Design Your Ideal State-based Polling Survey

Ever wish you could bang pollsters and professional survey outfits over the head and say "dammit you're doing it wrong!" like I have?  Seems like it, since right here on MyDD, there's been a lot of discussion around Bowers' Inflated Clinton Poll Theory and the Democratic Strategist even picked up on it.  Well what if we could actually design what a survey looks like?  Check below the fold for more...

We all wish there were more state-based polls put out on the presidential race, and I'm sure we all wish their methodologies were a little more 'on.'  For example, here in Wisconsin, the only "regular" state polling, done by a combo of Wisconsin Public Radio (best in the country for public radio our state is) and Saint Norbert's College in Green Bay.  The other day, they released a pretty uninformative poll that included numbers on presidential preference.  The entire survey sample was 400 randomly-selected adults - not very large, and certianly not representative of the voting population or "likely voters" in either a general or primary election.

Since we have open primaries here in Wisconsin, they polled first on peoples' inclination to vote for a Democrat or Republican in the general before asking for a candidate preference.  Pleasantly and suprisingly, 47% of the respondents said they would vote for a Democrat and only 29% picked a generic Republican.  On my blog, Madison For Edwards, I speak a little bit about how this is either a fundamental re-alignment of Wisconsin politics or a big 'oops' on the part of the surveyors (and I just got ahold of the full results of the survey so I will provide some more analysis on this on my blog - including on specific issues and Wisconsin-focused items).  But the takeaway is that the sample size to get a candidate preference read on Wisconsin voters for the Democratic candidates slimmed down to less than 200 voters.

Not exactly a really good sample size, not a good sampling of likely voters, and not a good sample of likely Democratic primary voters.  But at least we're getting a poll.  There's a decent chance that I might be able to push someone doing polling in my state to do a survey that reflects a methodology more in-tune with the kind of things we've talked about here at MyDD.

So I ask you, the assembled members of those MyDDers that read the diaries...

What questions would you want to see on a state-based poll?  What kind of sampling would you want to see?  What other aspects of surveying should a pollster keep in mind?  Comment on this and I'll compile it and see if we can get a good state-based poll done here.

Tags: polling, Presidential Race, states, Wisconsin (all tags)



Re: Design Your Ideal State-based Polling Survey

I'll get things started off myself...


- Get ahold of the easily available data on those who voted in the Democratic primary in 2006 and only poll them as likely Democratic primary voters for 2008.  This year, we had an incumbent governor and senator at the top of the ticket.  The former did not face a challenge (and interestingly, has moved back to his progressive roots in the absence of a challenge) and the latter faced a nominal challenge from a joker.  But the incumbent Democratic attorney general (the amazing Peg Lautenschlager) was challenged by the Dane County executive in a hotly contested race.  The latter won because of large turnout outstate from Madison and Milwaukee.  There were enough Democratic primary voters in 2006 to provide a large pool of people that can yield a decent sample.


- Ask about name recognition of the top tier candidates (exclude Al Gore, please) first, then favorability ratings for those where there is actual name ID, and then finally ask about preference of candidates.  That way, Hillary Clinton, Barack Obama, and John Edwards do not get skewed results because of the very high name ID of Clinton and lesser name ID of Edwards and Obama comparatively.  

Then, we can also check the crosstab of Clinton's favorability ratings with the actual preference - to see if there is an "inevitability dynamic" at play.  That is to say that even with 99% name ID, she could have 50-50 positives, but still have one-third of the vote, meaning that some are choosing her as their preference merely because she seems like the presumed nominee.

What else do you guys have?

by Peter from WI 2007-04-27 11:21AM | 0 recs
Crafting an Ideal poll

Large sample size so sub-groups and sub-sub-groups are still statisically relevant. Only poll those that are actually running.

Breakdowns based on:

region of state
rural, urban, suburban, exburbs
age, male, female, race, income, education
union member in household
level of attention paying to race
falling behind economically, struggling to stay put, getting ahead
don't push leaners/unsure too hard to decide
issues most important to voter, who they support and their second choice
really dig into leaners/unsure to see what issues are most important to them and get their full demographic breakdown as above.
really dig into supports to see how strong thier support is with break down for issues important to each within each level of support (strong, moderate, weak), level of attention paid (a lot, moderate, not much), all the other demographics above.

Ideally I want to be able to see how many male, rural, union, unsure or weak supporter, whose major issue his job creation, who feels they are slipping behind or struggling there are. And other such deeply defined sub-sub-sub-groups. It would be fantastic to have that sort of information, with such a poll you could really get the lay of the land in states and nationally.

The inital required sample size to keep the sub-sub-groups statisically relevant and the complexity involved would probably make such a poll either very expensive or very difficult, or both.

But you asked me to design what I wanted :-P

by Quinton 2007-04-27 02:41PM | 0 recs
Re: Crafting an Ideal poll

I guess you have to be careful what you ask for...

I think a couple things you point out are very relevant.

First, geographic location.  In Wisconsin, there are a few centers of population where there are lots of Democratic votes - like in most states.  Here, they include:

Milwaukee (city)
Madison area
Milwaukee (inner suburbs)
La Crosse area
Eau Claire
Stevens Point/Wausau/Wisconsin Rapids/Marshfield
Green Bay
Fox Cities + Oshkosh

They are not monolithically Democratic there, with the notable exceptions of Milwaukee (city) and Madison, but they are mostly Democratic there.  But the rest of the state, which is largely small communities and highly-rural areas will be more Republican in the general elections.  However, in primaries (especially since we are an open primary state), roughly 80% of the Democratic vote comes from the Madison area, Milwaukee (both city and inner suburban), La Crosse, the Fox Valley-Green Bay corridor, and the "Northwoods."  So sub-sampling for rural, suburban and urban will be instructive.  

Wisconsin is pretty white-bread.  There is a lower than national average minority population - and the overwhelming bulk of the black vote is in Milwaukee (city).  But there is an increasing Hispanic population and a small but burgeoning Hmong vote as well.  Unlike in neighboring Minnesota, the Hmong population is not politically active.  The black vote in Milwaukee is fairly well-organized, but nearly to the extent of other Democratic areas (that will soon change as there are plans within the party proper in the works to re-organize Milwaukee).  So for the time being, I don't think that there will be much to see in racial sub-groups.

Wisconsin is a fairly unionized state.  AFSCME, which was started here and has its national headquarters in Madison (I ran part of the Democratic coordinated campaign out of there this past fall), is very strong.  In all 72 counties of the state they are pretty strong, but again, much moreso in Madison (where state government means lots of members) and Milwaukee.  And the retirees group here is potent as well.  Milwaukee has a growing SEIU membership and the Fox Valley, like similarly-sized and economically-structured communities around the state has always had a good-sized labor-involved population.  I think union households would be a very interesting sub-group in Wisconsin.  

As fascinating as it would be, cross-tabbing preference for candidate with issue concerns would be very hard to do at this point.  My sense of the state though is that it is similar to the rest of the country.  Iraq is #1 easily, then healthcare is easily #2, then the economy at-large.  The rest is just details.

Let me add another wrinkle that is Wisconsin-specific (very important notions in state-based polls, especially as per Democratic ID).  Our governor, who was re-elected on the strength of progressives and establishment-types in the party coming together is enjoying something of a renaissance of public support.  A bold healthcare agenda is driving this, as he takes a big lead on pushing our state to continue to be a national leader in healthcare.  Likewise, he is known as an education-advocate, in all aspects from pre-K on up to higher ed and technical colleges.  Were I doing an absolute ideal poll, I would get a 1-10 approval rating for Jim Doyle from likely Democatic voters, and then ask for presidential choice as well.  

Doyle is not likely to come out for anyone in particular right now (though much of his network is committed - actually outside of where I think he would align), as it might be disastrous to the liberal/Democratic coalition he has on-board with him at present.  But where his strong and relatively-strong supporters break will be instructive on where the state goes in 2008 in the primary.  As an aside to this, Doyle's continued health as a relatively popular incumbent Democratic governor and his prospective re-election in 2010 without a primary challenger (which almost came about this past year from the progressives) will hinge on whether or not he works hard for the Democratic nominee in 2008.  

Anyone else have other thoughts?

by Peter from WI 2007-04-28 09:19PM | 0 recs
I'm just taking notes


by IVR Polls 2007-04-27 03:40PM | 0 recs


Advertise Blogads