Is Obama Really Going to Win? (Part II - Big Data Baseline)
In my first post, I asked what baseline we’re working with this year . . . is it President Obama’s election to lose, or Governor Romney’s?
Instead of starting with current polls that try to guess at who in the sample are going to show up in November (some “likely” voters stay home, some “unlikely” vote), I think it might be more clarifying to begin with who we know actually showed up in 2008 and who likely showed up in 2010, then see what 2012 looks like if we put those two together.
There is a spectacular political science resource called the Cooperative Congressional Election Study (CCES) with more than 30,000 respondents from the 2006, 2008, and 2010 elections. The 2006 and 2008 data have been validated against voter file information, which means we don’t have to rely on self-reported voting (people lie).
The 2008 data match the actual vote totals for the election almost perfectly, and for about 19,000 people, we know that they really voted and that they cast their ballots for the person they said they voted for.
This is addictive data, overdose-grade for a political junky. It is Far more detailed, accurate, and useful than the exit-poll data that the mainstream media’s drowning in. And although the 2010 data aren’t vote-validated, pegging it to those who said they voted in 2008 and 2010 gives a close match to the known 2008 vote.
I think these two elections are the best place to start when we ask what 2012 is likely to look like ... after all, Obama was on the ticket (literally or figuratively) in both years, and it's likely that 2008 and 2010 are the high/low points for Democrats and Republicans.
In 2008, the electorate according to the CCES data was 77 percent white, 11 percent black, and 7 percent Hispanic (exit polls put the white share around 74 percent, but other data sources are closer to CCES). In 2010, the electorate looks like it was about 80 percent white, 10 percent black, and 4 percent Hispanic.
And, not surprisingly, more white 2008 Obama voters “defected” and voted Republican in their House election than the reverse for McCain voters. The House Republican share of the vote increased about 4 points over McCain when you include Obama “defectors” and (net) new Republican turnout.
Let’s give Romney that 4-point bump among white voters in 2012 (concentrating on the white vote just makes things easier and clearer, since the Hispanic vote is relatively small and Democratic, and the black vote, while larger, went so overwhelmingly for Obama).
The minority vote was way down in 2010, but President Obama wasn’t on the ticket . . . I think it’s reasonable to assume that most, but not all, minority voters from 2008 make it to the polls in 2012. Let’s add back in almost all (75 percent) of those Obama voters who stayed home in 2010.
So, what we have is an electorate that looks like a blend of 2008 and 2010, with turnout closer to 2008, and the net partisan shift in the white vote looking like 2010. I’m assuming that the minority vote stays the same as 2008 (for simplicity -- although I think it’s quite possible these voters will shift somewhat Republican).
This is my best estimate of the baseline for 2012 . . . Romney ever so slightly ahead with about 50.4 percent of the two-party vote. The starting point for this race looks exceedingly tight. Let’s hope voters are ready to shift in the right direction.
- Comment (15)
- · Quote
- · UnfollowFollow (7)
Comments:
Sep '12
Re: Is Obama Really Going to Win? (Part II - Big Data Baseline)
This data doesn't take the shift in partisan self-identification and the differential in intensity this time. The conclusion is headed in the right direction compared to the MSM polls, but still far under reports Romney's potential. Still good news!
Jun '10
Re: Is Obama Really Going to Win? (Part II - Big Data Baseline)
Blacks and Hispanics are defecting or choosing to stay home over Obama's stand on gay marriage. Both groups lean conservative on this particular issue. I wouldn't assume Obama gets the same turnout in 2012 that he did in 2008.
Jun '11
Re: Is Obama Really Going to Win? (Part II - Big Data Baseline)
No numbers? White Vote = 77%-80%...78.5%? If your assumptions are a less than 23% Non-White Vote, then you are going against 97% of all the segmentation models utilized by the big pollsters. You might be skewing farther white than Rassmussen.
Thank you for sharing your expertise with us!
Jul '11
Re: Is Obama Really Going to Win? (Part II - Big Data Baseline)
I still believe in America's ability to figure out when someone just plain sucks no matter how much they lie to the public about their illusionary awesomeness.
May '11
Re: Is Obama Really Going to Win? (Part II - Big Data Baseline)
Very interesting. Can you break the data down to the key swing states, say Florida, Ohio and Virginia, and come up with a prediction for just those states ?
Dec '11
Re: Is Obama Really Going to Win? (Part II - Big Data Baseline)
Strategoist
This will be, if it is, due to the "none of the above" vote coming out saying "enough." The MSM will paint this in horrible colors and a lot of time remains between Nov 6 and late January in which Obama, with the MSM blessing, can salt the wells. I hope there is a plan to counteract this.
Nov '11
Re: Is Obama Really Going to Win? (Part II - Big Data Baseline)
Adam Schaeffer: . . .[2008 and 2010] elections are the best place to start when we ask what 2012 is likely to look like . . . it's likely that 2008 and 2010 are the high/low points for Democrats and Republicans. . . .
Let’s give Romney that 4-point bump among white voters in 2012 . . .
. . . it’s reasonable to assume that most, but not all, minority voters from 2008 make it to the polls in 2012. Let’s add back in almost all (75 percent) of those Obama voters who stayed home in 2010.
. . . we have is an electorate that looks like a blend of 2008 and 2010, with turnout closer to 2008, and the net partisan shift in the white vote looking like 2010. . . .
. . . best estimate of the baseline for 2012 . . . Romney ever so slightly ahead with about 50.4 percent . . .
I agree that this race is more like 2008 than 2010, simply because 2010 was midterm.
You need to "add back" more than 75% of 2010 minority no-shows.
Similarly, the 4% white voter bump you extrapolate for Romney from 2010 does not reflect Obama white no-shows in 2010 who will reappear in 2012.
Re: Is Obama Really Going to Win? (Part II - Big Data Baseline)
Strategoist
No numbers? White Vote = 77%-80%...78.5%? If your assumptions are a less than 23% Non-White Vote, then you are going against 97% of all the segmentation models utilized by the big pollsters. You might be skewing farther white than Rassmussen.
Thank you for sharing your expertise with us! · 1 hour ago
Strategoist . . . thanks, I overlooked the final numbers . . . I have 78 percent white, 10.5 percent black and just over 6 percent Hispanic. And that is a very high for the white share compared to what polling outfits and apparently both campaigns assume. It seems the media at least is working with their exit poll data from 2008, with just 74 percent white. All I can say is that I trust the CCES data more, and it's validated against voter files.
Oct '10
Re: Is Obama Really Going to Win? (Part II - Big Data Baseline)
Without an underlying model of voter behaviour, isn't this just magical data-trawling?
Re: Is Obama Really Going to Win? (Part II - Big Data Baseline)
Astonishing
I agree that this race is more like 2008 than 2010, simply because 2010 was midterm.
You need to "add back" more than 75% of 2010 minority no-shows.
Similarly, the 4% white voter bump you extrapolate for Romney from 2010 does not reflect Obama white no-shows in 2010 who will reappear in 2012. · 2 minutes ago
Astonishing . . . completely reasonable points. Again, these are assumptions. On the minority turnout, I think it's reasonable to think one out of four 2010 dropouts stays home. After all, this was an historic turnout in general and for minority voters in particular. That means even more marginal voters in the pool than normal. At the very least, the excitement has faded tremendously. Add to that very high unemployment, potential wedge issues like gay marriage, and I think getting 3 out of 4 of those marginal voters to the polls again is a fairly tall order.
On the 2010 no-shows, fewer McCain than Obama voters dropped out in '10, I didn't count them in the vote share shift for 2012. I only counted net "defectors" and new Republican voters (turnout will probably be even higher).
May '10
Re: Is Obama Really Going to Win? (Part II - Big Data Baseline)
I think the most interesting possible dynamic is the college-or close age crowd. They turned out in large numbers in 2008, most in blue states (or purple). My sense is that many of those will take a pass.
This really explains the media poll-lying. If the trend were basically "Mitt by a bit" or more, most of the youth and minority tend to say, "fuggetaboutit." They need their 2008 numbers to have a chance.
Edited to add: The voter fraud bills (ID laws) are key. I know that many of the tens of thousands of college-age youth in Wisconsin drove I-94 to vote in Madison AND Milwaukee because Wisconsin had "same day registration."
Edited on September 17, 2012 at 10:15pmFeb '11
Re: Is Obama Really Going to Win? (Part II - Big Data Baseline)
DaTechGuy has a positive analysis based on party ID. He backs it up with specifics. I'd like to see how this looks in the battleground states.
I still say, I'd like to see a campaign walk and chew gum at the same time. The new ads from Romney today suggest that his campaign is doing that.
Edited on September 17, 2012 at 10:27pmJun '11
Re: Is Obama Really Going to Win? (Part II - Big Data Baseline)
Dr. Schaeffer,
Thank you for responding. :) I'm curious if you've encountered intense pushback on your segmentation model. I can imagine howls of recrimination concerning a 78/10.5/6 split across white/AA/hispanic vote nationally. That means every single national poll being regularly shared via the MSM are dead wrong, and so wrong as to be a kind of fraud.
I am personally skeptical of black voters defecting to republicans. Romney is polling somewhere between 7% and 0% among that segment. How do you come to your conclusion that there will be defections? What percentage of black voters do you estimate will defect to the Republicans?
Edited on September 17, 2012 at 10:55pmRe: Is Obama Really Going to Win? (Part II - Big Data Baseline)
Genferei . . . thanks for the comment . . . this isn't a sophisticated model of voter behavior or turnout, it's a baseline derived from a large amount of solid data on who voted in '08 and (likely) '10. The vast majority of voters who voted in 2008 will vote in 2012. And those who voted in 2008 and 2010 are very, very likely to vote in 2012 (consistent past turnout is a very good predictor of future turnout). Nothing magical here, just a handful of assumptions and data on 2008 and 2010.
Re: Is Obama Really Going to Win? (Part II - Big Data Baseline)
No Caesar, thanks for the link, and very interesting stuff on partyID . . . makes me more hopeful.
Strategoist, my pleasure (and no need for Dr. unless you just like to address folks formally). I've not had any pushback on these numbers because this is something I've looked at for our internal use, but I doubt the MSM or most pollsters would take note in any case . . . they have their methods and standard procedure, and their own datasets.
On the minority vote, I didn't mean to suggest I thought there would be a big shift. But I do think 2008 is probably an historic high-water mark, an extreme turnout and margin for the minority Dem vote, which means there is more room (and a greater chance) to decline slightly rather than increase.