|Saturday, February 17|
|PS3 Poster Session 3 and Continental Breakfast||
Sat, Feb 17, 8:00 AM - 9:15 AM
Exploring Data Quality and Time Series Event Detection in 2016 US Presidential Election Polls (303685)
*Kaelyn M. Rosenberg, Reed College
Keywords: Surveys, Polls, Mode Effect, Time Series
Public opinion regarding political polling after the last election has declined. We will explore the 2016 US Presidential Election Polls from 196 pollsters from November 2015 to November 2016. There is evidence to suggest that the data quality, particularly the mode of survey administration (paper, web, in-person, or telephone), may have influenced polling results. Literature supports that social desirability and acquiescence bias is significant in interview type modes rather than in self-administered modes. We hypothesize that estimates from self-administered surveys may be more similar to election results. We propose modeling and adjusting for survey mode effect in order to account for inconsistency between poll estimates with R. After adjusting our estimates we model these data with a time series and explore possible event detection.