Thursday, August 21, 2008

Party ID Apples and Oranges

In the next few weeks, election polls will be coming fast and furious. And that also means an increase in professed skepticism from party leaders whose candidate ends up on the short end of the results.

One of the most common, and misused, attack strategies is to claim the poll doesn’t have the “right” partisan mix. Usually, these partisan critics will point out how the offending poll’s party identification is out of line with the state’s party registration figures. Party leaders know this is a red herring, and so should you. Here’s why.

Party affiliation can be viewed in two ways: the party you register with and the party you “identify” with (e.g. the standard polling question “In politics today, do you consider yourself…”). Getting party registration numbers is pretty straightforward. Their relation to how voters think of themselves is less so.

On February 4, 2008, the majority of New Jersey voters – 58% in fact – were unaffiliated with either of the two main parties. Among the remainder, Democrats had a 6 point registration advantage over Republicans (24% to 18%). However, 2 days later, party registration in the Garden State experienced a sea change. As of February 6, the majority of New Jersey’s voters were suddenly registered with a party, 34% Dem to 21% Rep, while 45% remained unaffiliated.

If you accept the premise that party self-identification as measured in polls should match party registration figures, then you would also need to swallow as the notion that 600,000 previously independent New Jersey voters awoke on the morning of February 5 with the sudden revelation that after years of political independence, they would now be partisan – a sort of party ID road to Damascus.

Or you could accept the more reasonable premise that these voters were always partisan – i.e. they tended to vote consistently for one party or the other in general elections. However, due to the nature of New Jersey’s voter registration system, they weren’t recorded as partisan on the registration rolls until they decided to vote in a primary (which they did in record numbers for the Feb. 5 presidential primary. I’ve already posted about that phenomenon here.)

Let’s look at a direct comparison of party registration and party ID. Based on estimates culled from voter list services, the 2.3 million New Jerseyans who turned out to vote in the November 2006 U.S. Senate election were registered as follows: 34% Democrat, 27%, Republican, and 39% unaffiliated/other. (As a side note: turnout is much lower among unaffiliated voters than it is among registered partisans, so you should expect the actual electorate to have fewer independents than the registration figures show). The 2006 Exit Polls showed party ID as 41% Democrat, 28% Republican, and 31% Independent. While Democrats had a 7 point registration advantage in the 2006 vote (similar to their overall registration advantage), they enjoyed a larger 13 point “identification” advantage.

Turning back to the current cycle, what appears to have happened on February 5 is that partisan voters turned out in proportions more similar to their self-professed party ID than they had in prior state primaries. However, our polling indicates that the Democratic party ID advantage has continued to grow since the presidential primary.

I’ve compiled 7 month rolling averages for party ID from Monmouth’s past year and a half of polling. Because party ID is an attitude and not a stable demographic like gender or race (see this Gallup report from January 2007), averaging partisan self-identification helps to smooth sudden fluctuations in party ID that may be due to either sample variation or reactions to major political events.

New Jersey Voters’ Party Self-Identification
(7 month rolling averages)
DateDemRepIndDem adv
Jul 0842%25%33%+17
Apr 0841%24%35%+17
Mar 0840%25%35%+15
Feb 0838%24%37%+14
Jan 0837%24%39%+13
Oct 0738%24%38%+14
Jul 0737%25%39%+12
Apr 0737%24%39%+13
Feb 0736%25%39%+11

A number of things are worth noting here. First, while the Democrats’ edge in party registration jumped from 6 points to 13 points because of February’s presidential primary, their advantage in party ID remained fairly stable in the run up (ranging from 11 to 14 points during the prior year).

Second, the Democrats’ ID advantage has grown since the February 5 vote. The percentage of voters identifying themselves with the Democratic party has increased from 38% to 42%, while Republican identification has remained stable at 25%. (This is also supported by a partisan “enthusiasm gap” noted in our July poll).

Third, while both Barack Obama and frank Lautenberg are maintaining leads in our latest poll, they are both underperforming the partisan ID split. Among registered voters in our July poll, Obama was ahead of John McCain by 14 and Lautenberg led Dick Zimmer by 10, while the Dem party ID advantage was 17 points.

The bottom line is that the apples of party ID as measured in polls are not the same as the oranges of party registration. If you see a poll with party ID that matches New Jersey’s registration, the pollster likely gave too much weight to unaffiliateds who are unlikely to vote. Otherwise, when you hear a party spokesperson say something like “We know in New Jersey that’s not exactly how the makeup goes” to shoot down a poll’s party split, consider the source and check the facts.