Publish or Probability?
The perilous journey of academic research
From studies about the effects of coffee on your health to education on earnings to the performances of mutual fund managers, every day the news seems to be filled with new research conclusions. But just how reliable are the results we see and hear in the media?
Justin Tobias was named a distinguished author by the Journal of Applied Econometrics in 2009 and currently serves as the publication’s associate editor. He also contributes to and has edited numerous other academic journals and co-authored a popular textbook on Bayesian economics. (Photo by Mark Simons)
That’s one of the questions Justin Tobias, head of the Department of Economics at Krannert, tries to answer with his research in econometrics, the application of mathematics and statistical methods to the analysis of economic data.
Specifically, Tobias studies the theory and application of Bayesian econometric methods, a branch of econometrics and statistics named in recognition of Thomas Bayes and a simple mathematical formula he produced for calculating conditional probabilities and updating beliefs about events as new data arrives.
“Econometric models can be investigated with either Bayesian or classical (frequentist) methods,” Tobias says. “Bayesians make inferences that are conditional on the data observed. The classical method provides a fundamental role for data that could have been observed, but were not, and involves averaging over different possible data outcomes that could have happened.”
Tobias often uses this example with his students — imagine you are editing a journal, and are torn between using Referee A or B to review a paper for potential publication. You flip a coin, and decide to send the paper to Referee A. After A returns a report, do you make a decision on the paper’s suitability based upon that report, or do you also imagine what Referee B might have said?
“I think most would render a decision based on the information provided by Referee A, and I say with confidence that authors would not be happy with me if I rejected their papers for publication because I believed a referee who was not assigned as a reviewer would not have liked their work,” Tobias says. “Admittedly, it’s a bit more complicated than this, and the principle is somewhat controversial, but this example serves to illustrate the idea of making decisions based on the observed data.”