SEC uses algorithms to see billions in errant bond trades
NEW YORK - Jesse Litvak kicked off a widespread investigation into the opaque world of bond trading by accidentally sending a client a spreadsheet detailing how he was allegedly ripping them off. Wall Street's main regulator is trying to make sure future probes aren't relying on similar stumbles.
The Securities and Exchange Commission has been building its own algorithms over about the past two years to better police trading of securities tied to mortgages and auto loans. In doing so, it says, it's uncovered billions of dollars in trades that may raise red flags - from excessive markups on bond sales to possible kickbacks for brokers who act as middlemen.
"We've identified billions of dollars of potentially problematic trades," Mike Osnato, head of the SEC's complex financial instruments unit, which helps examine over-the-counter bond markets, said in an interview. "We have opened promising investigations thus far based on these efforts and expect more to follow soon."
The push is part of the SEC's wider effort to harness big data to regulate markets after facing lawmaker criticism that it failed to detect conduct that led to the 2008 financial crisis. The agency began making its technology strategy more of a focus when it started the Center for Risk and Quantitative Analytics in July 2013. Staffed with computer programmers and mathematicians, the group is spotting suspicious trading and starting investigations, rather than waiting for complaints or a plunge in prices to spur action.
By combing through mortgage and auto loan trading data from the Financial Industry Regulatory Authority, which is the most available among securitized debt, the SEC's proprietary software is discovering new connections across Wall Street. For the first time, the SEC can find out which banks trade most often with each other in those markets. While this type of information doesn't necessarily represent securities violations, it gives the agency a more complete view of market players.
The software is helping the SEC advance its investigations of bond parking, which is when traders place their bonds with accomplices with an agreement that they'll buy them back at a higher price at a later date. By keeping tabs on individual bonds, the regulator can see whose balance sheets they wind up on. Traders often park bonds to increase their compensation and circumvent bank rules that restrict holding debt for an extended period of time. To cover up bond parking, some buyers and sellers are routing trades through an interdealer, who then may receive a kickback.
Historically, the SEC and the industry-funded regulator Finra have monitored stocks and options more closely than the world of bond trading. Before the credit crisis, the SEC lightly regulated these markets, in part because the agency viewed market participants as sophisticated, requiring less protection than individual investors. That assumption came undone when plummeting prices in the market kicked off the crisis.
Relying on data has its pitfalls. The government is trusting dealers across Wall Street to report accurate information on trading. Data-driven software often uncovers what looks to be suspicious behavior, but it can't ferret out what is a securities violation or not.
"You'll always get a large number of false positives no matter what algorithm you're running," said John Reed Stark, a consultant who previously worked as an SEC enforcement attorney and helped lead technology projects at the agency. "Once you get the extraordinary behavior, there is still quite a lot of work to do."
The trading scrutiny comes as the agency is probing further into credit markets, where securities often lack transparent pricing. The SEC has discovered potential collusion in the pricing of credit-default swap indexes and potential conflicts of interest in how issuers create bonds made up of company loans.
Investigations into securitized debt markets have yielded some of the agency's highest profile cases. Litvak, the former Jefferies Group LLC bond trader, was convicted of criminal securities fraud for misleading clients about how much he was making on bonds he was selling.
That conviction was overturned Dec. 8 because the lower court improperly excluded testimony from Litvak's expert witness that misrepresentations in the bond market are widespread. The probe started when Litvak inadvertently sent an email in November 2011 to a client with a spreadsheet containing the actual prices he paid for the securities.
Even with the loss, the Justice Department, along with the SEC, is moving forward with as many as a dozen criminal cases that are modeled after Litvak's conduct, two people familiar with the cases have said. On Dec. 21, former Royal Bank of Scotland Group Plc bond trader Adam Siegel pleaded guilty to lying to buyers. The agreement comes with a big caveat: He can withdraw his plea if Litvak is found not to have broken the law.