By Michael Bitzer
With 2025 concluded and the start of what will be an intense 2026, I was reviewing the past year’s Catawba College-YouGov surveys of North Carolinians (1) to see what interesting patterns or trends occurred over the year.
One thing that stood out to me (because of its potential implications for electoral volatility in a presidential-to-midterm transition year) was a fairly consistent pattern of the state’s partisan identification—that is, until the final survey of 2025.
As part of a standard set of questions asked in each Catawba-YouGov poll, respondents can initally say what their partisan identification is: Democrat, Republican, Independent, other, or not sure. Another question asks for the respondents’ ‘strength’ of their initial partisan identification, meaning for partisans ‘are you strong or not very strong’ in your identity, while among those who initially say they are ‘Independent,’ do they lean to one party or the other, or do they consider themselves a ‘pure’ independent.
Starting with the first Catawba-YouGov Survey in August 2024, the partisan self-identifications of the surveys showed a pretty consistent pattern over the past year, until the last survey.
Figure 1: Catawba-YouGov 2024-2025 Surveys with initial partisan identifications among North Carolina Respondents.
In our first survey of August 2024, the state was evenly divided. But entering 2025, January’s Republican identification increased, apparently at the expense of Independent identification, likely due to the pending inauguration of Donald Trump back into the presidency. Following a short-lived honeymoon period for Trump, partisan identification returned to a roughly three-way division from March to August—until October, when Independents spiked sharply and Republicans fell.
Of course, this could be a ‘blip’ on the public opinion radar, an outlier dynamic that may return when we go into the field this month for the first 2026 survey of North Carolinians. But I was intrigued by this dynamic, and dug into the data across the 2025 surveys to see what might be driving it.
I first reviewed partisan self-identification by race, with the expectation that White non-Hispanic North Carolinians would be the most Republican of the three major racial groups (the other two are Black non-Hispanic and a group made up of Hispanic and all other races (non-Hispanic), the third group being the smallest number of respondents typically).
Indeed, something shifted in October’s survey among White non-Hispanic respondents.
Figure 2: Catawba-YouGov 2025 Surveys with initial partisan identifications among White non-Hispanic North Carolina Respondents.
From January’s high of 45 percent of White North Carolinians identifying with the GOP, the mid-year surveys held steady across the identifications until October, when four out of ten Whites identified as Independent, while GOP identification dropped to a little over a third.
In comparison, Black respondents held their partisan identification fairly steady over 2025:
Figure 3: Catawba-YouGov 2025 Surveys with initial partisan identifications among Black non-Hispanic North Carolina Respondents.
Among White North Carolinians, something spurred the defection from Republican to Independent identification. This pattern could be a classic “label avoidance” dynamic, where folks are free to step away from a party identity without necessarily embracing the opposition.
To potentially understand one aspect of it, I looked at the “strength” of partisanship to see where the shift in identification may have been among White respondents.
My working hypothesis was straightforward: if this shift reflected mild dissatisfaction rather than a deeper realignment, I thought it would be concentrated among “not very strong” Republicans drifting into Republican-leaning independence. This would be reflected in the 7-point partisan identification that focused on ‘strength of partisanship.’
Figure 4: Catawba-YouGov 2025 Surveys with strength of partisan identifications among White non-Hispanic North Carolina Respondents.
As the above chart shows, that hypothesis wasn’t the case. Instead, it was the drop among those who identified as ‘strong Republicans,’ from a third of respondents in January down to a quarter in October. And ‘not very strong Republicans’ dropped as well, from 13 percent to 10 percent.
Conversely, Independents increased across the board from January to October: 9 to 13 percent among those Indies who lean Democratic, 12 to 17 percent among those pure Indies, and 11 to 16 percent among the lean Republican indies.
Between the final two waves of polling (August and October), there was a noticeable shift in party identification among White non-Hispanic North Carolinians, but the movement was less about a big pro-Democratic surge and more about a step away from the Republican label.
But is this “real,” or could it be polling noise?
To analyze this, a simple robustness check was conducted: a weighted bootstrap.(2) Think of it like this: you repeatedly “re-draw” thousands of simulated samples from the original poll (while resampling in a way that preserves the survey’s weighting to reflect the target population), recompute the August–October change each time, and see how often you’d get a change this big just by chance due to normal sampling variability. In other words, this asks whether the October pattern looks unusual relative to what random sampling alone would normally produce.
On the 3-way partisan measure, the Independent increase and Republican drop hold up under that test. Here is the clearest snapshot of what changed between the late-summer and fall waves:
Figure 5. White non-Hispanic North Carolinians: PID3 change from Aug to Oct 2025. Dots show the estimated change, with 95% bootstrap confidence intervals.
The horizontal line around each dot is a 95% confidence interval (built using a bootstrap method). The vertical “0” line is the key reference point: if a category’s confidence interval stays entirely on one side of 0, that change looks real rather than just random survey wiggle or noise.
From August to October 2025, the share of White North Carolinians calling themselves Republican drops by about 7 percentage points, with the error bar on the chart staying entirely below zero. That means, statistically, this looks like a real decline rather than just sampling noise.
Over the same period, the Independent share rises by roughly 9 points, with its error bar sitting clearly above zero—again, consistent with a genuine increase. The Democratic share nudges down a bit (about 3 points), but its error bar crosses zero, so that piece of movement is too small and noisy to interpret confidently.
For readers who want the statistical detail behind that snapshot, here is the full bootstrap uncertainty graphic for PID3:
Using a straightforward within-wave weighted bootstrap, the partisan shift from August to October among White non-Hispanic respondents is robust (notably the Independent rise and Republican decline).(3)
With the more detailed 7-point party ID scale (Strong/Weak/Lean partisans separate from pure Independents), the story becomes more nuanced. The movement looks like a softening: fewer “Republican” identifiers, with more people drifting into “lean” categories or independence. But once you collapse those seven categories into big blocks like Any Republican or a single Net Republican number, the uncertainty gets big enough that you can’t responsibly claim a precise electorate-wide swing from August to October using PID7 nets alone.
Notice that all of PID7 lines cross the zero line (unlike with PID3 where Independent and Republican lines didn’t cross), potentially saying that there is ‘noise’ within this data that we can’t discount.
The bottom line: October looks like a month when more White non-Hispanic voters in North Carolina stopped calling themselves Republicans and started identifying themselves as Independents. The fine-grained 7-point scale suggests softening more than flipping, and big net measures on PID7 are directionally consistent but not definitive in this sample.
And while October marked a moment of measurable partisan softening among White non-Hispanic North Carolinians, the evidence does not point to benefiting Democrats in terms of partisan identification, but rather to a potential meaningful step away from the Republican label and toward independence—interestingly, among those who previously identified most strongly with the GOP.
These bootstrap results indicate that this pattern on the three-way party identification measure is unlikely to be pure sampling noise, even as the finer-grained seven-point scale counsels restraint about claiming a precise net partisan swing.
Of course, a few points of caution to this analysis. First, as just noted, this shift does not indicate a sudden surge in Democratic identification among White non-Hispanic North Carolinians. The data show movement away from the Republican label, not toward the Democratic one—basically to a ‘neutral’ independent political zone. And it doesn’t mean that groups moved beyond Republicans. Aggregate changes can mask important differences across partisans, demographic groups, or regions. Some groups may have shifted substantially, while others did not move at all.
Second, these surveys are cross-sectional snapshots, not a panel tracking the same individuals over time. While the October shift appears statistically meaningful on the three-way party identification, it could still reflect a temporary reaction to political conditions rather than a lasting change in identity—that is, a temporary blip on the polling radar.
Third, this does not tell us why the shift occurred. The data could be potentially consistent with hypotheses such as dissatisfaction with national leadership or “label avoidance” during periods of political turbulence, but these surveys did not directly measure the motivations behind individual respondents’ choices.
Finally, do not try to assign any of this to November’s general election. Partisan identification is related to vote choice, and many voters who identify as independents—particularly those who lean Republican—are likely to continue to support GOP candidates, especially in a polarized general election.
Whether this shift proves to be a temporary reaction (4), an adjustment due to the president’s unpopularity, or an early signal of broader volatility heading into the midterms remains an open question.
What appears is that North Carolina’s party identification—especially among White voters—could be more fluid than for most of 2025, a dynamic worth watching closely as the state enters the politically consequential cycle of 2026.
-------------------
Dr. Michael Bitzer is a professor of politics & history at Catawba College in Salisbury, NC, where he directs the Center for North Carolina Politics & Public Service.
1. The January Catawba-YouGov survey had 1,500 weighted respondents, while the August 2024 and the four other 2025 Catawba-YouGov surveys had 1,000 weighted respondents. Each of the surveys’ methodologies and findings can be found at the Center’s webpage. Each survey was written and paid for by Catawba College’s Center for North Carolina Politics & Public Service and administered by YouGov, which matched down to a representative sample of adults who are 18 and older. Each survey has an overall margin of error (adjusted for weights); where the results of subgroups are reported, the margin of error will be greater.
For each survey, the respondents were matched to a sampling frame on gender, age, race, and education, which was constructed by stratified sampling from the full 2023 American Community Survey (ACS) 1-year sample with selection within strata by weighted sampling with replacements (using the person weights on the public use file). The matched cases were weighted to the sampling frame using propensity scores. The matched cases and the frame were combined, and a logistic regression was estimated for inclusion in the frame. The propensity score function included age, gender, race/ethnicity, years of education, and region. The propensity scores were grouped into deciles of the estimated propensity score in the frame and post-stratified according to these deciles. The weights were then post-stratified on 2024 presidential vote choice as well as a four-way stratification of gender, age (4-categories), race (2-categories), and education (4-categories), to produce the final weight.
Since additional factors such as question wording and other methodological choices in conducting survey research can introduce additional errors into the findings, survey results should be viewed as informative and not determinative.
2. FAQs / Methodology on the bootstrap analysis:
What data is this based on? These results compare the August 2025 and October 2025 waves of the Catawba-YouGov survey series.
Who is included in the subgroup? The analysis shown here focuses on White non-Hispanic respondents, coded as Race_Ethnicity=1 by combining the self-identified race=White and hispanic=not responses among the survey respondents.
What are PID3 and PID7? PID3 is the three-way party ID measure (Democrat, Republican, Independent, others). PID7 is the seven-point scale that separates strong/weak partisans and partisan leaners among independents.
What does weighted mean? The survey provides weights so the estimates better reflect the intended population; all percentages here are computed with those weights.
What is the bootstrap and why use it? A weighted bootstrap repeatedly resamples respondents within each wave (with replacement), recomputes the weighted shares, and then recomputes the Oct minus Aug difference. The 95% confidence interval is taken from the simulated distribution of differences.
How should I interpret the confidence intervals? If the interval crosses 0, the observed change could plausibly be sampling noise. If the interval sits clearly above or below 0, it is stronger evidence of a real shift in the population.
What are the big caveats with this analysis? These are cross-sectional waves rather than the same individuals over time, so shifts may reflect real change, sample composition, or both (weights help, but do not eliminate this completely).
3. While the bootstrap accounts for sampling variability (and survey weighting), it does not capture all possible sources of error—such as question wording effects, nonresponse bias, or changes in who is paying attention to politics.
4. We’ll find out in a few weeks from the January 2026 Catawba-YouGov Survey if this was just a ‘blip’ or something that has momentum going into the mid-term election cycle.
