Tracking 2024: What's Behind Democrats' Post-Debate Battleground Surge

Harris on the upswing in several key states

Vice President Kamala Harris hit a record-high number and lead over Republican Donald Trump following what voters viewed as a decisive victory in their Sept. 10 debate. (Pro subscribers can read more about the pre- and post-debate trends in my latest Election Watch report, published yesterday.) 

A big question was how that shift would play out in key states. Our latest state-level tracking — conducted Sept. 9-18 in 14 states, mostly after the debate — reveals a pretty rosy picture for Harris in most of the pivotal battlegrounds.  

What you’re looking at above is the shift in Harris’ margin against Trump in each state. So for example, in Michigan Harris led Trump by 3 percentage points (49% to 46%) ahead of the debate, but she now leads him by 8 points (52% to 44%), yielding a 5-point shift in her favor.

She also picked up 5 points on net in Texas, where she nonetheless continues to trail the former president, 46% to 50%. Harris saw sizable bumps in Nevada, Arizona and Wisconsin as well, where she now leads Trump 51% to 47%, 48% to 47% and 50% to 44%. (Pro subscribers can see all of these numbers here.)

To be clear, all of this movement is within the margin of sampling error between each survey, and of the four states listed above where Harris is leading (Arizona, Michigan, Nevada and Wisconsin), it’s only in Michigan that her advantage is currently outside of the survey’s margin of error. Nonetheless, it’s movement in a positive direction for her, and given the timing, it feels safe to infer that the shifts are influenced by the debate.

Before I get into the why, let’s touch on the Senate races we’re tracking, where we saw some similar momentum for Democrats down ballot.

The chart above shows the same shift as the prior one. The biggest beneficiaries of the latest trends were lesser-known Democratic candidates in open-seat races. 

Rep. Colin Allred of Texas actually managed to grab a lead over incumbent Sen. Ted Cruz, although the former NFL player’s 45% share of support doesn’t exactly inspire confidence that he could ultimately win the race in November. County executive Angela Alsobrooks in Maryland and Reps. Ruben Gallego of Arizona and Elissa Slotkin of Michigan look to be in more commanding positions following their positive swings.

The worst news is reserved for Sen. Sherrod Brown of red-leaning Ohio, who saw his advantage over Republican Bernie Moreno decline by 1 point since the debate. The second-most endangered Democrat facing re-election this cycle remains at 46% support among likely voters, a fairly scary place for an incumbent to be this close to Election Day.

Non-response bias much?

Most polls of electoral races this time of year, ours included, are among likely voters. We determine these voters based on a self-reported basis, meaning we rely on respondents to be honest about their intention to vote. It’s not used by all research firms, but is a pretty popular approach. 

But it does add an additional layer of risk to a common survey phenomenon called non-response bias, which is what can happen to a voter sample when some Americans don’t feel like responding to surveys. Shifts in perceived U.S. voter sentiment can often be explained by which Americans are choosing to take a survey. 

For example, if you’re a Trump supporter, you may have been less motivated to take a survey over the past week than in recent months given the news environment surrounding the debate. And even if you are motivated enough, you may be more likely to say that you’re not feeling particularly inclined to vote in November. The inverse could be true for a Harris supporter.

As I’ve noted before, Morning Consult weights on 2020 presidential vote — not political affiliation. That means the partisan splits in our surveys are fairly consistent, but not quite as consistent as if we were weighting on actual political affiliation. It also means that we can look at how our sample evolves from fielding period to fielding period to see what’s potentially behind the movement in support for certain candidates.

That brings me to this next chart, which shows the shift in the partisan composition of the sample over the two surveys discussed above.

Thinking about the shifts in support I highlight above, there are three pivotal states that especially stand out as potential sources of non-response bias: Michigan, Nevada and North Carolina. In each of these states, we see the Democratic-identifying share of the sample increase by 3 points and the GOP-identifying sample decrease by at least 1 point.

There are also signs of depression on the Republican side in two other key states, Wisconsin and Pennsylvania. In these so-called potential tipping-point states (i.e. the states that look primed to deliver the 270th vote to either candidate if won), the GOP-identifying sample decreased by 3 points.

All of this suggests to me that we’re looking at a relative high-water mark for Harris in these states, especially Michigan and Nevada, and it’s more likely that any polling error on Election Day would push her numbers downward a bit — and correspondingly, Trump’s upward.

At this stage of the presidential race, you’d rather be Harris than Trump. But it’s very likely that Harris’ leads in these states are smaller than our latest round of survey results suggest, which means this race very much remains a toss-up.