White: ‘Sea Change’ Created in SEC Exams By New Quantitative Analytics Unit Programs

A Securities and Exchange Commission unit whose members include a former computer game developer and a rocket scientist is leveraging quantitative analytics to strengthen the agency’s examination and enforcement efforts as it pushes to keep pace with advances such as high frequency trading based on algorithms.

Housed within the agency’s Office of Compliance Inspections and Examinations (OCIE), the Quantitative Analytics Unit (QAU) created a sophisticated software tool that is currently used by approximately 1,000 SEC employees—about one of every four workers. The unit, established in April 2012 to assist the SEC exam function, has supported more than 70 examinations and participated in at least 12 enforcement investigations.

The unit has created “a sea change” at the agency, SEC Chairman Mary Jo White told Bloomberg BNA on Nov. 18. “It basically allows you to deal with big data,” and in the process has revolutionized the SEC’s examination function, she said.

Projects in development include one that would improve SEC examinations of high-frequency trading firms and another to help rank registrant potential compliance risk, helping the agency select exam targets more efficiently and effectively.

‘A Touch of Idealism.’

While the SEC traditionally hires individuals with backgrounds in accounting, law or affiliated areas, “modern finance is built on and driven by the tools of financial engineering,” Erozan Kurtas, QAU head of unit, said at a Regulatory Compliance Association securities conference held in May in New York. The unit is designed to address what Kurtas called “this mismatch between the backgrounds and skill sets of people at modern financial institutions and regulatory bodies.”

While QAU personnel develop software tools to boost examination techniques and outcomes, they also accompany more traditional SEC examiners to on-site examinations. “We don’t sit in the office and write computer programs all the time,” Kurtas said. “We are examiners at the end of the day.”

“We sit side by side with examiners in the field,” Elcin Yildirim, SEC Quantitative Analytics Unit senior science adviser, told Bloomberg BNA in November. “We are trying to make the exams higher quality,” he said.

Similar to others who work at the SEC, quantitative analytics unit workers display a noticeable sense of public service and intent to safeguard investors from harm. “There’s obviously a touch of idealism in the group,” Yildirim said. “Some of us have had careers in the industry, so we may not be looking for that anymore. But I would say everyone shares some sort of quantitative background, communication and some idealism.”

“It’s kind of a new field,” he said. “You could call it quantitative compliance.”

Rocket Scientists at SEC.

Kurtas described his recruiting and hiring process as “very, very vigorous.” Four or five rounds of interviews are accompanied with required presentations, mathematics puzzles and computer science questions—a process similar to interviews conducted by top-notch hedge funds, he said. “I interviewed more than 100 people to hire a handful,” he said.

One recent hire is a former National Aeronautics and Space Administration rocket scientist with a background in studying black holes, regions in space where the pull of gravity is so strong that light can’t escape. “I told myself if he is able to deal with black holes, he should be able to deal with dark pools,” Kurtas said, referring to unregistered securities trading venues.

Another QAU team member worked in proprietary trading groups at top investment banks and a hedge fund before joining the SEC. Other QAU employees include:

• a former California Institute of Technology (Caltech) researcher who worked in exotic options and volatility arbitrage,
• statistics and machine learning specialists,
• a software developer with a Ph.D. in mathematics who designed and sold a computer game before working for the SEC, and
• quantitative analysts who have specialized in arbitrage trading and risk management.

While about 70 percent of the QUA team members have Ph.D. degrees not everyone does, “and they don’t need one,” Kurtas said. “What I’m looking for is curiosity, idealism, loyalty and problem-solving minds,” he said.

The unit has doubled in size in about six months, from eight people in May to 16. An SEC spokeswoman wouldn’t comment about the agency’s plans for hiring additional workers for the unit. “This is a good start but we need dozens and dozens more of these technical people to catch up to the industry,” Kurtas said.

Trading Activity Analysis.

The unit’s current projects have memorable names such as HAL, the name of a computer in Stanley Kubrick’s 1968 film “2001: A Space Odyssey.”

One QAU project that has been implemented is the National Examination Analytics Tool, or NEAT. The program, which supports examinations by analyzing very large amounts of trading activity, or big data, was built in-house by QAU team members in less than a year.

Before NEAT’s implementation, examiners would typically use off-the-shelf business software to scrutinize a finite amount of registrant trading data. With NEAT, examiners can now translate data into a common format and run a series of tests on that data to detect indications of fraud and other securities law violations, as well as weaknesses in a firm’s internal controls.

“With NEAT, you can literally look at 17 million transactions in 24 hours,” White said. “And you can look for insider trading, you can look for front running. It used to take you months and months and months to do that.”

“This will drastically increase the chance of detecting violations and manipulative patterns,” Kurtas said.

The SEC is already at work on its next, improved version of NEAT.

HAL and MARS.

Another QAU project, the High-frequency Analytics Lab or HAL, is intended to allow SEC examiners to collect large amounts of data from high frequency trading (HFT) firms and analyze hundreds of thousands of orders a day. When implemented, HAL, which is still in the development stage, could be used to learn about the purposes and motivations behind trading activities as well as to measure the impact HFT firms have on U.S. financial markets more generally.

A project known as MARS—Machine Analyzed Risk Scoring—when implemented will quantify different risk levels of registrants by using machine-learning techniques on a wide range of data sources. While it is it only one input, if the program is successful, it will be able to help the agency determine those firms most at risk of not complying with securities laws.

The MARS project’s importance to the agency increases as the number of entities the SEC regulates increases, because the program will help the agency utilize its examination resources by improving the process used to select registrants for examinations.

Next Steps.

A common thread to the projects is not only making the examination process more robust, but making it more efficient. Using the analytics tools, SEC examiners will be able to focus future exams on those issues posing particular compliance risks rather than conducting more general examinations, reducing both the time needed to conduct a more general exam and the administrative burden on registrants, SEC employees said.

“We want them figuring out how to empower our examiners in the field,” OCIE Director Andrew Bowden told Bloomberg BNA in an October interview. “Give us the ability to do sophisticated, rigorous analytics in the context of every exam we do.”

What’s next for the unit? It is currently developing a program designed to detect anti-money laundering violations. An initial version of the program was used in a 2014 examination conducted out of the SEC’s Los Angeles regional office.

In the midst of early successes and anticipated future project rollouts, Kurtas and his colleagues continue to think about new ways of improving the SEC examination function. “I and my team, we prefer to fail trying to solve very hard problems rather than succeed by going after very easy ones,” he said.

To contact the reporter on this story: Stephen Joyce in New York at sjoyce@bna.com

To contact the editor responsible for this story: at @bna.com