My post on weighting by Party generated by far the most comment, and has left a few good questions unanswered. I’ll try to answer them in the next several posts.
For now, let me add some additional data released by the Pew Research Center last week that I had missed (thanks for the tip to in-the-know emailer Mr. X). Their report covers a lot of the same ground as I did, but adds some very pertinent data. Here are the highlights, but read it in full:
They include the most complete summary I’ve seen yet on changes in Party ID before and after the Republican convention as reported by Pew, ABC/Washington Post, CBS NYTimes, Gallup/CNN/USAToday and Newsweek among registered voters (including data not included in the public releases). Each survey shows relatively small increase in Republican identifiers (+1 to +5) and similarly small shifts away from the Democrats (-2 to -6). If you average the results for the five organizations, we see a five-point Democratic advantage in early August (37%-32%) shift to a two point deficit in September (33%-35%).
Pew also reports data from three “panel” studies – from 1988, 1992 and 2000 — that interviewed respondents from the same random sample twice: once several weeks or months before the general election and again a few weeks after.
· In 2000, 18% of respondents changed their answer between September and November
· In 1992, 26% changed their party identification between June and November
· In 1988, 16% of the sample changed its party identification in just two weeks, shifting from a 5% Democratic edge (35% to 30%) to an even division (33% to 33%). Note the similarity in both time and magnitude to the shifts seen this month.
Finally, Pew includes a graphic showing trends in Republican identification as measured by Gallup over the last two years, that shows frequent spikes and drops drops that often follow major events, including a roughly 10-point GOP gain around the capture of Saddam Hussein.
All of which suggests that the slight increase in Republican party ID evidenced on polls last month is entirely plausible, but may also be quite short lived. Pollsters weight by long term averages in Party ID at their own risk.
Bonus: Gallup’s Frank Newport’s blog covers much of the same ground (hat tip to Dalythoughts). Thanks to Newport for the link.
[Continue with Why & How Pollsters Weight, Part I]
Corrected Name of Pew Research Center (10/29)
I have a question I don’t think you’ve covered.
I read somewhere last week–and now can’t find the source–that large cities, which vote Democratic, vote more consistently Democratic than do areas that vote Republican.
When a Republican candidate wins in a suburb, he wins by a lower margin than a Democratic candidate enjoys when he wins a city.
I believe I read this is also true of the “exurbs.” Areas that go Republican are somewhat or fairly mixed; areas that go Democratic often tend to be homogeneously Democratic. (Or, at least, large cities are homogeneously Democratic . . . )
Is this true?
If so, what are the implications? (The implications for anything at all—-the November election, shifting demographics in the parties, the effects of gerrymandering, etc.)
I’d also be very interested to know your take on the most-polarized-public-ever meme.
I’ve now read three different positions:
–America is more polarized than it’s ever been, at any time, with fewer swing voters & more intense & angry identification with each party
–ordinary Americans are not particularly polarized; it’s the elites who are polarized. The higher your degree (B.A., M.A., Ph.D.) the more likely you are to be fiercely committed to your party’s candidate (I think this is the Alan Wolfe point of view, though I wouldn’t swear to it)
–No one is especially polarized; it’s the parties themselves that are polarized, thanks to gerrymandering. Voters must vote for one of the candidates offered them, which makes voters look more polarized than they are. A vote for Kerry is quite different from a vote for Bush, but the people casting those votes aren’t necessarily all that different in POV, and might prefer more centrist candidates.
I have no idea whether there are more positions on this question—-if there are, that would be interesting, too.
Thanks!
Thanks for your site, and for the good explanation of confidence intervals.
I agree with you that party ID is more fluid — at least in the contemporary period — than the re-weighters assume. However, could using 2000 data on the partisan composition of the voters be just as valid a way of estimating the partisan composition of “likely voters” as standard forms of likely voter screening? Obviously, assuming that 2000 and 2004 will be alike in terms of turnout is a big “if”, but it strikes me as no more or less arbitrary than many likely-voter screening questions. If one has reason to believe that 2000 and 2004 will be comparable, then using partisan weighting to reflect that assumption is certainly plausible.
Dear Mr. Geraghty
I exchanged an email or two with Jim Geraghty of the Kerry Spot today. At one point he asked me if I had noticed any of the newer polls having “whacked-out samples.” My reply to him rambled into a territory I had been meaning to blog about but had ye…
All Cheap Cruises in one click away
Bit Torrent Links
BitTorrent allows users to download a file from multiple different people, rather than a central location.