Princeton University Library Catalog

TRUMPING THE PREDICTION COMPETITION?: Analyzing U.S. Treasury Yields as a Potential Predictor of the 2016 U.S. Presidential Election

Yong, Elliot [Browse]
Senior thesis
Bhatt, Swati [Browse]
Princeton University. Department of Economics [Browse]
Class year:
Summary note:
The 2016 U.S. presidential campaign and election were unprecedented for, among other reasons, featuring the first female nominee for a major political party as well as a winner who had no prior military or government service. The final result was a surprise to many, especially political pollsters, the majority of whom predicted that Hillary Clinton would win by a comfortable margin. But might financial predictors have done better? Economists like Ray Fair have long touted the virtues of economic predictors in political spheres, especially in election prediction. Building o their work, I use a somewhat novel approach, analyzing U.S. Treasury yields as a potential election predictor. In addition to interest rate and inflation expectations, Treasury yields can also capture more general U.S. growth expectations. And from evidence including yield changes following Election Day, it appears that Trump had a positive effect on yields and the slope of the yield curve. This has been attributed primarily to Trump’s protectionist and domestic growth-focused policies, which would understandably lead to higher growth expectations and thus higher yields and yield curve slopes. With this in mind, I perform high-frequency event studies on certain campaign events to examine whether yield changes during such events may have indicated higher probabilities of a Trump presidency. I also examine long-term trends of yields and yield curve slope over several months to see if those may also have indicated a Trump presidency. My results suggest that none of the campaign events I examine strongly indicated higher probabilities of Trump’s victory. And while yields and the yield curve slope appeared to have systematically increased over the 3 months prior to the election, there are potentially confounding factors that make it difficult to attribute this result solely to election expectations.