Colorado College Logo

  DigitalCC

Use AND (in capitals) to search multiple keywords.
Example: harmonica AND cobos

Explaining Election 2016

by Nunez-Garcia, Emilio A.

Abstract

The 2016 presidential elections were a mess. Despite high confidence in the polling industry, as in years past, the polls failed to accurately predict the next president. As improving the accuracy of polling methodology is key to improving the accuracy of polling forecasts, this paper intends to examine the root causes for the polling miss. Specifically, this paper hypothesizes, and will examine, whether there was a statistically significant larger polling error in states in which higher proportions of the electorate are made up of uneducated white voters. In order to provide a more wholesome regression, this model includes of other possible variables that may have contributed to the polling miss. This study uses FiveThirtyEight’s database on polls conducted in 2016, the CEPR transcribed CPS ORG 2016 survey, and the 2016 Cook Political Report’s county election results. Demographic data from the CPS is regressed against a Trump margin of error that is calculated by subtracting average polling results from his election day results. Regressions of independent variables against this measure intend to identify demographic groups that may have been excluded or underrepresented from polling results due to some unforeseen bias. This study concludes that polls tended to underestimate Trump in states with high uneducated white voter populations, as well as those states that have experienced an increase in unemployment rate. The results from this study were statistically significantly. Although these findings can’t not pin down the exact bias that caused these discrepancies within the polls (given that they just measure which groups were missed and not why) they can be used to guide further research to determine exactly where pollsters encountered survey bias in 2016.

Note

The author has given permission for this work to be deposited in the Digital Archive of Colorado College.

Colorado College Honor Code upheld.

Includes bibliographical references.

Administrative Notes

The author has given permission for this work to be deposited in the Digital Archive of Colorado College.

Colorado College Honor Code upheld.

Copyright
Copyright restrictions apply.
Publisher
Colorado College Tutt Library
PID
coccc:27370
Digital Origin
born digital
Extent
46 pages : illustrations
Thesis
Senior Thesis -- Colorado College
Thesis Advisor
Esther Redmount
Department/Program
Economics and Business
Degree Name
Bachelor of Arts
Degree Type
bachelor
Degree Grantor
Colorado College Tutt Library
Date Issued
2017-05