After running our national conjoint survey in August, we built issue-level support and salience models to find voters who might respond more positively to specific issue-based messaging. 

In brief, the support models predict how conservative or progressive a voter is in a specific issue area, while the salience model predicts how much they care about that issue area relative to others when evaluating a candidate. The conjoint survey design allowed us to better understand how voters might make trade-offs between different positions a candidate could hold on issues related to abortion, crime, immigration, protecting workers, and “wokeness” in schools.  You can read our detailed discussion of these scores here. 

In this blog, we’ll explore how voters were distributed across these different issue areas and how the different issue models might be correlated.

We’ll start with transparent density plots of issue sentiment. 

  • Abortion: Our model follows a bimodal distribution with most voters either only being okay with modest restrictions or completely opposing any restrictions. 
  • Crime: Our model shows that voters are slightly conservative on crime on average.
  • Immigration: Our model shows that voters are slightly conservative on immigration on average, but there is more spread than there was for crime.
  • Party: Unsurprisingly, our model shows that voters are largely concentrated at the two ends of the distribution with some people in the middle.
  • Wokeness: Our model shows that most voters are in favor of letting local school boards decide curriculum on racism and sexuality or mandating schools teach those subjects. 
  • Workers: Our model shows that voters either want very aggressive tariffs against China or are very supportive of strengthening unions as ways to support workers in the US, resembling the density plot for party affiliation.

How do those sentiment scores correlate with each other? Do some issues just cluster together? We plotted out a correlation matrix to find out. Greener squares indicated more positive correlation between two issues, while dark orange squares indicate stronger negative correlation. 

Some key takeaways:

  • Across each issue, the more progressive positions were positively correlated with each other – except for wokeness in schools
  • Abortion and party were strongly correlated – perhaps unsurprising given the high salience of abortion and the recent histories of each party.
  • Wokeness is effectively uncorrelated with all other issues. This might be due to the very high concentration of voters in one place in the distribution, leaving very little variation.

What about salience? Do we see a lot of variation in how much people care about these issues? Interestingly, it seems like most voters care a little about each issue.

What do the issue salience correlations show? We see little evidence of any correlations across issue salience models. There might be a weak correlation between caring about abortion and wokeness and caring about crime and immigration. But, overall, there isn’t much evidence that voters who care a lot about one issue will end up caring a lot about another issue. This might suggest that while we can see patterns in the aggregate (such as abortion being much more salient than wokeness), each individual’s personal politics can be highly idiosyncratic.