More than 17 million people have voted in the 2014 election. As the early vote pulls into the station, it is time to interpret the meaning of these numbers. I track early voting statistics here.
It's now possible to see how states are faring in their early vote compared to 2010. I've found the early vote is a decent calibration tool in addition to comparable past elections, to draw upon when making national and state turnout forecasts. Nationally, I expect a little over 90 million people to vote, or 41% of those eligible. In the competitive Senate races, turnout will be higher, averaging 46% of those eligible.
I believe we are on track to have about 27.5% of the votes cast prior to Election Day, up from 24.9% in 2010, as reported by the Census Bureau's Current Population Survey.
The early vote numbers are much higher in some competitive Senate and Gubernatorial states, and there is a wealth of data to pour over. The statistics in these states provide some clues as to what to expect on Election Day.
This is a deep dive into the data, so take a deep breath. If you want to stay in the shallow end, my take on the early vote data -- where there are enough statistics to be informative -- is that the Republican sweep screaming in the headlines is overblown. Senate control is up for grabs and Democrats have a decent chance to defy the polls. I expect that the election will be so close that we won't know who won until all ballots are counted and the vote is certified several days following the election, not to mention highly probable run-off elections in Georgia and Louisiana.
As of Saturday, Nate Cohn of the New York Times -- who has access to the daily updates from the Secretary of State -- reports registered Republicans enjoy a 40.5% to 32.5% or a +8.0 point lead among the 1,343,059 ballots cast.
Despite those favorable numbers, Gardner does not have the election sewn up. This election is going down to the wire. I'm not saying Udall will win; I'm saying this election is truly too-close-to-call despite poll averages showing a small Gardner lead.
In 2010, pre-election polls showed Bennet trailing. But, Democratic Senator Bennet won a tight race by +1.7 percentage points even though Republicans enjoyed a 39.8% to 33.0% or +6.8 point edge among all voters (not just early voters). If we add together 1.7 + 6.8 = 8.5, this might reasonably be an estimate of the registration margin that would equal a tied race.
In 2012, Obama won Colorado by +5.4 points even though Republicans enjoyed a +1.8 point lead among early voters. If we add together 5.4 + 1.8 = 7.2, which serves as another estimate that excludes Election Day voters in 2012. Although I don't know the composition of Election Day voters, surveys indicated that they supported Romney more than early voters, so it is reasonable to assume 7.2 is an underestimate on the typical Republican registration advantage needed for Republican candidates to win.
From just Friday to Saturday, Democrats trimmed the Republican lead in the 2014 early vote from +8.8 to +8.0. With three days of voting left, if this trend holds the Republican advantage will stand at +5.6 points. Garnder would have to do much better than Romney among unaffiliated registered voters to win.
There are three reasons why Democrats can be optimistic.
First, Bennet won even though a Green Party candidate won 2.2% of the vote. There is no Green Party candidate on the ballot in 2014. Without a Green Party candidate on his left, Bennet's margin might have been larger.
Second, the Republican lead has been narrowing as Election Day nears, and is almost certain to narrow further.
Democrats have more of an upside among the ballots to be returned. As of October 1, Republicans have a 32.5% to 30.7% or a +1.8 point lead among Colorado's 2,916,145 active registered voters. Since so many Republicans have returned ballots so far, Democrats and unaffiliated voters are a larger share of those who have yet to return ballots.
Among the 1,573,086 persons who have not returned their mail ballots, Democrats lead 29.1% to 25.8% or a +3.3 point margin. There are even more unaffiliated voters, 45.1%.
It is common to see unaffiliated voters turn their ballots in later than party affiliated voters, who tend to be the most committed to their candidates and thus most comfortable about voting earliest. At the very least, the margin between Democrats and Republicans will diminish as more unaffiliated voters return ballots. Polls show Udall leading among unaffiliated voters -- past election results like the 2010 Senate race demonstrate these voters break towards the Democrats, too -- so more of these voters should also help Udall.
As Election Day nears, a typical pattern in states with large mail ballot usage is that more people will return ballots as Election Day nears. Given the composition of these late voters, Democrats have a larger pool of potential votes to add to their column. It's not just Colorado where we've seen Democrats return their mail ballots later than Republicans, this trend has also happening in other states with high mail ballot usage and competitive statewide elections: Florida and Iowa.
Third, in addition to vote-by-mail Colorado implemented election day registration. Unregistered people who might take advantage of Election Day registration tend to be younger, persons of color, poorer, and less educated, i.e., Democrats. How might a person register to vote who doesn't have a mail ballot? Eligible unregistered Coloradoans who wish to vote can register and vote at special polling locations before or on Election Day. Inactive registered voters -- voters have not participated in a recent election and were not mailed a ballot -- may also vote at these polling sites.
Udall has a realistic path to victory. He needs Democrats and unaffiliated voters to continue to return their ballots as they currently are and unregistered voters to register and vote at special polling locations. The Democrats made a large investment to mobilize these very voters, an organization that became famous for their efforts in 2010 from their headquarters on Bannock Street in Denver, Colorado. The moment of truth has arrived which will reveal the efficacy of this effort. We will get a final read on Monday's party registration composition of the early electorate on Tuesday (Election Day data will not be processed, obviously). If Democrats continue to narrow Republicans' party registration, a surprise defying the polls and pundits may be in the works, as occurred in 2010.
If the election enters or is close to a recount situation there are two reasons why more votes may be tabulated after Election Day.
First, according to the United States Election Assistance Commission in 2010, 35,694 provisional ballots were counted in Colorado. Election officials rejected 3,667. Although we don't know why voters needed to cast a provisional ballot, a reason may be a voter who was issued a mail ballot but cast a second ballot in-person because the mail ballot was lost. Election officials require voters in this circumstance to cast a provisional ballots to ensure they cannot vote twice; provisional ballots are only opened once election officials confirm voters' mail ballots are not returned. With an all-mail ballot election, the number of provisional ballots to adjudicate may be higher. In states that report election results among provisional ballots, Democrats are typically favored.
Second, in 2010, 4,378 overseas civilian and military voters cast ballots. In Colorado, mail ballots from these voters received by Nov. 12 are counted. Likely most of these 2010 voters returned ballots by Election Day, but some were likely returned afterwards.
A simple reading of Florida's early vote is that Crist is poised for a big win. In 2010, Gov. Scott squeezed out a +1.1 point win while Republicans enjoyed a +12.7 registration advantage in the early vote. As of Saturday, the Republican advantage stands at +4.1.
There is a an important complication. In 2012, Florida adopted a quasi-permanent absentee status, whereby voters who vote a mail ballot can choose to automatically receive a mail ballot the next election. The Obama campaign, anticipating a reduction of in-person early voting due to a change that year, encouraged their supporters to vote a mail ballot; a good number automatically received a mail ballot in 2014. The consequence of this law is evident in the mail ballots. In 2010, 1,124,431 Floridians cast a mail ballot; in 2014, the number is already 1,733,014. Republicans led mail ballots in 2010 52% to 34%, they lead in 2014 by a narrower 45% to 37%.
In-person early vote is up more modestly, too, 1,184,395 have voted early Saturday with Sunday voting remaining in a few of the larger counties compared with 1,042,759 in 2010. Republicans led the 2010 in-person early vote 45% to 40%, Democrats lead 2014 42% to 41%.
The polls show a narrow lead for Crist over Scott. The outstanding question from the early vote is how many Democrats who voted in 2010 on Election Day have shifted to casting mail ballots in 2014. The in-person early voting, which seems of comparable magnitude, may thus be a better indicator of a Crist lead. I thus believe Crist indeed leads, but would have more confidence in my belief if not for the law change.
As of Thursday, registered Democrats have a narrow 6,829 ballot lead over Republicans out of 391,772 votes cast, a 40.9% to 39.2% or a +1.7 point lead.
Republicans have been crowing (pun intended) about this Democratic narrow lead because Democrats usually do better. In 2010, Democrats had a +5.7 point lead and in 2012 they had a +9.9 point lead. There were no statewide competitive elections in 2010, so the 2012 election may serve as the better benchmark. Obama won by +5.8, meaning that we might expect Braley to under-perform the early vote by 4.1 points, putting Ernst at a +2.4 point advantage.
Most likely, the Democratic lead will grow. Iowa continues accepting mail ballots postmarked no later than Monday, Nov. 3 as late as Monday, Nov. 10. Among the 111,154 mail ballots yet to be returned, Democrats are 42.3% compared to 27.5% for Republicans. Indeed, Since a week from last Wednesday, more Democrats have returned ballots every day, taking Republican's brief and small one-day ballot return lead.
If all the outstanding mail ballots are returned, Democrats will lead 41.2% to 36.6% or a +4.6 lead. I think it unlikely that all these ballots will be returned, so the margin from the mail ballots alone will be between +1.7 and +4.6, perhaps splitting the difference.
If the early vote margin ends at +3.6, the Ernst might have a slight projected lead of +0.5, but this lead would not be definitive for two reasons.
First, local election officials have an unknown number of satellite in-person early voting polling locations open through Monday. These votes are classified by Iowa as mail ballots. There may be some additional upside for Democrats from these polling locations as Democrats typically like to vote in-person early. However, where these polling sites are located will also affect the partisan composition of these voters.
Related, Iowa also has Election Day registration, but unlike Colorado, this convenience is only offered to voters on Election Day. Still, Election Day registration may provide more votes from those unregistered, which usually favor the Democrats.
Second, the Iowa early vote has already exceeded the 2010 total of 367,350. With several days of ballots yet to be returned and counted into the mix, the 2014 total is going to continue its increase. Democrats claim that Republicans' improved early vote performance over past elections is a consequence of shuffling their voters from Election Day to early through a more intensive ground effort. It is highly implausible all of the new early voters are people who never voted before, so this must be true for at least some voters -- on both sides. "How many?" is the outstanding question of importance to the forecast. I'm cautious about my Florida projection for this reason -- with a larger margin of new early voters favoring Democrats -- and I am similarly cautious about Iowa, especially considering the slim margin.
In sum, I think Ernst will end early voting with a slight projected lead from the early vote onto the final vote count. It will be a narrow lead with some uncertainty, and a lead much narrower than the poll averages indicate. Granted, there are outliers like the Des Moines Register poll with an Ernst +7 lead that are driving the poll averages -- I don't buy that poll for a minute. The lead is narrow enough that Braley may be able to overcome it with a better than usual Election Day performance, which could depend on how many typical Republican Election Day voters have voted early.
I suspect the most likely outcome is that this election will be so close that we won't know on election night who the winner is. There are two reasons why Iowa may end in overtime.
First, there are the unreturned mail ballots. The number of outstanding ballots will decrease by Election Day, but there will still likely be thousands of ballots in the mail when the vote is tabulated on election night. If the election margin is within a few thousand votes, the media may not call the election pending the disposition of these ballots.
Second, in 2010 Iowa had 1,999 counted provisional ballots and 349 rejected provisional ballots. These ballots add additional uncertainty to the vote beyond the outstanding mail ballots.
North Carolina ended in-person early voting on Saturday with 1,155,124 votes (including mail ballots). Registered Democrats enjoy a +15.7 point or a 47.6% to 31.9% lead. This lead will diminish a little over the next few days as the 35,060 outstanding mail ballots are returned, which are more Republican by a +9.1 margin or 41.4% to 32.3%.
North Carolina Democrats are breathing a sigh of relief now that the early voting period is in the books. The total early vote to date exceeds 2010's 962,647 by 20%. The legislative made a number of changes to early voting laws that Democrats were fearful would result in lower turnout. Although, I believe that some of these laws, such as expanded early voting hours the week before the election, may have actually facilitated more early voting.
North Carolina Democrats typically dominate the state's early vote. The good news for Democrats is their North Carolina early vote margin looks more like the 2012 +16.1 point performance than the +9.9 point 2010 performance. The bad news is that Obama lost North Carolina by 2.0 percentage points. Looking strictly at the 2012 projection from the early vote, Tillis would be expected to win by 2.4 points. The poll averages point to a Hagan win by about the same margin.
There are two potential explanations for the discrepancy between the polling averages and my projection.
First, the Libertarian candidate Haugh is averaging 4% in the polls, while the Libertarian candidate in 2012 received only 1%. The 3 point difference mostly explains the difference between the 2012 projection and the 2014 poll averages. The poll margins for Haugh have been decreasing over time. Haugh's share of the vote will be important to watch on election night. If it slips below 3%, that may spell trouble for Hagan.
Second, the early vote level is much higher than 2010, signaling an improved Democratic performance that may carry through to Election Day. Democrats have thus banked a sizable number of early votes, which could help them in the lower turnout 2014 election, compared to the 2012 election.
North Carolina has a good chance to go to overtime, too. North Carolina continues accepts mail ballots postmarked Tuesday, Nov. 4 as late as Friday, Nov. 7. If Hagan has a small enough lead, and enough ballots are still outstanding (which currently are more Republican in character), the media may not call the election.
There are also provisional ballots to consider. In 2010, North Carolina accepted 15,964 and rejected 10,563 provisional ballots. The number of rejected provisional ballots may be higher since the state no longer allows voters who cast a ballot outside of their home precinct.
In sum, the early vote confirms the polling averages that Hagan has a narrow lead, but there is much uncertainty here as elsewhere that could lead to overtime. Pay attention to the Libertarian Haugh's vote share; if it is below 3%, that signals danger for Hagan.
In-person early voting concluded with an impressive 93% increase over 2010. Registered Democrats are up 117% compared to 54% for Republicans. Those Democrats are not just old school Southern Democrats who have yet to change their party registration. African-American turnout is up 196% compared to Whites' 64%.
In 2008, Sen. Landrieu barely staved off a run-off garnering +2.1% above 50% with +28.9 point or 58% Democratic registration share to 29% Republican among 292,213 early voters. Despite the impressive increase over 2010, Democrats only have a +18.8 or 52.6% to 33.8% lead. Maybe some old school Democrats have finally changed their voter registration in the last six years, but the early vote appears to confirm the polling that this election is going to a run-off.
934,484 Georgians have voted early, with 37,034 ballots still outstanding. Georgia is yet another competitive Senate state where the early vote has easily surpassed the 2010 level of 783,253. According to the NY Time's Nate Cohn, African-Americans are 32.8% of early voters and Whites are 61.1%. With an older voter file I calculate a comparable 32.7% and 61.2% (I'm glad I didn't shell out $500 for a new file to get statistics the Georgia Secretary of State used to publish for free).
I unfortunately don't have good historical data to compare against to make a projection based on the racial composition of the early vote. So, I'll turn to adjudicating the polls.
The unknown is how much Whiter will the Election Day electorate be? This is the crux of the polling. The polls that show Perdue under 50% have the overall electorate around 30% African-American. Those that show a much bigger Perdue lead, like the Monmouth poll which has Perdue leading by +8, have African-Americans at 25% of the overall electorate. With perhaps a third of the vote already cast at 32.8%, it seems unlikely that there will be so few persons of color on Election Day. My best guess is that this election goes to a runoff.
The Georgia wildcard is the Libertarian candidate, Swafford, who is polling 3%. If Swafford's vote share drops, the odds of an outright Purdue win increase. Like Haugh in North Carolina, her polling share has been decreasing as Election Day nears.
An addendum on Alaska: We don't have much to go on with the Alaska early vote data, as the state only reports publicly the total number of mail ballots returned. Alaska is yet another state where there will be a generous number of late ballots to be counted after Election Day. Alaska accepts ballots postmarked on Election Day as late as Nov. 14. In 2008, Stevens led by +1 in the election night reporting, but Begich eventually won by +1.3 points. If Sullivan is leading by only a few points on election night, the media may not make a call until more ballots are counted.
More early voting statistics are here, and follow me on Twitter for continued commentary on updated statistics through Election Day.