Electoral Mobilization in the United States

James W Endersby, John R Petrocik, Daron R Shaw. Handbook of Party Politics. Editor: Richard S Katz & William Crotty. Sage Publications, 2006.

Simply put, Americans are not politically active. Thirty-five percent reported an attempt to persuade someone to vote a particular way in 2000; about 11 percent displayed a campaign button or sticker on their person, lawn, or car; 11 percent reported a contribution to a candidate, party, or some group that worked on behalf of a candidate or party; 5 percent remember attending a political rally or meeting; only 3 percent did any work on behalf of a party or candidate. Each activity requires energy, spare time, disposable income, or a level of interpersonal aggressiveness that many people simply do not have.

Voting is much less demanding and civic norms encourage Americans to vote, but even this activity is not that attractive to many Americans. In the 1996 American National Election Study (ANES) survey, 50 percent agreed that it is acceptable to stay home if one doesn’t care how an election comes out. Only 41 percent thought that a person had an obligation to vote even if the person didn’t care who won. This expressed indifference is corroborated by turnout rates. Just over 17.5 percent took part in the primaries that selected the candidates who ran in the 2000 general election. The low participation rate of Democrats (less than 14.7 million) might be explained by the virtually uncontested nomination of Vice-President Gore, but only 20.7 million Republicans turned out—and no one should describe George Bush’s nomination as uncontested. Americans did better in the general election where turnout exceeded 51 percent, about 15 points above the turnout rate for the 1998 congressional elections. But ‘better’ still leaves the United States well below the norm for almost any comparable country. The turnout rate for our presidential elections has averaged about 55 percent of the voter-age population (VAP) since 1980; a figure which is about 25 points below the average of other Anglo-American democracies, and about 30 points below the average turnout rate for the nations of Western Europe, with whom we share a substantially common political culture. As Figure 26.1 indicates, only the Swiss have been less likely to participate.

Interestingly American turnout is not lower because Americans have unusually low interest in public affairs or weak partisanship. On average, Americans display interest levels and party attachments that are similar to those of citizens of countries with much higher turnout rates. Since both interest in politics and partisanship have some natural ceiling, a program to increase interest or party attachments is not likely to do much to increase turnout levels from their current 55-60 percent or more in presidential elections (Powell, 1982; Teixeira, 1992; McDonald and Popkin, 2001).

The differential is likely to be found among institutional variables—registration requirements, voter canvassing and election day mobilization, and fewer elections—not in American citizenship norms (which are not so terrible in comparative perspective). This chapter reviews the role of voter canvassing and election day mobilization in promoting turnout. The working hypothesis that drives this chapter is that a significant contributor to low turnout is the style of campaigning in America.

Figure 26.1 Election turnout in 20 comparable countries (percent)

The Partisan Consequences of Electoral Mobilization

American campaigns do relatively little to shape turnout. They spend the bulk of their resources on attempting to shape the direction of the vote, effectively assuming that turnout will reach a level ‘typical’ of turnout of the type of election at hand and that the best way to win an election is to ‘improve’ the vote share among the expected voters. Political science has encouraged this by demonstrating repeatedly that non-voters typically share the preferences of voters. The argument that turnout tends to have an inconsistent partisan bias is, however, not uncontested.

Lijphart’s prominent (1997) essay arguing that a bias exists reflects a substantial literature linking turnout and electoral mobilization to partisan outcomes. Campbell’s early (1966) study of shifts in party fortunes between higher-turnout presidential and lower-turnout congressional elections provided a basis in survey data for the biased turnout hypothesis. Subsequent analyses of election results provided additional empirical support (recent examples include Tucker and Vedlitz, 1986; Nagel, 1988; Avery 1989; Radcliff, 1994; Pacek and Radcliff, 1995; and—more tentatively -Nagel and McNulty 1996). Sporadic popular political commentary often reported circumstances that reinforced the notion that turnout has a direct effect on election outcomes (Duncan, 1991; Freedman, 1996). Commentaries that attributed the success of the GOP, especially in presidential elections, to a decline in turnout since the 1960s (Edsall, 1984; Burnham, 1987, Piven and Cloward, 1988) created a chorus of support for schemes that might increase turnout—same-day registration, ‘motor voter’ bills, and so forth.

The literature on party system change provided important ancillary support for the biased turnout thesis. It identified a similar turnout/mobilization process as a critical ingredient in party realignments, arguing that any party system with a low rate of participation is vulnerable to substantial change because the segments of the society who are involved are virtually never a representative sample of those who do not participate. When these peripherals are mobilized they typically transform the party balance because the newly mobilized rarely find attractive choices among the parties of the limited electorate (some examples of this literature include Lubell, 1952; Eldersveld, 1949; Key 1955; Burnham, 1970; Przeworski, 1975; Andersen, 1979; Petrocik, 1981a; Petrocik and Brown, 1999).

Figure 26.2 The presidential preferences of voters and non-voters

The 1980, 1992, and 1996 preferences of non-voters are calculated only for those interviewed during October in order to account for a clear shift in preferences in the last half of the September-October campaign period.

Despite its intuitive appeal and fragmentary supportive evidence, the turnout bias thesis is not well supported by direct evidence. Kernell (1977) was among the first to demonstrate that the on-year/off-year oscillation in party fortunes (described by Campbell, 1966) was less a function of turnout differences than it was a retrospective reaction to the incumbent president and, perhaps, the loss of coattails in the off-year contest. Wolfinger and his colleagues (Rosenstone and Wolfinger, 1978; Wolfinger and Rosenstone, 1980; Highton and Wolfinger, 2001), Crewe (1981), Erikson (1995a, 1995b), DeNardo (1980), Petrocik (1981b, 1987), Petrocik and Shaw (1991), Petrocik and Perkins (2003), Teixeira (1987, 1992), Calvert and Gilchrist (1991), Gant and Lyons (1993), and a host of textbook authors who included a chapter on turnout (e.g., Beck and Hershey 2001) have found little or no support for the bias thesis. Recent research by Citrin et al. (2003) which reports a partisan bias to turnout in Senate elections in the 1990s, nonetheless finds it to be inconsistent, varying in size and party beneficiary according to election-specific factors that may include incumbency, campaign resources, etc. Notably, the bias, even when it was observed, would almost never be large enough to change the outcome of the elections even if 100 percent of the non-voters were persuaded to turn up at the polls.

Recent Patterns: Voters and Non-Voters in National Elections

Figure 26.2 summarizes how closely non-voter preferences match those of voters. It plots two lines: the reported vote for the winner in presidential elections from 1952 through 2000 and the expressed preferences of those who reported not voting in the ANES surveys of those years. The preferences of non-voters were expressed prior to the election (eliminating some of the bandwagon effect that is observed when candidate preference is obtained after the election); their voter status is determined by their post-election report. The pattern is clear. Whether the election is lopsided or close, a two-party struggle, or one with a significant third candidate, non-voters expressed the same candidate preferences as did voters—and sometimes a bit more so. The pattern is not perfect. Non-voters preferred Stevenson in 1952 and seem to have preferred Carter in 1980 (but see Petrocik, 1987). Overall, however, whether the election is lopsided or close, a two-party struggle, or one with a significant third candidate, non-voters express the same candidate preferences as do voters—and sometimes a bit more so (see below for more on surge effects). A 100 percent turnout rate would have produced the same winner, at either the same or a slightly greater margin.

Figure 26.3 Turnout and the presidential vote, 1948-2000

Of course the similarity of the preferences of voters and non-voters that is observed in survey data appears in aggregate election results. Figure 26.3 plots the Democratic presidential vote against turnout; Figure 26.4 presents plots of the Democratic vote for Congress against turnout for presidential election years (the points) and for mid-term elections (the untilled diamonds).

The data are clear. Democrats won and lost with high turnout among the 14 presidential elections (Figure 26.3). House elections in the on years and the off years show the same pattern (Figure 26.4). In off years the Democratic share of the congressional vote was actually slightly higher (by about 1 percent), although the typical turnout rate for off-year elections from 1950 through 1998 averaged approximately 15 points below the turnout typical of presidential elections. The pro-GOP tides of some presidential elections were absent, so there was a less depressing effect on the Democratic vote in the off years. Petrocik and Perkins’ (2003) analysis of turnout effects on election outcomes in congressional elections by district through time confirms this pattern. Non-voters tend to echo the preferences of voters, and perhaps even exaggerate them, depending on the magnitude of the forces influencing the vote. A close contest among voters and the more involved produces a roughly similar division among the less involved peripherals. Short-term forces sufficient to boost turnout are likely also to favor a candidate. They tilt the candidate preference of the core electorate and have an even greater influence on the peripheral electorate. Consequently not only are lopsided elections unlikely to be undone by higher turnout, they are likely to become even more lopsided as turnout increases beyond a normal level since the entering voters create an electorate with a larger than normal proportion of peripheral voters.

Figure 26.4 Turnout and the congressional vote, 1948-2000

The Basics of Turnout

The most defensible estimates of turnout, based on the fraction who vote as a proportion of the VAP that is not legally disqualified, are that about 55 percent vote in presidential elections (53 percent in 1996 and 51 percent in 2000) and about 40 percent show up in off-year national elections (40 percent in 1998 and 39 percent in 2002). But these figures are only estimates. We know (with considerable accuracy) how many votes are cast, but we are unsure about how many might be cast if all those who are eligible actually turned up at the polls. The most commonly reported turnout rate is calculated as a proportion of the VAP. It has the virtue of making easy comparisons among jurisdictions within the United States, and between the United States and any given country. But these strengths are dominated by a weakness: it significantly deflates the turnout rate by including many ineligibles—resident aliens, the criminal population, and various other institutionalized individuals (Burnham, 1987; McDonald and Popkin, 2001). What is the ‘correct’ turnout rate? Including aliens and others who are legally barred from voting in a count of the eligible electorate is unreasonable since by custom and law we specifically bar their participation. Similarly, a turnout rate that is calculated as a proportion of the registered electorate misses many millions who could have voted had they satisfied the technical requirement of officially registering an intention to vote. Put differently, the unregistered eligibles have, by intention or inadvertence, indicated they will not be voting before the polls open—when we find out how many of the rest will abstain.

Who Votes: The Social and Demographic Basis of the Vote

Turnout rates vary by (in rough order of importance) race, age, education, income (and similar SES markers), marital status, and geographic region (the data are in Table 26.1). Type of employment matters: government employees, especially those in states with a tradition of patronage, turn out at higher rates than comparable citizens. Unionization also matters. The difference in the voting rates of union versus non-union households in 1988 is quite striking, given the slightly more down-scale status of union households. The direct mobilization efforts of unions (more on that below) probably deserve credit for this difference. Employment, per se, is also associated with turnout. The unemployed do not vote at rates comparable to those who are employed, retired, homemakers, or students.

Table 26.1 US election turnout by demographics, 1988 and 2000
1988 validated turnout (n = 2040)
Percent of population in 2000 Reported turnout in 2000 (n = 1807) Voted Reg’d, didn’t vote Unreg’d
Total 100% 75% 58% 11% 32%
   Male 48 75 59 11 30
   Female 52 71 56 11 33
   White 76 74 62 10 29
   Black 12 73 40 18 42
   Hispanic 6 56 49 10 41
   Other 3 67 41 1 7 42
   18-24 13 51 31 11 58
   25-39 29 69 51 13 36
   40-54 29 77 66 8 26
   55-64 13 84 67 7 26
   65+ 17 81 68 12 20
   Low 39 62 43 14 44
   Middle 30 70 59 10 31
   High 32 86 71 9 20
   Less than high school 15 46 40 11 49
   High school 33 66 51 12 37
   Some college 28 77 64 10 26
   College degree 16 92 79 10 11
   Post-graduate 8 94 84 9 8
Marital status
   Married 60 79 66 9 26
   Formerly married 17 67 51 12 37
   Never married 24 61 46 14 40
   Protestant 54 72 57 12 31
   Catholic 27 79 62 9 29
   Jewish 2 89 66 6 28
   Other 10 64 45 11 44
   Agnostic/Atheist/None 14 60 42 21 37
Born again
   Yes 40 73 56 12 33
   No 60 76 61 10 29
Church attendance
   Every week 37 83 71 9 20
   Almost every week 16 83 63 11 26
   Once or twice a month 23 75 59 8 33
   A few times a year 24 73 54 12 34
   Never 1 80 43 12 45
Length at residence
   0-11 months 12 51 45 13 42
   12 months-2 years 18 63 51 9 40
   3-5 years 17 71 52 12 36
   6-10 years 17 80 58 10 32
   11-25 years 22 83 66 9 25
   25+ years 14 81 76 10 15
   Executive, management 13 87 69 8 22
   Professional 18 85 79 7 14
   Craftsman, skilled worker 11 66 48 15 37
   Sales 12 72 60 9 31
   Administrative worker 16 75 63 12 25
   Service 11 58 50 13 37
   Machine operator 5 66 38 12 50
   Transportation 4 72 48 15 38
   Agriculture 2 69
Working status
   Employed 64 74 59 10 31
   Temporarily laid-off 1 65 44 16 41
   Unemployed 4 55 28 14 58
   Retired 17 82 69 9 22
   Homemaker 8 66 55 13 33
   Student 3 62 50 11 39
Union household
   Yes 15 77 62 9 29
   No 85 72 56 11 32
   New England 7 82 72 1 0 18
   Mid-Atlantic 16 80 53 9 38
   East North-Central 21 77 67 8 26
   West North Central 9 81 73 7 20
   Deep South 26 67 43 16 41
   Border South 6 75 46 11 42
   Mountain 5 61 55 6 39
   Pacific 11 62 68 10 22
Population Size
   Major city 10 80 57 15 28
   Minor city 14 75 54 13 33
   Major suburb 8 79 63 1 0 26
   Other suburb 21 71 58 9 33
   Adjacent area 28 72 57 11 32
   Rural 19 68 55 9 36

Notes: Data are taken from the American National Election Studies, of 1988 and 2000. Validation of the self-reported vote was not undertaken in the 2000 ANES. Explanations of the region and population size variables can be found in the technical appendices of the ANES codebooks.

Catholics and Jews vote at slightly higher rates than do Protestants, but most of this religious difference is a proxy for region and race-ethnicity Turnout differences by religion are trivial or non-existent among northern whites. Lower overall turnout among Protestants reflects the Protestant traditions of African-Americans and southern whites. Catholic turnout is partially suppressed by the concentration of Hispanics among Catholics. Religiosity, however, is correlated with turnout: 59 percent of those with a religious identification voted in 1988, while only 43 percent of atheists, agnostics, and those with no religious preference cast ballots. Among believers, church attendance is also related to turnout. The turnout in 1988 was 71 percent of those attending church weekly, compared to 43 percent for those who never attend services. Living circumstances, from length of residence in a given community to the part of the country in which that community is located, also correlate with turnout. Among these factors, the length of residence is the most significant. Newcomers (those who have been at their current residence 6 months or less) exhibited a 41 percent turnout rate in 1988, compared to a 76 percent rate for those who had been at their residence for over 25 years. This 35 point effect is obviously related to structural factors, such as registration, as well as to other demographics, such as age and marital status.

There are also slight turnout differences according to the size of the community and its regional location. Americans living in the suburbs of major cities turned out more than those living in smaller cities or rural areas. Distances may also matter. Gimpel et al. (2004b) have found that turnout rates are higher for those living within a mile of the polling place than for those who live farther away. New Englanders (at 72 percent) and those living on the Pacific coast (68 percent) are the most likely voters, while southerners turn out at the lowest rates (44 percent).

Who Votes: The Attitudes and Beliefs of Voters and Non-Voters

Although demography can sometimes be plausibly and directly related to turnout (consider ethnicity: in the past African-Americans were systematically denied the right to vote), demographic differences matter because they produce circumstances that shape attitudes that are the direct influence on individual turnout. For example, Americans who are interested in politics vote; those who are not, do not. A closely related attitude is whether one cares about the election outcome: 67 percent of those who cared ‘a great deal’ voted, compared with 43 percent of those who ‘do not much care’ (Table 26.2). Voters who think that an election is close are more likely to vote (at 61 percent) than those who foresee a lopsided result (53 percent). Popular commentary and recent research have focused on declining confidence in government as a factor in low turnout (Hetherington, 1999). The evidence here is contradictory and mixed, and findings depend on fine distinctions between the electorate’s trust in government and their sense that government is effective. On average, however, there is evidence that those who regard government as ineffective and confusing have low turnout rates.

Strong partisans (Republicans and Democrats) are much more likely to vote than are weak partisans and independents (although Republicans have a higher turnout rate than Democrats at any given level of partisan intensity). But issue orientation, in general, does not seem to matter. The relationship between ideology and turnout is complicated, and while there is a small correlation, it is only small. Self-described conservatives have slightly higher turnout rates, but conservative preferences on specific issues are not necessarily associated with high turnout. Republicans have slightly higher turnout rates than Democrats; and there will be a turnout difference between conservatives and liberals if the issue is closely tied to party identification.

The Influence of Political Institutions and Systemic Processes on Turnout

No one is required to pay a fee in order to vote, but it is not completely costless. Trooping to the polls requires time that could be spent doing something—watching television, going shopping, painting a bedroom, or reading a book -that is personally rewarding, and at least represents another use of one’s time. A person who will be out of town might decide to vote absentee, a choice which might require even more energy: the registrar of voters must be asked for an absentee ballot application, the application must be completed and mailed, and the ballot must be completed and mailed. Deciding whom to support can be costly if only for the time it takes to become informed. The huge number of offices, many non-partisan, or propositions for which we must often vote can stymie even the conscientious and informed since they might require just that much more commitment to the vote than a person feels, causing them to abstain. This ‘cost’ feature of voting is represented by political institutions that structure campaigns, the way parties mobilize, voting requirements, and so forth. The United States has many such institutions and practices. They affect the turnout of individuals and groups in any given election, they play a role in holding down turnout levels in the United States compared to similar countries.

Table 26.2 US election turnout by attitudes, 1984-88 ANES
1984-88 validated turnout Did not vote Registered?
% of Pop. % of VAP. Voted Yes No
Party identification
   Strong Republican 14 18 74 7 19
   Weak Republican 14 15 61 10 29
   Lean Republican 14 13 54 12 34
   Independent 11 7 39 9 52
   Lean Democratic 12 11 53 13 34
   Weak Democratic 18 16 54 13 33
   Strong Democratic 17 20 67 11 23
   Conservative 60 65 64 10 26
   Moderate 10 8 55 10 35
   Liberal 30 27 55 13 33
Efficacy and trust attitudes
   Trust government just about always* 4 3 61 11 29
   Trust government most of the time* 41 42 72 9 19
   Trust government only some of the time* 54 54 70 9 21
   Government too powerful 63 62 67 8 24
   Government not too powerful 36 37 69 10 22
   Government wastes a lot* 66 70 74 8 15
   Government wastes some* 3 28 65 10 25
   Government wastes hardly anything* 4 3 48 10 42
External efficacy
   Government cares what I think* 57 64 68 10 22
   Government doesn’t care* 43 36 53 12 35
Internal efficacy
   Government not too complicated* 71 78 71 9 20
   Government too complicated* 29 22 57 11 32
   People like me have a say* 68 76 69 10 21
   People like me have no say* 32 24 46 14 40

Notes: Asterisks indicate data come from 1984 ANES; all other data are from the 1988 survey. Turnout estimates are higher for the 1984 results because turnout was higher and the questions were only asked of post-election survey participants.

Voluntary Pre-Election Registration

The Federal Election Commission reports a registration percentage (based on the VAP) of 76 percent for 2000 (up about 2 points from 1996). If the VAP is reduced to eliminate ineligibles, the Commission’s estimated registration rate is over 80 percent, a figure that is surely too high and may only reflect a failure to purge deadwood from the rolls (a possible result of state attempts to meet the spirit as well as the letter of the 1994 ‘motor voter law’). Like turnout, therefore, the exact registration rate is unknown, but a reasonable estimate would be in the low seventies.

Registration is important because those who are registered have a high turnout rate (Kelly et al., 1967; Erikson, 1981; Highton, 1997). A plausible (and not too conservative) estimate of the turnout rate among the registered (based on the voter validation portion of the 1988 ANES) is about 85 percent. If everyone was registered does that imply a turnout rate of about 85 percent -and a participation rate that would put us in the middle of the pack of comparison countries (see Figure 26.1)? Probably not, although research has concluded that liberalized registration rules would have a large effect on participation rates.

Table 26.3 Registration and turnout in 1988 (validated data)
Turnout Registered Turnout of registered
Years of education
   Less than high school 40 50 80
High school 51 62 82
College 79 87 91
Post-graduate degree 84 90 93
18-24 31 42 74
25-39 51 63 81
40-64 67 72 93
65 and above 68 85 92
   Time at current address
Less than 1 year 41 53 77
1 year 48 59 81
2 or more years 61 70 87
   Interest in politics
Very little 34 45 76
Some 58 68 85
A great deal 78 85 92

The largest estimate of the effect of registration is that American turnout is 14-15 percentage points below where it would be with election day registration. The most conservative estimate, based on turnout changes that occurred when Minnesota and Wisconsin introduced election day registration, pegged the increase in turnout at 1-2 points. A more systematic study that examined turnout and registration data for the period from 1960 through 1986 concluded that election day registration would increase turnout by about 4 percentage points. Still other studies have estimated that election day registration would increase turnout by 7-9 percentage points. This last estimate seems to be the most likely consequence of eliminating prior registration requirements. The methodology used to derive this estimate is the most defensible, and at least three separate studies have produced registration effects in the range of 7-9 percentage points (see Teixeira, 1992).

Who is Affected by Registration?

Registration depresses the turnout of those who are less interested in politics or must exert noticeable effort to become registered: the less educated, younger, or the geographically mobile. The less educated tend to be less involved in public and community affairs and less likely to think of the need to register in the absence of considerable stimulation; younger citizens are much more likely to change their address frequently, and every change of address imposes a requirement to reregister. These conditions—mobility, education, etc.—often occur together (see Rosenstone and Wolfinger, 1978; Wolfinger and Rosenstone, 1980; Nagler, 1991). Further, registration deadlines are often earlier than the arousal of interest in the election for many Americans. Only the drum beat of the election campaign will stir many to consider voting, but an interest in the election is likely to be peaked only when the campaign is at full tide—and that is often too late for many to recognize that they are unregistered, discover how to become registered, and actually do it.

Table 26.3 summarizes the effects of registration the registration on voters who are unlikely to solve the registration puzzle. The first column reports the difference in the turnout rates between those who are the least and best educated, the younger and the older, those who have no interest in politics and those who pay at least some attention, and those who recently changed their place of residence and those who have had the same address for at least 2 years. On average, those who have some college education, are 40 years of age or above, and have some interest in public affairs are about 30 percentage points more likely to vote than their counterparts who, respectively, never completed high school, are under 24 years or age, or have no interest in public affairs. Among those who are registered (the second column) the differences between these educational, age, interest levels, and mobility levels are substantially reduced (and age differences completely disappear).

Registration does not affect all groups uniformly. The largest effects (of about 10 percentage points) are found among the younger and less educated, minorities, and southerners. Better educated and older voters, white voters, and those who live in areas where participation is already reasonably high (the Midwest, for example) are the least affected (Wolfinger and Rosenstone, 1980).

Registration reform in the past 30 years has had limited effects on turnout. The estimated 7-9 point turnout reduction imposed by prior registration requirements translates into a projected 61-62 percent turnout—20 points below the turnout rates of the currently registered—if election day registration had been the law of the land in recent elections. The shortfall is a reminder that registration is only part of the story. Registration is a burden for those who have enough intrinsic interest in public affairs to vote only when the effort to vote is minimal. ‘Enough intrinsic interest’ is key, and it is lacking among many, perhaps most, of the unregistered.

A useful illustration of the importance of individual motivation is given in Cain and McCue’s (1985) study, which compared the turnout rate of individuals who registered on their own initiative with the turnout rate of those who were registered as a result of an organized effort by voter mobilization groups. Fifty-six percent of the former turned out, only 41 percent of the latter voted. Democratic identifiers registered by Democratic-organized groups were the least likely to vote after they were registered. Only a follow-up with organized get-out-the-vote (GOTV) efforts yielded high turnout among the group-registered. The difference between group-registered and self-registered is that the self-registered wanted to vote and, when they were eligible, did vote. The group-registrants were less interested in voting. They registered only when external pressure was applied; they voted only when external pressure was applied. Voting simply was not a habit for them (Gerber et al., 2003; Green and Shachar, 2000; Plutzer, 2002).

The System of Representation

In proportional representation systems, which attempt to ensure that parties win offices in proportion to the number of votes they receive, each party has a strong incentive to get every potential supporter to the polls. When the electoral system is less committed to proportionality the incentive to mobilize every prospective voter is less compelling. Single-member/simple plurality representation systems (typical throughout the United States where we elect one office-holder from each district) are likely to generate high turnout only when the election is competitive. A district (for a school board, city council, county supervisor, state legislature, or the House of Representatives) that regularly supports one party by a wide margin (the typical condition for America’s artfully gerrymandered districts) is not likely to be subject to vigorous registration or GOTV drives. The loser is usually pleased just to ‘show the flag’ and the winner, facing no challenge, rarely feels a need to mobilize voters since those who turn out for their own reasons can be relied upon to elect the candidate of the district’s majority party. Presidential elections, which aggregate votes by state, have a similar operational dynamic.

Off-Year Elections

The drop-off in turnout from presidential to congressional elections runs to about 12-17 percentage points, and turnout for municipal elections commonly runs as low as 20-25 percent (and even lower turnout rates are not unusual). Americans see a considerable difference in the importance of these different types of elections, and the money, effort, and media attention they draw reflect this assessment. Those with a lower sense of civic duty and less interest in public affairs are much less stimulated to vote by the mild attention off-year elections receive compared to the presidential year benchmark which attracts the most attention. Lower turnout is the result.

Long Ballots

Long ballots with candidates for justice of the peace, road inspector, and so forth depress turnout by encouraging ‘roll-off—the tendency of voters to stop voting as the choice gets farther down ballot. The longer the ballot, the greater the roll-off will be because it occurs with even relatively short ballots.

Multiple Elections

Although it cannot be proven, and the magnitude of the effect is unknown, there is a predisposition to believe that the frequency of our elections tends to depress turnout. The idea is that a consolidated ballot would probably produce a greater overall level of participation because elections would be less common and, thereby more intrinsically interesting to voters and commentators of all stripes. The evidence for this is indirect, but more than impressionistic.

In one study of multiple election effects, the citizens of Middlefield, Connecticut, had the opportunity during a 4-year period to vote in three general elections (1974, 1976, and 1978), two town elections (1975 and 1977), and a referendum in 1976 that was scheduled apart from the presidential election (Boyd, 1981). Turnout varied from a high of just under 80 percent for the 1976 election to a low of 41.3 percent for the referendum. Excepting the referendum, turnout was 69 percent or greater in every election. The interesting fact, for getting a sense for who votes, is that only 11 percent did not vote at any time in this period. Put differently, almost 90 percent voted in at least one of these elections. The core rate of non-participation was only 11 percent. Sixty-two percent voted at least five times; a third voted in all six elections.

A similar study across multiple elections in Kentucky produced results that are probably more similar to the national pattern. Sigelman et al. (1985) looked at ten elections, including primaries, state contests, a presidential race, and two congressional elections between 1978 and 1982. While the highest turnout recorded during the period was 50 percent for the 1980 presidential election, that fraction undercounted the share that participated at least once in the period. In Kentucky, very few were high-propensity voters: only 13 percent turned out for seven or more of the ten elections for which they could have voted. One the other hand, only 29 percent were unregistered and another 10 percent were registered but never voted. In other words, upwards of 60 percent voted at least once during the 4 years. This seems like only a modest improvement until one realizes that, as Middlefield had unusually high turnout rates, the people of Kentucky participate less than the national average. If the difference between national turnout and the turnout rate in Kentucky is adjusted, the result suggests that consistent non-voters may be as few as 30 percent—and the proportion of Americans who vote at least occasionally may be as high as 70 percent.

A plausible conclusion: core turnout in national elections may be about 35-40 percent, while another 30-35 percent may move in and out of the electorate depending upon the appeal of the election, canvassing efforts by parties or candidates, and institutional and personal restraints.

A caveat is in order: these data do not prove that turnout would be higher if Americans were called upon to vote less frequently. That acknowledged, it seems possible that the frequency with which Americans are called upon to vote may allow those with a weaker sense of civic duty or interest in politics to abstain in an election without feeling that they have been remiss in their duty as a citizen—producing, in any given election, a lower turnout rate than might be observed if elections were less common.

Postal, Absentee, and Holiday Voting

Sixteen states have conducted voting by mail since 1977. Four states—Alaska, Minnesota, Utah, and Washington—have expanded the practice to include partisan elections. Oregon is the only state to have used it for significant statewide special and partisan elections. In principle, postal voting should reduce the cost of voting to a near minimum and thereby increase participation. Oregon’s much publicized mail ballot to elect a senator had a very high participation rate for a special election: 57.9 percent for the December 1995 primary election and 66.3 percent in the January 1996 general. Similarly, Oregon turnout for the 2000 presidential election was 67 percent. These ‘by mail’ elections have been too varied by type and have occurred over too long a period to provide a good data-set for reliably estimating the impact of mail balloting on turnout. That granted, there is some reason to believe that mail ballots may increase participation in elections. Among two recent odd-year special elections conducted by mail the participation rate was 41.4 percent, compared to 39.7 percent for three equally recent odd-year special elections conducted at polls in the conventional manner. Whether that effect would generalize to other states, or produce higher than normal turnout in high-stimulus general elections, is unknowable.

Absentee ballot requirements have been sufficiently liberalized in most states and communities that the impact of absentee voting on participation rates has been dwarfed by the sheer numbers who frequently choose it as an alternative to in-person voting, not as a substitute for abstention. In some states (California is a good example), absentee voting is approaching 40 percent of the total vote cast in general elections. It is so common that absentee voters are almost representative of the electorate—or at least less distinctive than when absentee voting began to grow in the 1980s. Its probable effect on overall turnout at this time is modest to the point of insignificance. Early and ‘no-fault’ absentee voting in states such as Texas, Nevada, and New Mexico, for example, had no discernible effect on turnout in the data examined thus far.

Table 26.4 An illustration of party alignment
Groups Christian Labor Liberal Farmer Total
High-alignment party system
Religiously observant 80 5 5 10 100%
Working class 10 85 5 0 100%
Middle class 40 20 40 0 100%
Rural 5 5 5 85 100%
Language minority 10 60 20 10 100%
Low-alignment party system
Religiously observant 55 40 5 0 100%
Working class 30 55 15 0 100%
Middle class 35 25 40 0 100%
Rural 45 45 10 0 100%
Language minority 20 60 20 0 100%

Note: Table entries are hypothetical only, illustrating patterns common to highly aligned Western European party systems. No country or party system in particular is represented.

In several countries where turnout is substantially higher than it is in the USA, voting is done on weekends, or election day is a national holiday, and some have suggested that a similar arrangement in the USA would increase turnout. The effect seems unlikely. Countries with election holidays are different from the United States in so many other ways related to turnout (the registration of voters) that it is unlikely that American turnout is depressed by our practice of holding elections on weekdays.

The Party System

This has several dimensions. Most are fairly trivial, and some (the number of parties) are strongly related to the feature which does have an impact on voter turnout: the alignment between social groups and parties. Alignment refers to the degree to which (1) a party draws heavy support from particular groups and (2) supporters of a given party are homogeneous in their religion, class, ethnicity, or place of residence. An example of a party system with the maximum degree of alignment would be one in which the supporters of each party were drawn from a single religion, and all members of a given religion supported the same party. The parties in an unaligned party system draw supporters from every religion equally, and each religion is represented in each party in proportion to their occurrence in the society. Table 26.4 presents a hypothetical example.

High-alignment party systems have higher turnout rates because highly aligned parties stimulate social as well as political identities, and elections are occasions to support parties which exist as a political expression of a salient social distinction—being Catholic, French-speaking, working class—with which people identify. Disputes around social identities -and they will exist as parties coincide with social groups—draw people to the polls more easily than simple political identities such as being a conservative, a Republican, or a supporter of limited government (Powell, 1986; Lijphart, 1984; Teixeira, 1992). The American electorate has an even lower turnout rate than one would have expected from the weak alignment of groups with the parties, an ‘underperformance’ which reflects the multiple turnout depressing institutions and practices characteristic of the United States.

Campaign Styles

While civic leaders of every stripe encourage Americans to vote, candidates and political strategists occasionally make decisions which depress participation. In some cases the pressure to hold turnout down is intentional. For example, a special election is scheduled for mid-March because strategists have decided that the small number who are the most likely to turn out at such an odd time are more likely to support the issue or candidate than the electorate which will participate in a higher-turnout election. The long-term consequence for turnout rates of a history of such calculations is probably quite small, but it may help to depress turnout rates on average by making non-voting a more common experience.

Table 26.5 Campaign styles and turnout in Senate elections, 1992
Campaign Style 1 2
Predominantly Negative 49.7% 51.8%
Mixed 52.4% 50.3%
Predominantly Positive 57.0% 58.9%

Notes: Table entries are percentage of the VAP turning out for the election. The data for study 1 are from Ansolabehere and Iyengar (1996). Study 2 is from Martin Wattenberg (1996).

Potentially much more consequential is the attack style of campaigning that has become standard in American elections. Two recent studies show a difference in the turnout rates of aggressive attacking compared to ‘positive’ campaigns for the US Senate. However, as Table 26.5 indicates, the studies do not agree on the nature of the effect. Using ‘mixed campaign style’ states as a benchmark, one study shows lower turnout when the campaign style of the candidates is ‘negative’ and higher turnout when the campaign style is ‘positive’. The second study did not show an unusual decline in turnout associated with negative campaigning, but it did show higher than average turnout when the campaign was predominantly positive. Further, the studies disagree about the robustness of the findings. It is possible that the differences are trivial or even completely absent when other features of these campaigns are considered for their impact on turnout. In brief, the effect of campaign styles remains unclear, but there are data indicating an effect. What we know about the predisposition of voters toward political debate makes it at least plausible that attacking negative campaigns can create enough disaffection to depress participation.

The Media

The media became more critical of government, politics, and politicians after the mid-1960s than it had been in previous decades. Today, its tendency is to be more critical of incumbents than challengers, more critical of stronger candidates than it is of weaker candidates, and more critical of candidates who try to limit opportunities for reporters to watch them carefully.

We have only limited data on the effect of this media style on turnout, and no solid evidence that the critical tone of media coverage depresses turnout or that it has played a part in depressing turnout in recent years (although the decline in turnout and the rise of a critical media do correlate in the aggregate). However, we do have data demonstrating that higher media users are more likely to be critical of public officials and less likely to have a positive evaluation of the candidates of their party. Positive feelings toward candidates play a role in the vote choice; media-induced ambivalence toward the candidates (both—or all—candidates) in an election may influence the decision about whether they vote at all.

Office-Holding, the Separation of Powers, and Divided Government

The separation of legislative and executive elections in the United States (both nationally and in the states) is also believed to suppress turnout. Legislative candidates tend to their own election, usually assured because of incumbent-protecting district lines, and often do not consider driving turnout in their district to its maximum in order to provide more votes for a statewide candidate. The result, some speculate, is turnout levels which reflect the individual citizen’s enthusiasm for voting, a result which will always be lower than turnout which results from individual willingness to vote that is boosted by coordinated GOTV efforts by candidates.

When the separation of powers yields divided government, there is some evidence that the division may further depress turnout between 2-6 percentage points, depending upon how long the government has been divided between the parties. Only a few studies have been done on this feature of American politics, and the processes are not well understood, but it seems likely that separated offices reduces turnout-enhancing coordination between executive and legislative candidates. A period of divided outcomes may encourage legislative and executive candidates to separate their campaigns even further and, thereby, reduce the ability of mobilization efforts to identify and turn out potential supporters.

Table 26.6 Party canvassing and turnout, 1988 (percent)
Percentage of: Validated turnout and registration
Contacted by Sample Voters Voted Registered but did not vote Not registered
A party or candidate 24 31 78 8 13
Not contacted 76 69 54 11 35

Source: 1988 ANES.

Mobilizing Voters

This relative lack of coordinated mobilization of potential supporters seems particularly important, given evidence that voter mobilization has a substantial impact on turnout (evidence of this can be found in research from the 1960s, but for recent examples, see Huckfeldt and Sprague, 1992; Gerber and Green, 2000, 2001; Green et al., 2003). Of the three dimensions of campaigns -creating candidate awareness, creating candidate positivity getting voters to the polls—most contemporary campaigns place the least emphasis on getting voters to the polls whether measured by the amount of money or workforce effort expended on it. For instance, in the state legislative election study described below, on average, less than 10 percent of campaign expenditures are invested in GOTV but over one-fourth pays for mass media advertising. The best recent data on turnout (found in the 1988 ANES) show that only about 24 percent of voters remember any contact with a campaign worker (Table 26.6), while almost all remember some exposure to a campaign message.

But contact by campaign workers increases turnout. Among those who remember no contact, 54 percent voted—a figure that was virtually identical to an a priori estimate of who was the most likely to turn out (the turnout estimator as in Petrocik, 1991). By contrast, turnout was about 10 percentage points higher than expected (according to the prediction of the turnout model) among those who were contacted by some campaign. Other studies, specifically designed to estimate the effect of canvassing voters, found total turnout effects of as much as 6 percentage points from in-person canvassing and about 4 points from telephone canvassing.

In any given election, therefore, turnout will be a product of three factors: the aggregated individual equilibrium-level interest in voting, the enthusiasm for the race generated by the attention it receives, and tailored efforts to mobilize voters who are not sufficiently motivated to participate by (1) the excitement of the election and (2) their intrinsic interest in voting.

Competitiveness will not (by itself) increase turnout. Rather, competitiveness creates conditions—more candidate events, advertising, party and candidate contacting, and GOTV efforts—which increase the likelihood that people will be exposed to and drawn into the election (Cox and Munger, 1989; Gimpel et al., 2004a). Voters are analogous to sports fans: there are some diehards, but many only follow the game casually and these casual fans are much more apt to watch the games that generate the most publicity (i.e., the Super Bowl). If this game is supposed to be competitive, it will generate additional media coverage and social attention, which, in turn, create attention and interest among the public.

A Case Study: Missouri Legislative Elections

A sample of Missouri legislative elections offers an opportunity to further examine the effects of mobilization efforts on voter participation by assessing the effects of expenditures principally intended to get out the vote on turnout compared to the effects of campaign expenditures that are more intended to persuade. The data in Table 26.7 are drawn from the periodic reports of campaign receipts and expenditures filed by candidate committees. All campaign expenses over $100 must be itemized. Many House campaigns also identify the purpose of expenditures below this limit. Expenditure items were coded into one of six categories according to whether they were mobilization efforts, attempts to enhance name recognition, or attempts to control the campaign agenda and persuade partisans and independents to support one candidate over another (Endersby and Petrocik, 2001).

Table 26.7 Summary statistics for the Missouri case study (two major parties only)
Variable Mean Std. dev. Minimum Maximum
Individual Campaigns (n = 193)
Turnout for candidate (%) 20.746 6.853 6.386 49.907
Expenditures per capita ($) 0.468 0.422 0.000 03.970
Canvassing (%) 9.442 15.711 0.000 100.000
Direct mail (%) 34.703 23.856 0.000 89.137
Advertising (%) 25.508 24.055 0.000 93.087
Signs and appearances (%) 11.475 13.403 0.000 70.033
Campaign support (%) 15.065 15.990 0.000 94.565
Miscellaneous (%) 2.252 6.758 0.000 49.481
Off year Election (0,1) 0.492 0.501 0.000 1.000
Two party Elections (n = 104)
Turnout for election (%) 38.501 8.982 15.576 59.579
Expenditures per capita ($) 0.868 0.685 0.001 4.414
Canvassing (%) 10.255 15.527 0.000 100.000
Direct mail (%) 35.113 20.478 0.000 81.893
Advertising (%) 25.147 21.689 0.000 80.968
Signs and appearances (%) 10.522 10.646 0.000 69.185
Campaign support (%) 16.468 16.960 0.000 94.565
Miscellaneous (%) 2.495 6.822 0.000 49.481
Off-year election (0,1) 0.500 0.502 0.000 1.000

The category of campaign expenditures designed specifically to mobilize voters is canvassing. This includes expenses for traditional GOTV activities. Activities coded as canvassing include phone banks and door-to-door visits—interactions between the candidate or campaign staff and potential voters. Relevant costs include those for voter registration lists, door hangers, transportation expenses identified for GOTV, phone banks and charges, campaign staff to man them. Canvassing efforts are those intended to identify potential voters and encourage their participation.

Other campaign expenditures are less clearly identified as mobilization. Direct mail efforts to reach voters are less personal. Candidates may attempt to encourage partisans to vote; but they may also use direct mail to distinguish their records from those of their opponents. The literature on the effects of direct mail is not clear as to whether direct mail should influence turnout. Advertising includes expenditures for newspaper and radio advertising, as well as cable or other mass media. Messages intended for the general populace, however, typically boast of the relative merits of one candidate over another and usually cannot be directed to partisan supporters exclusively. Signs and appearances are more passive forms of communication. Yard signs, bumper stickers, and billboards can be used to enhance name recognition, but these are unlikely to either mobilize inactive supporters or control the public agenda. Campaign support includes any other identifiable expenditures to mount an election campaign. Most of these expenses pertain to maintaining a campaign office. Any remaining expenditures that cannot be identified fall into the miscellaneous category. Small costs, which are not itemized, are included as well as other unusual expenses, which do not fit into any other category.

A measure of a campaign canvassing effort can be calculated as the percentage of canvassing expenses to total expenditures (excluding transfers and cash on hand after the election). The percentage is a proxy for the type of campaign mounted by a House candidate. For instance, campaigns that try to mobilize voters should spend a higher percentage of funds on canvassing than those emphasizing name recognition or persuasion of independent voters.

The analysis is limited to major party candidates from 26 districts for the House of Representatives for four election cycles: 1992, 1994, 1996, and 1998 (two presidential and two off-year elections). The total sample of all elections, then, regardless of competition, is 104. The sample was created to be representative of House districts and reflected regional divisions of the state. A total of 193 Democratic and Republican candidates appeared on general election ballots in the sampled districts. Although many minor party candidates appeared on the ballot, these candidates have little likelihood of winning a House seat in Missouri and attract relatively small amounts of campaign contributions. Turnout and party competition vary significantly by House district.

Table 26.8 Turnout for candidate as a function of campaign expenditures, all Democratic and Republican state House candidates
All races (0-100% margin) Contested races (70-30% margins) Close races (60-40% margins)
Canvassing effort 0.115∗∗∗ 0.075∗∗∗ 0.083∗∗∗
(% of expenditures) (0.028) (0.023) (0.024)
Total expenditures 0.344 1.623∗∗ 0.904
(dollars per capita) (1.044) (0.792) (0.707)
Off-year election −5.967∗∗∗ −6.023∗∗∗ −5.761∗∗∗
(1 = Yes, 0 = No) (0.877) (0.678) (0.644)
Constant 22.436∗∗∗ 21.272∗∗∗ 21.466∗∗∗
(0.761) (0.590) (0.560)
R2 0.245 0.360 0.432
(Adjusted) (0.233) (0.348) (0.418)
S.E. 6.00 4.23 3.54
F 20.42∗∗∗ 29.82∗∗∗ 31.70∗∗∗
n 193 163 129

∗∗∗ Significant at the 0.01 level
∗∗ Significant at the 0.05 level
* Significant at the 0.10 level

The Influence of Expenditures on Turnout

There are several theoretically viable ways to measure an effort to mobilize the number of supporters in an election. The best compare the number of partisans to some measure of a normal vote. However, for state legislative elections, with low levels of partisan competition, no measure of a normal vote is at hand. Turnout, however, can still be evaluated. First, if the campaign’s mobilization efforts are successful, the number of voters casting ballots for the candidate should be higher. So one measure of campaign effects is the number of votes cast for a candidate divided by the population (eligible voters). Second, campaign efforts mounted by one or both campaigns may influence turnout on behalf of both candidates. Another measure, then, is the variation in turnout for both parties in the election. Below, the effects of one campaign’s expenditures are considered first on the ratio of votes cast for a candidate. Next, the percentage of campaign expenditures within each category for both campaigns is compared to two-party turnout.

Tables 26.8 and 26.9 report the results from multiple regression models predicting voter participation as a function of canvassing effort, considering the overall size of the campaign—total expenditures per capita. Table 26.8 shows results for individual campaigns; Table 26.9 does the same for the two-party election. These models are considered in three different contexts that may influence mobilization, recognition, or agenda-setting effects.

The first includes all campaigns, including uncontested elections. The second narrows the number of campaigns to ‘contested’ races; that is, weakly competitive elections where the candidate receives between 30 and 70 percent of total votes. The third reduces the range to ‘close’ elections, here defined as those in which a candidate receives between 40 and 60 percent of the vote.

Table 26.8 reveals that mobilization is significantly and positively related to greater numbers of partisan voters, controlling for the influence of per capita expenditures—which has a mild effect. This strong result holds across all classes of election environments—from all races to close races. If expenditures on canvassing increase a mere 10 percent, candidates should expect around a 1 percent growth in the number of supporters casting ballots at the polls. Although the level of significance falls slightly as elections become closer, the explanatory power of this simple model grows, accounting for less than a quarter to over 40 percent of the variance.

Table 26.9 Turnout in election as a function of campaign expenditures, all state House elections (two-party candidates)
All races (0-100% margin) Contested races (70-30% margins) Close races (60-40% margins)
Canvassing effort 0.125∗∗∗ 0.125∗∗ 0.162∗∗
(% of expenditures) (0.041) (0.052) (0.070)
Total expenditures 1.821* 0.349 −0.465
(dollars per capita) (0.939) (0.999) (1.447)
Off-year election −12.468∗∗∗ −11.660∗∗∗ −10.739∗∗∗
(1 = Yes, 0 = No) (1.282) (1.336) (1.617)
Constant 41.869∗∗∗ 43.777∗∗∗ 43.828∗∗∗
(1.189) (1.289) (1.641)
R2 0.502 0.519 0.487
(Adjusted) (0.487) (0.500) (0.462)
S.E. 6.43 5.84 6.08
F 33.61∗∗∗ 28.04∗∗∗ 19.01∗∗∗
n 104 82 64

∗∗∗Significant at the 0.01 level
∗∗Significant at the 0.05 level
∗Significant at the 0.10 level

Table 26.9 provides similar models for two-party elections. The canvassing coefficient remains positive and significant. The absolute magnitude of the effect grows, although the level of significance is lower than for corresponding models for individual campaigns. For general elections, an overall increase in canvassing efforts equal to 1 percent of total expenditures generates 0.125 percent more turnout. This seems particularly influential, given the multitude of variables affecting voter participation in general elections. Per capita campaign expenditures in the aggregate, however, do not appear to lead to higher turnout. The off-year indicator and canvassing explain approximately half of the variance in turnout in state legislative contests. In short, despite numerous difficulties with measurement error, we find strong evidence supporting the theory that mobilization efforts give rise to higher turnout.


Political science research typically shows that while both attitudes and institutions matter for turnout, institutions—such as election laws and organized efforts to mobilize voters—are more often causal. High turnout rates throughout the 19th century were maintained by election schedules, which coincided with the schedules of county courts and local fairs. Public balloting, in which voters verbally declared their preference before local registrars, also pressured those who were eligible to discharge their public duty and (often) satisfy promises and obligations to candidates and important figures in the local community who organized for the parties and the candidates. Most of this institutional facilitation of turnout came to an end with the Progressive era reforms of the 1890s, which, we generally believe, eroded the ability of the parties to mobilize voters.

The shift from ‘retail’ to ‘wholesale’ campaigning—coincident with the rise of ‘personal’ campaigns, television, and the reputed decline of the American political parties—has also mattered. Media campaigns broadcast to the masses shape perceptions of candidates but their contribution to getting out the vote on election day is modest.

Perhaps surprisingly, given the amount of scholarly interest in and attention to turnout, there is much that remains controversial or vague. Given numerous problems with measurement of the variables—difficulties with classifying campaign expenditures, measuring turnout and changes in levels of voter participation, and the lack of controls for party (and interest group) campaign efforts—significant results are difficult to attain.

The finding that campaigns that emphasize canvassing efforts, regardless of total expenditures, produce higher levels of turnout is discernible through the cacophony of white noise. The empirical evidence for the notion that mobilization efforts, at least canvassing efforts, can identify partisans and encourage them to show their support at the polls is strong.

This research into mobilization effects also suggests that the direction in which many contemporary campaigns have been oriented—that is, to greater expenditures on advertising and name recognition—may have exhausted its potential. Traditional means of increasing the number of votes by identifying and encouraging supporters to turn out may be resurgent. Parties have discovered that they may significantly increase their vote (regardless of whether the opposing party responds and there is no concomitant change in vote share). Also, more participation by voters satisfies a growing concern about civic responsibility and the legitimacy of election results.

These practical and normative concerns were noticed at the highest levels of Democratic and Republican campaign planners in 2004. Both the Democratic National Committee (with its ‘5104’ plan) and the Republican National Committee (with its ‘72 Hour’ plan) invested substantial resources in GOTV, canvassing, and grassroots programs for 2004, partly reflecting recent research of Gerber and Green (2001) and Huckfeldt and Sprague (1992), as well as by the success of Democratic group outreach programs in the 1998 and 2000 elections. The ultimate impact of these efforts, and the possibility that they will be institutionalized as we proceed into the 21st century, remain an open question.