HuffPollster: Some Republicans Less Likely to Support Donald Trump Out Loud
September 18, 2015
'MODE EFFECT' APPARENT IN TRUMP POLLING - The involvement of an interviewer appears to dampen support for Donald Trump slightly in surveys on the 2016 Republican race, according to a chart prepared by Democratic research analyst Jon Robinson. “Trump support among polls without an interviewer has been anywhere from five to ten points higher than in polls with interviewers since Trump’s rise in the polls began in July,” Robinson writes. No other Republican or Democratic candidate sees their support affected in this way.
Robinson’s chart begins with all national public polls on the Republican race compiled by HuffPost Pollster and divides them into two categories: Those that used live callers for some or all interviewers and those that “had no interviewer whatsoever involved in the process,” including both internet and automated telephone polls. The chart uses a loess regression to plot smoothed trend lines for each category of poll weighted to reflect their sample sizes (with each dot representing a poll and the size of the dot proportional to the number of interviews conducted).
Natalie Jackson, HuffPost’s Senior Data Scientist, modified Robinson’s chart to display bands showing the potential for variation based on random chance alone. The lack of overlap between the two bands throughout much of August indicates a statistically meaningful difference between live interview and automated or internet polls in measuring Trump’s support. (Robinson also produced similar charts for other candidates, although the mode differences for these were not statistically significant).
Click here to read on HuffPollster >