The Winston Group is a strategy and research firm dedicated to making ideas matter.

Don’t Rush to Any Early Conclusions on Election Night

by David Winston

Four years ago, early exit poll data was leaked midday to the Drudge Report, showing Sen. John Kerry (D-Mass.) with a huge lead in Pennsylvania — so large, in fact, that Republicans were in despair and the Kerry folks were celebrating their “victory.” The problem was the exit poll data reflected a partial sample with a significant number of interviews yet to be completed and, thus, was not yet statistically sound.

Until a full sample is completed, some of the underlying demographics may be disproportionate. So, in the interests of electoral integrity and to help avoid a similar rush to an erroneous conclusion next Tuesday, here is a short primer on how to interpret Election Day exit polls and results.

Exit Polls: The first official exit polling data will be made available to the networks at 5 p.m. It’s risky business, however, to rely too heavily on these first numbers because in most states they will reflect only two of three waves of interviews. It would be like looking at the results of 667 interviews out of a 1,000 national sample and expecting the final numbers to be exactly the same.

Drawing conclusions from the 5 p.m. exit poll data is further complicated by the fact that the data has not yet been matched to actual precinct voting returns, which can produce some changes.

Key States: When elections break one way or another, categories of states tend to move decisively. Four key states will come in early and could be predictors of the final outcome: Indiana, Virginia, North Carolina and Ohio. Ohio is clearly a swing state, but the other three have been reliably Republican in recent presidential elections. As the exit polls and actual results begin to come in, if Sen. Barack Obama (D-Ill.) is even or close in any of those states, especially Indiana, it could represent a potential significant shift.

Key Groups: Independents will be crucial to the outcome of the election. From 1994 through 2002, Republicans won this important voter group. Democrats edged out a small victory with independents in 2004, 49 percent to 46 percent. In 2006, Republicans lost this voter bloc by 18 points, 39 percent to 57 percent, and along with it the House and Senate.

To have a chance at winning the White House and making any headway in Congress, Republicans must be in a position of near parity with Democrats in attracting independent voters. The Catholic vote is another group to watch.

In the past two presidential elections, Republicans and Democrats split this key bloc. But in the 2006 election, Republicans lost Catholics by 11 points, 44 percent to 55 percent. If the margin next week is closer to 2006 than 2000, Sen. John McCain (R-Ariz.) and Congressional Republicans face a hurdle not only nationally but in key states like New Hampshire, Pennsylvania, Wisconsin, New Mexico, Florida and Ohio.

The fate of the middle class has been a central issue throughout the election. Republicans won middle-income voters ($50,000-$75,000) from 1992 to 2004, losing them in the 2006 elections. While the voting behavior of this group tends to be more evenly divided in presidential election years, it remains a key voter segment for the Republican coalition.

Married women with children have also been a major factor in past Republican victories. In 2002, the GOP won them 55 percent to 43 percent; in 2004, 54 percent to 45 percent. However, in 2006, Democrats made gains, splitting this vote at 49 percent each. To have a good night Nov. 4, Republicans need to win this group by a margin of close to 10 points.

Party Identification and Turnout: The wide disparity in national presidential election polls has created much controversy and speculation in recent weeks. The significant differences in the ballot test in various polls reflect uncertainty about current voter registration and how it will translate into actual voters.

In every election since 1992, exit polls showed party self-identification for Democrats and Republicans ranged from 36 percent to 40 percent each. In only one of those elections, 2002, did Republicans outnumber Democrats, and then by only a small margin. In fact, Republican self-identification numbers remained relatively constant in six of the past eight elections at 36 percent.

One of the biggest questions concerning voter turnout is the participation of younger voters and African-Americans. As a percent of the final election turnout, African-Americans made up 10 percent and 11 percent, respectively, in the 2000 and 2004 presidential elections. Seventeen percent of the electorate in each of those elections was made up of young voters, ages 18-29.

The question isn’t whether more of these voters will turn out, but whether they will represent a larger portion of the electorate than in previous years. Yes, more young voters turned out in the 2004 elections, but it’s important to remember that the size of the total electorate also increased by 20 percent, from 106 million to 122 million.

This year may be different. Because of the Obama factor, we may see unusual increases in the turnout of both groups. Some pollsters have created survey samples based on that assumption rather than past voting behavior. This adds an element of uncertainty to results and explains some of the disparity in presidential ballot test numbers in national surveys and tracking polls.

Last, on election night, if the margins for either candidate get big enough, it will overwhelm state-by-state Electoral College dynamics. Remember, in 1992, Bill Clinton won by 5.6 percent and got 370 electoral votes. In 1980, Ronald Reagan swept the election by a margin of 9.7 percent and won 489 electoral votes.