Date of Graduation

5-2022

Document Type

Thesis

Degree Name

Bachelor of Arts

Degree Level

Undergraduate

Department

Political Science

Advisor/Mentor

Shields, Todd

Committee Member/Reader

Robinson, Samantha

Committee Member/Second Reader

Maxwell, Angie

Abstract

In an effort to increase the accuracy of statewide political polling in Arkansas, we explore the statistical strategy of weighting with a focus on one yearly opinion poll: The Arkansas Poll. We conduct over 70 weighting experiments on the 2016 and 2020 Arkansas Polls using a variety of variables and opinion questions. From these experiments, we find that while some weighted variables tend to create larger changes, weighting typically results in a single-digit percentage change that does not substantially shift or “flip” the majorities. Due to a greater rate of change through weighting in the 2020 Poll compared to the 2016 Poll, we recommend that The Arkansas Poll perform in-house weighting from now on as a measure of accuracy where greater changes indicate less accurate raw results. Additionally, we encourage The Arkansas Poll to print these weighted results within the yearly summary report should the change be substantial enough to shift answers and perspectives. Finally, these findings and recommendations can and should be applied to other statewide political polls to create a more effective state polling system.

Keywords

Political Science; Statistics; Survey Weights; Polling; Political Polling; Arkansas

Download

Share

COinS