Political junkies view their party affiliations as key parts of their identities. Most American don’t.

As a result, the number of people who consider themselves Republicans, Democrats, or Independents tends to shift over time. Generally, when a party wins a presidential election, the number aligned with that party goes up (that’s part of the reason they won). Usually, the number affiliating with that party begins drifting downwards shortly after the new president takes office and begins governing.

Most of the shifting is on the margins. For example, someone who is independent might begin leaning towards the winning party during a campaign. And some who previously leaned towards the losing party shift back to independent status.

For the majority of voters who talk politics once a week or less, these shifts are not terribly significant. However, for pollsters, they are very important. There is no better indicator of how a person will vote than his or her partisan allegiance.

The Challenge

Some pollsters treat the party identification breakdowns as a result from the survey rather than a weighting target. If the study is carefully conducted and other variables properly accounted for, the party identification numbers should work out.

That’s a reasonable approach in many ways, but it creates some challenges for interpreting trends. For example, in a survey with a 3-point margin of sampling error, one survey might find 34% Republican, 32% Democrat, and 34% Other. The next survey by the same firm might find 32% Republican, 35% Democrat, and 33% other.

On a net basis, that’s a shift from R+2 to D+3. So, it’s hard to know whether any other shifts in the survey results are real or just a reflection of a different partisan mix. The challenge is highlighted by Gallup party ID trends in 2024. In the first half of October, they showed Republicans with a 7-point advantage (including leaners). In the second half of the month, they showed Democrats with a 1-point advantage.

Some pollsters address this problem by assigning a specific partisan mix as a weighting target. This approach makes it easier to identify trends, but has two potential problems. First, the partisan mix set by the pollster may be wrong. Second, this approach will miss any gradual changes in the partisan mix.

The RMG Approach—Dynamic Partisan Weighting

At RMG, we conduct our surveys to ensure a sample that is broadly representative both geographically and demographically. We then lightly weight the results to match the population in terms of DMA size; racial mix of counties; population density of zip code and of county; gender and age; race; education by age, race, and gender; marital status; employment status; and internet use.

Following a preliminary weighting according to these parameters, we determine a preliminary partisan mix for the sample. We then average that total with results from our five most recent national surveys to create a six-poll rolling average. That becomes our partisan weighting target for final survey processing.

No system is perfect, but we believe that this dynamic approach provides us with broadly stable results to measure trends while still allowing us to capture the ongoing shifts in partisan identification.