Alcon
Atlas Superstar
Posts: 30,866
|
|
« Reply #1 on: September 15, 2008, 12:05:58 PM » |
|
|
« Edited: September 15, 2008, 12:07:45 PM by Alcon »
|
It is worth noting one thing about SUSA, though:
You can't just look at 2006 exit polls and dismiss their net partisan affiliation. Why? Because SUSA undercounts independents, almost entirely independents with partisan leans -- they tend to be forced into partisan groups. I think this is because their question is phrased in a way that asks about partisan registration, as opposed to self-identification. In either case, they get fewer self-identified indies than other organizations. In the current climate, that can cut two ways:
First, it can inflate Republicans, because it "assigns" Republicans who are reluctant to identify as Republicans. The effect of this vis a vis other pollsters should decrease over time, as partisan identification becomes more galvanized, as it already has.
Alternatively, it can inflate Democrats, because among independents leaning to either party, Democrats (I believe, as of recently) outnumber Republicans. Self-identified "pure independents" tend to lean McCain. This results in a smaller, more Republican, group of independents. I'm not convinced that this will, or will not, decrease as we get closer to election day.
Effects combined, my general guess is that SUSA assigns more independents as Democrats than Republicans, and McCain-supporting independents are less likely to be assigned as Republicans, than Obama-supporting independents as Democrats. The net effect: a higher Democratic partisan affiliation, and a lower Democratic support among independents.
Does that make D+3 reasonable? Probably not, but it comes closer to explaining it. I have qualms with SUSA outside of partisan identification, but it shows that SUSA results can't just be given a "dumb weight" to exit poll results.
Another note: Perennial caveat about margin of error on partisan sub-samples. Here, it's +/-6.0% to 7.6%.
|