Sorry, you need to enable JavaScript to visit this website.

P1-15: 2016 Presidential Election Predictions in the Undergraduate Statistics Classroom

With Alana Unfried (California State University, Monterey Bay)


The recent Presidential election provides a plethora of statistical learning opportunities for undergraduate students. In particular, utilizing real data related to current events demonstrates the importance of rigorous statistical practice to our students and increases student buy-in (GAISE College Report). Last fall, students in an upper-level applied regression course used real-time raw data from the USC Dornsife /LA Times Presidential Election “Daybreak” Poll to create their own predictions of the Presidential election outcome. This is usually the 2nd or 3rd statistics course for the 30 students enrolled. Students utilized the statistical model of their choice, and also considered the implications of the Electoral College. The students gained practical experience with messy data that needed to be cleaned, manipulated, and analyzed. The dataset and election outcomes also gave rise to discussions on the ethical use of polling data, and statistical pitfalls such as p-hacking. Student comments revealed that although many found this to be one of the more challenging projects of their academic career, they also found it to be the most rewarding. For non-election years, USC Dornsife also publishes raw data from several other current surveys that they are conducting. Further, the data can easily be used in introductory courses as well for confidence interval examples and more. The poster will discuss the merits of using real, current election data in undergraduate courses, samples of student methodology, tips for smooth implementation (handouts will be available), and student feedback on the project.