Garrett Bauer had been on the radar of securities regulators for years. The stock trader made big profits by betting on dozens of corporate take- overs but investigators could not figure out where he came by such reliable and accurate information.
Using a fairly crude computer program that analysed billions of trades collected over 25 years, investigators at the US Securities and Exchange Commission started running queries to find out who else traded the same stocks. A name came up: Kenneth Robinson, a stock and mortgage broker. The trading suggested the men knew each other but investigators could not prove illegal activity.
Finally in 2009, a breakthrough for the SEC led to Federal Bureau of Investigation agents knocking at Mr Robinson’s door. That year, he had bought shares in 3Com just before it was acquired by Hewlett-Packard. It was the only time both men had traded in advance of a deal where Wilson Sonsini Goodrich & Rosati, a law firm, acted as an adviser.
The trade confirmed the SEC’s suspicion that the two men were working together. After agents persuaded him to co-operate, both pleaded guilty along with Matthew Kluger, a senior associate at Wilson Sonsini who linked the men together. In 2011, at that time the biggest ever insider trading case, they admitted operating a $37m scheme that lasted 17 years.
The SEC caught the men by switching to a sophisticated, technology-based strategy that allowed them to focus on the traders themselves rather than following the movement of the stocks. Known as parallel trading, the tactic is one of several tech-focused methods the SEC is deploying to keep pace with high-speed traders and fast-growing markets.
As Wall Street pours billions into high-speed trading platforms – along with big salaries for top software developers – the question is whether regulators can keep up. Mary Jo White, SEC chairman, insists it can.
“This is not your father’s SEC – or your mother’s or even your older brother or sister’s,” Ms White told regulators and industry officials in January on the first anniversary of her nomination to be the US’s top securities regulator. “In this rapidly changing environment, we must stay on top of advances in technology.”
The SEC’s struggle to keep pace with the well-funded securities industry is not new. But Ms White inherited an agency still recovering from a particularly low period, marked by embarrassing failures to detect fraud and other wrongdoing in the run-up to the financial crisis.
In 2008, the SEC was faulted for missing numerous red flags that could have exposed Bernard Madoff’s Ponzi scheme much earlier. The convicted fraudster sent account statements with suspiciously consistent returns to investors while markets gyrated.
The most attention the agency received for its tech-savvy ways was an embarrassing one: 31 employees were criticised by its own inspector- general for watching porn at their office PCs during the financial crisis.
In 2009 the budget for investments in new technology fell to half of what it was in 2005. The following year, the SEC was caught flat-footed during the “flash crash” when the stock market fell more than 700 points in minutes. It took nearly six months for staff to trace the trades to determine what went wrong, taxing the agency’s aged computers and shaking investor confidence in the markets.
Since then, the SEC has hired quantitative analysts, built new products and partnered with external data providers and software groups to speed up investigations. It has also passed rules to collect more trading data that it can plug into these systems for deeper analysis. Congress allocated a $50m fund geared for technology spending.
The SEC is launching a multi-pronged approach involving a vital new tool for its examinations programme, which conducts routine checks on hedge funds, brokers and exchanges. The national exam analytics tool, known as Neat, analyses very large volumes of trading by hedge funds against the broader market to identify insider trading and the front-running of stocks.
Another analytical tool developed by the SEC’s chief economist will look for outliers in financial statements, or unusual trends such as changes in auditors, to detect accounting fraud.
The hub of the makeover is the SEC’s enforcement division. “We are developing very advanced systems to detect insider trading,” says Andrew Ceresney, director of enforcement at the SEC. “They will allow us to see patterns of traders trading in unison.”
Mr Ceresney is part of Ms White’s strategy of targeted recruitment, hired from Debevoise & Plimpton, a law firm where Ms White was a partner and his former supervisor when they were both federal prosecutors.
Two pioneers developing the “parallel trading” approach to investigations were Dan Hawke and Sanjay Wadhwa, who in 2007 investigated insider trading. Mr Hawke, a 15-year veteran of the agency, comes from a regulatory family: his father, John Hawke, is a former Comptroller of the Currency, who was also an undersecretary of the US Treasury and former general counsel to the Board of Governors of the Federal Reserve System.
Mr Wadhwa emigrated with his family from India to the US as a teenager. He joined the SEC in 2003 after working as a tax attorney. Mr Hawke is more tech-minded while Mr Wadhwa’s desk is covered in stacks of papers and sticky notes.
“We began to see an uptick in the number of cases involving licensed securities industry professionals,” Mr Hawke says. By flipping the investigative focus to people instead of stocks that “opened up the possibility of making connections between traders across multiple securities”.
With a shoestring budget, staffers built an Excel spreadsheet that pooled billions of trades collected from old investigations via “blue sheets” into one big database. The SEC devised algorithms to sift through that data and “in a matter of hours matched billions of trading records to see what kinds of patterns emerged. That was an epiphany,” says Mr Hawke.
The unit has recently hired nine specialists who have designed more sophisticated ways to search the data sets. “If you had asked somebody four years ago what is the relationship between the skills of a quant and the way we conduct law enforcement investigations into securities trading, we could find no template in the federal government. We had to basically invent it,” he says.
After the technology provides the leads, the SEC scours phone and bank records and works backwards from there. “You still need a lot of hard work to make the case, but it makes it quicker and easier,” adds Mr Ceresney. The system, he says, was cumbersome and ran on an antiquated version of Windows software.
A more sophisticated version expected to be rolled out this year will identify “heat maps” and eventually, he hopes, pull information from social media websites to identify connections. The regulator’s newest weapon comes from its Quantitative Analytics Unit, led by Erozan Kurtas, who was recruited in 2010. Mr Kurtas has a PhD in electrical and computer engineering and owns 20 patents on algorithms and statistical techniques.
In a recent examination the SEC said that in a few hours Neat analysed 17m transactions by one investment adviser. The program is still new but SEC attorneys say that over 36 hours it will be able to examine all the trades made by a hedge fund manager in one year and make comparisons against broader market data. For example, officials say the SEC could run the program to detect any aberrations, such as how many times a stock traded by the manager rose or fell by 10 per cent over a week.
If the hedge fund manager bought at $10 and sold at $9.50, it may not stand out to the human eye as suspicious. But with the new system the SEC will be able to see that the stock later fell to $5, giving the hedge fund manager who anticipated negative news the chance to avoid a big loss.
Ms White said “examiners will be using the Neat analytics to identify signs not only of possible insider trading but also front-running, window-dressing, improper allocations of investment opportunities and other kinds of misconduct”. It could be expanded to brokers to identify excessive fees, one former official said.
Other initiatives include an “accounting quality model”, which staff can use to mine annual reports and other financial disclosure forms for red flags that could indicate accounting fraud.
The agency is also institutionalising an “aberrational performance” initiative started in 2010. The aim is to identify hedge funds that do better than their peers or make steady returns even in down markets. Eight cases have been brought so far.
The SEC is paying for some sophisticated external software. It is piloting a project with Palantir, the software company backed by the Central Intelligence Agency, to further foreign bribery investigations and the parallel trading initiative.
Mr Ceresney says it has identified links by accessing multiple data sources. For example, it can match stock trades made one day with money transfers executed three days later. “It’s not stuff we couldn’t have done before. It would have taken weeks and now it takes minutes.”
The SEC is betting the new programs will make it smarter in policing the markets and deploying its investigators.
Much effort has been put into reshaping the image of the SEC as a modern tech-savvy agency. But current and former officials caution that while the new technology will speed up investigations it is not a panacea. Old-fashioned legwork is still paramount.
Some former SEC officials say that as well intentioned as the initiatives are there are doubts about their effectiveness. Others suggest the current efforts will be limited until the real-time surveillance system, which was designed in the aftermath of the flash crash, is implemented. The system has been delayed until 2016.
“Everything they’re doing now is meaningful but it’s time-intensive and has limitations. Until the data are granular, complete and clock-synchronised, the SEC will have challenges in this complex market with nanosecond trading, in figuring out whether there is an unfair playing field,” says Thomas Sporkin, a former SEC enforcement attorney. “It will be great to see what they can do in 2016 when they get the data.”
Regulators also have to contend with the political flows of Washington and its effect on hiring and operations. In January, two weeks before Ms White addressed the California crowd, Congress approved $1.35bn for the securities watchdog, about $324m less than it requested. It halved the reserve fund, the purse for multiyear technology projects, to $25m.
“One of the dangers for the SEC historically – because of the budget volatility and uncertainty – is that it’s either feast or famine on technology projects,” says Mary Schapiro, a former SEC chairman.
“Technology really requires concerted, consistent spending to make sure systems stay up to date as well as developing new tools when they’re needed.”
终于在2009年，SEC取得的一项突破让联邦调查局(FBI)的探员们敲响了罗宾逊的家门。那一年，罗宾逊在3Com公司被惠普(Hewlett-Packard)收购前夕买入了前者的股票。这是罗宾逊和鲍尔两人仅有的一次双双在并购交易公布前买入相关公司股票，而担任此次交易的顾问是威尔逊-松西尼-古德里奇-罗萨蒂律所(Wilson Sonsini Goodrich & Rosati)。
随着华尔街投入数十亿美元资金开发高速交易平台——顶级软件开发商也因此获得了丰厚报酬——问题的关键在于监管当局能否跟上这一趋势。SEC主席玛丽?乔?怀特(Mary Jo White，见右图)对此给出了肯定的回答。
怀特制定了定向人才招募战略，塞雷斯尼就是她根据该战略从德普律所(Debevoise & Plimpton)挖过来的。怀特曾是德普律所的合伙人，而当怀特和塞雷斯尼都担任联邦检察官时，怀特还曾是塞雷斯尼的上级主管。
开发出“平行交易”调查方法的两位开拓者是，曾于2007年调查内幕交易的丹?霍克(Dan Hawke)和贾伊?瓦德瓦(Sanjay Wadhwa)。霍克是一位在SEC工作达15年之久的资深人士，来自金融监管世家：他的父亲约翰?霍克(John Hawke)曾任货币监理官，并曾担任美国财政部副部长以及美联储理事会总顾问。
一个有望于今年推出的更复杂的版本将能识别出“热度图”，而塞雷斯尼希望，这套系统最终能够从社交媒体网站中抽取信息，进而识别个体之间的联系。SEC的最新武器来自其量化分析小组(Quantitative Analytics Unit)，该小组由2010年加入SEC的埃罗赞?柯塔斯(Erozan Kurtas)领导。柯塔斯是电气与计算机工程领域的博士，并拥有20项有关算法和统计技术的专利。