Important legislation in the twentieth century, most notably the Voting Rights Act of 1965, has led to a long-term increase in the ability of Americans to participate in elections. The effects of other legislation intended to increase turnout, such as the National Voter Registration Act, have been more limited. Because Americans who vote tend to have different demographic characteristics than those who don't, it is often assumed that if turnout were to increase, election results would be significantly altered. However, research indicates that the political attitudes of non-voters are not much different than those of voters. Therefore, successful efforts to expand the electorate may have less impact on election results than many people expect.
What is 'voter turnout'?
Because high voter turnout is considered a mark of a thriving democracy, policymakers and citizens often support electoral reform measures based on whether they will increase turnout, either overall or for particular groups.
Although the idea of voter turnout is simple, measuring it is complicated. And even if the number of people who voted in an election is accurately counted, it's often unclear what turnout should be compared to—the number of eligible voters? Registered voters?
Political debates often rage over whether particular reforms will raise or lower turnout, either overall or for particular groups. Despite the political heat generated by such debates, academic research suggests that in most cases, policy change itself usually has little or no effect on voter turnout.
Turnout can be measured in the aggregate by simply counting up the number who vote in an election. Not every state does this, however. In 2016, four states (Mississippi, Missouri, Pennsylvania, and Texas) did not record how many people turned out to vote. Thus, the closest we can come in these states is to count the number of people who cast a ballot for a candidate, usually the one "at the top of the ticket." This comes close, but still leaves out the 1%-3% of voters in an election who abstain for the highest office on the ticket. Data from the United States Elections Project (USEP) indicate that 138.6 million voters participated in the 2016 presidential election.
Once we've settled on how to measure the number of people who voted, we need to compare it to the size of the eligible voting population, by calculating the turnout rate. The easiest comparison is with the voting age population (VAP)-that is, the number of people who are 18 and older according to U.S. Census Bureau estimates. However, VAP includes individuals who are ineligible to vote, such as non-citizens and those disfranchised because of felony convictions. Thus, two additional measures of the voting-eligible population have been developed:
- Citizen Voting Age Population (CVAP), which is based on Census Bureau population estimates generated using the American Community Survey.
- Voting Eligible Population (VEP), which is calculated by removing felons (according to state law), non-citizens, and those judged mentally incapacitated.
The denominator one chooses to calculate the turnout rate depends on the purposes of the analysis and the availability of data. Usually, VEP is the most preferred denominator, followed by CVAP, and then VAP. The estimated VEP in 2016 was 230.6 million, compared to an estimated VAP of 250.1 million.
Figure 1 shows the nationwide turnout rate in federal elections, calculated as a percentage of VEP by the USEP, from 1980 to 2016. In addition to the variation across time, the most notable pattern in this graph is the difference in turnout between years with presidential elections ("on years") and those without presidential elections ("off years"). Elections that occur in odd-numbered years and at times other than November typically have significantly lower turnout rates than the ones shown on the graph.
Figure 2 shows turnout rates in the 2016 election for each state. Although there are exceptions, states with the highest turnout rates in presidential elections tend to be in the north, while states with lower turnout rates tend to be in the south.
Studying voter turnout
Sometimes we want to measure the turnout rates of groups of voters, or study the factors that lead individual citizens to vote. In these cases we need individual measures of turnout based on answers to public opinion surveys.
The chief difficulty in using public opinion surveys to ascertain individual voter turnout is the problem of social-desirability bias, whereby many respondents who did not vote will nonetheless say they did to look like good citizens. As a result, estimates of turnout rates based on surveys will be higher than those based on administrative records. (For example, 78% of respondents to the 2012 American National Election Studies survey reported voting, compared to the actual turnout rate of 58% as reflected in the graph above.) To guard against over-reporting turnout in surveys, some studies use voter registration records to independently verify whether respondents voted, but few do.
Even with the problems of over-reporting, public opinion surveys are usually the only way we can study the turnout patterns of subpopulations of voters, such as regional or racial groups. It would be safe to use these surveys if all groups over-report on whether they voted by equal amounts, but there is evidence they don’t.
One survey often used to compare turnout rates for different groups is the Current Population Survey (CPS) conducted by the U.S. Census Bureau. In the Voter Registration Supplement, which is administered every November in even-numbered years, respondents are asked whether they voted in the most recent election. Because the CPS already has a rich set of demographic information about each voter and has been conducted for decades, this is often the best source of data.
The dominant theory for why turnout varies focuses on a type of cost-benefit calculation as seen from the perspective of the voter. Voters balance what they stand to gain if one candidate beats another, vs. their economic or social costs of voting. Other scholarship has challenged this approach by showing that going to the polls is largely based on voting being intrinsically rewarding.
A long history of political science research has shown that the following demographic factors are associated with higher levels of voter turnout: more education, higher income, older age, and being married (see table below). Women currently vote at slightly higher levels than men.
|High School Grad or less||44%|
|More than high school||77%|
|Family income less than $50,000||50%|
|Family income more than $50,000||69%|
For decades, African-Americans voted at lower levels than whites, but as Figure 3 illustrates, blacks and whites have now reached rough parity in recent national elections. Hispanics and Asian-Americans still vote at lower levels than whites and blacks.
Can reforms increase turnout?
Can particular election reforms such as Election Day registration, vote-by-mail, early voting, photo ID, etc., have an effect on voter turnout? Research results in most of these areas have been mixed at best. The one reform that is most consistently correlated with higher levels of turnout is Election Day registration (EDR), although even here, there is disagreement over whether EDR causes higher turnout or if states with existing higher turnout levels are more likely to pass EDR laws (it’s probably a combination of the two).
What about the roles that campaigns play in stimulating voter turnout? This is most visible in presidential elections, where candidates pour disproportionate resources into campaigning in battleground states—those that are closely divided along partisan lines and thus are most likely to swing the result of the Electoral College vote. In 2016, the average turnout in the 11 states where the presidential margin of victory was 5 percentage points or less was 67%, compared to 56% in the seven states where the margin of victory was greater than 30 points. (For the states in the middle, the average turnout rate was 61%.)
Field experiments to test the effects of campaign communications on voter turnout have shown that personalized methods work best in mobilizing voters and mass e-mails are virtually never effective in stimulating turnout.
What if everyone voted?
We care about turnout levels for two reasons. First, they're considered a measure of the health of a democracy, so higher turnout is always better than lower turnout. Second, if we believe that lower turnout levels exclude citizens with particular political views, then increasing turnout would “unskew” the electorate.
Early research seemed to justify skepticism that increasing turnout in federal elections would radically change the mix of opinions among those who actually vote. However, more recent research suggests that voters in national elections are more likely to be Republican and to oppose redistributive social policies than non-voters. Differences between voters and non-voters on other issues such as foreign policy are much less pronounced.
When it comes to local elections, overall turnout rates tend to be much lower than elections held to coincide with federal elections, and the demographic characteristics of voters are much more skewed compared to non-voters.