Using HPC to explore competitive balance

Image courtesy Colin Harris at www.flickr.com. (CC BY-NC-ND 2.0)

The financial services and medical insurance industries in the US account for 6-8% (more than a trillion dollars) of the gross domestic product annually, according to the US Department of Commerce Bureau of Economic Analysis. The power wielded by both industries influences the lives of most Americans.

The magnitude of such power suggests the significance of better understanding the hidden forces involved in determining the quality of service-provider competition and the design of contracts. Investigating those forces are researchers Victor Zhorin of the Computation Institute at the University of Chicago in Illinois, US, and Robert Townsend of the department of economics at the Massachusetts Institute of Technology in Cambridge, US.

By examining economic fundamentals without making any strong assumptions, says Zhorin, he and Townsend are attempting to connect industrial organization and contract theory — two fields that are largely considered separately. To explore a contract-based model of industrial organization, the researchers have employed XSEDE (Extreme Science and Engineering Discovery Environment) advanced digital resources.

The researchers are considering information frictions — phenomena that are not directly observable, but that affect business profits. One friction, ‘moral hazard,’ pertains to the risk that contracts are not always entered into in good faith.

In the medical insurance marketplace, moral hazard may be exemplified by an insured person's laissez-faire attitude and by their taking unnecessary risks. In the financial services sector, institutions that behave irresponsibly in their quest to reap profits, troubled banks that accept bailouts, and homeowners that receive reduced mortgages to avoid default or foreclosure, are all examples of moral hazards.

Another friction factored into the study, ‘adverse selection,’ occurs when the bulk of medical insurance policies are purchased by those with the most health risks, while healthy people don't see a purchase incentive. In banking, when the creditworthiness of a customer is unknown — something that occurs more often in developing countries, says Zhorin — adverse selection is in play.

The project also examines the economic influences of various market structures — monopoly, imperfect competition, perfect competition, and strategic interaction. While the definition of monopoly may be familiar, what is less well known is that even a competitive market situation can function monopolistically. A case in point is when a number of different businesses sell similar products, yet the products are unique in their own rights, thus representing a form of ‘imperfect competition.’

3D graphs represent out-of-equilibrium contract space under imperfect competition market structures. The contract space encompasses both geography and customer product preferences. Colors represent value levels (profits). The graphs reflect how businesses anticipate their competitors' responses to possible strategies. For example, financial institutions can expand into either urban or rural provinces, demand more or less collateral for loans, and set varying insurance premiums and deductibles. Image courtesy Victor Zhorin, University of Chicago in Illinois, US.

Other imperfect-competition scenarios include situations where there are only a few sellers of a product (oligopoly), or numerous sellers of a product and only one buyer-demographic (monopsony). Zhorin explains that even though perfect competition does exist, "each case has to be tested and verified with empirical data behaving [or not] according to perfect market predictions."


Connecting industrial organization and contract theory is complicated. Zhorin explains that the project's simulation model encompasses several layers. From what takes place on the micro-level between households and the financial or insurance firms, to industry service providers engaged in strategic interaction and competition throughout markets, enhancing market competition requires a close look at the industries and their contracts with customers. “At each layer, a different computational method with a different optimization approach is used to compute both the equilibrium and out-of-equilibrium outcomes,” Zhorin says.

A key aspect of Zhorin and Townsend's research is examining economies that are in transition — where companies operate under imperfect competition — with consideration of both geography and customer product preferences. The results, Zhorin says, are directly applicable to service-oriented industries including medical insurance and financial services.

According to Zhorin, the project has revealed ways that government policymaking — relative to promoting healthier competition — could be enhanced. He and Townsend have isolated data that will help regulatory agencies make better decisions about reducing regulations, promoting a wider choice of services across society, and possibly thwarting monopolistic actions in service-oriented industries.

The research is also illuminating the nature of imperfect competition and related dynamics. “We implemented a general framework that provides insights into the mechanics of imperfect competition, the role of obstacles, and the impact of financial-service providers on small businesses and households,” Zhorin says.

The researchers are testing developed algorithms on many-core computer architecture provided by XSEDE, including Nautilus at the National Institute for Computational Sciences (NICS) and Stampede at the Texas Advanced Computing Center (TACC). "XSEDE provides a variety of resources that allows us to choose the optimal combination for a particular research project," he adds.


Zhorin and Townsend are employing MPI (message passing interface) and OpenMP hybrid parallelization. Zhorin says that future work — with more complex contract models and estimation of model parameters with real data — would require a computing architecture combining large memory nodes with thousands of worker nodes attached.

“Workshops organized by XSEDE were very helpful in getting up to speed on the Intel® Xeon Phi™,” Zhorin says. "Those workshops provided both expert advice and hands-on training on emerging technologies and computing architectures that are being deployed on a massive scale."


The theory explored in Zhorin and Townsend’s project will be used in the creation of information products to enhance industrial contracts. “This work lays the foundation for a novel approach in industrial organization with application to the banking and insurance industries,” Zhorin says. “This has been mostly theory exploratory work so far, that will be extended to a variety of real data-based applications.”

The authors researched not only empirically observed contracts, but also the entire set of contracts that could be put in place to improve welfare. “In our framework with optimal contracts in place,” explains Zhorin, “the competitive environment always benefits the households.”

He believes it may be possible to provide a utility-based value pricing for complex products, such as insurance and other financial contracts, that would account for consumer’s precise preferences. This would also stimulate an industry to develop a larger variety of products that would appeal specifically to particular consumers.