Thursday, October 27, 2005

Party ID

FDM fires off an attack on my critique of their innumerate swipe at the SUSA tracking polls.

Survey USA didn’t ask a question about party identification.

The poll did ask a question inquiring about a voter's party affilation. Evidently, Josh doesn’t understand that questions about a respondent’s party affiliation and a respondent’s party identification are completely different.


Then they propose a bunch of questions that no actual pollster would actually ask, all under the (false) assumption that pollsters treat "party affiliation" as something based on voter registration, and wholly distinct from "party ID."

Not The Harris Poll. The statistic they report as "Party Affiliation" is generated by asking:
"Regardless of how you may vote, what do you usually consider yourself – a Republican, a Democrat, an Independent, or some other party?"


If you want to know what someone's party registration is, you look at the goddamn voter registration card. If you want to know how people "consider" themselves, "identify" themselves or "affiliate" themselves, you ask a question like Harris's and then you use party ID and party affiliation interchangeably, as Harris does in that release.

If FDM had data, maybe a link to an actual pollster explaining this, or if they had bothered to contact SUSA and ask what question they use, and whether they feel party ID and party affiliation were interchangeable, I might be impressed. But FDM decided to hang their entire defense thus far on the fact that SUSA reported "affiliation" not "ID." Had they Googled in those terms, they'd see that pollsters seem to use the terms without distinction, though "affiliation" is a more formal term than "ID."

Consider the question the Pew Center for People and the Press uses to compute party ID (or party affiliation):

In politics TODAY, do you consider yourself a Republican, Democrat, or Independent?
Still don't believe me? Here's the LA Times's polling director on national party ID/affiliation (scroll down):

Party ID is a moving variable that changes from one election to another, and weighting by party registration makes no sense nationally because many states don't have their voters register by party and some states don't have voters register to vote until the day of the election.

"Here is the breakdown of party affiliation in Times polls going back to September 2001:












DATEN/WAYDEMINDREPS/ELSED/KNOW
06/04 4 38 24 25 7 2
03/04 4 33 26 25 10 2
11/03 5 31 25 26 12 3
04/03 3 38 19 26 10 4
02/03 4 28 30 26 10 4
12/02 4 33 28 27 9 2
08/02 3 35 26 28 8 3
02/02 6 31 26 27 11 2
11/01 5 34 28 25 10 4
09/01 5 38 20 26 8 3


Two things worth noting. He uses "party ID" and "party affiliation" interchangeably. And "party affiliation" is a moving target. Those shifts are consistently outside or near the margin of error for the polls. In aggregate, these numbers are too fluid to reflect an unchanging underlying trend, the idea FDM advocates.

Still don't believe me? Here's SUSA's explanation of their methodology (my emphasis):

SurveyUSA doesn't "weight" to party ID.

…Every pollster asks the Party ID question differently. Every pollster places the Party ID question in a different place in its poll; depending on how close to the "who will you vote for" question the party question is, the more interaction there is between the "who will you vote for" answer and the "Party ID" question. (Meaning: a few people who told you in Question #4 they are voting for Bush may feel hypocritical telling you in Question #5 that they are a Democrat).

Here is how SurveyUSA asks the party question (we ask it identically, for a general election, in every state):

Do you consider yourself a Republican, a Democrat, or an independent?

Republican? press 1.

Democrat? press 2.

If you consider yourself an independent, or a member of some other party, press 3.

If you are not sure, press 4.

(Here is how we do NOT ask it: Are you registered as a Republican? Registered as a Democrat? Or are you not registered with either party?)

How you ask the question affects the answers you get. We like the way our question is worded. We like our data (in Oregon, and in the 35 other states we are polling).



Summarizing: fluidity in party identification is a natural part of the political process, and an expected part of a campaign, in my opinion, in today's world.
But not in the world of FDM.

Why did I put this last? Because it took the most work to find. I probably spent 15 minutes finding it. But FDM doesn't have that kind of time. Better to be wrong than learn something.

I won't ask FDM to retract their puerile title in the face of evidence. But they owe it to their readers to inform them how badly they fucked this up.

This is my beef with FDM. They don't understand the things they talk about. My conservative commenters know that I engage in thoughtful dialog, and all I ask in return is the same courtesy back. I respect thoughtful conservatives, and I try to understand them and help them understand me. I think that's valuable, and I love the way the Internet, especially blogs, makes it easy to reach across lines and across distance to have that sort of honest, open dialog. I want to argue about Dennis Moore with someone who has the cojones to take a stand on an issue and fight for it, not just work to get person X out of office.

I think politics is about ideas, not people. FDM thinks it's about party. And that's a policy that rarely does anyone any good.