This is what we work through Election Night for (my colleagues more than I): an exclusive, preliminary youth turnout estimate. It shows at least 9.9 million young Americans (ages 18-29), or 21.3%, voted in Tuesday’s midterm election, according to national exit polls, demographic data, and current counts of votes cast.

In a wave election for the GOP, youth still tended to vote Democratic. In the national exit poll data on House races, 18-29 year-olds preferred Democratic candidates by 54% to 43%. In many close Senatorial and gubernatorial races, youth preferred the Democratic candidate, and sometimes they were the only group that did (e.g., in Florida).

In terms of both turnout and vote choice, 2014 actually seems quite typical of a midterm year as far as youth are concerned. Young people made up a similar proportion of voters, and with some exceptions, were more likely to cast ballots for Democrats in tight races.

However, the Senate class of 2008 was not elected in a midterm year. They were elected in 2008, an exceptionally strong year for Democrats, when youth support for Barack Obama set the all-time record in presidential elections. The change from an extraordinary presidential year to a rather typical midterm year hurt the Democratic Senate incumbents. Their advantage among youth voters shrank compared to 2008 in some key states, such as North Carolina (down from 71% in 2008 to 54% in 2014) and Virginia (down from 71% to just 50%). And in some states that had been expected to be competitive this year, the Republican Senatorial candidate won the youth vote along with all older groups–Arkansas and Alaska being examples.

For Republicans, the lesson is they can be competitive among younger voters, although nationally, they still lag behind with that group, and in some states, the Democratic tilt of young voters may pose a problem in years to come.

For Democrats, the message must be to re-engage with young people, who had provided more support in 2008 Senate contests.

National Youth Turnout

According to our preliminary analysis, an estimated 21.3% percent of young Americans under the age of 30 voted in Tuesday’s midterm elections. That’s very close to our early estimate of 20.4 percent at this time in the last midterm election (2010).

This day-after youth turnout estimate, based on exit polls, the number of ballots counted, and demographic data from the US Census, is subject to change. In past years the National Exit Polls (NEP), conducted by Edison Research, have adjusted their data after an election; for example, its estimate of the proportion of youth in the 2010 electorate was adjusted twice after the election. Additionally, in three states, less than 95% of precincts have reported. As the number of ballots counted increases, so will youth turnout unless the share is adjusted downward.

Sources: The percentages of voters, ages 18-29, are obtained from national exit polls conducted by Edison Research. The numbers of votes cast are obtained from the media the first day following the election. Estimated voter turnout is obtained by taking the estimated number of votes cast and dividing it by the estimated population of 18 to 29-year-old citizens from the Census Current Population Survey 2014 March Demographic File.

When voting data from the U.S. Census (its Current Population Survey, November 2014 Voting Supplement) become available next year, it will be possible to see with greater certainty whether turnout rose, fell, or stayed the same. It is already clear, however, that turnout was in the typical range for a midterm election. See our note below for more information on these estimates.

*All estimates of youth turnout are subject to bias and error. The Exit Polls use a complex sampling method whose main purpose is not to estimate the ages of voters. If the Exit Polls report an inaccurate proportion of young voters, that will introduce error in our turnout estimates. Another estimate will become available during 2015, from the Census Current Population Survey 2014 November Supplement, which is a survey of a random sample of Americans conducted shortly after the election. The CPS is also subject to bias (for example, people may say they voted when they did not), but it has the advantages of a large sample and consistent method from year to year.