As society adjusts to the knowledge that our climate is changing, policy makers are faced with a difficult question: how can they use policy to help prevent and cope with climate change, while minimizing the damage to their nation or organization’s economic health?
For example, “one policy that has been considered and proposed is carbon taxes and taxes on carbon-emitting industry,” said Ian Foster, director of the Computation Institute at University of Chicago/Argonne National Laboratory. “The question is: will that tend to drive dirty industries offshore in a way that will perhaps increase the total emissions and harm US industry at the same time?”
Two years ago, Foster and several other computer scientists joined forces with economists, climate scientists, and geophysicists to create a computer modeling framework that could help decision makers answer these sorts of questions.
Foster and his colleagues called the first version of their model CIM-EARTH (a Community Integrated Model of Economic And Resource Trajectories for Humankind). Already, several related papers have been published, and in February 2011, a larger group of researchers launched the Center for Robust Decision-Making on Climate and Energy Policy (RDCEP).
Computational methods have been used to simulate a variety of complex systems, from black holes to human biology. But all of these have one thing in common: they are all models of physical systems, governed by physical laws.
The same cannot be said of the economy. It’s true that the economy can be constrained by physical limits such as the quantity of a resource that exists on our planet, or the rate at which we can extract it. But with those few exceptions, the economy is governed by collective human behavior – and we can certainly rely on humans engaging in irrational and unpredictable behavior.
“On the other hand, economists do know a lot about how society responds and the economy responds to various pressures,” Foster said.
As an example, he described an approach to economic models called “rational expectations.” The simplest economic models assume that the public will behave as though they knew nothing about the future. But that’s wrong; in reality, we modify our behavior based on what we expect to happen. Rational expectations models attempt to emulate that by assuming that the public has perfect knowledge of the future. The result is a more accurate, effective model – one that will likely only improve as it evolves to assume imperfect knowledge of the future.
The computational economics community is small, according to Foster, but growing. Historically, they’ve had access to very limited resources. But with the CIM-EARTH model, and other projects launched by RDCEP, economists have the opportunity to work hand-in-hand with experts in optimization, numerical methods, and more.
“Our models have some new features relative to the old ones,” Foster said. “They can run with greater detail, with smaller time steps, they can incorporate factors that are not in the previous models because we have more modern numerical methods.”
The RDCEP team also has access to a wider variety of grid and high-performance computing systems. These make it possible for them to do the larger studies needed to look more rigorously at how much the model output depends on various assumptions. Currently, their model is a software stack designed to serve as a framework for climate and energy-related decision-making models. But the Center’s researchers have ambitious plans for that framework, such as offering it as a service that emphasizes user-friendliness.
“The goal is to allow people to express on a fairly high level what the model is that they want to solve and then have software that can map that into the computational activities that are required,” Foster explained.
A user-friendly interface is helpful for economists, but essential for decision makers.
“Traditionally, the decision makers have been reached by people writing reports and papers that describe the results of a set of computations,” Foster said. “I think they would be much more interested in something that’s more interactive so they can do what-if scenarios.”
The RDCEP team has high hopes of attracting and engaging a large community of contributors. But to do so, they will have to surmount some cultural barriers.
“There has not been a tradition of open source modeling and open models in this community and so we’re trying to have an influence in that regard as well,” Foster explained.
As many programming communities have already discovered, an open source approach is extremely helpful in encouraging a wider variety and higher quality of contributions. Indeed, a successful open source project can inspire and anchor a thriving collaborative community of contributors.
But in a scientific context, the open source approach becomes essential. Research relies on the concept of peer review and duplication of results. Part of the peer review involves reviewing the method that was used during the research, and the model is the most important tool a computational economist can employ. If the peers who are reviewing a paper cannot examine the model that was used, they cannot properly review it.
In a similar vein, results are generally considered suspect until two or more independent research groups have achieved the same result. But to reproduce the experiment, they need to be able to use the same model. With a closed model, that isn’t possible.
So far, the RDCEP team’s outreach efforts have gone well.
“A lot of people seem very enthused,” Foster said, adding, “There is a set of existing models that have been closed. Some of them actually are now starting to open up, I think partly because of pressure from our suggestions that that would be desirable.”
RDCEP, which was launched using a five-year $6 million National Science Foundation grant, has nearly five years remaining to continue to hone, improve, and add to their modeling framework.
“We did one first study that was looking at the sensitivity of the core economic model to some of the input parameters,” Foster said. Next, they plan to do a large study of scenarios in which crop land use changes.
“What we want to do there is actually take an agricultural model and build up a big database characterizing the agricultural output that would result for different land types under different climate scenarios and that would then be used in other studies,” Foster explained.
Another avenue they are exploring is how an increased demand for biofuel might affect land use. For example, will the demand be sufficiently high that forests will be cut down to make room for more crop lands.
Ultimately, Foster is optimistic about what the team might accomplish.
Said Foster, “If we build the best model possible, I think there’s a strong likelihood that we’ll be able to do a better job of preparing for the future.”