Tag Archives: Party identification

Individual change in party identification, 2016-2020

Posted on by

Recently I tweeted about shifts in aggregate party ID, following Gallup’s release of shifts in their polls. I find 4 other pollsters showing the D-R margin tightening, though not as much as Gallup. Blog version: pollsandvotes.com/?p=217

But what about individual change?

For tracking individual change we need panel data. Thanks to Democracy Fund and Voter Study Group we have a public panel from 2011 through 2020.

I prefer fresh cross sections for tracking aggregate shifts, but panels are THE thing for individual change.

See (or rather hear) the @FiveThirtyEight podcast of Jan 31 for a discussion that includes issues about party leaners that we’ll see in about one tweet here. How much more movable are leaners? Here strength of partisanship matters.

VSG uses the “Michigan” party id item: “Generally speaking do you think of yourself as…” followed by “strong or not so strong” or by “lean to Dem or Rep” to make a 7 point scale.

Below is the 2016-2020 panel turnover. Rows 2016, columns 2020.

Strong partisans are different from any of the 5 middle groups. Strong partisans are 90% likely to still be in same category after 4 years. The middle 5 groups are about 64% likely to be same category.

If you think party is forever, you are thinking about strong partisans.

Weak partisans (“not very strong”) either stay there (62% or 66%) or are more likely to shift into strong (19% or 20%) though 15-20% scatters toward the other party. These are pretty partisan but 1/5 weaken or shift direction. Similar for both D & R.

Leaners D (61%) & R (66%) stay leaners but some (D 12%, R 14%) jump to strong for the party. About as many (D 14%, R 13%) shift to pure independent. Add weak and about 20% become stronger partisans, 12-14 pure independent and 5%D and 2%R shift to other party

And pure independents also remain 62% unchanged. Those who shifted in this 2016-20 period were a bit more likely to shift toward R, though shifting into lean was most common (D10%, R16%)

This seems consistent w Trump attracting previously less involved voters, esp pure Inds.

Party is sticky & especially so for strong partisans. Leaners are more apt to move toward their party but a few move away. Inds also stick as much as weak & leaners, but were drawn to GOP a little more in 2016-20

There was a lot of symmetry in the movements but a slight GOP edge.

Party ID Trends, Jan. 2022

Posted on by

So I, as many of you, woke up Monday to Gallup’s latest party ID numbers with a sharp move towards GOP. The full article is here and I strongly recommend reading it all. It is more nuanced than Twitter headlines might sound.

Let’s look at 4 other high frequency live phone polls from Marist, Quinnipiac, Kaiser, and NBC/Wall Street Journal since 2014.

First Ds and Rs w/o leaners, same scales.

Now the Dem minus Rep margin since 2014, again without leaners.

Finally, the percent who are neither Dems nor Reps. This is 100-Rep-Dem, so it includes Inds, other, dk, refused. Polls aren’t consistent in reporting these, so just the Not D and Not R seems the most consistent practice here.

As with Gallup, the Dem minus Rep margin has tightened in all 3 sets of polls here. Gallup has Ds & Rs both at 28% unleaned in both 3rd & 4th quarter, and Ds had 30-25 and 31-26 in 1st 2 qtrs. Their leaned party has Rs ahead in 4th qtr.

Leaned party is not readily available for some polling organizations, so I’ve used the unleaned which are comparable across all. Shifts among leaners are not uncommon but can clearly tilt the balance. It would be nice if all reported both unleaned and leaned every time.

The big headline is right: The balance of Ds vs Rs has shifted over 2021 to a smaller D advantage. We see this in all 4 sets of surveys.

Do note that inds+other rise and fall with the election cycle, so both parties tend to decline between elections as the non-partisans rise.

But the parties aren’t losing supporters at the same rate. In 2021 it was the Dems who lost support a bit faster than the Reps.

Bottom line is the 4 polls I’ve collected for 2014-2022, QPoll, Marist, KFF and NBC all agree the Dem-Rep margin has tightened but all still have at least a small D advantage. Trending down, so that could change but it hasn’t yet, though for Gallup it has crossed over.

Some technical details

There are two different wordings that are most often used for measuring party identification. The “Michigan” wording is from the UM Survey Research Center work used in The American Voter, a cornerstone of political science:

Generally speaking, do you usually think of yourself as a Republican, a Democrat, an independent, or what?

In contrast the long-standing wording in Gallup polls is

In politics, as of today, do you consider yourself a Republican, a Democrat or an independent?

(Modern surveys randomize the order of parties in the questions.)

Both items are often followed by a strength question (especially for Michigan wordings for partisans) and a lean question for independents (both styles do this most of the time).

Quinnipiac and NBC/WSJ use the Michigan wording and Kaiser uses the Gallup wording. I’ve not been able to find the wording used by Marist as they don’t publish the full survey instrument including demographics on their website.

In the 1980s and 1990s there was a debate in political science about whether party identification moved in response to party performance or issue positions or other “short term forces”. These debates, among other things, considered the different dynamic properties of measures from the two question wordings.

This point was raised in my Twitter thread by https://twitter.com/bcburden and https://twitter.com/drjjdyck provided a pointer to one important article that compared the dynamics of the two measures:

Abramson and Ostrom 1991 argued that the Gallup wording produced more short-term variation and should be used with extreme caution. Seems like this still holds. https://t.co/gSpsqqH2wQ— Joshua J. Dyck (@drjjdyck) January 18, 2022

I would say that research that is crucially dependent on the dynamic properties of the different measures should consider Abramson and Ostrom’s warning, though I might dissent from “extreme caution” and say “with full awareness of possible differences.” Their work was part of a debate over the responsiveness of partisanship and which measure was “really” capturing it. I’d say we aren’t too wrapped up in that issue these days. As the charts above show, both measures are showing similar trends, and for my purposes that seems the fundamental point.

Given the rise in partisan polarization it might be time to update the comparative analysis of these two wordings, but that isn’t my task today.