Princeton Election Consortium

Innovations in democracy since 2004

Outcome: Biden 306 EV (D+1.2% from toss-up), Senate 50 D (D+1.0%)
Nov 3 polls: Biden 342 EV (D+5.3%), Senate 50-55 D (D+3.9%), House control D+4.6%
Moneyball states: President AZ NE-2 NV, Senate MT ME AK, Legislatures KS TX NC

Massive early voting

October 31st, 2008, 6:36am by Sam Wang

Here’s a great resource – a compilation of early voting numbers. It’s amazing. Compared with the total number of people who voted in 2004, about half that number have already voted in seven states: CO, GA, NV, NM, NC, OR, and TN. Many more states are close behind. Hats off to Prof. Michael P. McDonald at George Mason University for assembling this information.

Among early voters, so far the fraction Democratic voters leads Republican voters, in some places by double digits (for instance, 22% in North Carolina and 28% in West Virginia). Don’t get too excited. These voters are probably counted in last-minute polls. Pollsters have been asking people if they have voted early, and if so for whom. It’s more likely that Obama voters are more enthusiastic, rather than there being some giant trove of them that polls are missing.

Tags: 2008 Election

35 Comments so far ↓

  • William

    Another thing is that the Obama campaign is really *pushing* Early Voting(at least in North Carolina) because Obama supporters are of the demographics least able to stand in long lines on Election Day.

  • Geoff Cohen

    Of course, we don’t know for whom these early votes are cast. In West Virginia, ballots cast by Democrats outnumber ballots cast by Republicans by 28%, but we don’t have direct evidence about Obama vs. McCain numbers there.

  • DR

    What impact does early voting have on exit polling? It seems to me that early voters will be left out of the sampling frame for the exit polls that the network news stations rely on for making election day predictions about specific states. If the early voters are more likely to vote for Obama, it could potentially bias estimates toward McCain, resulting in false state-level results from media outlets.

  • Paul

    DR: Good point. I believe that in some states at least, officials start counting early ballots before election day. Is that correct? If so, can the media use those numbers to adjust their polls?

    Or is this going to set the Republicans on fire with cries of voter fraud driven by bogus exit polls?

  • Jeff

    My understanding is that the exit-pollsters conduct phone surveys in the days before the election to obtain information on early voters – and adjust their sample/results accordingly.

    Also, I think that there is an important element being missed by people claiming that the early voting doesn’t matter if these are people who would vote anyway. If a large portion of Obama voters vote early, then in states where voter lists are provided to campaigns (which I think is many of them) they can remove early voters from their election-day contact lists thus freeing up resources to contact voters who might have been less likely to turn out. If this happens, it could further magnify Obama’s ground game advantage.

  • Nicholas J. Alcock

    “We do not know how individuals voted, we only know if a voter is registered Democrat or Republican. This may be an indicator of how they have or will vote, but voters do not always cast ballots for the same party with which they are registered.” Source:EVIC. Further, voters can only vote once so although votes in the bag are very useful the party behind can still mobilise its greater pool of non-voters on election day e.g. Iowa 2004. However, it is clear from several sources early voting will be higher in 2008 than 2004 as a proportion of total votes at a time of greater registration. There are indications Dems are voting early more than Reps. But, one caveat young voters in FL, a key Obama target group, lag behind.

  • Michael

    I have also read that at least some of the pollsters are doing exit polling at early voting locations, and will include this data on election night. It’s become quite popular to trash pollsters (at least the ones whose numbers we don’t like), but I think these folks are smart enough to realize that they can’t systematically exclude 15-40% of the voters from their exit sampling and expect to get reliable results.

  • David Scheinker

    A conservative, back of the envelope calculation about Georgia.
    If black voters give Obama 25% of Georgia’s total vote ( and McCain 2%) the remaining voters would have to break Red 48% to 25% just for McCain to tie. This is exceedingly unlikely.
    Most “likely voter” polls are severely underestimating the black voter turnout and even so and already have Georgia as a toss up.
    The black early voter turnout in Goergia is over 35% of total early turnout, even though less than 30% of the population is black. This demonstrates the enthusiasm of Democratic operatives on the ground, you can also ask Google to find individual stories. Since neither campaign is advertising in Goergia, Obama’s national advertising advantage is translating to a greater TV presence.
    If Obama’s campaign throws in some TV money (and dare I hope a visit) Georgia is going BLUE in a big way. You heard it here first, I hope.

  • Matt

    Re: Obama campaign pushing early voting

    They are doing that everywhere where it is an option and they are on the ground. The main reason has to do with banking votes. The more supporters they get to vote the smaller the universe of supporters they need to turn out on election day and the more focused the voter contacting becomes.

    Suppose you have 10,000 supporters with 100 volunteers to contact them. If 5,000 of them vote early, you have 50 voters per volunteer on election day as opposed to 100. It makes it that much easier to get the remaining voters to the polls and increases your turnout rate – which really is the goal.

    Every field organizer and state campaign manager has an exact number of voting supporters that the campaign thinks they need to win. The more supporters you get to vote early, the easier it will be to hit your number on election day. Having your supporters avoid long lines helps increase your numbers, but even in precincts where there aren’t long lines we still want our supporters to vote early.

  • Al

    PPP released a poll of WV just last night, in which they found that Obama was up by just one point in early voting there. It seems that lots of WV Democrats are voting for McCain.

  • Sam Wang

    Al – In addition to your point, it’s also worth noting that this result differs drastically from polls showing McCain leading by 10% among likely voters. This is consistent with the idea that Obama voters are more mobilized by the campaign and/or more enthusiastic.

    Michael is probably correct that exit pollsters are sufficiently good at their craft to avoid a rookie error!

  • Evans

    About the get out the early vote Obama strategy: that’s not the important part. The important part is that, using the same system, the Obama team plans to make a drastic number of reminder calls on election day. Estimates are that they will be able to call nearly 2% of the electorate on Nov. 4. Excluding voters in the big blue states, which won’t be the focus, and assuming they call only dems and independents, I’d estimate that 10-20% of Obama voters who did not vote early or absentee will receive a reminder call from a volunteer on election day.

    That’s why I’m predicting the polls will be off by 2-3% in every state, leading dems to 62 senate seats and Obama to 390+ EV. Enthusiasm is a big deal this year.

  • DFS

    I know you don’t focus on individual polls, but Politico/Insider Advantage has a poll out showing Missouri as a dead heat, with Obama getting 65% of the African-American vote. Clearly that can’t be correct….

  • Mark


    Prof. Wang has shown that aggregating polls will minimize the effects of outliers and lead to greater accuracy.

    Looking at subgroups from a single poll has the opposite effect. The margin of error may be an order of magnitude higher on small subgroups from a single poll. A notorious example of this is a recent IBD poll showing McCain beating Obama 75-25 among under 30 year old voters. Nate Silver explains the problem here:

    So I wouldn’t make much of the percentages on AA vote in a single MO poll. The poll you mention had 814 LV and MOE of about 3.5%. We don’t know how many AA voters were in the sample, but AA% in Missouri is about 12%. So you really can’t draw any significant conclusions from a sample of less than 100 people.

    Personally, I am ignoring the individual polls at this point and just looking at Prof. Wang’s aggregated numbers. I believe they are much more accurate than any individual poll.

  • Steko

    MoE for small subgroups can approach 20%. I think Sam has also said that observed MoE has been about 1.4 times larger then the states MoE.

  • Mark

    Sorry I wrote accuracy when I meant precision. A poll with a very large MOE may technically be accurate as the truth is within the MOE, but such a poll is meaningless.

    We won’t know how accurate the final polls are until we get the election results.

  • John B.

    Early voting seems to be just coming into its own…here’s a somewhat humorous take on it:

  • Michael

    And as luck would have it, Mark Blumenthal addresses this issue over at pollster today:

  • Eddie

    I know (I think I know) you don’t feel that the polls are being disrupted very much, but I’ll throw this out there for fun:
    I’m thinking that if Obama’s heritage and coolness is disrupting the polling data in ways that other folks talk about as possibilities, then maybe he would lose all of the eastern swing-states plus PA, but then win GA, and the election with 272 ev.

  • Andre Washington

    Early voting from the Black perspective involves not wanting your vote to be not counted. It is a matter of political survival. Vote suppression is alive and well in the GOP. In fact it is the norm. Do you ever hear of the republicans questioning the veracity of votes in lily-white suburban precincts. If you have please post that instance so that I might be corrected.

  • Liberal Chris

    David – you are unfortunately vastly underestimating the degree to which white southern voters vote republican. I believe Kerry got 26% of the white vote in Georgia and maybe 17% of the white vote in Mississippi. This is why the Republicans usually crush the Democrats even in southern states with very large black populations. I mean Kerry got something like 90% of the black vote, didn’t he?

    That said, if black turnout ends up above 30% of the total voters in Georgia, I think Obama has a shot. Then we’re talking that Obama needs only about 30% of the white vote, which would be tough but doable.

  • William

    Actually, if I remember correctly, Bush got something like 20% of the black vote in some states, something of a Republican record if you only count from Goldwater on.

  • Andrew Foland

    All this early voting is presumably biting, at least somewhat, into the election day totals. But who knows how much?

    On Election night, I don’t think I’d want to be a statistician trying to call any states based on exit polls, having no good model of how much cannibalization there’s been.

    Has anyone heard anything about how early voting cannibalization is likely to get taken into account Tuesday night, as networks try to call which way states have gone?

  • Michael


    See my earlier post(#18) and link.

  • 4degreesnorth

    Looking at this from Europe, this site is very useful and informative. Thank you.

    This being said, I am always amazed at how undersized polling samples are in the US compared to Europe. I keep seeing samples of 500 to 750 voters, which leads to huge margins of error. Our samples are usually 1000 and sometimes 2500 !

    It looks like our polling institutions in Europe – or is it our media – have less stringy purses.

  • DFS


    I dont recall bush getting 20% of the AA vote in any state. You have a cite for that?

  • William

    A simple Google of “bush black vote” suggests it happened in Ohio.

  • Behnam

    Professor, I have a question unrelated to this post.

    From the maps I see that all the blue and red states are geographically contiguous except Colorado and New Mexico. Demographically speaking, what makes these two states different from the states surrounding them?


  • Andrew Foland

    Thanks Michael, I didn’t know what the antecedent for “this” was on your post so didn’t click through. Actually, Blumenthal then points to a nice collection of early-voting data at Xcurmudgeon. If I were McCain’s pollster, after reading those numbers I’d tell him to just take the weekend off and think about what he wants to do next.

    In any case, it occurs to me that on election night, exit polls + precinct-reported numbers will give a good guess of the total turnout by party affilition. Summed with early-voting numbers, you could probably make a good model of how much of the early voting was simply sloshing back and forth versus how much was truly “new” GOTV. So maybe it’s not as bad as I was thinking at first.

  • Sam Wang

    William – In 2004, 11% of blacks voted for Bush. In Ohio, it was said to be 16%.

    4degreesnorth – I rather doubt that European polling organizations are much different from those in the US. US national pollsters often survey 1000 persons per day. At a state level, in the last week a total of about 14,000 people were surveyed in Ohio alone. Ohio is one of 50 states, though it is more popular than most for polling because of its swing state status. Its population is about 11 million, comparable to Belgium or Belarus.

  • Michael


    I do get a bit sloppy with my antecedents at times. Sorry. Sister William Mary would be very disappointed in me. Then again, she already was pretty disappointed in me as I recall.

  • wilderwood

    The numbers on early voting may be far more impressive than first realized. Consider NC, and the most recent (yesterday( DailyKos poll (from R2K), showing a 2% Obama lead, It only shows 19% of the poll respondents were early voters, whereas it is clear from the NC records that (a) the early voting (well over 2 million) is already 66 percent of the 2004 total turnout and (b) it has a considerably higher Black and Democratic percentage. It may therefore be the case these more Obama-supporting early voters are escaping the reach of polling phone calls and that there is much bigger Democratic vote building up–much bigger than the polls are reflecting. You didn’t hear it first here, but you’re still hearing it before the election itself!

  • Paul

    I notice that Gallup’s “traditional” and “expanded” models have converged — mostly due to movement in the traditional towards Obama.

    I realize I’m cherry picked a data point, and there’s not enough to draw a solid conclusion. Still, I can’t wonder: could this be due not just to random movement, but to the effect of early voting? Is the traditional model finding more and more people who already voted?

    Can’t wait for the post-hoc analysis on this whole election. It will be something to behold!

  • David

    Isn’t there a chance of a backlash as a result of the release and analysis of early voting data?

    Presumably the Republicans in the states where the black share of the early voters is running ahead of their share of the population will be using this information to galvanize white voters to vote on Nov 4.

    Similarly where the Democrat share of the vote is running well ahead of the Democrat share predicted by the polls. (This of course is pretty much everywhere!)

  • Sam Wang

    David – Who but total nerds like us is going to care about the details of early voting? We’re not exactly a key demographic.

Leave a Comment