Election 2008: what really happened

After a quick look at the election results and exit polls (from www.cnn.com), some thoughts:

1. The election was pretty close. Obama won by about 5% of the vote, consistent with the latest polls and consistent with his forecast vote based on forecasts based on the economy.

2. As with previous Republican candidates, McCain did better among the rich than the poor:


But the pattern has changed among the highest-income categories:


3. The gap between young and old has increased–a lot:


But there was no massive turnout among young voters. According to the exit polls, 18% of the voters this time were under 30, as compared to 17% of voters in 2004. (By comparison, 22% of voting-age Americans are under 30.)

4. By ethnicity: Barack Obama won 96% of African Americans, 68% of Latinos, 64% of Asians, and 44% of whites. In 2004, Kerry won 89% of African Americans, 55% of Latinos, 56% of Asians, and 41% of whites. So Obama gained the most among ethnic minorities.

5. The red/blue map was not redrawn; it was more of a national partisan swing. See this state-by-state scatterplot of Obama vote in 2008 vs. Kerry vote in 2004:


The standard deviation of the state swings (excluding D.C. and the unusual case of Hawaii) was 3.3%. That is, after accounting for the national swing in Obama’s favor, most of the states were within 3% of where they were, compared to their relative positions in 2004.

By comparison, here’s the 2000/2004 graph:


The standard deviation of these state swings was 2.4%. This was even less variation–2004 was basically a replay of 2000–still, the relative state swings of 3.3% in 2008 were not large by historical standards.

Again, Obama didn’t redraw the map; he shifted the map over in his favor. (Or, to put it more precisely, the economy shifted the map over in the Democrats’ favor and Obama took advantage of this.)

Here’s the map showing where Obama and McCain did better or worse than expected based on 2004:


6. Finally, how did the pre-election polls do? Unsurprisingly, they pretty much nailed the national vote. And what about the relative positions of the states? The pre-election polls did well there too, at least using Nate Silver’s aggregations. Here’s the scatterplot:


Pretty damn good. The standard deviation of the discrepancies, again excluding D.C. and Hawaii, is 2.5%, which is a big improvement on the 3.3% using Kerry04 alone.

I see some systematic patterns: Obama underperformed where the polls had him way down, and he outperformed where the polls had him up. We should go back and look at these patterns from earlier elections and see if this is consistent. If so, it suggests a way to improve forecasts for next time.

P.S. Age graph fixed from first posting; thanks to Andy Guess for pointing out the error.

100 thoughts on “Election 2008: what really happened

  1. Just out of curiosity did you use the “R” programming language for that plot?

    If so, I covet your ability to generate character labels on the data points.

  2. Pingback: A Bit Of Analysis at futuraprime:blog

  3. How did you get all that so quickly? One more question.

    “According to the exit polls, 18% of the voters this time were under 30, as compared to 17% of voters in 2004. (By comparison, 22% of voting-age Americans are under 30.)”

    In 2004 what percentage of voting-age Americans were under 30? Perhaps a percent or two less?

  4. Pingback: What just happened. « The Edge of the American West

  5. Lots of interesting information here — the most surprising to me is the youth vote information (both in terms of party shift and in terms of relative turnout).

    I hope by the time you read this you’ve had a nice long nap!

  6. I can’t help but noticing that (a) Obama underperformed relative to Kerry the most in Louisiana and Arkansas, and (b) polls underpredicted Obama’s actual vote the most in (what from the graph seems to be) Oklahoma, Arkansas, Kentucky, Louisiana, Wyoming and Alaska. So the main evidence of the so called “Bradley Effect” seems to be coming from in states which our Bayesian priors suggest hold more white racist voters.

  7. A very nice, succinct analysis written in the wee hours of the morning. A question: What individual-level factors changed the relationship between income and the vote? Was it wealthier voters frightened by a possible presidency by Governor Palin? Is it a simple relationship between education and voting?

    Thanks, Fabrice Lehoucq
    Department of Political Science
    Greensboro, NC

  8. (1) It would be interesting to see fractional Obama support among whites by state.

    (2) You comment in (6) about Obama doing better where he was expected to do well and more poorly in where he was expected to do poorly. Couldn’t this reflect a kind of Bradley Effect in which pollees in relatively racist areas are reluctant to admit they won’t vote for Obama?

  9. Thank you so much for an excellent analysis – sophisticated yet understandable and not overly technical – just the right amount of “jargon” to “layman’s language.”


  10. You say it’s the economy but then you point out that minorities and age were important factors. Granted that all these factors (economy, age, race) are correlated but, could you just conclude equally from what you explain that a shift in voting patterns by age and race nationwide made the change over ’04? Maybe I’m wrong but I don’t see how the data presented disproves this.
    The level of Democratic registration can be a symptom not a cause for this shift.

  11. Pingback: Boîte noire – Not Fiction Anymore

  12. Robert: In R, use the text() function.

    Tom: I got the data from the CNN website and just typed in the numbers myself. I actually used 2006 census numbers. I expect 2004 and 2008 weren’t much different.

    Bill R.: I was actually frustrated with how little information they were putting on TV. I’ll have to save that rant for another post.

    Jhu, Martin: Maybe yes on that.

    Fabrice: Good question. I wonder what the income and voting graph would’ve looked like if it had been Clinton vs. Romney.

    Bill B.: Thanks for the comment. I put in some jargon to maintain my street-cred.

    RR: Race was part of the story, to be sure. But, according to the exit polls, Obama outperformed Kerry by 3% even among whites. A 3% shift would’ve given Obama the White House. It would’ve been a narrower victory to be sure, but a win is a win.

  13. Was turnout much lower than expected? Using the CNN.com number of votes v. 2004 FEC data for Bush and Kerry, it seems that either turn-out was as high as expected or nth parties did really well (compared to 2004). My math shows Obama getting under a million votes more than Bush in 2004 but McCain getting over 3 million votes less than Kerry in 2004.

  14. Pingback: No Revolution, No Bipartisan Reconfiguration | The Global Sociology Blog

  15. Pingback: Income and voting: what if it’d been Clinton vs. Romney? | Red State, Blue State, Rich State, Poor State

  16. Pingback: Matthew Yglesias » A Nationwide Swing

  17. Jhu, I don’t see any Bradley effect here. Alaska and Wyoming are Western states with very few blacks, and no tradition of racially divisive politics. The Southern states you mention are outliers even in the South, and at any rate, standard Bradley effect theory predicts it will be found in the Northeast, where racism is veiled, more than in the South, where it is open.

  18. Does the Obama 08 residual (predicted from Kerry 04) correlate with the Kerry 04 residual (predicted from Gore 00)? If so, that suggests that the map is changing over time (in terms of the relative positions of states). Even if the changes only amount to a couple percentage points per election cycle (relative to the nationwide vote), they’ll accumulate over time.

  19. This makes sense to me, but, following up on jhu, I’d also like to suggest one tweak, based on an observation from here in Tennessee. If one were to draw a regression line through these data points, it would lie roughly parallel to, but to the left of, the line indicating equal Democratic percentages in 2004 and 2008. The vast majority of the states lie above the equal-percentage line, but a cluster of states on the lower end lie on or below it: OK, LA, AR, TN, WV, AK, and AZ. AK and AZ are self explanatory; the others form a cohesive cluster that, along with nearby AL, MS, and KY can be called the Central South. These states were basically immune to Gelman’s Obama shift, or in the case of LA and AR actually went the other direction. Katrina probably explains LA in part, but it appears to this Nashvillian [Nashville and Memphis being blue islands in a sea of red] that this is a zone in which cultural politics retained its salience [Republicans have taken over the TN General Assembly for the first time since Reconstruction]. If those who see the Republican Party as an increasingly southernized rump are correct, one might expect Republicanism to intensify in these places even as it weakens elsewhere. The differing behavior of these states from VA, NC, and FL [and even, on this evidence, SC and GA] suggests that Obama may have in fact succeeded in fracturing the South, but mainly by revealing a difference between southern states [mainly on the East Coast] that follow national shifts and those that don’t.

  20. Re: your point 6. How did the pre-election polls do?” ”

    I was led to believe that the polls may underestimate Obama’s performance for two reasons:

    1. Most polls excluded cell-phones and thus missed cell-phone only voters who heavily favored Obama. Adjusting for this effect across relevant polls would increase Obama’s expected performance by 2-4% (though average of polls would be affected less).

    2. An excellent GOTV “ground game” is supposedly capable of improving performance over polls by 2-3%. Obama reportedly had one of the best ground games in history, and McCain’s was believed to be terrible.

    If the above two points are accurate, shouldn’t we have expected Obama to reach (conservatively), say, 56% of the popular vote? If so, what kept him at 52%?

  21. A few thoughts on ethnic voting…

    1. As RR said, an ethnic voting story might generate similar effects to the economic voting story (since the two are probably heavily correlated will be hard to pull them apart).

    2. Ethnic voting (Hispanics becoming solidly Democratic in particular) would explain NV and NM outliers (and I suspect is part of the story in CO as well – 73% support for Obama by Hispanics in CO according to MSNBC exit polls)

    3. If we consider the electoral college, if NV (~55%), CO (~52.6% currently), and NM (~57%) are now “safe” Democratic states because of Hispanic support, then the Democratic party could win in 2012 (278 electoral votes) losing Indiana, Ohio, Florida, Virginia, and North Carolina (all marginal states less <52% this year)

    4. Therefore, as a policy matter, I suspect the Obama administration will be VERY sensitive to the needs and interests of Hispanics. If they can maintain CO, NV, and NM, they will be almost bullet-proof to a bad economy in 2012. Especially CO; CO looks like the marginal state.

    So I agree that economics explains why this was a very strong year (in the electoral college) for the Democrats, but I suspect that racial politics is going to have a lingering impact on both policy and long-run presidential politics.

  22. One note: Do we really care how the models perform in the 10 bluest and 10 reddest states? You could make an argument that we should throw those out in evaluating Silver’s state level predictions. First, they aren’t polled much so you know the predictions will be bad. Second, a prediction should be evaluated on the basis of it’s utility. Getting the close states right is a lot more important, imop.

  23. Silvers Aggregation was no doubt well within the electoral voted standard deviations consistent with the prior polling. Absolutely no surprises. I guess thats the buggiest surprise

  24. What my frank opinion? What a bunch of hooey.

    First of all, what is the difference between a “redrawn red/blue map” and “a national partisan swing”?

    Sounds like a distinction without a difference to me.

    In any case: Nevada, Colorado, New Mexico, Indiana, Ohio, Virginia, Florida and North Carolina all turned blue (plus Missouri is essentially tied). Obama massively outperformed Gore and Kerry in these states.

    But the red/blue map wasn’t “redrawn.” Oh. Okay. Rrright.

    You also say “the election was pretty close” at a 5% difference.

    First of all, 5% is *not* remotely close in this country’s elections. Maybe you missed the past couple. Those were close. Perhaps the workings of the Electoral College are a mystery to you? Any national margin above 2% results in huge disparities in the EC.

    Oh, and anyway: The margin was 6%. Not 5%.

    Going to go yell at the person who recommended this site to me. Later.

  25. When you compare Nate Silver and Intrade in the individual state probability markets it makes a difference whether you use absolute or squared errors. With absolute errors, he kills Intrade even after commissions are accounted for — although the error is comparable in the markets that were in the 30-70% band. With squared errors, Silver has a much smaller edge. Squaring the errors effectively removes the markets’ penalty for being less confident near extreme prices. I wasn’t sure if that was justified, but mpowell’s comment is interesting on that point.

  26. I don’t quite understand the contention that the youth vote once again failed to materialize. No, one percentage point increase from ’04 to ’08 doesn’t look very big, but that’s in the context of an overall turnout boost of, what, 18%? So it looks like ~20% more under-30s voted in 2008 than in 2004 – that’s not a small increase.

    IOW, they managed to increase their participation rate (albeit slightly) even against a background of generally increasing participation. I know it’s not inspiring like them becoming 22% (or more) of the electorate would be, but it’s not nothing.

  27. Re: turnout levels

    California still has an unknown number of ballots left to count, but it’s somewhere between 1.5 and 3 million ballots.

    Given the state’s 61 – 37% split, that will add between 1 million and 1.8 million votes to Obaman’s national count, plus 500,000 to 1.1 million to McCain’s count.

    Per comment #17, there’s your missing 2 million votes and then some. The remainder, plus the votes remaining to be counted in the other 49 states represent the increase in turnout over 2004.

  28. Pingback: US 2008 Presidential Election Breakdown | John Barrdear

  29. Andrew: Where did you get comparable age breaks on CNN?

    I have noticed CNN 2008 breaks down at 45-64 and then 65+, whereas CNN 2004 was broken down 45-59 and then 60+ (yet also giving 65+).

  30. Pingback: Three Links For The Election » The Ethereal Voice

  31. Pingback: Spread My Wealth!

  32. Pingback: Social vs Economic Issues: US vs NZ Elections « The visible hand in economics

  33. Pingback: The emerging Democrat majority « Amused Cynicism

  34. How else would you define “redrawing the red/blue map” other than “a nationwide partisan swing”?

    By your definition, Reagan didn’t redraw the national map, but if Mondale’s lone state in 1984 had been Alabama instead of Minnesota, he would have.

  35. Pingback: Will Wilkinson Spreads the Wealth | Contrarian Musings

  36. Pingback: Etl World News | The election: what really happened

  37. very good analysis. Using two colors in your map will make it easier to read (light to dark red for Mac, light to dark blue for Obama).

    If you don’t have much gis see geoda, brewer color, and QGIS for some free software and advice.

  38. Well done, sir. Thanks, and…

    Going out on a limb putting the national shift on the economy, isn’t it? That requires discounting, among other things: a complete retooling of Democratic electoral strategy (50 state), a large disparity in money, the biggest and best run ground game in decades, the most disciplined campaign in decades, the most transparently sleazy, cynical and erratic campaign I can remember, and a public increasingly fed up with endless war, torture, gulags, and the cost of health care. Finally, there is the dramatic shift in youth support that you so aptly illustrated. That’s a lot to discount.

    re the shift in minority support: Obama’s share among either white or A-A voters increased by 7-8% over Kerry, while support among other races increased by double digits. It would be interesting to see racial breakdown among first-time voters.


    A one percent difference between exit polls is probably inconclusive. But the poster’s point is that youth remains underrepresented (while the “older” demographic are overrepresented), thus underwhelming. On the other hand, I would not be surprised if youth were over-representing on the ground game (canvassing, registering, getting out the vote, etc.), and had the kind of impact exit polls don’t track.

  39. Pingback: 4 Lazy Cats’ Backyard : Political Laffer Curve

  40. Pingback: Gotten Chillier; some post election thoughts (Palin: Africa is a country?) « blueollie

  41. A couple of people here are really not clear on the concept, are they?

    “A national swing” = everybody moves a little one way, to about the same degree. This pushes some swing states over the line.

    “Redrawing the map” = some states move left, others move right. Or at the very least, some states move further or less far than others.

    What Andrew is saying is that no states changed “against the tide”, i.e. from D to R in a year when D won. Also, few states voted less for the Democrat in 2008 than did in 2004, and few voted substantially more or less than the other states. It was a small uniform movement, not a redistribution of strongholds.

  42. Pingback: Stones Cry Out - If they keep silent… » Things Heard: e39v4

  43. Pingback: More on the shift from Republican to Democrat | John Barrdear

  44. Pingback: Another interesting thing about young voters | BlogWonks

  45. Thanks for crunching the numbers.
    I was telling a co-worker on election day there were only four kinds of reasons anyone would vote for Obama. Looks like Obama leveraged the ‘youth’ vote’s Ignorance and Irrationality and the ‘extreme rich’ segment’s Evilness. The Stupidity factor is probably spread through all the demographics a little more evenly and will be harder to winnow out of the bulk.
    Thanks Again!

  46. Pingback: Real Time Economics : Secondary Sources: Response to Crisis Myths, Election, China Slump

  47. Pingback: Do the numbers « Zeno’s Parcel Service

  48. Pingback: the next battleground states « Save it for the blog

  49. Pingback: hughstimson.org » Blog Archive » Plus Ca Change

  50. Andrew Gelman sees a pattern in the accuracies of the pre-election pollsters: Obama underperformed poll-based expectations in states where the polls had him way down, and he outperformed where the polls had him up. To this it should be added that states where Obama was way down, or way up, weren’t polled as heavily as states where the contest was close. Thus, it’s nice to find the state polling average to be most accurate in the states that were most polled. I think the pollsters as a group are entitled to take a victory lap. It was not an easy election to poll. The exit poll conducted on election day is considerably less accurate than the polling averages of the days leading up to the election. It seems the exit poll is “a piece of trash” (to quote Karl Rove) because the outcome indicated by the straight-up exit poll is way off from the truth in a great many states. This in turn means that information based on the exit poll, such as Andrew Gelman’s information about youth turnout above, should be read with a lot of reserve, doubt, and circumspection.

  51. Pingback: Blue America « Upturned Earth || John Schwenkler

  52. Here’s Karl Rove today at online.wsj.com: “for the third election in a row the exit polls were trash. The raw numbers forecast an 18-point Obama win, news organizations who underwrote the poll arbitrarily dialed it down to a 10-point Obama edge, and the actual margin was six.”

  53. The rational independents and republicans made the difference for Obama. We are practical above all. When a party or candidate veers from reality, we switch. It is clear that ideological free trade, deregulation etc. are not working. Time for something new, we will see if Obama has the answer.

  54. Is the income in #2 charts household income or individual income? Charts such as these consistently omit that information. IMHO it is important to be explicit and not leave it for assumption.

  55. Although a completely red state this year, Obama gained 8% in Oklahoma County (most populous of the 77 counties) over Kerry in 2004. This is fascinating to me.

    Thanks for all the graphs; data organization is cool!

  56. Pingback: Rich States still Vote Democratic | Red State, Blue State, Rich State, Poor State

  57. I disagree with your assessment that the youth vote “didn’t turn out.” You say this based on the statistic that the youth vote was a similar _percentage_ of the total vote. In a year when the participation of all groups rise, how is the percentage supposed to change? Please use the statistic of absolute numbers – this many more hundred thousand or so. I will believe you then. There was too much excitement on my college campus for me to believe that the youth vote “didn’t turn out”.

  58. I’m confused about the map at the end of point 5. Kerry lost Ohio, but Obama won it. Wouldn’t that mean that Obama did well, not McCain? Or is that grey a “light” color? Not a very well designed graph, but very interesting overall.

  59. Pingback: More on the Youth Vote « Thumbspeak’s Blog

  60. Pingback: The Fire of a New Generation

  61. I would like to see a state-by-state turnout comparison from 2004-2008. In other words answer the question, “which locations have marginally higher turnouts”.

    I’d like to see a national district-by-district map of the country visualizing the same data down to the finest level of resolution possible. I’d be interested both in hard numbers (% of eligible voters in a locality) and relative numbers (change in % since some previous election).

  62. Andrew, quite a few states did move against the tide (LA, TN, some other states at the extreme South), and quite a few states had large swings – Indiana and North Carolina, for example. It was not as big a redrawing as, say, 1964. But there’s a trend here in which the Republican party becomes increasingly based in the South and minorities, which are rapidly growing, are becoming increasingly Democratic. This is an important trend that may make Republican the minority party for decades to come.

  63. Pingback: Pinko Magazine » Friday! Moving right along.

  64. Pingback: Links - November 7th 2008 « Curiously Persistent

  65. Pingback: Jacob Christensen › links for 2008-11-07

  66. Pingback: The Moderate Voice

  67. Pingback: The Culture War is over. The Liberals won. « Amused Cynicism

  68. Pingback: Fruits and Votes » Prof. Shugart's Blog » The swing

  69. Pingback: What does non-uniform partisan swing look like? | Red State, Blue State, Rich State, Poor State

  70. Pingback: Left Flank

  71. Pingback: What Really Happened « The State of the Union

  72. Andrew-

    Thanks for this analysis. You base your conclusion that there was “no massive turnout among young voters” on the fact that, as a proportion of all voters, young voters’ share increased by only one percentage point (from 17% to 18%) from 2004 to 2008.

    I’m wondering, though, if this is the correct measure. Might it have been the case that there was, in fact, a massive increase in youth turnout, but also a massive increase in turnout across the board so that young people’s relative vote share remained stable? I’m willing to bet that if you took the absolute difference between the raw number of young people who voted in 2008 and the raw number who voted in 2004 and divided the difference by the number of young people who voted in 2004, you would see an increase that could reasonably be termed “massive”. If other demographic segments exhibited similarly “massive” increases, though, young people’s share of total turnout would be stable.

    Just wondering.

    David Crow
    Survey Research Center
    UC Riverside

  73. The correlation of age to vote is striking, in virtually every state. Even in the red states, the youngest group of voters tended to come in for Obama. Generational change, or susceptibility to respond to the issues of the time?

    Also… are the youngest voters disproportionately non-white?

  74. To various commenters: Yes, approximately uniform is not the same as exactly uniform. See some later entries for examples.

    Fr: I’d post the R code but it’s just so ugly I’m embarrassed. I do have some slightly cleaner R code in a recent entry at my other blog.

    David: There’s still some debate on what the total turnout was in 2008.

    Jon: I want to get my hands on the raw exit poll and pre-election poll data to answer such questions.

  75. Pingback: I believe the children are our future « the nef triple crunch blog

  76. Andrew — I’ve read your wonderful code in the other post, that’s precisely why I want the code for these graphs too. I know too little about R and am eager to learn from these graphs!

    I’m especially interested in the code for the first graphs, so that I can build my own using the exit polls I have collected on my side (same as yours, except I did not find the same data on age groups: CNN exit polls for 2000, 2004, 2008 do not use the same intervals).

    Data here, with some French:
    PDF: phnk.com/files/Briatte_ObamaExitPolls2008_3.pdf
    XLS: phnk.com/files/Briatte_ObamaExitPolls2008_3.xls

  77. Pingback: The myth of poor Democratic performance in House races | Red State, Blue State, Rich State, Poor State

  78. I would like it if you would look at something other than annual salary to define “rich” v. “poor”. Rather if possible look at net worth.

    Someone who is earning a high income but is maxed out on credit cards may be saving less that someone with a smaller salary but higher savings ability. As a result that “high-income” earner is disproportionately effected but changes in the income taxes. Small increases in income taxes will push these high-earners/high-spenders into negative territory. Thus, these people are very sensitive to the anti-tax argument.

  79. Pingback: on the left side….. » Blog Archive » Obama & Reiche

  80. Pingback: Dataset of the Day: How Did the Rich Vote? | Off the Map - Official Blog of FortiusOne

  81. Pingback: Predicting the election outcome months ahead of time: discussion and link to revised paper with Kari Lock | Red State, Blue State, Rich State, Poor State

  82. Pingback: Election 2008: What Really Happened | The Morningside Post

  83. Pingback: Matthew Yglesias » Elections Getting Duller

  84. Pingback: Get Defunct Outta My Face [Dan Collins]

  85. Pingback: Yes, we can « alfanje

  86. Pingback: In Mala Fide | Inter-class conflict: the rich versus the super-rich

  87. Pingback: 20 October 2010 (pm) « blueollie

  88. Pingback: In Defense of a Man and a Movement « The Harvard Political Review

  89. Pingback: Social vs Economic Issues: US vs NZ Elections » TVHE

Comments are closed.