Brokers Wary About ‘Kill Switch’ Design

Brokerage executives are anxious about mechanisms being designed by stock exchanges that will automatically shut off a firm’s incoming orders, to prevent a market disruption or a spate of erroneous orders.

Although the exchanges are considering a “layered” approach, whereby a brokerage receives alerts before a complete shutoff, the Securities and Exchange Commission appears to favor a more abrupt solution. Chairman Mary L. Schapiro, at a roundtable on market stability last month said that, in automated trading, orders need to be shut off within two minutes, roughly, to be effective.

And while the brokerage houses themselves will be permitted to determine the thresholds that would determine the point at which the exchanges would cut off their orders, some brokerage executives worry those thresholds will be set too liberally to be of any use.

“We worry about automation kicking in at the wrong time and perhaps destabilizing the system,” TD Ameritrade chief technology officer Lou Steinberg warned participants at a roundtable sponsored by the SEC last month.

The so-called “kill switches” were proposed by an industry working group in September in reaction to the mayhem of Aug. 1, when systems operated by Knight Capital Group flooded the New York Stock Exchange with thousands of money-losing orders.

The SEC, concerned about the integrity of the marketplace, is pushing the exchanges to make sure such an event doesn’t happen again. The regulator wants the exchanges to install simple and speedy trip wires that automatically cut brokers off.

At the roundtable, the SEC did not appear to want too much wiggle room built into the setup. Chairman Schapiro expressed concern that alerting a brokerage before cutting it off may defeat the purpose.

“Is there time for that?” Schapiro asked about alerts. “It doesn’t take very long in these markets—a couple of minutes—for an enormous amount of damage to be done.”

The brokers, on the other hand, want to be alerted before they are cut off. For example, Steinberg wants a five- or 10-minute warning from an exchange before a cutoff. That way, he can explain whether the trading is normal or unusual. “We believe a layered approach with a human discussion probably makes the most sense,” Steinberg said at the roundtable.

The TD Ameritrade executive isn’t the only brokerage official concerned about a hasty cutoff. Chad Cook, chief technology officer at Lime Brokerage, a firm that acts as a conduit to the public markets for trading firms, is worried about the “downside risks to the people we just cut off,” he said during the roundtable. “How do we work around those things?” he asked.

Cook suggested it might be necessary to build other technologies to “help offset some of the downside issues” involved with a kill switch.

Vaishali Javeri, an attorney with Credit Suisse, also fretted about getting cut off without warning. Speaking at the annual market structure conference sponsored by the Securities Industry and Financial Markets Association last month, Javeri was apprehensive about the “consequences of a shutoff” and the effect on clients. She wondered whether a “multi-layered kill switch that takes more time to implement” might be a better approach. “The details must be thought out,” Javeri said at the conference. “It’s not as easy as we might believe based on the reports in the media.”

The “layered” approach to the kill switch is getting the endorsement of both brokerage and exchange officials, if not the SEC. The approach means that any attempt to cut a broker off could involve multiple warnings first—telephonic, electronic or both.

“Kill switches need to be part of multiple layers,” Nasdaq OMX Group chief information officer Anna Ewing said at the roundtable. “We need to ensure we don’t think of it as the Big Red Easy Button. It’s layered. It’s complex. There’s decision-making criteria. There’s that human element involved. You do an outreach. You make a phone call.”

Whether the industry eventually adopts a multi-layered approach or a single kill switch, the purpose of the mechanism is to prevent a brokerage from sending too much volume to a particular exchange over the course of a single day.

The details of the proposed kill switch were first spelled out in a letter sent by the working group to the SEC on Sept. 28. The working group consists of five exchange operators, the Financial Industry Regulatory Authority, the Depository Trust & Clearing Corp. and 16 broker-dealers.

Under the proposal, according to Lou Pastina, an NYSE Euronext executive vice president who participated in the SEC roundtable, a broker would determine its “Peak Net Notional Volume Threshold” for every symbol for a given day of trading.

This is not average daily volume. Rather, the metric is the sum of the firm’s net long positions and its net short positions. If, for example, a firm bought 1,000 shares of IBM and sold 600 shares of IBM, 1,600 shares would count against the threshold. Its net long position, however, would only be 400 shares.

Under the proposal, if the broker exceeds this net notional threshold, the exchange would be obliged to cut it off. Importantly, it is the broker, and not the exchange, that determines this metric, based on its historical trading records.

In addition, Pastina explained, the exchange could incorporate functionality that would automatically send an alert to the brokerage if it were to come within a certain percentage of the volume threshold.

The question of kill switches first arose in August after the Knight debacle. Traders wondered why NYSE didn’t unilaterally cut off Knight’s algos gone wild. NYSE would not comment on the Knight case, but officials at other exchanges told Traders Magazine that to unilaterally cut off a member would amount to second-guessing the broker’s intentions. The exchange could be wrong and cause the broker to suffer losses. That might open the exchange up to a lawsuit.

Thus, under the working group’s proposal, brokers decide how much volume is too much. That way the exchanges don’t assume any liability for loss if the kill switch triggers at the wrong time, sources said.

Still, placing the responsibility of determining the volume threshold on the broker isn’t expected to make the broker’s life any easier.

“It’s a ‘kill switch’ if somebody does it to you,” TD Ameritrade’s Steinberg noted at the SEC roundtable. “It’s a ‘suicide switch’ if you do it to yourself. There’s a big, big difference. And people are going to be reluctant to systemically cut themselves off from the market.”

Steinberg explained that the fear of unnecessarily ejecting themselves from the marketplace could cause firms to apply a wide margin of error to their volume threshold calculations.

“If we put in pre-defined limits, that suggests that we can, in advance, figure out all the combinations and permutations of all the ways things might go wrong,” he said. “I think our fear of getting it wrong will lead us to use artificially high limits—in which case we will have done a lot of work to not much effect.”

In addition to the overall threshold, brokers are going to have to agree on the level at which they receive an alert. The tentative plan is to alert a firm when its volume reaches, say, 70 or 75 percent of the peak notional amount.

David Bloom, UBS’s head of group technology for the Americas, said at the roundtable he was in favor of the kill switch, but that “we need to set those levels intelligently. That’s where a lot of the debate will come from. Because the last thing we want is a well-intentioned kill switch disrupting proper market activity.”

Despite the working group’s relatively simple threshold concept, arriving at those levels may not be as easy as picking a single number, roundtable participants indicated.

Saro Jahani, the chief information officer at Direct Edge, noted that the kill switch thresholds would depend on various “scenarios.” He told the SEC that “some of the scenarios can obviously be automated, but the question is: ‘How many?’” Jahani wants to see a working group come up with a list of “clear scenarios.”

For his part, Steinberg is skeptical that a single threshold based on one variable would be adequate. “You would need a combination of thresholds in place before you could start to trust the automation,” he said. “I doubt that any single factor would make sense.”

Possibilities for thresholds include the number of fills, the size of fills and overall volume of orders.

For the SEC, this level of detailed analysis and sophisticated construction may be too much. “It feels like we are adding a lot of complexity,” Schapiro said. “I worry about too much complexity. It might make it sophisticated, but harder to build and manage. It would be harder for everybody to know at what point their trading could be cut off. I don’t want it to be a surprise, but perhaps ‘blunter, but simpler’ might be a better way to go,” she told the assembled technology execs.