As the early signs of the financial crisis began showing up in 2007, the Federal Reserve’s research economists were so busy studying the problems cropping up in the mortgage markets, and then in student loans, and then in asset-backed securities, that they couldn’t leave the building for dinner. The Fed’s cafeteria closes at two in the afternoon, so a staffer at its building in Washington, D.C., installed a mini fridge on the second floor and kept it stocked with sandwiches from Subway. “There was a flowering of research in 2009,” says Andreas Lehnert, one of the Fed economists. “It’s the economist’s equivalent of a war memoir.”
Much of that research attempted to understand why banks failed in unison, sapping the credit markets and crashing the economy. It was a chain of events that the Ph.D.s at the Fed had failed to foresee. Their subsequent research has given them a better grasp of why they missed the collapse, but it’s also revealed how poorly traditional macroeconomic methods predict systemic bank failure and the domino effect it can unleash on the economy. A survey published last fall by an economist with the Federal Reserve Bank of Richmond showed that macroeconomic models typically assume that there will always be enough credit in the financial system. “In the post-crisis environment, the single most important macro event of our lifetime, our models are totally unable to comprehend it,” says Lehnert. The Fed’s economists realized that in order to do their jobs, they needed to have a better understanding of finance, something most macroeconomists are trained to ignore in graduate school, says Lehnert. They needed to start thinking like bankers, and using bankers’ risk-management tools for their own purposes.
The research Ph.D.s at the Fed in Washington use macroeconomic modeling—complex computer simulations of the economy—to provide guidance for the board’s monetary policy decisions. Before the crisis, they had little to do with other divisions of the Fed, like bank supervision. In 2010 the Fed created the Office of Financial Stability Policy and Research, where Lehnert is now deputy director, as part of an effort to change the agency’s culture, as well as its research methods.
The new office works closely with the Fed’s bank supervisors, looking at worst-case scenarios for the economy, which could start with the banks. The office is also sifting through a deluge of new models, many created by finance academics, that measure the risk widespread bank failures pose to the economy. According to a January survey by the Treasury Department, there are 31 of these models. The Fed is trying to figure out which ones it likes best.
Though they’re cagey about models, Fed officials confirm that, among others, researchers are using the Marginal Expected Shortfall approach developed by Robert Engle, a Nobel laureate and finance professor, along with his colleagues at New York University’s Stern School of Business. Like a lot of the new models, it relies on market information to make predictions about bank defaults, something the Fed has steered clear of in the past. The Fed’s bank supervisors have access to confidential data they get directly from the nation’s banks, numbers typical bank investors don’t have, which provide insight into each bank’s internal accounting. The Fed has always considered itself better informed than the market, and has been reluctant to rely on market-based yardsticks such as debt-to-equity ratios, a crucial measure of financial risk measurement. Even though investors don’t have access to banks’ internal accounting, their behavior can provide important information about the risk a bank poses to the rest of the system. And that behavioral data is available more regularly than the Fed’s quarterly requests for bank information.
Using data gleaned from Engle’s model, which compares the equity value of a bank to its liabilities, NYU Stern’s Volatility Laboratory publishes a weekly update showing which bank is contributing most to systemic risk. (For the week of Aug. 29, it was Bank of America, which declined to comment.) Engle says his model is “quick and easy and noninvasive.” It works without the private data the Fed collects from banks. Engle says he’s been asked by the Fed how his model would perform if he had access to the Fed’s supervisory information. He suggests that regulators might want to start digging where market data and a bank’s internal accounting disagree.
Edward Kane, a professor at Boston College who’s been studying bank risk for decades, has come up with a way to measure systemic risk using the Merton model, a staple of finance that bankers use to assess the risk that a corporation will default on its bonds. “Isn’t it ironic,” says Kane, “that the Federal Reserve was not using the same tools that their opponents were using? The regulators were playing against the financial institutions. The latter have better tools, better-trained personnel. It’s like a powerhouse team in college playing one of these Division II programs.” A person inside the Federal Reserve system, speaking on condition of anonymity because of the sensitivity of the topic, says the Fed can’t compete with banks and universities for top financial academics.
The Federal Reserve Bank of New York, with a talented core of financial economists, is the exception. Tobias Adrian, a vice president at the New York Fed, and Markus Brunnermeier, a financial economist at Princeton University, have developed a model they call CoVaR. They based it on the value-at-risk model that banks and hedge funds use every morning to calculate the maximum amount of money they could lose if all their positions went bad. CoVaR applies this method across banks to measure how each institution contributes to the risk in the entire banking system. The financial stability office in Washington is now regularly using it along with Engle’s model, say Lehnert and other Fed officials.
This post-crisis research shift toward finance is not limited to the Federal Reserve. Dale Gray worked at the International Monetary Fund during the Russian and Asian crises in the late 1990s. He noticed that countries hold what he calls a “hidden option” on banks: If the value of a bank sinks low enough, politicians will feel compelled to buy it, rather than letting it fail. He priced this option, also using the Merton model, then left the IMF after a colleague told him “no one is interested, and no one will be interested, in this idea.” The fund, like the Fed, tended to hire macroeconomists. “They assume away default,” says Gray. “It doesn’t fit in their model.”
After licensing his model privately, Gray is now back at the IMF, which has adopted it as well. Gray says the European Central Bank is interested, as are the central banks of the U.K., Israel, Sweden, plus the Fed. In August 2008, two weeks before Lehman Brothers filed for bankruptcy, Gray presented his ideas to the central bank of Iceland. “We’re very interested in your models,” he says he was told. “We should have been looking at them a lot earlier.”