Presenting Data: Tabular and graphic display of social indicators
 
Gary Klass
Illinois State University
2002

Note: The website will be discontinued shortly, to be replaced by the Just Plain Data Analysis site

Home Good Tables Good Charts Divide Analyzing Budgets Election Data Education Poverty References Course page Chart of the week your comments

Election Data. (under construction)

Voter Turnout.

 

One would think that calculating a basic political statistic like voter turnout would be straightforward.  It's not. Consider the many different ways one could go about defining the concept of voter turnout:

  • V/RV:  Number of voters as a percent of registered voters. 
  • V/VAP: Number of voters as a percent of the voting age population.
  • V/VAC: Number of voters as a percent of the number of voting age citizens.
  • V/VEP: Number of voters as a percent of the population eligible to vote.

The registered voter turnout statistic (V/RV) is of concern only in very specialized applications; when using it note the problematic nature of the concept: registration standards vary from state to state, some states have same day registration and states and localities vary greatly in terms of how often they purge voters who may have moved or died from the voter registration rolls. 

The most common voter turnout statistic is the measure of voters as a percent of the voting age population. This is especially useful as a measure related to characteristics of democratic society in a political system.   See International data.  Especially when comparing US states, measuring citizen voting percentages may be a more valid measure of political participation, but it depends on what one means by political participation.  Whichever concept one is trying to measure, however, there are omplications.

Counting the voters.

How, for example does one count the number of voters?  Surprisingly, not all American states actually count the number of people who vote in elections.  The most common solution is to count the number of votes cast for president (in presidential election years) and the number of votes cast for the house of representatives in congressional election years.  The 2000 presidential election provided a stunning reminder that not all voters who turned out cast votes for president or had their votes counted as having been cast.  In Congressional election years this is an even greater problem; especially because some Congressional elections are not contested, a significant number of voters probably do not cast votes for Congress.  In addition the "votes cast" measure has a significant disadvantage: the data are derived from local election records which are reported, in most cases, to the state secretaries of state (and then compiled by the non-governmental Congress Quarterly).  Except for a few Southern states that are required to report the race of registered voters, election officials collect almost no other demographic data on their voters.  Thus it is not possible to use this statistics to measure voter turnout for different racial and ethnic groups, different ages, or any demographic category other than geographic residence. 

Potentially a better measure of the number of voters cast is derived from the November Current Population Survey (CPS) conducted by the Census Bureau in presidential and congressional election years.  The monthly CPS is based on interviews of a sample of approximately 55,000 households, providing for a sampling error of something less than half a percent.  Because the survey also includes information on a wide range of demographic characteristics (citizenship, age, race, income, but not religion or political party).  The Census Bureau is careful to always refer to this statistic as a measure of reported voting (RV), implicitly acknowledging that not everyone answering the survey may be telling the truth. For the 2000 election, the votes cast measure (CQ /VAP) indicated that 51.2% of the voting age population cast votes that were counted, while the CPS measure (CPS /VAP) indicates that 54.7%. This would indicate that approximately 3.5% of the voting age population either:

  • voted, but not for president,
  • voted for president, but their votes were not counted or,
  • did not vote, but told the Census Bureau that they did. 

The American National Election Survey (NES) conducted by the University of Michigan is a frequent source of voting participation data used in political science research and is conducted before and after each presidential and congressional election. Based on a survey of approximately 1,500 persons (sampling error about 2.5%), the NES data consistently report a much higher level of voter turnout -- 72.7% in the 2000 election.  This would indicate that there is something wrong with the University of Michigan sample (the NES response rate fell to 52% for the 2000 election while the CPS maintains a response rate higher than 90%) or that people are more likely to lie to the university pollsters than they are to the Census Bureau.  The advantage of the NES data is that the survey includes a great many other questions that the Census Bureau does not ask.  If you want to compare the voter turnout rates for Democrats and Republicans, different religions or people who have different opinions on different social issues, the NES is the best source of data (see some of these data here).  Of course, nobody knows how much respondents lie on those other questions.

Deciding on the denominator in the turnout equation also presents difficulties.......

Figure E1: Five Measures of Turnout

Things to watch out for:

Although the US has a lower record of voter turnout that most other countries, Americans have more elections and vote for more elected officials than any other country. Since 1945, France and England have each had 15 national elections, the US has had 26.  Few parliamentary democracies have elections as often as every two years, few Democracies  have primary elections (fewer, still, have open primaries).  The practice of holding elections for offices such as County treasurer, County recorder, County assessor, and County coroner (as is the case in Illinois) is a peculiarly American phenomenon.

When interpreting the time series trends in American voter turnout, be aware that much of the post-1972 drop in voter turnout had to do with the 26th Amendment which provided 18 year olds the right to vote (a few states allowed persons under 21 to vote before 1972).  This accounts for the decline from pre-1972 to post 1972 and most of the decline in voter turnout since 1972, since the younger voters less likely and increasingly unlikely to vote.

References:

NES /VAP:  American National Election Survey
McD /VEP:  Michael McDonald US Elections Project
CQ /VAP:    Congressional Quarterly Inc. "America Votes"; see US Statistical Abstract 2002, table 395
CPS /VAP /VAC:  Current Population Survey

Frances Fox Piven and Richard A. Cloward, Why Americans Still Don't Vote: And Why Politicians Want It That Way (Beacon Press, 2000)

David Glass, Peverill Squire, and Raymond Wolfinger, "Voter Turnout: An International Comparison," Public Opinion 6 (December 1983/January 1984): 52.

Thomas E. Patterson, The Vanishing Voter (Alfred A. Knopf, 2002).

Raymond E. Wolfinger and Steven J. Rosenstone, Who Votes? (Yale University Press, 1980)

_______________

Exercises:

a. Examine these data from the National Election Survey.  Which variables have the strongest effect on voter turnout?

b. Prepare a chart showing the relationship between age...