PHOTO GALLERY
Building the next generation of macroeconomic models is a challenge that requires the energy and ideas of the next generation of economists. The Macro Financial Modeling (MFM) Initiative recruited nearly 50 of them and, in an intensive four-day summer session, passed on the expertise they need to join the effort.
The 2016 MFM Summer Session held June 11–15 gave doctoral students and other early career researchers a thorough grounding in techniques, data sources, evidence, applications, and methods to assess how the financial sector impacts the economy as a whole. Organized by MFM codirectors Andrew W. Lo and Lars Peter Hansen, the program brought together leading academics and experts who build and use models to manage system risk in financial and policy settings. Speakers presented models and methods in contexts focused on understanding the last financial crisis—and preventing the next one.
Several talks on methods and evidence gave participants insights and approaches they could use in their own work. In the first talk, Yuliy Sannikov reviewed the computational techniques behind a model he and Markus Brunnermeier developed. The model is designed to simulate crises, assess the resilience of the financial system, examine the system’s response to various policies, and estimate spillover and welfare effects.
David Rappoport, a researcher from the Federal Reserve Board who presented his work on a discrete time version of the Brunnermeier-Sannikov model at a poster session during the program, said he appreciated the opportunity to talk directly with Sannikov and to many other participants about his work. “I also had interesting exchanges with Hansen and Jaroslav Borovička about how our discrete time version, which exhibits the same qualitative properties of the original model, could help in understanding properties of the model necessary to apply their methodology to analyze macroeconomic models through the dynamic behavior of asset prices,” Rappoport noted.
Frank Schorfheide and Christopher Sims continued the focus on methods with sessions exploring the importance of applying Bayesian estimation to correct for errors in models. Sims, co-recipient of the 2011 Nobel Prize in economics, used a hot topic—the question of whether rapid credit expansion lowers future economic growth—to argue for the use of Bayesian techniques and multiple-equation models.
Christopher Sims
“Many people say there is evidence that rapid credit growth is dangerous; I’m not so sure,” Sims told participants, presenting new evidence that bears on the dynamic linkages between credit and the macroeconomy. His preliminary findings suggest that innovation in household credit did not have an effect on GDP, but firm credit did.
Hansen and Jaroslav Borovička each gave sessions exploring the dynamics of asset prices influenced by risk and uncertainty, and reviewed ways to incorporate these factors into macroeconomic models. Simon Gilchrist modeled the behavior of inflation during the recession recovery.
Managing Massive Data
Researchers now have vast arrays of transaction data to help trace the precise ways financial institutions and financial behavior affect the economy as a whole. Several sessions touched on computational methods and the challenges of extracting valid lessons from massive datasets.
Antoinette Schoar
One important lesson Antoinette Schoar stressed is that aggregating data incorrectly can tell a misleading story. By taking a close and careful look at mortgage issuance between 2002 and 2006, she found evidence to refute the common notion that the financial crisis was precipitated by subprime lending to unqualified borrowers. When you decompose aggregate data more carefully to look at individual loans, it becomes clear that the share of mortgages issued to the lowest-income borrowers stayed stable throughout the run-up to the crisis, and the share of defaults in the low-income quintile actually declined to 11 percent.
It was actually the top two quintiles by income that accounted for half of the loans and, by 2006, half of the defaults. “This doesn’t support the view that what went wrong is a supply-side distortion—a misallocation toward the type of people who got credit,” she said. Her analysis reveals a demand-side phenomenon driven by expectations of rising home prices, causing people to take out bigger mortgages, take out equity, or refinance more quickly.
“Zip codes are not people,” Schoar reminded participants. “Whenever you’re doing analysis, you have to think about the level at which you’re aggregating. When you actually decompose these different channels, it looks quite different from the story we’ve heard,” she said. “I’ve seen this in so many papers, and I don’t want you guys to do this anymore.”
The challenges of working with big data cropped up repeatedly throughout the program, particularly in a panel where practitioners from the financial industry discussed how they use models and big data to run scenario analyses, forecast the impact of weather on commodities markets, identify potential business growth opportunities, and estimate political risk.
Technology’s Strengths and Limitations
Sanmay Das of Washington University in St. Louis outlined the basic framework of machine learning and the state of practice. He pointed to growing cross-fertilization between the machine learning and statistics/econometrics communities. Machine learning’s strength is using algorithms to sift through data and generate useful predictions—not to determine causation. “Machine learning has a lot to learn from economics on causation. Prediction is what we’re good at. How can MFM use it? How do we marry causation with machine learning?”
He walked students through prediction problems with applications in finance. Afterwards, participants said they found discussions of machine learning and big data techniques particularly interesting, because their studies had given them little exposure to these topics.
Technology can be a positive force in financial markets, automating repetitive tasks, but it also can contribute to systemic risk, as Andrei Kirilenko of Massachusetts Institute of Technology demonstrated in his talk on high frequency trading. He led the investigation into the momentary “flash crash” of May 6, 2010, where a glitch in an automated trading algorithm caused markets to drop as much as 9 percent—and then bounce back up seconds later.
“This was a systemic event that affected the entire US financial system, and led to regulation in market function and design,” he said. Reviewing the causes, response, and safeguards implemented, he noted that the crash leaves a useful lesson for researchers to keep in mind.
“Study of institutional details will be very important in your work. These details underlie data generation processes. If you don’t understand the processes and where data comes from, you will misunderstand what you’re looking at.”
Practitioner and Regulatory Insights
To give students a practical view of modeling challenges, the program included panel discussions with economists and researchers from the financial industry and from policy and regulatory agencies. Both panels offered advice on career opportunities in their sector, and students took advantage of the opportunity to get feedback on their own work between sessions.
“Views and questions from the policy world are always important components for academic work,” noted Chang Ma of Johns Hopkins University. “Connecting these views with my latest research agenda makes my work relevant. My conversation with the International Monetary Fund policymaker Laura Kodres was fantastic. She offered her views towards optimal macroprudential policies on economic growth, which is what my job market paper is about.”
Building a Network
Speakers were pleased to have an opportunity to encourage new research on issues they worry about. “I think everyone who is speaking is really excited because we have an opportunity to influence young minds,” noted Deborah Lucas, who gave a talk analyzing government programs as a source of systemic risk.
Many speakers wove advice and encouragement through their talks, and students took full advantage of opportunities to learn from and talk with available experts, frequently clustering around speakers with follow-up questions after talks. Between sessions and after meals, they eagerly took the opportunity to discuss their own work with faculty and practitioners, garnering advice and ideas for new avenues to pursue.
“This event has been tremendously beneficial for me in terms of getting a bigger picture of how academia, policy circles, and the real sector can collaborate, and how much they indeed need to,” noted Selman Erol, who is completing his PhD at the University of Pennsylvania. “The sessions and audience were carefully crafted to include a perfect mix of scholars in early and late stages of their career—both academics and other professionals. The contacts I have acquired during this event will certainly benefit me a lot in the future, whichever direction I choose for my career.”
Students presented and discussed their research
But students learned from each other as well. Nine students gave brief formal talks on the closing day; poster sessions throughout the program gave 17 more participants a chance to present their work. They valued both the feedback and the experience. “Attracting people from their breakfast to hear about your research is difficult; it adds further difficulty if you are required to compete with your talented [peers] at the same time,” Chang Ma said. But the challenge paid off, he added. “The benefits exceed expectations. The skills to market your research in a few words and be able to answer questions smartly and promptly will benefit me in the long run.”
“Interacting casually with each other was very beneficial,” added Christopher Lako of University of California, Berkeley. While sharing meals, lodging, and non-stop conversation about research, students discovered peers working in their areas and traded views with students working with other topics. They made lasting connections and came away feeling like part of a growing and vibrant community addressing an important problem.
“I had the chance of meeting an incredible number of people who are as passionate as I am about these issues and learned so much just by constantly discussing these four days,” said Quentin Vandeweyer, a participating graduate student from Sciences Po in Paris.
“All in all, it was a very positive experience, and I am very grateful for the opportunity to be part of the MFM group,” concluded Igor Cesarec of New York University.