analyze a massive amount of data to identify distinct patterns of order routing behavior.

Using data from the Order Audit Trail System (OATS), FINRA completed a detailed analysis of order handling information over the life of 330 million institutional orders routed by 43 active institutional brokers for a size stratified sample of 273 stocks, or 2.3 billion orders, in October 2016.

FINRA used the FINRA ATS Transparency data, which had 32 ATSs on the list during the sample period.

Buy-side firms have long suspected that their orders may have been exposed to many venues during an order’s life cycle, but they have lacked the industry-wide data necessary to prove it.

In FINRA’s sample, the institutional top order size was 1,371 shares with the average share in the order routed approximately 10 times in the order’s life cycle.

Orders coming to the brokers were received either electronically or manually. They could also be generated by the broker’s order routing algorithm, according to Jonathan Sokobin, FINRA’s Chief Economist and Senior Vice President in a follow-up email.

As the paper notes, it’s been difficult for institutions to assess the performance of their brokers, even though they utilize transaction cost analysis “especially since routing data is either unavailable or hard to decipher.”

Opaque reporting has prevented the buy side from accessing this data, notes the paper. And, with 13 stock exchanges and 45 dark pools, US equity market structure is complex to assess.

Results

While FINRA’s paper states that not all brokers show a preference for routing institutional client orders to affiliated ATSs, some brokers demonstrated a persistent pattern of routing orders to affiliated ATSs.

The regulator divided brokers into three groups based on their average deviations from a sample wide benchmark. Brokers with the highest proportion of affiliated ATS routes were placed in the third group T3, while those in the lowest and middle groups were T1 and T2, respectively.

• T3 brokers routed 64% of the total quantity of client orders to ATSs, of which 50% is largely attributable to affiliated ATS routes.
• Brokers in the lowest (T1) and middle (T2) groups, routed 10.2% and 25.3% shares to ATSs, respectively.
• The average share received by T3 brokers is routed 16.6 times whereas the statistic is 4.5 times for T1 brokers and 9.4 times for T2 brokers.
• In terms of the connection between routing to affiliated ATSs and low fill rates, T3 brokers with the highest proportion of affiliated ATS routing had lower fill rates (17%) as compared with T1 brokers (44%) and T2 brokers (30%).
• Even though T3 brokers sent a high percentage of orders to ATSs, FINRA’s report found: “The majority of T3 brokers’ executions occur on exchanges. Notably, exchanges account for only 30.1% of T3 brokers’ routed quantity but 70% of executed quantity. “

Industry participants were quick to react.

“This paper is nothing short of jaw-dropping,” remarked Themis Trading’s blog post “FINRA Working Paper on Order Routing.”

“While anecdotally many of us may have sensed that brokers are routing in ways that are not in their client’s best interests (and we are not only talking venue costs here, we are also talking broker routing that feeds non-client short term trading alpha), this paper documents it with FINRA OATS data.”

As Themis Trading’s blog points out, the results of FINRA’s analysis are similar to a previous study by Babel Fish Analytics, a venue and order routing analysis firm which examines the routing sequence of smaller orders as they are sent to venues.

In a May 2017 blog post “The Truth Behind Broker Routing: Don’t Believe the Hype,” which is also cited in FINRA’s paper, Babel Fish provides an example of a broker’s conflicted routing behavior for an institutional order using an algorithm. In the example, 3,100 shares of an order were filled as a result of sending 1,200 shares to Venue A and another 1,900 shares were fully filled in Venue A two milliseconds later.

The concerning part is that 19 routes to other venues took place before the algorithm went back to Venue A.  Of these 19 routes, 7 were to the algorithm provider’s own venue, which resulted in only 100 shares executed.

Later in the routing sequence, even after getting full fills in Venue A, the router goes back to check the broker’s own pool and a new Venue H.

Affiliated ATS Routing & Trading Costs

In addition, FINRA investigated whether the lower fill rates were offset by favorable transaction prices or less information leakage.

FINRA estimates transaction costs based on the implementation shortfall method, which accounts for fill rates, the bid-asked spread, market impact and the drift in price during an order’s life cycle.

After controlling for intraday market conditions and other order characteristics, brokers with high affiliated ATS routes had fill rates that are 5.6 percentage points lower and implementation shortfall costs that are 1.13 to 1.95 basis points higher relative to other matched brokers, states the working paper.

It found that higher levels of affiliated ATS routes are associated with higher implementation shortfall costs.

These estimates translate to implementation costs that are 5.3% to 19.7% higher for brokers with high affiliated ATS routes.

“The evidence from this study does not support the idea that when a broker sends a high proportion of orders to an affiliated ATS, these venues necessarily offer a lower-cost option that avoids other market participants from learning of an institutional order,” said Sokobin in FINRA’s announcement of the paper.

Venue Economics

There are two reasons why brokers trade on the venues they own: It saves them money since they don’t have to pay another trading venue and second, brokers have an interest in maximizing the volume they do on their own ATS because higher amounts of trading attract other traders to the venue, which increases their trading fees.

“Historically, the argument that brokers made for preferring their own ATSs is that it reduces information leakage. By going to their own ATS, they’re improving trading performance on behalf to their clients, so there isn’t a conflict of interest.  This paper is saying there is no evidence of that,” commented Alistair Cree, product manager for TCA, Analytics and Algo Wheel at FlexTrade Systems.

“What this paper argues is that by constraining where an algo can trade to prefer the broker’s own ATS, this is actually going to harm the client’s trading performance,” said Cree. “People who are more likely to trade in their own ATS have worse performance overall,” he said referring to the paper.

By limiting the algo, there is an opportunity cost from sitting in the broker’s ATS, he explained.  “If your order spends five minutes sitting in the broker’s ATS, and not trading in the wider market, someone is going to take that liquidity,” said Cree.

If an order is routed to the general market, it will have a higher probability of getting filled, said Cree

However, many institutions prefer to have less aggressive orders routed to dark pools first to avoid information leakage before routing the unfilled portion to the general market.

It remains to be seen if the working paper will impact brokers routing a high percentage of orders to affiliated ATSs, but it certainly has raised awareness for the buy side.

Despite the enormous dataset used in this analysis, this is an observational study, and so it’s possible that other factors are causing the differences in performance, which were not considered by the authors. Rather than leap to conclusions based purely on observational data, this paper provides further evidence that experiments in broker selection and routing behavior are worthwhile for firms seeking to boost trading performance.

“It’s worth talking to brokers about their routing and their algos,” said Cree. “The brokers may not want to disclose that data because it provides a proprietary edge, but it’s up to brokers to weigh that decision on what to disclose.”

“Ultimately, brokers determine the routing of algos and it is brokers who can run these experiments on their algo routing and really establish causality,” said Cree.

Meanwhile, regulators are increasing disclosure requirements on order routing decisions given the focus on rebates, payment for order flow and other incentives that can clash with the broker’s best execution obligations.

FINRA mentioned best execution decision making as a top priority  in its 2019 Annual Risk Monitoring and Examination Priorities Letter to ensure that customer orders are directed to the best market.

“In particular, FINRA will review firms’ best execution decision-making where the firm routed all or substantially all customer orders to a small number of wholesale market makers from which they received payment for order flow or an affiliated broker-dealer or an alternative trading system (ATS) in which the firm had a financial interest,” stated the letter.

While the focus on order routing practices and disclosure is ongoing, brokers that show a preference for routing orders to affiliated ATSs can expect to receive a higher level of scrutiny, noted the regulator.