More Charts:
Voters By Age | Voters By Religion | Independent Voters
White Voters | White Men | White Women
Hispanic Voters | African-American Voters

National Journal has taken an unusually detailed look at the American electorate over the past 20 years. The project examines the results since 1988 from the general election presidential exit polls conducted outside polling places across the country by a consortium of newspapers and television networks. (The organization performing the exit polls has changed over the years. Edison Media Research/Mitofsky International conducted the most recent one, called the National Election Pool survey.)

Ron Brownstein explores The Hidden History Of The American Electorate in National Journal magazine. Results for key voting groups will be released regularly on NationalJournal.com through Election Day.

Surveys used in this analysis

• 1988 CBS News/New York Times: National Election Day Exit Poll, 11,645 respondents
• 1992 Voter Research and Surveys: National Election Day Exit Poll, 15,490 respondents
• 1996 Voter News Service National Exit Poll, 16,637 respondents
• 2000 Voter News Service National Election Day Exit Poll, 13,225 respondents
• 2004 National Election Pool National Election Day Exit Poll, 13,719 respondents

Margin of error

The huge sample size of the exit polls means you can look at smaller subgroups of voters than in a typical poll. However, there is still a margin of error (MOE) associated with exit poll estimates and the smaller the number of respondents included in a given subgroup, the greater the MOE. The MOE, as typically defined for polls, is a number you add and subtract from a poll finding to get the 95 percent confidence interval for the poll finding. For example, if 30 percent of a group voted for the Democrats in the survey and the MOE is +/-3 points, you can have 95 percent confidence that the true figure for the entire voting population in that group was between 27 and 33.

In the exit polls, the MOE ranges from +/-1 point when the sample size for a subgroup is over 8000 to a minimum of +/-6 points when the sample size is 100. We flag subgroups with sample sizes under 100 with an asterisk to indicate that their MOEs are relatively large and those particular estimates should be viewed with some caution.