(Note: Virginia Tangel of the Rutgers-Eagleton Poll staff prepared this post.)
In our December 9, 2010 Rutgers-Eagleton Poll release, we included a question that asked about the public’s support for the proposed ARC (Access to the Region’s Core) tunnel under the Hudson River to Manhattan. Governor Christie had canceled the project in October citing cost overruns, but we still wanted to measure the public’s support or opposition to the project, as well as attitudes on Christie’s decision. In order to do both of these things while attempting to avoid biasing our results, we employed a common methodological tool of the split ballot experiment. In this scenario, a random half of sampled respondents (N=447) first received a question asking their personal view of the importance of the ARC tunnel to economic development in New Jersey, followed by a question measuring their support or opposition to Governor Christie’s cancellation of the project. The other group received the question about their support or opposition to Christie’s decision first (N=436), followed by the question about their own opinions regarding the importance of the tunnel. Setting up question order experiments offers survey researchers ways in which we can see effects of context in a survey.
Here are the results of the question about the importance of the tunnel:
Note that when asked first about the importance of the proposed ARC tunnel under the Hudson River, 37% of New Jerseyans reported that they thought the tunnel was “very important.” But, respondents asked this question after offering their opinion about Governor Christie’s cancellation of the tunnel had significantly different views. In this sample, only 22% -- 15% fewer than those without the Christie context – say the tunnel is very important. Note also that the “Don’t Know’s” go down. Providing information that Christie canceled the tunnel helps people come up with an answer on its importance. Clearly, asking first about Christie’s decision had an effect on respondents’ opinions—a sign that survey respondents are willing to rationalize their own views to get behind the decision of the governor.
This pattern becomes even more interesting when breaking down by political party identification.
For respondents who were not asked first about supporting or opposing Christie’s decision, the tunnel seems pretty important – even more than a third of Republicans say it is very important, and few people say it is not at all important.
But when we ask about Christie’s decision first, the effect is huge for DEMOCRATS! Only 18% who get to think about Christie’s decision first say the tunnel is very important, a drop of 24 points among Democrats. Republicans and independents are also less supportive, but the effect is smaller. In general, opinions of the importance of the transit tunnel drop precipitously when placed in the context of Governor Christie’s decision – especially among Democrats.
Question order effects are the result of what survey researchers call “consistency effects” (Schuman and Presser 1981 in Wilson et al. 2008), in which a norm of reciprocity in responses comes into play. In this case, the comparative context exists when framed in terms of Christie’s cancellation of the ARC tunnel project. If a respondent first learns of Christie’s cancellation and is asked to provide their opinion of that, some adjust their response in the following question about the importance of the ARC tunnel because they feel obligated in light of their previous answer. So, this norm of reciprocity is why we see New Jerseyans’ opinions of the importance of the ARC tunnel decline when placed in the Christie context.
Wilson, David C. et al. 2008. “Affirmative Action Programs for Women and Minorities: Expressed Support Affected by Question Order.” Public Opinion Quarterly 72:514-522.
Schuman, Howard and Stanley Presser. 1981. Questions and Answers in Attitude Surveys: Experiments on Question Form, Wording, and Context. New York: Academic Press.