WASHINGTON (AP) — Voters who cast ballots in Tuesday's election had a strikingly similar demographic profile to those who voted in 2010 and weren't all that different from those voting in 2006. But how they voted this year was very different from years past.
Exit polls are generally a bad way to measure changes in turnout. In addition to the usual concerns about sampling error in any poll, exit polls could show additional bias on some traits, such as race or population density. That's because the sample of precincts where voters fill out the surveys are chosen based on geography. If this year's exit poll includes more precincts located in an area where a cluster of voters share characteristics, such as race or income, it might overestimate the size of that group. Estimates of characteristics that aren't clustered this way, such as gender, are not subject to this type of bias, because men and women live in all precincts.
But exit polls do a better job of estimating how different groups voted.
Here's a look at some of the ways shifts in turnout and voter preferences played out in the 2014 midterms:
OVERALL TURNOUT DOWN
Michael McDonald, an associate professor at the University of Florida, estimates that about 37 percent of eligible voters cast a ballot in the midterm elections. If that projection holds, it would be the lowest turnout since 1942. With Americans busy fighting a war, the level dropped to 34 percent.
In 2010, the last midterm election, the turnout was about 41 percent of eligible voters, McDonald said.
Lots of votes are still being counted, including many absentee ballots and provisional ballots. And some states won't officially certify their vote count until January.
But preliminary estimates suggest turnout was higher in those states with more competitive races, such as Wisconsin, Maine and Iowa, and the overall national drop in turnout may be attributable to a lack of high-profile competitive contests in well-populated states like California, New York and Texas.
WHITE SUPPORT FOR DEMOCRATS FADES
Exit polling shows racial polarization of the electorate has begun to cross party lines, with whites less likely to back Democratic candidates than they have been in the past. Across 21 states where Senate races were exit polled, whites broke for the Republican by a significant margin in all but four — Michigan, Minnesota, New Hampshire and Oregon. None of those four states has backed a GOP candidate for president in the post-Reagan era except when New Hampshire went for George W. Bush by 1 point in 2000.
The Senate seats on the ballot this year were last up for re-election in 2008, a presidential year. Democrats typically rely on greater turnout among their core voters when the presidential race tops the ticket. But still, Democratic Senate candidates lost ground among white voters by an average of 10 points compared with 2008. White voters abandoned Democrats in droves in places with heated contests as well as those without much action. The exceptions were Minnesota and Oregon — where Democratic incumbents improved their overall support across the board — and Mississippi — where Travis Childers managed to grow the Democratic share of the white vote from 8 percent to 16 percent.
The shift is particularly acute in the South, where some of the last white Democrats in the House of Representatives lost their seats on Tuesday.
— In North Carolina, Sen. Kay Hagan carried just 33 percent of the white vote, down from 39 percent in 2008. White voters under age 30 backed Hagan decisively in 2008, 60 percent for her to 36 percent for her opponent, as they helped to sweep Barack Obama into office. But this year, younger white voters who cast ballots in North Carolina broke just as decisively for Thom Tillis, with 56 percent to 32 percent for Hagan. Twelve percent backed Sean Haugh, the Libertarian.
— In Louisiana, Mary Landrieu captured just 18 percent of the white vote, a sharp decline from the 33 percent she garnered in 2008. Younger whites there broke for her Republican opponent in 2008, 68 percent to 30 percent, and they were even more likely to back one of her GOP opponents this time around — 22 percent voted for Landrieu while 74 percent went for Bill Cassidy or Rob Maness.
— In one surprisingly competitive Senate race Tuesday, whites in Virginia voted 37 percent for Mark Warner, 60 percent for Ed Gillespie. In 2008, Warner won the votes of 56 percent of whites. Younger whites broke heavily this year for Ed Gillespie in Virginia, 57 percent to 31 percent for Warner. In 2008, Warner carried 59 percent among this group.
— Even winning Democrats aren't immune to the drop-off in white support: Illinois Democratic Senator Dick Durbin captured 43 percent of the white vote in his successful bid for re-election, that's down 18 points from his support among whites in 2008.
FEW REPUBLICANS HAVE REACHED BEYOND WHITE VOTERS
But Republicans haven't minimized racial polarization in the other direction either.
The coalition behind Republican Senate candidates was predominantly white, 90 percent across all 21 states with Senate races that were exit polled, ranging from 79 percent white Alaska to 98 percent white in West Virginia. Dan Sullivan in Alaska managed to pool the most diverse electorate with a strong showing among Alaska natives, and more than 10 percent of those backing both John Cornyn in Texas and Cory Gardner in Colorado were Hispanic.
Those three — Sullivan, Cornyn and Gardner — were the only Republicans to assemble a coalition that was less white than Mitt Romney's in the 2012 presidential election.
Exit polls of voters nationally and in 27 states were conducted for AP and the television networks by Edison Research. Most interviews were conducted among randomly selected voters at a random sample of precincts nationwide and in each state. In addition, nationally and in 12 states with a high percentage of absentee or early voters, telephone polls were conducted between Oct. 24 and Nov. 2 to ensure that the views of those who voted early or absentee were reflected. Results are subject to sampling error.
Associated Press writers Connie Cass and Hope Yen contributed to this report.
Exit poll methodology information: http://surveys.ap.org/exitpolls