Opinion polls may inadvertently affect public opinion itself as people change attitudes after learning what others think. A disconcerting possibility is that opinion polls have the ability to create information cascades or spirals of silence where the majority opinion becomes increasingly larger over time. Testing this hypothesis on attitudes towards Syrian refugees and mandatory measles vaccination, survey experiments are performed on a population based web panel using a novel automated procedure that measure the influence of an initial poll over subsequent polls. No indications of spiraling opinion gaps over time between the treatment and control groups are identified. The polls do however trigger a cognitive response as the treated respondents become more opinionated and alter their justification for their answers.
When people form their preferences on political issues, other people’s opinions matter to them. There is thus a potential self-reinforcing mechanism at play when individuals form their political preferences, where the aggregate level of support and opposition shaping public opinion. In this paper we present an experiment where we inform the treatment groups with a skewed distribution of opinions about a political issue, and then track the treatment effect over a sequence of polls with the aim of exploring the effect of the treatment over time. The research design of the presented experiment gauges to what extent knowledge of opinion polls as such have a lasting impact on people’s political preferences. Can polls themselves change the dynamics of public opinion formation among the citizens? To facilitate the implementation of the experimental design, we introduce a technical innovation which we label the Dynamic Response Feedback. This procedure automatically generates a pie chart with poll distributions from previous responses and uses it as information to respondents who are taking the survey in present time. By repeating this procedure several times a series of poll distributions are created within one survey wave. The survey experiment is conducted on 2500 respondents in The Norwegian Citizen Panel, which is a probability based web panel of Norwegian residents. The paper presents results from an experiment on the question of whether or not the respondents think that measles vaccination of children should be compulsory. We find that when respondents are presented with a poll that is slightly skewed in the sense that it displays a higher share of supporters for the issue than the true share (as found in the control group), the aggregate mean of the treated respondents also becomes higher. Moreover, the effect of the initial poll distribution has a diminishing effect over time which resembles a fading autoregressive (AR1) process becomes a factor of its own.