On July 18, 2024, Rajiv Sethi, professor of economics, published new research in Proceedings of the ACM Collective Intelligence Conference (CI ’24), alongside several researchers, including Barnard affiliates Olivia Bobrownicki ’24, Yuqi Cheng ’25, Anushka Kumar ’26, and Anusha Wanganoo ’25. Cheng and Kumar are part of the Bridgewater New Pathways Scholars program; Wanganoo is a Laidlaw scholar. All students worked with Sethi to help conduct this research, and all three current students attended the CI’24 conference. Two other Bridgewater Scholars, Ramya Subramanian ’26 and Sachi Patel ’27, worked on the project as research assistants and also attended the conference.
The paper, “Evaluating Prediction Mechanisms: A Profitability Test,” proposes a method of assessing the accuracy of any forecasting model by examining the profitability of a virtual trader in a prediction market. The trader is endowed with a budget, a level of risk tolerance, and beliefs inherited from the model. Updates to the model result in portfolio adjustments over time, based on the prices prevailing in the market. To implement this test, the authors used predictions for the 2020 U.S. presidential election and the 2020 and 2022 congressional elections, using data from The Economist, FiveThirtyEight, and PredictIt. The proposed method can be applied whenever models and markets referencing the same events exist.