## How did white people vote?

I posted the maps at 538.

And here’s what we did:

For the left column of maps, we fit a multilevel logistic regression fit to Pew Research pre-election polls, with intercepts and slopes for individual income (categorized as -2, -1, 0, 1, 2) with indicators for individual income, state income, and region of the country as predictors. Here’s what the R code looks like:

M1a <- glmer (rvote ~ z.inc2*z.state.income.full + (1|inc2) + (1 + z.inc2|region.full) + (1 + z.inc2 | stnum), family=binomial(link="logit")) Here, inc2 are the income categories, z.inc2 is income on the continuous -2 to 2 scale (rescaled to have center 0 and standard deviation 1/2), stnum are state indicators, region.full are region indicators (expanded from a state-level vector to a respondent-level vector), and z.state.income.full is state income (rescaled to have center 0 and standard deviation 1/2, and expanded from a state-level vector to a respondent-level vector). Withing each state, we then did the following: (a) We used the estimates from the above model to compute McCain's and Obama's estimated shares of the two-party vote in each of the five income categories; (b) We then summed these, weighting by the estimated proportion (from the survey) of each income category to get total McCain and Obama proportions; (c) We compared this to the election outcome in the state, computed the difference, and shifted the estimates within each income category by that amount so that our estimates are consistent with the actual votes. For the "whites" maps, we fit the model just to the survey respondents who described themselves as white and non-Hispanic, doing all the steps above, except using the shifts for each state computed using all the data (since our election data is only in aggregate, not by ethnicity). P.S. Alaska and Hawaii are not included in these surveys. P.P.S. More discussion here.

1. Shylock says:

Whence comes this trend of systematically excluding Alaska and Hawaii from such maps? Are they not States? If you prick them, do they not bleed?

2. Evan says:

And more to the point, do they not bleed red or blue?

3. Daniel Habtemariam says:

Was nonresponse a problem?

I only ask because, in my experience, race and ethnicity are notoriously troublesome variables…what with all the "other" and "more than 1 race" categories out there.

We've had to do tricky things with our denominators in the past in order to to simplify/accommodate things…I wonder if you and your coauthors had the same problem.

And call me crazy, but I love standard errors.

I know there's no easy to way to display them on your maps, but I feel naked and lost staring at results without them.

I can feel the Type I error staring back at me.

4. Eric Tassone says:

Nice maps. One possible suggestion that I forgot to mention to you in person during your Google visit was this: Have you considered placing the corresponding number of electoral college votes for each map, as if that map were the only one determining the allocation of the electoral college votes? (I suppose Hawaii and Alaska could simply be allocated in all cases (even if not technically true) to the Dems and GOP respectively, just to make it add up to 538.) Might this be yet another interesting bit of info to add to the titles or next to each map or something?

5. mclever says:

I believe Hawaii and Alaska are often excluded due to lack of sufficient information. Polling is notoriously difficult and unreliable in Alaska, for example.

6. The Groundhog says:

Do you have a sum map of the white vote? In other words, a map that shows how ALL white non-Hispanics voted, (regardless of income level)? P.S. Interesting data on Connecticut and New Hampshire :-)

7. The Groundhog says:

Q: Nate Silver suggests–rather infers–that, where whites live with blacks outside the South, Obama's white vote goes up–but the inverse is true in the South (more black people, more white votes for McCain). Like much of Nate's analyses, this may be true in the aggregate–but the electoral college is not an aggregate. Are there enough black people in Maine, New Hampshire, Vermont, Iowa and Oregon to make this relevant? (Granted, that's only 25 EVs).

8. The Groundhog says:

I still maintain that the Obama coalition is: northern educated whites, all nonwhites, college students and enough union voters to put him at 52.87%.

The question is, how much of Obama's 4.6% improvement over Kerry was due to:

–bringing in new (young and/or ethnic) voters;
–Bush voters changing their minds;
–Bush voters not voting in 2008.

Any way to break this down? I know Nate Sivler points out that Obama improved among blacks by 6% (88 to 94 %). How about among whites? How much improvement?

Thanks.

P.S.

Nate is grasping at straws about Appalachia. Yes, it did not turn en masse towards McCain–but unlike Nate (and my fellow Dems), it did not drink the Obama koolaid, either (his metaphor).

Think of it: A Democrat wins soundly, yet gets thumped in Arkansas AND West Virginia? Were all these erstwhile CLinton voters only voting for CLinton because he was their homeboy?

9. Antonio says:

Who is white in US?

10. Chris says:

@ The Groundhog:

Think of it: A Democrat wins soundly, yet gets thumped in Arkansas AND West Virginia? Were all these erstwhile CLinton voters only voting for CLinton because he was their homeboy?

…actually, yes.

Clinton's a Southern Democrat and would likely be considered a Blue Dog in today's terminology. H. W., I believe, hails from Massachusetts (unlike his son, who pretends he's from Texas). In fact, one of the biggest critiques I've heard about Clinton was that he was pretty much Republican-Lite.

Obama, on the other hand, is a VERY urban president, even excepting his ethnicity. His appeal – his airy rhetoric, then cool-headed pragmatism – was really pitched more to the coasts and to the cities than it was to rural areas…like much of Arkansas and West Virginia.

Further, the South as a whole had been sliding towards the Republicans since the Civil Rights Act. I believe that LBJ said, at the time he signed it, that he'd lose the South for the Dems for a generation.

And just because a state votes for one party at the presidential level doesn't mean that they'll follow suit at lower levels. My own state has a House delegation that's about 60/40 Dem/Repub. (8 D's, 5 R's.) I'm from New Jersey, which has been pretty reliably blue since 1992 for the Presidency and has had two Democratic senators since 1978, I think.

Delaware's at-large rep, Mike Castle, is a Republican. And, of course, Maine's senators are both Republicans, and Maine is as blue for presidential elections as the rest of New England.

At the state level: A lot of blue states have – or have recently had – Republican governors. Off the top of my head, I can name Connecticut and Hawaii, and – of course – California with sitting R governors. My own state will almost certainly have a Republican as a governor by next year (I can't see Corzine getting re-elected since no one likes him), and George Pataki was governor of New York for 12 years.

Conversely, Louisiana, which is considered blood red for the most part, is home to quite a few federal Democrats, and Mary Landrieu just won re-election there. Kansas has had a hugely popular Democratic governor. So on and so forth.

tl;dr – Party affiliation and party voting is not transitive between candidates. States may be, as a whole, liberal, conservative, or centrist (due to either a truly moderate state ideology or extremes), but due to state parties differing and candidates differing, they may do things that look pretty weird from a pure party-line perspective.

11. Kevin says:

Two points:

1. Can you do it with shades of red/blue? Binary is only so helpful for understanding the state of play at a deep level.

2. What's going on in with NH and RI whites? You see a reversal of economic roles? Is this simply a sample size issue? Are similar issues in play across New England (but only not flipping the binary switch)? Degree of change might help here too.

12. BorisG says:

Since this is all compiled from pre-election Pew polling data, isn't it pretty much useless? I mean, why not use exit polling data from the *actual election results* than pre-election telephone polling?

I'd have to say that this is not informative at all. The methodology chosen is baffling.

13. Andrew Gelman says:

Thanks for all the comments. In response:

– Alaska and Hawaii were not included in the surveys.

– Race/ethnicity are defined by what people say they are. Only 6% of the respondents declined to answer this one.

– I agree that the maps oversimplify. I played around with gradual color schemes but then it seemed like the maps were losing some of their impact.

– Yes, I could sum over the whole country and make maps similar to what I did for 2000 and 2004 here.

– No, there aren't a lot of black people in Maine, New Hampshire, Vermont, Iowa, and Oregon.

– I think most of Obama's improvement over Kerry came from a national swing: just about all groups moved toward him. See here.

– No need to over-analyze New Hampshire. We estimate McCain as getting 50.2% of the two-party vote among whites in the lowest income category, 50.3% of the next income category, 50.1%, 50.0%, and 49.8%. The colors just happen to change at 50.0%.

– I don't have individual-level exit poll data so I can't make these graphs with exit polls. In any case, exit polls have big problems too. We do our best with what information we have available. And if you really think these maps are "pretty much useless" . . . just remember, you didn't pay anything for them!

14. The Groundhog says:

Thanks, Chris. Let me elaborate: I am well aware of LBJ's comment (which I believe he said to BIll Moyers). He was, of course, correct (even before that, the whole reason for JFKs visit to Texas is because they saw the future…)

1) However, West Virginia and Arkansas still present interesting cases. The political scientists Galston and Karmenck discuss what they call "the great sorting out"–the phnenomenon of northern Rockefeller Republicans trending blue, as well as southern working class whites trending red (I am aware of the data that it is rich, churchgoing southern whites who vote the reddest, but the GOP has made inroads among all southern whites). Think of the map of 1976: not like today, eh? However, West Virginia voted for Humphrey and Dukakis, and Arkansas was long considered the "least Republican" of the southern states (defined here as the old Confederacy). Maybe these trends were merely delayed or disguised. To use an another example, I remember that, for years, Minnesota was considred by many to be the most Democratic state. Um, no. It only looked that way, because there was always a Minnesotan on the ballot! (1964, 68, 76, 80, 84).

It is absolutely true that a historical trend does not equal inevitability (it is like this nonsense that "no foreign army has ever succeeded in Afghanistan"–well, the same thing could have been said of Japan–until August 2,1945…)

My hypothesis is this: Nte Silver is right when he says that there was no "Bradley effect"–or a white backlash in this election–IN THE AGGREGATE. But Appalachia did not trend our way. To put it another way, the Arkansas and West Virginia results are what the Carolinas and Georgia would look like sans black people. I only ask this because I am curious as to where the best regions are to broaden our calition–the upper South, or out West? I know what Tom Schaller would say, but it is a fair question.

2) Similarly, is New Jersey really BLUE, or just not red? In other words, if the GOP ever nominated a sane Republican, oh let's say A Weld/Whitman ticket (I know, fat chance), would it not beat Sanders/Frank? THis is only an academic question, because I do not see Obama losing NJ to Palin, Romney, Jindal, Huckabee, etc.) I live in one the bluest cities in the country–and we elected Giuliani twice…

15. The Groundhog says:

Addendum: I did not think Obama would win Arkansas or West Virginia (my only missed calls were Indiana and North Carolina), but I thought they would be closer, similar to Georgia (46.5–52). And this is what many polls showed, as well.

16. William Ockham says:

Here's my advice if you want to expand the Democratic coalition (in Presidential elections).

1. Concentrate on registering and mobilizing young (age 18-29) Hispanics. This group went nearly 80-20 for Obama. They are the key to moving Colorado, Nevada, and New Mexico from battleground to solid Dem. Eventually (12-16 years from now), they'll make Texas a battleground state. I would not want to be a Republican when that happens.

2. Concentrate on 'light Red' states where Obama leads among youth voters and McCain leads among the oldest voters (e.g. Missouri and Montana are best candidates). Those are the states that the millenials are turning blue.

3. In general, the only Southern states within reach are those on the Atlantic coast. I would expect South Carolina to be closer than Georgia in 2012.