We happened to be in the field the month prior to election day, and we included questions on vote intention in the gubernatorial races in Virginia and New Jersey. Splitting the data by (1) 2024 Presidential preference and (2) household income (above/below $100,000 a year), we see evidence of a mix between persuasion and mobilization that largely occurred among the less affluent. Combining Virginia and New Jersey, we see:

Among households with incomes greater than $100,000:

79% of 2024 Trump voters supported the Republican candidate, 2% the Democrat, and 18% supported neither (meaning that they either said that they did not plan to vote, or did not indicate a preference, or supported a third party candidate; we take it as a signal that they did not vote for either candidate).

83% of 2024 Harris voters supported the Democratic candidate, 2% the Republican, and 13% supported neither.

In short, Republicans lost 18% of Trump voters, and Democrats lost 13% of Harris voters; perhaps there was a small net shift to Democrats among households with $100k income or more, but if so, it was beyond the precision of our survey to be confident.

The picture for households with incomes less than $100,000 is very different.

53% of Trump voters supported the Republican candidate, 10% the Democrat, and 34% supported neither.

78% of 2024 Harris voters supported the Democratic candidate, 2% the Republican, and 18% supported neither.

In this case, Republicans lost 44% of Trump voters, whereas Democrats lost 20% of Harris voters, with a bigger demobilization effect than conversion (but where conversion counts twice as much as demobilization, so the net effects on the election margins of these effects are roughly equal). 

Note that household income is correlated with multiple things here (age, education, race, gender, etc; where we see similar patterns), but the overarching takeaway is that there was substantial demobilization and conversion of Trump supporters, but almost exclusively among less affluent voters in Virginia and New Jersey.

All of this suggests that in coming elections we should see much bigger shifts to Democrats in less affluent areas relative to 2024 baselines in coming elections.

For context: our surveys had point estimates for the Democrats of 59% in New Jersey (vs 57% in the results); and 63% (also vs 57%) in Virginia. We thus overshot a bit, while most polls in Virginia and especially New Jersey undershot Democratic vote share. We note that our surveys do not aim to optimize on outcome prediction– for example, we do not incorporate any modeling of individual-level turnout. We do think, however, that for predictive purposes it is useful to report these shifts from 2024 voting baselines. 

‍