The Importance Trap: What Voters Say They Care About Isn’t What Influences Their Votes
Executive Summary
A survey conducted during the April 2026 federal by-election in the University-Rosedale riding demonstrates a stark disconnect between what voters say they care about and what actually drives their behaviour. The findings reveal that traditional issue-importance polling creates dangerous strategic blind spots for political campaigns.Key Findings
- Stated Priorities: Voters identify housing affordability, food prices, and the economy as their top three concerns, while ranking Artificial Intelligence and foreign policy among their lowest priorities.
- Latent Drivers: Foreign policy positions had the strongest influence on vote choice.
- The AI Surprise: Despite ranking near the bottom in stated importance, a candidate's position on AI regulation influenced vote choice about as much as housing and food affordability.
Implications for Campaigns
- Stated Importance Is Backward-Looking: Relying on traditional issue importance questions is circular, largely reflecting what is already dominating the current media cycle rather than what will drive future choices.
- Conjoint Analysis Reveals Actionable Drivers: Instead of asking voters what they care about, campaigns should use choice-based conjoint analysis to simulate real-world trade-offs. This quantifies the latent influence of policy positions on voting decisions.
Pollsters regularly present data on issues voters tell them are important. There are various ways of asking about issue importance, but, typically, pollsters provide respondents with a list of issues and ask them to identify the most important two or three. Candidates and parties often rely on such issue importance measures when they decide what issues to focus their campaigns on and how to position themselves in election campaigns.
Why Issue Importance Questions May Be Misleading
It’s surprising responses to such questions get so much attention. Scientific research suggests responses to such questions tell us little about what drives votes. Research in psychology tells us that humans struggle to tell us what motivates their behaviour.
We have also long known that survey responses reflect issues that are discussed by political elites. That means that one of the determinants of what people say is important to them in surveys is what political parties talk about. Parties and candidates relying on issue importance is circular. Does it really make sense for parties to determine their strategies based on priorities they themselves influence?
Prior studies have found that survey respondents’ self-reported issue priorities have little relationship to the issues that influence their votes. In this article, I run a similar test to compare self-reported issue importance to what motivated voters in a recent federal by-election in Canada.
Like previous research, I find that what survey respondents say matters to them tells us little about what drives their votes. Parties and candidates would be wise not to rely on subjective issue importance when deciding what issues to focus on and how to position themselves. Instead, it makes a lot more sense to use conjoint analysis to test reactions to a variety of issue positions.
The Study
I ran a small survey in the context of the federal by-election in the University-Rosedale Canadian federal riding held on April 13, 2026. The sample is admittedly small, 195 respondents. That’s what could realistically be collected at the riding level from online panels during a short period of time.
Despite the small sample size, the data are illustrative of the limitations of issue importance measures. Moreover, the main analysis relies on a choice-based conjoint study, which does not require as large a sample of respondents as more conventional survey questions. Because respondents go through multiple choice tasks, there is a total of 1,788 observations in the conjoint.
What Respondents Said Mattered
I first asked respondents a standard issue importance question. I asked them which three of a list of 21 issues are the most important to them. Figure 1 shows the results.
Figure 1: Issues Respondents Say Are Most Important to Them
Percentage of respondents selecting each issue as one of their top three priorities.
Housing affordability is the top issue, followed by food prices and the economy. Artificial Intelligence and threats to democracy from Artificial Intelligence are among the five least important issues. The two foreign policy issues, the war in Iran and the conflict between Israel and Palestine, are also some of the least important issues by this measure. Does that mean that housing and food affordability have the most influence on voting behaviour? Conversely, do AI and foreign policy issues have little or no influence?
The Choice-Based Conjoint
In the choice-based conjoint, respondents were asked to select a candidate from a list of three candidates with varying policy positions and party. Respondents went through 12 such choice tasks each. Here is an example:
Figure 2: Example of a Choice-Based Conjoint Task
Respondents were asked to choose between three hypothetical candidates with varying policy profiles.
Hidden Drivers: The Case of AI and the Iran War
The following figure shows how voters reacted to each issue position. One of the advantages of choice-based conjoints is that, by showing respondents multiple choice tasks, we can observe how the same respondents react to different combinations of issue positions. Because respondents evaluated many different candidate profiles, the conjoint allows us to estimate how policy positions affected individuals' vote choice. Individual-level measures are important because respondents likely reacted to different issues and had different responses to each policy position.
The conjoint analysis allows us to isolate exactly how individual policy positions shifted a candidate's support (measured in percentage points) relative to a baseline position.
The figure includes points showing each respondent’s reaction to each issue position other than the reference position. For reference positions, there is a single dot at 0. The darker a dot is, the more confident the data allow us to be about a respondent’s reaction to an issue position. It also shows how much each position mattered on average in a box.
Figure 3: Impact of Policy Positions on Candidate Support
Estimated change in the probability of supporting a candidate with each position relative to the least popular position (percentage points).
Even though housing and food affordability are the issues respondents said were the most important to them, they don’t appear to have mattered a huge amount to their vote choice. Respondents didn’t seem to react as strongly to any of the positions they saw on those issues as the issue importance measures would suggest.
AI regulation was one of the issues respondents said is least important to them. Nevertheless, it did make some difference to a strong minority of respondents. A third of respondents were at least one percentage point more supportive of candidates calling for more regulation of AI. On average, respondents were 1.1 percentage points more supportive of candidates calling for AI regulation. In a close election, an issue like AI regulation can make a difference.
More fundamentally, the war in Iran was also one of issues respondents said mattered least to them. When it came to behaviour, it was the issue that mattered most. On average, respondents were 10.9 percentage points more supportive of candidates who wanted Canada to stay out of the war.
Voters had a very strong preference for staying out of the war. I asked respondents whether Canada should join the war and only six percent said yes. Seventy-eight percent said no. If one of the parties had supported a role for Canada in the war, the Iran issue likely would have mattered to voters. Instead, the parties realized how unpopular that position would be and kept Canada out of the war. Similarly, the Israeli-Palestinian conflict appears to have considerable potential to influence votes.
Issue Importance Is About the Past. Conjoints Can Tell Us About the Future
This analysis is admittedly based on a small sample. However, the issues respondents tell us are important to them are not the same issues that influence their choices between candidates. That is consistent with decades of research showing that people aren’t good at explaining their behaviour.
One of the biggest problems with issue importance questions is that responses to them partly depend on what is already on the political agenda. The issues that aren’t currently on the agenda might influence voters more than those that are if candidates and parties start focusing on them.
Issues voters tell us are important are partly based on what they’ve heard candidates and parties talk about in the past. Conjoint analysis can tell us what will matter in the future.
Of course, the sample is limited to one riding and a relatively small number of respondents. However, the findings align with a broader body of research questioning the usefulness of direct issue-importance measures.