President Trump’s temporary “Travel Ban” has been the source of national controversy in the past few weeks. On January 27th Trump signed an Executive Order that halted all immigration from 7 predominantly Muslim countries for 90 days, halted refugee admittance for 120 days, and ended Syrian refugee entrance indefinitely. Opponents of this ban are concerned of the fundamental shift in American refugee policy that accompanies the EO, and criticize it for being “Unamerican.” In the first few days after the EO was signed thousands of protesters showed up at major airports to protest the ban. Proponents of the ban hail it for its efforts to keep America safe by restricting the flows of immigrants and refugees for a limited time so that the United States can re-evaluate their admittance process for refugees.
How divided are we? This is a more difficult question to answer than you might think. In the time since the “Travel Ban” went into effect there have been a litany of polls trying to gage public opinion of the “Travel Ban.” And…well…the results are inconclusive. Rasmussen has a poll that has net positive support at +24, but Gallup has a found it has net negative support at -13. What could possibly account for this kind of spread? Vox concluded the inconsistent polling came down to what the question was, and specifically what information you did or did not put in the poll. If the question frames the “Travel Ban” as being about stopping Terrorism people are a lot more willing to support the ban, but when you put the “Travel Ban” in the context of preventing refugees from entering the country then the favorability plummets.
We wanted to test this hypothesis in our own poll. Half of our respondent were asked:
“Do you approve or disapprove of President Trump’s executive order which temporarily bans refugees from Iran, Iraq, Libya, Somalia, Sudan, Syria, and Yemen from entering the US?”
The the other half was asked:
“Do you approve or disapprove of President Trump’s executive order which temporarily bans citizens from Iran, Iraq, Libya, Somalia, Sudan, Syria, and Yemen from entering the US in order to keep out potential terrorist threats?”
We found that there was a slight shift towards approval when terrorist’s threats were mentioned, but this is not nearly as dramatic a shift was we see from the Rasmussen to the Gallup poll we mentioned earlier. We found that when refugees were explicitly mentioned net disapproval was 42.5, and when terrorist threats were mentioned net disapproval was at 38.9.
This small change in opinion could be due to the latency of the poll. We asked respondents three weeks after Trump signed the Executive Order, whereas the Rasmussen and the Gallup polls were taken just a few days after the EO was signed. Additionally, the small change could be due to our similar question wording. In a long survey, we cannot assume that the respondents read every question fully. More likely, they skim over the questions. So, when it comes to the above questions it could be that the structures were similar enough to each other that the small variance flew under most people’s radar and they were just responding to the idea of the “Travel Ban” rather than the justification for it.
A Toplines report for all survey questions, and results, is available here: Big Poll Spring 2017 Toplines
NOTE ON METHODOLOGY: This semesters’ “Big Poll” took place February 14-18, 2017. The survey was administered over the internet to a random sample of 4,500 NCSU undergraduates and 876 completed the survey, generating a 19.5% response rate. Assuming a 50-50 division in opinion calculated at a 95 percent confidence level, the margin of sampling error for this survey is +/-3.25% when questions were answered by the full sample; sampling error increases for estimates when three or more choices were offered, when fewer respondents were asked a version of a question, or when analyzing the opinions of sub-groups, such as when looking just at how females responded to a question.
In addition to sampling error, other forms of non-sampling error occur in surveys, such as confusion about question wording or the order of questions, and non-response bias (low response rates), but these types of error are not precisely quantifiable. We do not apply post-stratification sample weights for the slight demographic imbalance in gender because weighting the data would not have noticeable effects on the results.