Economic model builders strive to do the impossible. Their architects aim to fully capture the range of signals, signifiers, indicators, and observations that our species—both as individual actors and a collective—use to make decisions about how to improve our lives and invest in our futures.
Thus, all economic models are strategically flawed. They omit, abstract, and simplify, in the hope that the parameters it can capture offer clarifying insights that outweigh the more ambiguous elements. However, speaking to colleagues, students, and alumni in Paris, Lars Peter Hansen urged a spirit of adventure in economic modeling. He called for extending the outer edges of what those models can reveal in order to more richly understand the complexities of the forces that shape business, policy, investment, and education.
“Models are not replications of reality. They are simplifications, and thus, are imperfect,” said Hansen during one of four lectures during his visit to France. He argued for confronting that reality and accepting that there is little we can know with certainty, but much we can understand about uncertainty itself. “Let’s push uncertainty to the forefront of economic analysis.”
The 2013 Nobel laureate—famed for teaching economists that they could “learn something without knowing everything” via his Generalized Method of Moments—traveled to Paris to speak at HEC’s Decision: Theory, Experiments, and Applications Conference and to receive an honorary professorship for his work advancing the use of statistical rigor to more tightly bind economic inquiry to empirical data.
Hansen’s efforts to move economic modeling beyond what HEC Professor Tomasz Michalski described as “what we can see in the light” is an area of economic inquiry much like the HEC campus itself. After an 83-year history of educating world business leaders in 1964 HEC transplanted itself from the protective surroundings of central Paris to a campus in Jouy-en-Josas embracing bolder, newer architectural styles juxtaposed against the French countryside.
HEC Campus in Jouy-en-Josas.
Likewise, Hansen’s work to improve model specification—how we establish the parameters of a model and fit them best we can to what we seek to know more about—builds on the foundations of traditional economic theory, but seeks to reinforce them with better accounting of sources of uncertainty. Specifically, he highlighted three types of uncertainty—– risk, ambiguity, and mismatched model design.
Hansen unpacked the intellectual history of these three forms of uncertainty in a talk to UChicago alumni in Paris, beginning with an examination of mathematicians Blaise Pascal and Jacob Bernoulli from 1654 and 1713, respectively. Pascal saw that, given large enough sample sizes, one could see patterns in the seemingly random outcomes of games of chance; Bernoulli sought to apply that insight to learning more about people and the patterns that emerge from their seemingly random individual decisions.
Bernoulli began a rich tradition of social scientists looking to the hard sciences for more rigorous tools to help learn more about human behavior; Louis Bachelier’s Brownian motion, Udny Yule’s attempts to model and predict the appearance sunspots, and Eugen Slutsky’s efforts to model the apparent cyclicality were indirect but important steps forward for economics as well as each man’s respective fields, even when the math didn’t precisely fit the data. “Cut [Slutsky] some slack, he was doing this in the 1920s with no computer technology to help him do this,” said Hansen.
Around the same time, UChicago’s Frank Knight complicated the practice of predictive models by noting that we can only predict what the future will hold without our own interference, a picture that will never capture the full breadth of what could happen.
By 1933, Ragnar Frisch was able to use Yule and Slutsky’s work to compare business cycles in a more rigorous, repeatable way, laying the groundwork for integrating more advanced statistical understanding of macroeconomics in the decades ahead. But economists continue to struggle with the complications wrought by various types of uncertainty to this day.
John Muth and Robert Lucas advanced their benchmark theory of rational expectations in very much the same vein, taking great first strides toward confronting that uncertainty, argued Hansen. Drawing on the Law of Large Numbers, their model used long histories of data to infer the model and its parameters. But even this landmark achievement contained critical simplifications: expectations of actors were formed inside of the model, ignoring ambiguity about how likely those expectations and preferences are to shift the predictions of the model.
Today, a rich literature about the factors outside of an economic model exists, but much more work is needed to better understand the individual actors within it. Hansen broke down the gaps in what we know about uncertainty with a simple thought experiment: suppose you have an urn filled with red and yellow balls. You can make an educated guess of the number of balls of each color when the proportion between red and yellow is known—that represents risk, the aspect of the problem we know how to estimate. But when those probabilities are unknown, and that proportion might change over time, things get much more complicated to predict.
That ambiguity and change of state more realistically reflects uncertainty around real world events, but models still need to capture them more richly in Hansen’s view. “If you’re examining uncertainty using real data, you’re kind of chasing a moving target.”
The rise of stochastic time series and other dynamic modeling techniques offer new possibilities of shining a light on that moving target, but Hansen cautions that we still have to reckon with the initial sin of economic modeling: that all models are imperfect. We can never trust one with absolute certainty. “Instead, we need to determine how to use models that are imperfect in ways that are sensible.”
Hansen enumerates the key issues for economic models when confronting uncertainty.
Some of that sensible thought comes down to what economists themselves can offer to the public. Hansen argued that “secular stagnation” and other explanations of slow macroeconomic growth facing the world today can’t be easily explained by a singular theory. The reality post-financial crisis is that we’re never 100 percent sure if a drop in growth is a anomaly from the trend line of our economic future or a new trend line established by permanent shocks to the economy, said Hansen. This complicates economists’ role as op-ed soothsayers, but Hansen said that perhaps that isn’t such a bad thing.
Take financial market regulation, a topic of particular interest to Hansen. If regulators aim to head off a future economic crisis, can we regulate systemic risk or systemically important financial institutions without actually defining what those terms mean? In Hansen’s view, richer macro financial models won’t tell us precisely what sound financial regulation will look like, but instead can better define the appearance of nebulous forces working between financial entities, based on probabilities rather than assumptions.
Hansen noted that the Dodd-Frank Act is written as though the financial world is predictable with absolute certainty, leaving regulators poorly equipped to reconcile the law with the realities of the market. Despite that ambiguity, the regulations within it run on for pages and pages, further adding to uncertainty among regulators working in the field. “Simple, robust policies have the advantage of not adding additional uncertainty to the economic environment,” says Hansen. “But we’re not making much progress in regards to complexity.”
Environmental policy could also benefit by accepting the limitations of what is knowable with absolute certainty. Former undersecretary for science in the Energy Department and director of the Center for Urban Science and Progress at New York University Steve Koonin wrote, “Any serious discussion of the changing climate must begin by acknowledging not only the scientific certainties but also the uncertainties, especially in projecting the future.”
Hansen cited Koonin’s view as one sorely lacking in discussions of public policy: a future where an engaged, informed public can be given probabilities—not certainties—when facing difficult choices as a nation. “There’s a fear within the scientific community that if we acknowledge the uncertainty, there will be an argument that we should do nothing,” said Hansen. “But this is the type of open discussion that I find valuable.”
Hansen didn’t bring any simple answers along with him to Paris. He, like all of us, stands under a small light of understanding amidst swirling darkness of uncertainty about what will happen to our economy in the years ahead. But instead of standing there in the limited light we have, Hansen seeks to push himself and his colleagues to the edge of that halo, finding the best ways that science can extend our vision into the darkness.
— Mark Riechers