What are citizens’ views about what governments should do in a democracy? More specifically, what should be the priority when an electoral promise does not gather much support in public opinion? Does it make a difference whether there is a strong or a weak majority opinion? These are the questions that we address in this research note. We address these questions through a survey experiment whereby a sample of Canadians was invited to tell us if a government should follow its promise even when a majority of citizens oppose the policy that had been promised in the previous election campaign.
There are good reasons to argue that the party in power should do what it promised during the election campaign. Election pledges play a central role in the promissory model of representation (Mansbridge, Reference Mansbridge2003). From that perspective, in an election the various parties reveal in their platforms what they would do if they win the election, the voters examine these platforms and then vote for the party whose pledges they like the most. As a consequence, the party with the most popular promises wins the election, which provides parties with the incentives to come up with policy positions that are in accord with public opinion. Furthermore, voters can reward or punish the incumbent party in the following election, depending on whether the promises have been fulfilled or not. The whole process ensures the representation of voters’ interests (Pomper, Reference Pomper1967); this is a crucial component of the Responsible Party Model (APSA, 1950).
There are also good reasons to argue that the government should do what a majority of citizens think it should do: that it should follow the majority point of view on a given issue. This corresponds to what Mansbridge (Reference Mansbridge2003) has called the anticipatory style of representation, in which the party in power attempts to maximize its chances of re-election and reasons that the best way to do so is to forget about past promises and follow present public opinion. This ensures a strong link between public opinion and public policy.
What do citizens think their government should do when there is a clash between these two types of representation—that is, when they find that the promised policy is opposed by a majority of voters? What course of action should the government take: follow its campaign promise or public opinion? This is our research question.
Our research is inspired by recent studies that have used an experimental design to determine what representational style citizens prefer (Campbell et al., Reference Campbell, Cowley, Vivyan and Wagner2019; Dassonneville et al., Reference Dassonneville, Blais, Sevi and Daoustforthcoming; Doherty et al., Reference Doherty, Dowling and Miller2016, Reference Doherty, Bryan, Willis and Witry2019; Werner, Reference Werner2019a, Reference Werner2019b). The basic idea is to present respondents with concrete situations and ask them what they think the legislator or party should do, with the various options corresponding to different styles of representation. We leave aside the trustee style of representation according to which legislators should decide on the basis of their own judgment about the common good (Pitkin, Reference Pitkin1967). We focus on whether priority should be given to campaign promises or public opinion.
Like Werner (Reference Werner2019a, Reference Werner2019b), we focus on the type of representation that the party should adopt. In a parliamentary system, as is the case in Canada, party discipline is strong; it is the parties that make the most important campaign pledges, and it is the leader of the party in power (the prime minister) who makes the final policy decisions.
However, contrary to Werner (Reference Werner2019a, Reference Werner2019b), and in line with Reference Dassonneville, Blais, Sevi and DaoustDassonneville et al. (forthcoming), we select a concrete issue that is highly salient and about which most people are likely to have formed prior opinions. We choose the issue of immigration. We present our respondents with the same specific proposal that was used by Dassonneville et al.—that is, to reduce the number of immigrants by 10 per cent.
Previous studies about the importance that people attach to public opinion show that views about what elected representatives should do shift, depending on whether a majority of the public supports or opposes a proposed policy (Dassonneville et al., forthcoming; Doherty et al., Reference Doherty, Dowling and Miller2016, Reference Doherty, Bryan, Willis and Witry2019; Werner Reference Werner2019a, Reference Werner2019b). This is an interesting finding. We suppose, however, that people take into account not only on which side the majority is (for or against the proposed policy) but also the relative size of that majority; that is, they react differently depending on whether there is a slight or a strong majority view. This is the main contribution of our study.
Finally, there are good reasons to believe that citizens also rely on their own policy preferences to determine what the government should do. This belief is in line with motivated reasoning theory (Taber and Lodge, Reference Taber and Lodge2006; Lodge and Taber, Reference Lodge and Taber2013).
Research Design and Data
We performed a vignette experiment among a national sample of 1,000 Canadians who are eligible to vote. The survey experiment was included in an online omnibus survey that was conducted by Ipsos between February 24 and March 5, 2020. The sample is nationally representative according to age, gender and region.Footnote 1
Respondents were first asked a question about their opinion on immigration: whether Canada should admit more or fewer immigrants (or the same as now). We asked this question before the treatment to make sure respondents’ personal preferences are not influenced by the treatment.
We subsequently showed the respondents a short text introducing the electoral promise, followed by a treatment, if any, and then a question asking what the party should do (Figure 1). Our experiment includes three groups: a control group with only the introduction, as well as two treatment groups with additional information about public opinion. The first treatment group was informed that the party conducted a poll showing that 55 per cent of the people were opposed to the proposal to reduce the number of immigrants. In the second treatment, the poll showed that 80 per cent of the people were opposed to the proposal. Finally, all respondents were asked whether the party should go ahead with the project. This question serves as our dependent variable.
Figure 1 The Vignettes
We expect respondents’ support for the project to be weaker in the experimental groups, as they are informed that a majority of people are opposed to the project. That impact, however, should be weaker when that majority is a mere 55 percent than when it is 80 per cent. Furthermore, a slight majority (55 per cent) could be viewed as ambiguous. In a way similar to Doherty and Wolak (Reference Doherty and Wolak2011), we thus expect people's personal preferences to play a bigger role in an ambiguous situation. Thereby, our hypotheses are the following:
H1: The two treatment groups are more likely to say that the project should be dropped than the control group.
H2: Those in the strong majority treatment are more likely to say that the project should be dropped than those in the slight majority treatment.
H3: The impact of respondents’ personal preferences is bigger in an ambiguous situation—that is, in the slight majority treatment than in the strong majority treatment.
Results
Our objective is to determine whether the size of the majority matters when there is a clash between public opinion and an electoral promise. We find that 64 per cent of respondents in the control group think that the party should go ahead with the project, compared to 51 per cent and 42 per cent, respectively, in the 55 per cent and 80 per cent treatment groups.Footnote 2 This simple comparison supports our first hypothesis.
Table 1 shows the results of multivariate linear regressions estimating the impact of the vignettes on respondents’ opinions. Our dependent variable is coded 1 if the respondent thinks the party should go ahead with the project and coded 0 otherwise. Model 1 assesses the impact of each treatment without controls. Everything else being equal, the predicted probability of believing that the project should go ahead is, respectively, 14 and 22 percentage points lower among those who had the 55 per cent and the 80 per cent vignettes. A post-estimation Wald test confirms that the treatments have significantly different effects (Prob > F = 0.0225). Thus, the 80 per cent vignette has a significantly stronger impact than the 55 per cent vignette, which supports our second hypothesis.
Table 1 The Impact of the Treatments on Supporting Going Ahead with the Project
Note: Standard errors in parentheses.
*p < .05; **p < .01; ***p < .001
We add in Model 2 respondents’ personal views on immigration. We create two dummy variables, respectively, for those who believe Canada should receive fewer or more immigrants (the reference category corresponds to those who believe Canada should receive about the same number). Respondents’ personal views help to explain their opinions about what the party should do, but the treatment effects remain unchanged.
In Model 3, we add interaction terms to determine whether respondents’ personal views affect their reaction to the treatments. Finally, we add in Model 4 controls for age, gender, income, education, whether the respondent is born in Canada, whether the respondent's parents are born in Canada, and region. All interactions between the treatments and respondents’ personal views on immigration have a positive sign, and three of the four interactions are statistically significant. This indicates that the negative impact of the treatments is weaker among both those who want more and those who want fewer immigrants than among those who prefer the status quo (the reference category). This suggests that the impact of the treatments is stronger among those who are more ambivalent (they want about the same number of immigrants) and more prone to be swayed by new information.Footnote 3 In line with hypothesis 3, there is an interaction between the treatments and respondents’ views on immigration, but what appears to matter is not the ambiguity of the situation but people's degree of ambivalence on the issue. As such, the evidence does not support hypothesis 3.
To give a better overview of our results, we present in Figure 2 the adjusted predictions of the propensity to think that the project should go ahead in each treatment group, according to the respondent's personal view. We can see that the impact of the two treatments is much stronger among those who think that the number of immigrants should remain about the same.
Figure 2 Predicted Support for Going Ahead with the Project
Discussion
Keeping campaign promises and following public opinion are two important considerations that parties should take into account when deciding to go ahead with a project, according to citizens. Our objective in this study was to assess whether it matters whether there is a strong or a slight majority view in the public. To do this, we ran a survey experiment on the issue of immigration. In our control group, no information was given about public opinion. In the treatment groups, respondents were told that 55 per cent or 80 per cent of the people are against the project.
Our results are straightforward: the size of the majority matters. When respondents are informed about the election promise but are not told about public opinion, a majority (64 per cent) say that the party should fulfill its promise. When they are informed that a slight majority (55 per cent) are opposed to the project, they are evenly divided (51 per cent say that the party should go ahead). But when they are told that a strong majority (80 per cent) are opposed, then only a minority (42 per cent) insists that the promise must be fulfilled.
Although these considerations help explain why people think that the party should drop or not drop the project, the respondent's personal view on immigration is also extremely important—that is, a respondent is much more likely to think that the party should go ahead with the promise to reduce the number of immigrants when that person personally favours having fewer immigrants. Furthermore, those who are more strongly affected by information about the state of public opinion are those with a “moderate” opinion on immigration, who think that the number of immigrants should remain about the same, and who are probably more ambivalent.
A limitation of the present study is that because we wanted to focus on the impact of majority size in public opinion, the electoral promise was kept present in all three conditions. As a consequence, we cannot compare the relative importance of a slight or strong majority versus that of a campaign promise. Furthermore, as there are big and small majorities, there are also big and small promises. Citizens may attach greater import to central than to secondary promises. We thus need to know how people react to different types of promises, as well as to different types of majorities. These are interesting avenues for future research, which would, of course, require additional treatment groups and larger samples.
In our study, people were asked if the party should go ahead with the project or not. This is the most direct way of tapping into views about what elected representatives should do. It could be argued that in a representative democracy, what matters is not what people think that parties or legislators should do but rather whether people are willing to reward or punish the incumbent government for its behaviour. From that perspective, the most appropriate question to be asked is if they would vote for the party if it had gone ahead with its promise even if a majority of voters were opposed. It would be interesting to see whether we get similar findings with such an approach.Footnote 4
Finally, our results raise questions about how people perceive what constitutes a majority in a democracy and also about whether people react differently to information about the majority or the minority. More specifically, we presented respondents with information that a majority of 55 per cent or 80 per cent opposed the campaign promise. We do not know whether the reaction would have been the same if, for example, they had been told that 45 per cent or 20 per cent supported the project. Would vignettes with a positive frame have a weaker effect (Tversky and Kahneman, Reference Tversky and Kahneman1981; Soroka, Reference Soroka2014)?
Whatever the case, our contribution to the literature is clear and simple. It may be misleading to ascertain how much importance people attach to public opinion in general, since the opinion depends on the size of the majority (and minority). Yes, public opinion matters a lot when there is a strong majority view in one direction. But it has less weight when there is only a slight majority.
Acknowledgments
This study was undertaken as a class project within the context of POL 2849 Les méthodes expérimentales en science politique, in winter 2020. We thank all the students enrolled in the class and who were involved in the initial stage of the study. The study would not have been possible without the generous collaboration of Ipsos, which agreed to fund the study. Special thanks to Sébastien Dallaire for his support.
Supplementary material
To view supplementary material for this article, please https://doi.org/10.1017/S000842392100007X
