Skip to main content
Menu
What To Make Of Polls That Show Americans Are Trending Toward The GOP

Are there really more Americans identifying as Republicans than Democrats now? For decades, more Americans have tended to identify as Democrats than as Republicans, even if the gap has sometimes been small. But in January, Gallup found that Republicans had taken the lead in party ID during the last three months of 2021: On average, 47 percent identified as a Republican or said they leaned toward the GOP, while 42 percent identified as a Democrat or leaned toward the Democratic Party. Given we’re headed into what will be a competitive midterm election year, this finding sparked a flurry of headlines in the days following the release of Gallup’s report.

But it’s still too soon to know whether more Americans are actually identifying as Republicans. Gallup’s finding could portend a lasting change, or it could also be a short-term reaction to an unpopular Democratic president, or it could be an artifact of lower response rates by Democrats frustrated with bad news for their party — or it could relate to all of the above. After all, political science research has found that an individual’s party identification can fluctuate. That said, party ID also tends to be pretty stable in the long run, such that we’d expect changes across the population to be gradual.

To that point, there are two ways of looking at Gallup’s party ID data: quarterly and annually. In the quarterly data, you can see that party ID is volatile. In general, it favors Democrats, but there are numerous spikes in the data as well as instances when Republicans have had the advantage. In the yearly data, meanwhile, it’s easier to see that more Americans have consistently identified as Democrats than Republicans, although there were periods in the early 2000s and early 2010s when party ID was more evenly matched — perhaps not coincidentally, those years featured strong Republican electoral performances — and in 1991 when Republicans even held an advantage, thanks in part to a very popular GOP president in the White House.

And that’s what makes interpreting party ID data so hard. Annual Gallup data suggests that the most recent quarterly numbers might have been a blip and that we should expect Democratic party ID to rebound soon. After all, in the last 30 years, Gallup’s annual data has found, on average, that 47 percent of Americans identified as Democrats versus 42 percent as Republicans (including independents who leaned toward each party). Then again, it’s possible that this quarterly data is pointing to a longer-lasting gain in party ID for Republicans, as was the case in the early 2000s and 2010s. This could have potentially major consequences for the upcoming midterm election, too.

In research published in 2018, political scientists Patrick Tucker, Jacob Montgomery and Steven Smith used longitudinal data from the American Panel Survey to study partisan identification from 2011 to 2016, finding that changes in presidential approval and favorability of the two parties often preceded changes in party affiliation in the short term, particularly among Americans who did not have strong ties to their party. Considering that President Biden’s job approval has suffered in recent months, it could be that the short-term changes we see in Gallup’s data indeed indicate the electorate’s movement away from the Democratic Party. 

Along the same lines, Charles Franklin, the director of the Marquette Law School poll, recently analyzed movement in party ID across a number of major pollsters, and he broadly found a decrease in those identifying as Democratic — although not the same degree of swing toward the GOP as in Gallup’s quarterly data. However, it is worth noting that Franklin was looking at party ID that did not include independents that lean toward a party, because many pollsters don’t publish party ID data with leaners.

However, that study by Tucker, Montgomery and Smith also found that changes in party ID tended to revert over time as voters’ evaluations of the parties and the president stabilize, so the initial movement we see in Gallup’s data may not last. In fact, Gallup found the two parties were running close to even in December, perhaps suggesting that the shift toward the GOP was already abating.

This tendency for people to shift their party ID in the short term only to then revert back is evident in the Pew Research Center’s party ID numbers, which are very similar overall to Gallup’s yearly numbers. For instance, between September 2018 and July 2020, Pew found that about one in five Americans in their American Trends Panel made some change to their party identification, but nearly nine in 10 identified or leaned toward the same party in 2020 as in 2018. In other words, nearly half of those who had switched parties ultimately switched back.1

And of course, as is the case anytime we’re talking about polling, there are methodological differences among pollsters that can further muddy the waters of how Americans see themselves in the political landscape. As Marquette Law School’s Franklin noted, differences in question wording can contribute to differences in partisan identification; Gallup’s party ID question asks respondents how they identify “as of today,” for instance, whereas some other pollsters ask how respondents identify “generally speaking.” And in a 1991 study, political scientists Paul Abramson and Charles Ostrom found that from 1952 to 1988, Gallup’s version of the question did produce more volatility in party ID than other pollsters’ versions of this question. That may still be the case, considering that Franklin found smaller shifts among other pollsters’ versions of this question in January.

It’s also possible that the large swing in party ID toward the GOP that Gallup observed could be partly due to Democratic-leaning Americans responding at lower rates. Democrats might have been more discouraged by what’s going on in Washington, what with Biden’s dismal approval rating and Democrats having failed to pass their much-ballyhooed social spending and voting rights legislation. There’s even a term for this — nonresponse bias — and we’ve seen it happen before: Lower response rates among Republican-leaning voters ahead of the 2020 election may have contributed to the polling error that underestimated GOP support. But it’s also possible that some pollsters are weighting their polls in ways that could diminish the impact of nonresponse bias or other short-term factors. For instance, YouGov weights its polls to match the 2016 and 2020 election results, which means that while they also found a Republican-leaning trend in party ID across 2021, the magnitude of that shift was much, much smaller.2

Overall, it’s probably too early to say how much of the Republican shift in partisan identification is real, or how long it will last. However, it’s also not clear whether this particular data point tells us much more about the electorate than we already know from Biden’s job approval, generic congressional ballot polling and other political polls. In their study, Abramson and Ostrom did show that party identification polls were moderately predictive of electoral outcomes for the House and presidency, but direct measures such as the generic congressional ballot and head-to-head polling are still much better indicators of what to expect in elections. All things considered, this may be just one more data point that directs us toward what we already expect: Things are looking pretty tough for the Democrats in November.

Footnotes

  1. This includes those who switched between the parties, and those who switched to independent and then switched back to their previous party.

  2. YouGov conducts online polls using nonprobability samples. This means their samples aren’t totally random, but it’s possible their approach of using past presidential results to weight results may be less error-prone than other methods. For instance, a 2016 Pew study compared the error of different nonprobability surveys, and it found that a sample weighted by party ID — similar to YouGov’s practice of weighting to past presidential vote — produced the smallest amount of polling error across a number of known benchmarks. But Pew did flag some concerns regarding online pollsters using such approaches — specifically, if done poorly, weighting by a predetermined result could unduly influence other findings in the survey.

Geoffrey Skelley is an elections analyst at FiveThirtyEight.

Mary Radcliffe is a senior research assistant for FiveThirtyEight.

Comments