Public Policy Polling for Daily Kos and SEIU (6/23-26, MoE: ±3.1, registered voters, Obama trendlines 6/16-19, all others 6/9-12):
Drops across the board for Democrats, but if you've been a close follower of our polling, by now you can probably guess the usual reason. Once again, it's our typical flux in party ID, a topic we discussed in detail
last week. Today I'd like to talk about something else entirely — a mysterious artifact of our polling that we're still trying to get to the bottom of. The issue is something I'll call "creative geography."
So what do I mean? Well, in every poll, we ask respondents to tell us where they live, and we give them one of four choices: Northeast, Midwest, South, and West. But we began noticing that the share of people in the West, in particular, seemed somewhat light compared to actual population distributions in the United States. We then ran an analysis comparing area codes to geographic responses and found that quite a few people were telling us they lived somewhere other than where you'd expect based on their area codes. (For landlines, which are the only numbers we call, studies show that region ID should be very high.)
Based on this, we decided we wanted more information, so we started asking respondents to punch in their ZIP codes as well. We also changed the wording of our question to emphasize that we wanted to know where people live now, not where they are originally "from." But the problem hasn't gone away:
The map above compares ZIP codes to self-reported region, and as you can see, there's quite a bit of mismatch. I'm convinced that a lot of people are still telling us where they're from, rather than where they live. For instance, I don't think there's a lot of confusion about where Texas and California sit geographically, yet respondents identify as living all over the place in those two states. Since both states are huge and have a lot of in-migration, I think at least some people are punching in their ancestral homes when they get to our regional question.
There are also quite a few areas where folks simply disagree with traditional regional definitions as proclaimed by the Census. For instance, quite a few people in eastern Ohio identify as living in the Northeast rather than the Midwest. And Coloradoans (to my surprise) also seem torn between the West and Midwest. The area right along the Mason-Dixon line is a huge jumble as well.
It's all a very interesting conundrum, and it points up a much larger issue with opinion polling in general: You always need to view what people self-report with a healthy dose of skepticism. At least in this case, though, by asking for ZIP codes, we have a pretty good cross-check on what respondents are telling us. One thing we might try doing in the future is breaking out our regional demographics using ZIPs rather than self-reported regions. I'll be very curious to see if there wind up being any differences, and I'll be sure to write it up either way.