In 1989, when Goldman Sachs started its computer-led investing group, the term big data hadn't entered the lexicon.
Yes, engineers were using modeling techniques to explore large data sets. But Silicon Valley had yet to come up with a way to store and parse the growing pile of data created as computers took over every corner of society. Goldman could cast its eye over wide swaths of the investing universe - 13,000 companies - but it couldn't dig too deeply.
Fast forward to today and the depth of analysis now possible is a fundamental shift on how Wall Street puts money to work. Gary Chropuvka, the co-head of Goldman's Quantitative Investment Strategies team, which uses statistical methods to invest in public equity markets, knows this better than anyone. Chropuvka's group has nearly doubled its assets under management in two-and-a-half years to $174 billion.
"As quants we've always played the breadth game," said Chropuvka, in an interview from his 35th-floor office overlooking the Hudson River, where a pile of portfolio management textbooks sat in a corner. "What big data has helped us do is really play the depth game and know more about a company, not just their fundamentals and what's affecting those, but also what's going on around that company."
Chropuvka joined Goldman in 1998, moved to QIS the following year and assumed his current role in July 2017. He shares management with Armen Avenanessians, a former Bell Labs engineer known for pioneering the role of Goldman's deskside strategists and launching its vaunted risk-management system SecDB. They meet to review portfolio returns each Monday.
One of Goldman's fastest growing businesses
The business has been on a tear of late, and now counts among Goldman's biggest and fastest growing. The group's 100 investment professionals managed more than 10% of the assets across Goldman's asset management unit at mid-year. The big data, or equity alpha, sleeve accounted for about $45 billion, nearly tripling over the same time period. Smart beta strategies commanded another $110 billion, while those that try to mirror traditional hedge fund strategies had another $19 billion.The figures mean QIS can stand shoulder-to-shoulder with the largest quant funds in the world: AQR manages $194 billion, Renaissance Technologies oversees more than $110 billion, while Two Sigma Investments supervises $60 billion.
That success will help Goldman seize on one of the few secular growth opportunities it's got right now: a surge in demand for alternatives investing. Goldman CEO David Solomon has big plans to raise money for the bank's private-equity, real estate and hedge fund strategies. QIS, with its long track record and torrid growth, has a good story to tell and stands to gain from the firmwide fundraising push, even as it's avoided some of the politics that have come with merging other investment teams.
Quantitative investing is one of Wall Street's hottest trends. Investment managers are increasingly employing enormous computing power and teams of engineers to make money from alternative data sources like satellite images, credit-card transactions, and information scraped off the web. For those in search of new sources of alpha, the potential is too good to pass up.
Inside QIS, the data-crunching has come up with more than 250 signals that can be exploited to make money in the markets. The tally a decade or so ago was just 15. The group sorts companies by topics in addition to industry or financial metrics; parses language used in written research reports to gauge analyst sentiment; and analyzes options pricing data collected over the years by Goldman's trading desks, to name just three.
Shadows of the quant quakeThe outlook hasn't always been rosy. Goldman's Global Alpha hedge fund, which once managed $12 billion in computer-led strategies, lost 23 percent in August 2007. It was caught up in what would be dubbed the quant quake, an event that led to an existential crisis for the industry, and shuttered several years later.
Industry flows didn't turn positive until 2010, according to Hedge Fund Research. But turn positive they did: Quant hedge funds managed $979 billion through June, more than double 2008's tally, according to HFR. They now account for 30% of everything managed by hedge funds.
Still, critics persist. JPMorgan quant guru Marko Kilanovic has blamed recent market volatility on the trend-following strategies of the quants. Other hedge fund managers have taken aim as well: Omega Advisers founder and former Goldman partner Leon Cooperman said in December the Securities and Exchange Commission should look into computer trading. Stan Druckenmiller said the same month quants were introducing volatility into the markets that could be harmful for other managers.The theory was tested again last week when a massive shift in the stock market from top-performing growth stocks to lower-performing names triggered a sharp shift in "momentum" strategies. While the sharpness of the move reminded some of 2007's quant quake, one quantitative hedge fund manager said the industry losses weren't nearly as severe.
While Chropuvka declined to comment on last week's move, he said in the earlier interview that he doesn't agree with the criticism.
Pushing back against the critics
"I struggle with this whole idea, especially around the alpha crowding piece," said Chropuvka, citing the diversity of strategies that have been brought about by more data. "People have very different decision-making processes. It's not just around momentum, quality, and low vol. Many quants have value-type metrics, which again, usually aren't trend-type strategies that will promote herding. So I think one of the things that quant strategies actually have is diversification."
That will only increase, he said, as new data sources come online.
So with all this data coming at them, how do the execs in QIS separate the good from the bad? They go looking for raw and unstructured information that hasn't been modified by others and then apply a systematic, three-step approach to evaluate its usefulness.
'A huge edge'Step one uses economic intuition - does it make sense. Questions asked at this stage include: Can the team develop an investment thesis or belief around why it will make money? Where does the data come from? How is it stored? How should Chropuvka's team account for missing database fields?
Once the data is acquired, step two involves conducting research to see if the thesis holds up. Step three is checking to see if the signal or trend has already been incorporated into prices. After all, the data isn't worth much if the market has already figured it out.
"It's economic intuition first and foremost," Chropuvka said. "Then it's, does the data prove something? And then lastly, which is really critical, is that information reflected in prices?"One of the biggest hurdles to incorporating all the new data sources into an investment process is simply its cost. Goldman spent more than $1 billion on communications and technology costs, where data and data license expenses appear, over the 12 months through June.
Even here, Chropuvka has a secret weapon: where possible, the group shares data licenses with Goldman's traders and investment bankers.
"To the extent we have a license that spans the firm, and we know people have done their due diligence, we take comfort and we use some of it," he said. "We think that's a huge edge."