Matthew Gentry, associate professor in the FSU Department of Economics, studies the design of auction markets by conducting mathematical simulations. Dr. Gentry’s research looks at patterns in specific data to learn about certain economic features in auction markets, such as bidders’ willingness to take on risk. Learning about these features provides more understanding of how a change to the design of an auction might change the auction’s outcome.
Much of Dr. Gentry’s research has focused on data from procurement applications for highway road work in Michigan and Texas. In these instances, the state hires a contractor to perform some road work. To choose the contractor, the state conducts an auction which awards the work to the bidder who offers to do the job at the lowest cost. Dr. Gentry develops statistical methods to recover economic features of interest from data for these auctions. He then runs Monte Carlo simulations to verify the statistical methods.
A Monte Carlo simulation is a computing experiment that replicates the same statistical method on many different sets of data points. These sets of data points represent situations similar to a set of real-world data where the outcome is already known. The outcomes of the replications can then be compared to verify that the statistical method consistently produces the expected outcome.
“As with many Monte Carlo studies, getting reliable results requires a large number of replications,” Dr. Gentry explains. These many different replications each require many data points, so using the high-performance cluster (HPC) at the RCC is vital for Dr. Gentry to perform simulations within a feasible time span.
In his words, “This is where high-performance computing shines. Each replication might take several hours to run. Having the option to distribute that across one hundred, or two hundred, or five hundred different computing nodes [in the HPC] allows me to get results in days that would otherwise take months if I were just to run them on my desktop.”