Democrats have a large turnout advantage in non-presidential elections

There are two different electorates in America. Who wins is largely a function of which one shows up.

In March, I wrote about how Democrats are doing as well or better in special elections this cycle as they did in 2018.

In that article, I proposed that what we are seeing now is two separate electorates driving election results in America. One electorate is less politically engaged, showing up in high-turnout races to deliver results that are friendlier to Donald Trump and MAGA Republicans. The other electorate is highly educated and engaged, showing up when overall turnout is weaker, and is much friendlier to Democrats.

This article revisits the subject using newer and deeper data on partisan voting habits from last month’s Wisconsin Supreme Court race. I also go more into the implications this could have for U.S. politics going forward.

Calculating the effects of turnout

In my March article, I estimated Democratic overperformance in congressional and state legislative specials this year suggests a national environment around D+8 or D+9 for the 2026 midterms, if past patterns in special elections hold. That estimate is based on applying the swing from 2024 to 2025 in the districts and states that have held specials so far to the overall nation as a whole.

However, new special election results suggest that this is, in fact, a very, very big "if." That’s because voters are more polarized now by education than they were in the last midterms, increasing the proportion of Democrats in the off-year electorate relative to eg 2018.

So let’s get to the new data. Previously, I estimated that around 80% of the swing in special elections from 2024-2025 was likely due to differential turnout among Democrats, with the remaining 20% due to people actually changing their minds. This estimate was pretty crude, but as it turns out, also pretty accurate.

In a new analysis of last month's Wisconsin Supreme Court race published at Split Ticket on April 21, 2025, elections analyst and friend of the blog Lakshya Jain analyzed precinct-level results and found that 73% of the swing in candidate margin in that race was due to turnout — and the remaining 27% persuasion. Of the 11-point swing to the left in Wisconsin, about 8 points were from turnout and 3 from actually changing minds.

Jain calculated this estimate by looking at the relationship between the change in (1) partisan vote margin and (2) aggregate vote totals in each precinct. With that relationship quantified, he could predict what the result in each precinct would have been if turnout had stayed stable from 2024 to 2025, and then add up the results statewide to come up with a counterfactual. He estimates that the electorate that cast ballots in the 2024 presidential race would have backed Susan Crawford, the liberal candidate, by 2 points, versus her 10-point margin in reality.

Jain includes this chart showing how his analysis works on X/Twitter:

(NB: Lakshya says this figure was deemed ill suited for publication at Split Ticket, so I’m happy to give it a home here.)

Dual (and duelling) electorates

Here’s Jain’s conclusion:

Many of the most pro-Trump Republicans have tuned out of politics following the presidential election, while Democrats stay engaged and vote at every opportunity they have.

Obviously, the less reliable coalition Republicans are leaning on now can still work out for a party when leaders are willing to quickly make huge changes to public policy and otherwise wield political power in unprecedented ways. As a corollary, it’s harder for Democrats to win higher-profile contests and rarer, then, to wield that power. But only winning races when (a) broader conditions are in your favor and (b) you can activate your hard-to-reach supporters is not a good strategy for establishing a broad base of power.

One consequence of this, as I pointed out in my piece last month, is that:

Political power in the U.S. [now] oscillates left and right as partisans show up for and tune out of elections. Trump showed that he can overperform when voters with a lower turnout probability actually show up to the polls. That may not be true forever, but so long as it is, the rest of the time, Democrats will be doing better.

The upshot of all of this is that we would inevitably end up with a funny mix of legislators elected in on- and off-years, representing different people with sharply different partisan political preferences simply by virtue of different contests having different turnout rates. This means, yes, special elections are less predictive of outcomes than before — but that does not mean they are useless or that we should ignore them.

If you forecast this trend out and take it to its extreme, and ignore that trends can change, you get a world in which Republicans have a durable advantage in federal politics, while in the odd years and Congressional midterms, engaged Democrats show up to act as a check on their rule.

That’s not to say that the trend will stay the same — I think something will likely interrupt it — but just an example of the weird type of government we could get if the correlation love triangle between education, turnout likelihood, and Democratic vote probability keeps increasing.

What this means for the midterms

So, assuming we will have free and fair elections next year, what does this mean for 2026?

Jain's estimates suggest Democrats would gain 3 points on vote margin in the midterms even if the electorate looked the exact same as 2024, which should be treated as a best-case scenario for Republicans. That would be enough to win the majority in the House of Representatives, and maybe pick up a few state legislatures and governors.

However, it would take a much bigger swing for Democrats to pick up the Senate. Maine and North Carolina are the likeliest flips, and possible at, say, a D+6 swing from 2024 to 2026, but Democrats would need two more of either Florida, Texas, and Ohio to win the Senate majority. That would require a swing relative to the 2024 presidential election well into the double digits.

Despite our calcified politics, the Wisconsin results suggest that some people really are changing their minds in response to political stimuli. From a normative point of view, that should provide some degree of reassurance that rationalism and democracy are not completely dead — that preferences are malleable and election results are not purely a function of who is mad enough to go to the polls.

But still, data is data, and here the data says that differential turnout nevertheless has an outsized influence on our politics. With outrage over Trump’s presidency leading news coverage daily, and education polarization favoring the left, Democrats are currently in the proverbial political driver’s seat for 2026 — IF, that is, they can keep their supporters highly engaged.