Institute addresses computational challenges posed by economic
models
ARGONNE, Ill. (Aug. 22, 2007) — Computer scientists from the Department
of Energy's Argonne National Laboratory joined with economists from the University
of Chicago earlier this month for a conference designed to bridge the existing
gulf between these two fields and teach young economists how to use state-of-the-art
software and computational methods.
NEOS finds a new application in computational economics
What do shopping centers, game theory, and applied mathematics have
in common? The answer is optimization – one of the fundamental modeling
paradigms of mathematics.
Optimization techniques have been used for more than a decade to study
price competition and marketing of new firms. But several key questions
have not been addressed. For example, which store types are more important
in determining the success of a shopping center? And how beneficial or
harmful is it for similar types of stores to be in close proximity?
Maria Ana Vitorino, a doctoral student at the University of Chicago
Graduate School of Business, is seeking answers to such questions. Key
to her work is the Network-Enabled
Optimization Server (NEOS). Developed
at the U.S. Department of Energy's Argonne National Laboratory in conjunction
with researchers at Northwestern University , NEOS allows users to solve
optimization problems over the Internet with state-of-the-art optimization
software and without downloading or linking code.
Vitorino's doctoral thesis examines the entry decisions of stores in
regional shopping centers. A typical estimation problem involves approximately
4,000 parameters. Vitorino is using two of NEOS's more than 50 constrained
optimization solvers to handle these parameters. With NEOS she can obtain
the maximum likelihood estimates of the parameters in her economic model.
“NEOS is a great tool for me,” Vitorino said. “It allows me to submit and
run several jobs simultaneously and has optimization solvers that can solve
my problem easily.” |
Economic models are vital for policy analysis, but some economists do not
understand the mathematical techniques used to generate them. Even more important,
many economists remain unaware of developments in computational science that
could greatly advance their work. The Institute
on Computational Economics (ICE2007), held at the University of Chicago
from July 30 to Aug. 9, attempted to redress these problems by showing more
than 50 economics graduate students, postdoctoral researchers and junior faculty
from the United States and Europe how to use new computational tools to solve
economic policy questions.
“We put great emphasis on helping these young scholars apply cutting-edge
software and techniques in computational science to actual economics research
problems,” said Sven Leyffer, Argonne computational mathematician and co-chair
of the workshop.
The institute consisted of morning tutorials on new analytical and numerical
methods in such areas as dynamic programming, stochastic modeling, structural
estimation and optimization problems with equilibrium constraints. Afternoon
sessions included software presentations and hands-on workshops in which participants
applied the new software to challenging economics applications.
Seminars featured during the second week were led by leading economists and
focused on recent advances in quantitative economic policy research, as well
as continuous time models, Baysian estimation, and the practical applications
of optimization technology.
The conference was highlighted by the introduction of state-of-the-art software
that the participants were able to use in the lab sessions. The free licenses
for this software were made possible through the support of AMPL, Ziena, the
Stanford Optimization Laboratory, the University of Dundee and the University
of Wisconsin.
“The solvers are complete versions, not just test versions,” said workshop
co-chair Kenneth Judd, senior fellow of the Hoover Institution and a visiting
professor of economics at the University of Chicago. “The ICE participants
were allowed to take these free solvers home and use them for the duration
of their graduate studies,” he said. “In that way, we hope to foster a worldwide
community in this emerging area of computational science.”
ICE2007 was the third in this annual series of computational economics workshops.
Previous institutes have been only five days. The expanded program provided
time for informal discussions as well as a poster session in which participants
presented their current research.
“ICE2007 provided an exciting opportunity to raise the level of sophistication
in economics by creating an interface between economists and computer scientists
so that they can address the computational challenges posed by modern economic
models,” said Leyffer.
More information on the institute can be found at ice.uchicago.edu/.
Argonne National Laboratory seeks solutions to pressing national problems in science and technology.
The nation's first national laboratory, Argonne conducts leading-edge basic
and applied scientific research in virtually every scientific discipline. Argonne
researchers work closely with researchers from hundreds of companies, universities,
and federal, state and municipal agencies to help them solve their specific
problems, advance America 's scientific leadership and prepare the nation for
a better future. With employees from more than 60 nations, Argonne is managed
by UChicago
Argonne, LLC for
the U.S.
Department of Energy's Office
of Science.
For more information, please
contact Eleanor Taylor (630/252-5510 or media@anl.gov)
at Argonne.
|