Predicting the Next Wall Street Disaster

Wouldn’t it be great if the U.S. had a heat map of the entire financial system that could alert it to vulnerabilities and approaching calamities before a global crisis struck?

While members of Congress squabble over a move in the House to erode provisions of the Dodd-Frank Wall Street Reform and Consumer Protection Act, a 64-year-old Wall Street veteran and MIT-trained economist named Richard Bookstaber is quietly working on an ambitious project that aims to do just that. “Most people in government don’t even know this is going on right now,” Bookstaber tells Newsweek of his work at the U.S. Treasury Department’s Office of Financial Research (OFR).

Using what are called “agent-based models,” Bookstaber’s project focuses on how the actions of individual agents—such as banks or traders on Wall Street—create chain reactions that cascade through a much wider ecosystem in ways that can threaten the global economy. “One of the earlier and simpler inspirations for agent-based modeling was looking at the migrating of birds,” Bookstaber says. “It’s not like they decide, ‘Hey guys, let’s make this V-shaped pattern.’ The birds are acting relative to the bird next to them. They’re in these patterns that can change in a second, and they just shift in a different direction. If you have a bird acting based on what the other birds are doing, you get this very complex, group-based behavior.”

In the wake of the financial crisis, agent-based models, or ABMs, were highlighted for their potential to help stave off the next disaster by the National Science Foundation and MIT, where Bookstaber earned his doctorate. However, their pecuniary applications remained in embryonic stages until recently. After more than three years of tweaking his prototype, Bookstaber says he’s hoping to start running live data by the end of this year: “This is a project we believe could have huge value not just for the U.S. financial system but, as we collect more data and test and improve our models, the global financial system as a whole.”

Often compared to a storm-warning system for the financial markets, the OFR was created by Dodd-Frank to be Washington’s brain trust for independent and rigorously researched financial data. Crucially, the OFR has subpoena power to seek out whatever information it needs, public or private—including that from banks and hedge funds—to build its models so it can alert the nation’s economic policymakers to impending threats.

Of course, models have been used before—and failed. Many financial models on Wall Street were treated as sacrosanct until the collapse of monster banks like Bear Stearns in 2008. But agent-based modeling has been effective in everything from tracking the spread of infectious diseases like SARS to forecasting the traffic patterns of automobiles and airplanes. And unlike the financial modeling that caused U.S. policymakers to miss the warning signs several years ago, agent-based models are designed to capture turbulent market features like bubbles and price crashes with built-in feedback mechanisms that can magnify even minor events when herd mentality takes over.

“Remember, some people initially saw the financial crisis as controllable and happening in hinterland markets,” Bookstaber says. “It was the storm in the Caribbean, with people saying, ‘What do we care? It’s so far away.’ And that is really the key lesson we learned from 2008. It wasn’t one thing, it was many things.… Mapping the feedback and how it worked its way through the system—this is why we need these dynamic agent-based models.”

Agent-based models are the antithesis of the models on which Wall Street, the U.S. government and central banks relied in the run-up to the financial crisis, which took for granted—wrongly—that markets are efficient and tend toward equilibrium, even after a panic. A House of Representatives hearing in 2010 found that a dependency on the models used by the Federal Reserve, called “dynamic stochastic general equilibrium” models, might not have been very shrewd.

So how do agent-based models work? Bookstaber’s simulation model uses parallel processing to map out possible scenarios in what he calls “crisis dynamics.” The model assigns “decision rules” to all the key players that reflect their priorities, financial positions and real-life behavior patterns, taking into account their interdependent relationships, the wide range of actions they might take under various circumstances, and the probable results of those actions. He says, “We’re looking to develop a kind of weather service, pre-shock, that asks, ‘Will this turn into a bigger storm and who is on that path? Is it a major funding system like a bank (a big deal) or an oil market (not as big a deal)?’”

Feeding live data into simulation models, researchers at the OFR can pose questions as broad as “What happens when interest rates go up?” or as narrow as “What is the exposure of Citigroup right now?” Bookstaber says he runs the model thousands of times under different scenarios, to get a full range of outcomes. “We do it again and again and come up with a general distribution of what might happen,” he says. “At a certain point, the distribution model is clear and you’re like, ‘OK, I get it.’”

Bookstaber says this modeling has yet to be fully embraced by Wall Street and academia, as people are accustomed to mathematical models rather than models based on the real-life behaviors of diverse individuals, groups and institutions whose actions might touch off “herding,” “second-order effects” and “non-linear dynamics” among market participants. “This is more about engineering and physics than mathematics,” Bookstaber tells Newsweek. “When a bank does something, it cascades across the system, but that’s also very difficult to look at mathematically.”

With agent-based models, no top-down assumptions (for instance, that markets are efficient and tend to move toward equilibrium determined by pricing) are imposed on the American economy. Instead, they follow the motivations of multiple agents, such as banks, hedge funds and investors, based on bottom-up rules of behavior that take into account how the players operate and how their behaviors may suddenly change based on abrupt shifts in the markets, as well as how they react to each other. Once Bookstaber begins to use live data, he can refine the models to make them increasingly true to life.

“Because I have had a long career on Wall Street, I can set up rules that are accurate and realistic,” he says. “With the economy, system issues grow slowly, and they tend to be obvious if you know where to look. The idea is to keep going and make it better and better. In the beginning, weather models were terrible, and now you get these 10-day forecasts.”

Before Bookstaber joined the OFR, he ran risk management at Moore Capital Management, a $12 billion hedge fund run by the famously curmudgeonly Louis Bacon. The complexity and opacity of the market and its innovations, even before the financial crisis, were of particular concern to Bookstaber. In 2007, he authored the book A Demon of Our Own Design, which laid out the urgent case for what might happen—and did happen—later that year. “The market is inhabited by people, heterogeneous and context-sensitive, who do not live up to the lofty assumptions of mathematical optimization and Aristotelian logic that underlie these approaches,” he wrote in Demon. “The nature of complexity also is different in the economic realm from that in physical systems because it can stem from people gaming, from changing the rules and assumptions of the system.”

Bookstaber joined the Securities and Exchange Commission as a senior policy adviser in 2009, working on the Volcker Rule, a part of Dodd-Frank that seeks to limit banks’ risk-taking so that taxpayers don’t get stuck paying the tab again. In 2012, he began working full time on agent-based models at the Treasury. “If I didn’t get involved, I would feel stupid about it for the rest of my life,” he says.

While the OFR already tracks and reports systemic risks to the Financial Stability Oversight Council (FSOC), Bookstaber’s models, combined with the OFR’s ability to subpoena data from Wall Street, would be a significant addition to what OFR Director Richard Berner likes to call the agency’s “prudential tool kit.”

Says Bookstaber, “By modeling this out, government can justify asking for specific information,” as opposed to going on what Wall Street has objected to in the past as fishing expeditions.

Berner is a nonvoting member of the FSOC and the OFR does not set policy. But it does advise Washington’s top economic policymakers, including FSOC chief and Secretary of the Treasury Jack Lew, along with more than a dozen other council members, such as Federal Reserve Chair Janet Yellen, Securities and Exchange Commission Chair Mary Jo White and Richard Cordray, director of the Consumer Financial Protection Bureau.

Launching a live-data model like Bookstaber’s won’t translate into results that can be acted upon by policymakers overnight, as agent-based models can take a few years to calibrate and perfect. But it could quickly begin providing insights into the nation’s economic health that chief economic policymakers can use in the near term.

The OFR and the FSOC have already begun using agent-based models to gauge how large-asset liquidations can affect the market and to evaluate the stability of financial networks, says OFR spokesman William Ruberry.

Some observers are doubtful. “I am very, very skeptical that the FSOC will be able to identify systemic risk,” says Paul Schultz, a professor of finance at the University of Notre Dame’s Mendoza College of Business who specializes in financial regulation. “This is the same group that missed the crisis last time around, and I think it’s hubristic to think that by just talking more they won’t miss it next time around.”

Read the entire article on the Newsweek website.