Voter Turnout as a Dynamic
Process·
Abstract:
While there is a vast literature on turnout in the United States, there are two problems that undermine it as the source of accumulated knowledge positivists strive for. First, many of the explanations for declining voter turnout rely on statistical models that do not properly account for the dynamic nature of the turnout time series. Second, various authors’ explanations for why turnout has declined are contradictory, and several authors claim to fully explain what drives voter turnout using variables that other authors have not considered (Rosenstone and Hansen 1993, Teixeira 1987, Abramson and Aldrich 1982). This analysis, then, contributes to the voter turnout literature in three ways. First, factors affecting the aggregate-level turnout time series are considered by employing a transfer function model. Second, three of the most popular claims for the decline of voter turnout are tested in this model: declining party mobilization, lower social capital, and increased cynicism. I conclude that declining social capital and increasing cynicism have not caused turnout to decline, but that instead more conventional factors affect turnout dynamically – including the youthfulness of the electorate, the electorate’s strength of party identification, and political party mobilization.
Gregory A. Pettis
Political Science
University of North Carolina at Chapel Hill
Chapel Hill, N.C. 27599-3265
gpettis@email.unc.edu
What causes American voter turnout to rise and fall across time? With falling turnout causing concern for the health of American democracy, this has been one of the more popular questions addressed by political scientists. The strength of this literature is that a substantial amount of theory has been developed and tested. One weakness of this literature, however, is that researchers have done a poor job of addressing the conflicting claims in one another’s work. The effect is that multiple models exist claiming to explain “100 percent” of the decline (Rosenstone and Hansen 1993, Teixeira 1987, Abramson and Aldrich 1982 all model the decline using different variables yet all claim to explain either all or nearly all of the decline). The result is that, to the degree there is a decline in voter turnout (a point to be addressed later), we do not know what is causing it. After every single biennial election we as political scientists are forced to endure another barrage of commentary and angst from political elites bemoaning the declining turnout rate and Americans’ further disengagement from political life. A research effort that allows causal explanations to compete will allow researchers to better address these concerns.
I proceed by first constructing aggregate-level models of turnout. While individual-level analyses have been invaluable, they do not answer a central question – what moves aggregate turnout over time? In other words, what dynamically rising and falling attributes of the electorate cause turnout to rise and fall in kind? A multitude of attributes have varied substantially over long time periods, but only some of these will move turnout. This is a central concern of observers of American politics – of the dynamic attributes of the electorate that we hypothesize affect turnout, which actually do?
Second, a review of the turnout literature reveals a variety of possible causes of turnout. These possible causes will be tested against one another in the same models to the degree data will permit.
“The Puzzle of Participation”
Richard Brody (1978) coined the
phrase “the puzzle of participation” as a general term for the question of why
voter turnout has been declining since 1960 when several factors that drive
turnout upward have been increasing.
The work of political scientists trying to solve this puzzle is in the
tradition of seeking to provide a complete explanation for variation in voter
turnout rates, frequently focusing on the sources of voter turnout decline.
There have been various attitudinal and sociological changes in the American electorate since 1960 that have contributed to the decline, and almost all researchers agree on what these factors are. The electorate has become younger (Rosenstone and Hansen 1993, Teixeira 1987, and Shaffer 1981). The electorate has a weakened sense of social involvement and social efficacy (Putnam 1995, Rosenstone and Hansen 1993, Teixeira 1987, Abramson and Aldrich 1982, and Shaffer 1981). The electorate has less attachment to the political parties then it once did (Rosenstone and Hansen 1993, Teixeira 1987, Abramson and Aldrich 1982, and Shaffer 1981).
In the analysis presented here I test these causal themes against one another. A review of this literature suggests there are six categories of possible causes of voter turnout. First, there are historical causes that are structural to the series. These include the expansion of the franchise to women and 18 to 20 year olds, the two world wars, and midterm elections. Socioeconomic changes may also affect turnout – rising rates of education and income, as well as the aging of the electorate. Third, a dwindling reservoir of social capital may be depressing the electorate’s involvement. Aspects of elections themselves may stimulate or depress turnout; close elections may generate heightened interest, as may a third-party candidate who garners many votes. The structure of the political party system in the electorate may stimulate turnout, with turnout increased when party systems are more deeply rooted at the mass level. Finally, the level of cynicism toward government may affect turnout rates, with increasing distrust toward government and disinterest in public affairs causing the electorate to become more disengaged. In different works, using different data and different methods, political scientists have tested these various causal explanations. Here the effects of each explanation will be estimated, controlling for a common set of causally prior factors.
Historical/Structural Factors
There are a variety of causes of turnout that I consider to be a structural part of the series. In other words, there is either an ongoing institutional intervention that affects turnout, or a historical event has occurred that altered the series. The first of these factors is the most obvious – the effect of the structure of biennial elections, with Presidential elections staggered every four years. As is well known, Presidential elections draw far more eligible voters to the polls than the midterm Congressional elections, and the effect is that voter turnout decreases every four years when Presidential candidates are absent from the ballot.
A second factor is the effects of World Wars I and II. These both involved a major commitment of the American people, sending many people overseas where they could not vote, as well as possibly affecting the priorities of many people who remained in the country. Both wars caused a decrease in turnout (Dugan and Taggart, 1995).
A third factor is the expansion of suffrage. Suffrage has been expanded twice by amending the Constitution – in 1920 and 1971 (Welch et al., 1999). The first expansion of suffrage dramatically decreased turnout as millions of previously non-voting women were considered members of the eligible electorate (Dugan and Taggart, 1995). In time women were socialized to vote at the same rate as men, but this took decades. The second expansion occurred in 1971 when the voting age was lowered from 21 to 18. There was a less dramatic turnout drop in this case because the number of previously non-voting adults was lower.
Socio-Economic Change
Changes in the socioeconomic composition of the electorate have also been found to affect turnout. The first of these is age. It takes time for new members of the electorate to be socialized to vote, and the greatest cause of this is simply the act of voting. The more elections a person has voted in, the greater the likelihood the person will vote in future elections. Some theorize that as people age they learn more about politics, thus increasing their participation (Jennings and Stoker 1999). To a lesser extent, aging produces life cycle effects which encourage voting (Highton and Wolfinger, 2001). Turnout thus increases into middle age (Milbrath and Goel, 1977, Verba et al. 1995). The aggregate-level implication of this is that the younger the eligible electorate is, the lower turnout should be.
The second is income. In the modern survey era, at the individual level, income has been found to predict turnout (Rosenstone and Hansen 1993, Leighley and Nagler 1992, Wolfinger and Rosenstone 1980, Milbrath and Goel 1977). However, research suggests that this is a product of the modern political era, and that in the late 19th Century and early 20th Century turnout was not stratified by income (Kleppner, 1982). The question for this analysis is: what are the aggregate-level implications for income, given that aggregate earnings have been increasing throughout the 20th Century, as well as the degree to which turnout is stratified by income? These two combinations may suggest that increasing aggregate income actually decreases turnout. The size of the effect will depend on the degree to which turnout is stratified by income. If at the end of the 20th Century a small number of people are in the upper income brackets and they turnout at a much higher level than other incomes, then the aggregate effect of increasing income would be to depress turnout. If, however, at the end of the 20th Century income is evenly distributed (that is, the number of people in the various income groups was evenly distributed across the groups, or even bunched in the middle) and the stratification was minor (higher income people turned out at only slightly higher rates than other income groups), then the aggregate increase in income would not be associated with lower turnout.
The third is education. Another common individual-level effect is that higher levels of education cause higher levels of turnout (Leighley and Nagler 1992, Tate 1991, Wolfinger and Rosenstone 1980, Milbrath and Goel 1977). As with income, the aggregate-level implication of the great aggregate increase in education depends on the distribution of education and the degree to which turnout is stratified by education.
Social Capital
In his 1995 address to the American Political Science Association, Robert Putnam (1996) argued that the declines in voter turnout and memberships in voluntary associations are linked. The decline in both, Putnam argues, is caused by declining social capital. Social capital is, “features of social life – networks, norms and trust – that enable participants to act together more effectively to pursue shared objectives…. Social capital, in short, refers to social connections and the attendant norms and trust” (Putnam, 1996, 664). Social capital and civic engagement go together, and they both cause political participation. Since participation in voluntary associations is declining, Putnam argues, then social capital must be declining, and this must be causing the decline in voter turnout.
What is causing the social capital decline? Putnam quickly reviews an array of possible causes (I won’t list all of them here, but a few of them are more time pressures, residential mobility, or the 1960s), and he concludes that the rise of the electronic media is leading to the death of Americans’ desire to participate together. He argues television has displaced people’s leisure time, transforming it from a time spent with others to a time spent alone. Also, the more people watch TV the more cynical, negative and passive they feel. Finally, television impacts children especially, causing them to feel more aggressively and underperform at school (Putnam 1996).
In his book on the subject (Putnam, 2000), he notes that just as turnout is declining, so are interest in politics, attendance at public meetings, and identification with the political parties. While he points out the covariance, Putnam’s theoretical link between social capital and voter turnout is tenuous. He states that not turning out to vote is analogous to withdrawing from your community, but this is simply a statement and not an argument for why declining participation in groups and declining voter turnout are the same thing. What is missing is an argument about why the decline in social capital, voter turnout, interest in politics, attendance at public meetings, and identification with the political parties should all be linked. He does point out that voters have higher levels of social capital than nonvoters, indicating that turnout is an indicator of social capital. This may be true, but the theoretical justification for this being a causal relationship needs to be better explained.
One can certainly imagine why these are different phenomena and not caused by the same factors. Verba, Schlozman and Brady (1995) make this point when they note that the difference between the rate at which the rich and the poor contribute their time and money to politics is different than the differences in the rates at which they contribute to social or charitable causes. As they note, in politics the wealthiest are four times as likely as the poorest to give time to politics, but they are only twice as likely as the poorest to give time to charitable causes. In terms of giving time to religious activities there is no difference between the two groups. The same pattern is apparent with donating money – the rich donate far more than the poor to political causes, and this gap decreases when considering charitable causes. The social capital argument is a sociological one: declining participation in group activities is eroding social networks which causes a decline in interpersonal trust. If the same factors that cause people to join social groups caused people to participate in politics, there should be no difference between the rate at which people participate in political versus non-political activities. But we do observe a difference because the factors are not the same. One key difference between political participation and other participation is that participation in politics is a means to exercise power (and this probably explains the differences found by Verba et al.) Frequently groups mobilize their members to vote in elections to influence election outcomes. Exercising power does not cause social capital, but it does cause voter turnout. Social capital is a sociological phenomena, and while it probably has political implications, these implications must be carefully explored and not merely asserted.
Other authors have addressed the question of the relationship between social connectedness and political participation as well. Allison Calhoun-Brown (1996) found that, among African-Americans, attending church was a weak predictor of political involvement. What was a strong predictor was attendance at a “political church” where the sermons contained political messages. Other researchers found that, along the lines of Putnam’s argument, community attachment and church attendance both cause higher turnout rates (Strate et al. 1989).
Generally speaking, there may be a relationship between declining group participation, social connectedness and political participation. What is missing is a strong theoretical explanation of what that connection is.
Election
Effects
Researchers have
theorized that idiosyncratic factors pertaining to some elections (that is,
factors that vary from election to election) may increase voter turnout. One of these factors is how close the
election is. A variety of studies have
found that closer election outcomes are correlated with higher voter
participation (Hanks and Grofman 1998, Rosenstone and Hansen 1993, Cox and
Munger 1989, and Milbrath and Goel 1977).
The argument for why this would be true can be made in rational choice
or psychological terms. The rational
choice argument is that voters are more likely to participate in elections in
which they most expect to be decisive (Hanks and Grofman 1998). This argument relies on the classic calculus
of voting as outlined by Anthony Downs (1957) which includes as part of an
equation predicting if an individual voter will vote the probability that voter
will cast the deciding vote. The
psychological/cognitive argument (Milbrath and Goel 1977) is that voters are
more interested in close elections because close elections, because they are
competitive, are inherently interesting.
One can also imagine
why close elections would not increase voter turnout, and this too can be
explained from a rational choice or psychological/cognitive perspective. The notion that voters will consider if
their individual votes will be decisive underlies almost all rational choice
models of turnout (Green and Shapiro 1994).
Assuming that to some degree this is true, and assuming voters have an
accurate understanding of this probability, then an election that is close in
conventional terms should not increase turnout. From the psychological/cognitive perspective, the degree to which
the electorate follows politics varies depending on a host of factors, and we know
that even the simple expectation that an election is supposed to be close will
not reach large numbers of eligible voters.
Also, how close an election outcome is will not be known until after the
polls have closed.
Another factor that may affect turnout is a popular third candidate’s ability to mobilize voters and bring them to the polls. Little has been written about this, but it is possible to imagine that a third party candidate could benefit by appealing to some disaffected interests of a portion of the electorate that does not participate. By bringing in new voters, voter turnout can increase.
Political Parties
Political scientists and elites have identified a number of political party-related factors that may cause levels of voter turnout. The first of these is the strength with which voters identify with political parties. As researchers have argued, party identification is the degree to which one identifies oneself as a member of a particular political party. This thinking of oneself as a party member integrates one’s self-identity into politics, transforming distant political events into personal experiences (Campbell et al. 1960). The greater the degree to which eligible voters do this, the greater voter turnout should be. One of the most replicated findings in voter turnout research is that this relationship holds true (Rosenstone and Hansen 1993, Tate 1991, Abramson and Aldrich 1982, Shaffer 1981, Milbrath and Goel 1977, Verba and Nie 1972, Campbell et al. 1960).
Another political party-related factor that has received much attention is the degree to which parties mobilize voters, and several studies have found a relationship between voter contact and higher voter turnout (Wielhouwer and Lockerbie 1994, Rosenstone and Hansen 1993). As Rosenstone and Hansen argue, political party contacts should stimulate turnout because such contact increases voters’ feelings of efficacy toward voting. It also increases voters’ concerns about the election outcome. Thus, mobilization works indirectly: it ξfosters perceptions and beliefs that cause people to participate. Rosenstone and Hansen find that over half of the decline in voter turnout is attributable to the decline in party contact.
However, one can imagine another, deeper notion of political party mobilization that transcends a party member merely contacting a voter. Compared to today, in the 1950s and 1960s it was much more common for political parties to have permanent precinct-level organizations established in communities. These organizations were composed of community members who were strong party adherents and who took the responsibility during an election to mobilize their friends and neighbors to vote. As NES evidence indicates (in the form of the great decline in the number of survey respondents reporting having worked for a political party in the last election), this kind of mass, ongoing, grass roots mobilization has largely disappeared.
Why might this more permanent, community-based mobilization have greater effects on participation than simply political party contact? Robert Huckfeldt and John Sprague (1992) answer this question when they describe why party mobilization should increase participation. Their general argument is that the most effective party mobilization brings distant partisan politics to the community level by transforming politics from a distant phenomena to a local one. First, political parties occupy an important role in the social flow of political communication. Party efforts at electoral mobilization depend upon a process of social diffusion and informal persuasion, so that the party canvass serves as a catalyst aimed at stimulating a cascading mobilization process. Local, community-based political mobilization creates a layer of political structure within the community (and hence the electorate) that parallels the community’s social structure.
Community-based mobilization should be more effective than other types of mobilization, then, because community mobilization utilizes far more of a community’s social resources in the act of mobilization. While normal political party phone-based get out the vote campaigns employ strangers who have no social connections with voters, community-based mobilization uses social ties and connections to make political participation more meaningful. The meaning of such action is transformed from an act taken for the purpose of acting on some abstract symbol (out of a sense of civic duty or to express support for a certain candidate or policy) to an act taken for the purpose of reinforcing a social relationship.
Other attributes of political parties may affect turnout as well. With 26 Presidential elections held during the last century, one can imagine that the electorate found some elections to be more exciting than others. One way to capture this is to measure how excited about the candidates the electorate is, and one can imagine that electorates that are more excited about candidates vote at higher rates. Rosenstone and Hansen (1993) found that voters with greater affect for Presidential candidates voted at higher rates, and the same process may operate in the aggregate.
On the same note, the electorate, at different times, may care more or less about which party wins the election. An especially partisan time, or a time when the country is under duress and is especially attentive to election outcomes, may produce an electorate that cares more about the election outcome. Rosenstone and Hansen (1993) find that voters who care more about which party wins turnout to vote more, and one can imagine this effect operating in the aggregate as well.
Finally, the popular
press frequently commentates on how divisiveness and rancor within the Congress
turns the electorate off and depresses turnout. Such fighting appears petty and cheapens lawmaking. It is argued lawmakers should be able to
transcend their political differences and cooperate to govern most
effectively. Thus, the greater the
ideological differences between the two parties in Congress, the lower turnout
may be due to partisan rancor.
Cynicism
By far the media’s favorite explanation for the plummeting turnout rate and the end of all political life as we know it is increased cynicism. I am referring to two distinct concepts when I refer to cynicism – declining trust in government and declining interest in public affairs. The argument made by the media and accepted as truth by the public at large is that Vietnam and Watergate transformed a previously innocent, trusting electorate into one that has now been betrayed and knows never to trust again. As a result, this lack of trust has caused people to disengage from politics because of its distasteful nature.
Political scientists have investigated the effects of trust in government to find that lower levels of trust either don’t effect turnout or actually cause it to increase (Citrin 1974, Rosenstone and Hansen 1993, Timpone 1998). Both Citrin and Rosenstone and Hansen find that as people trust the government more they do not vote at higher levels. Timpone finds the relationship to be tenuous as well, except that for young voters in the 1980s it appears discontent actually mobilized participation.
There are good theoretical arguments to be made for both why trusting the government might or might not affect turnout. Apparently, in the last thirty years, political events have occurred that have shaken the faith the electorate has in representative institutions. We know that trust in government has declined. However, would this decline necessarily affect voter turnout? If declining trust has caused members of the electorate who would have voted otherwise to become disengaged, then declining trust is causing turnout to decline. Compared to non-voters, voters are more interested in politics and identify more strongly with a political party, but if trust really drives people to the polls more so than, say, either of these other factors, then declining trust would cause turnout to decline. It is also plausible that this is not the case, however. Declining trust may affect different voters differently, such that core voters (that is, voters who turnout on a regular basis) remain engaged in politics. The underlying issue may revolve around which of these attitudes is more psychologically proximate to the act of voting – attitudes toward parties and candidates, or attitudes toward political institutions?
While trust in government has been declining, interest in public affairs declined too (although to a lesser degree). The argument here for why this decline should affect turnout is the same as above – the electorate is learning to turn away from politics because it is a distasteful business. But, just as this may very well be true, there may also not be a connection between declining interest in public affairs and declining participation (Shaffer 1981). There are many reasons as to why this decline may not affect turnout. While interest in public affairs may be declining, other factors that cause turnout to increase may be increasing. Just as with trust in government, the decline in political interest may not affect core voters to the same degree it would affect peripheral voters. The important question to ask is: would voters’ declining interest in public affairs be the deciding factor that keeps them from going to the polls?
What is Voter Turnout, Anyway?
Before proceeding to the analysis, there is a final
theoretical question to handle – what is voter
turnout? In
its simplest form it is a fraction: the
number of people who voted divided by the total number of people who could have
voted. This concept is deceptively
simple because exactly how turnout should be calculated is not a settled issue,
and the differences in calculating it matter.
The standard turnout fraction is the number of
voters divided by total voting age population (VAP). Calculating turnout this way gives us the turnout series as
reported by the media: turnout begins
falling after 1960, with a steady linear trend downward.
Notes:
VAP series is the original, unadjusted turnout series, calculated by
total number of voters / VAP. VEP is
the adjusted turnout series, calculated by total number of voters / VEP.
But calculating turnout another way produces
different results. As reported by
Michael P. MacDonald and Samuel Popkin (2000), the value for the denominator of
the fraction should not be the VAP but the voting eligible population
(VEP). The main difference between the
VAP and the VEP is that the VEP excludes two groups that are ineligible to vote
but counted by the VAP – non-citizen immigrants and convicted felons. As recently reported by the Census Bureau,
the size of the US population increased at its greatest rate ever during the
previous decade due to immigration.
Immigration into the US began to increase dramatically starting in 1970,
and the increase in non-citizen immigrants is responsible for the majority of
the voters excluded from the VEP component of the turnout fraction. Also, convicted felons in most states are
not allowed to vote, and the number of convicted felons has tripled since 1984
(MacDonald and Popkin 2000). Excluding
these two groups from the VAP produces the VEP and a new turnout series with
less of a decline since 1960. See
figure 1. The analysis that follows
employs the corrected turnout time series.
Method and Data
The following analyses employ either an ARIMA model or
a time series regression employing OLS, with the corrected turnout time series
as the dependent variable. In cases
where results for OLS are reported the residuals were white noise, and in other
cases the results from an ARIMA model are presented in which the autoregressive
error structure of the series was corrected for. The list of variables is presented in table 1.
Table 1: Variables
Variable
|
Timespan
|
Coding
|
Dependent
Variable:
|
|
|
Adjusted
voter turnout
|
1900-2000
|
% VEP
Voting, 0-100%
|
Core
Model:
|
|
|
Suffrage
|
1900-1960
|
See
text and appendix
|
26th
Amendment
|
1972
|
1 in
1972; 0 otherwise
|
World
war
|
1918,
1942, 1944
|
1
during a world war; 0 otherwise
|
Midterm
|
All
midterm elections
|
1
during midterm, 0 otherwise
|
Education
|
1900-1996
|
% of
high school graduates attending college
|
Income
|
1900-2000
|
Percapita
GDP per worker
|
Youthfulness
of electorate
|
1900-1998
|
Proportion
of population 15-34 years of age
|
Election
Effects:
|
|
|
Presidential
election margin
|
1900-2000
|
Inverse
of natural log of margin of President’s victory
|
Congressional
seat difference
|
1902-1998
|
Inverse
natural log of difference in seats held by two parties
|
Third
party candidate vote
|
1900-2000
|
Percentage
of vote garnered by most popular third party candidate
|
Social
Capital:
|
|
|
Membership
in 32 national chapter-based associations
|
1900-1997
|
Higher
values indicate higher membership
|
TV
viewership
|
1950-1998
|
Percentage
of homes with TV multiplied by average hours per day watched
|
Party
Variables:
|
|
|
Party
identification strength
|
1952-2000
|
% of
strong party identifiers
|
House
ideological polarization
|
1900-1996
|
Difference
in 2-party DW-nominate scores for each Congress
|
Party
contact
|
1956-2000
|
%
electorate contacted by a political party
|
Party
work
|
1952-2000
|
%
electorate worked for a political party during campaign
|
Presidential
candidate salience
|
1952-1996
|
Number
of totaled likes/dislikes about both major party candidates
|
Care
which party wins
|
1952-2000
|
(Presidential
election years) Proportion who say
they do care which party wins
|
Cynicism:
|
|
|
Trust
federal government
|
1958-2000
|
Higher
values trusting the federal government
|
Interest
in public affairs
|
1960-2000
|
Higher
values interested in public affairs
|
Big
interests run government
|
1964-2000
|
Proportion
who believe big interests run the government
|
Government
wastes tax money
|
1958-2000
|
Higher
values believing the government wastes tax money
|
See
appendix A for the list of sources for the series and explanations for how they
were calculated. Some missing values
were calculated employing statistical assumptions.
Variables were
organized into several explanatory blocks.
The most basic is the core model, consisting of variables that should be
controlled for before the effects of other variables are considered. These are the effects of expanding the
franchise (both to women and younger voters), effects of world wars, the
effects of midterm elections, the effects of increasing education and income
(measured as the percentage of high school students going on to college and the
average percapita gdp per worker), and the effects of the electorate becoming
younger or older (measured as the proportion of the population between the ages
of 15 and 34).
Then factors relating to
election outcomes themselves can be considered. The impact of close elections on turnout should increase as
election outcomes are closer and closer.
Because of this, the inverse of the natural log of election differences
has been included in the analysis to reflect that the effect of close elections
should not be modeled linearly. The
percentage of votes received by the most popular third party candidate in a
given Presidential election is also included, and this effect is modeled
linearly.
Social capital
variables are measured in two ways.
First, Putnam argues that the decline in social capital is a decline in
community-level group participation, and so the membership rates of 32 national
chapter-based associations is included.
Putnam also considers the role television viewing has caused, and so a
television viewership measure is included that reflects both the average number
of hours per day Americans watch television as well as the proportion of
Americans with at least one television in their home.
Variables relating to
political parties are measured in a variety of ways. The strength of party identification is modeled as the percentage
of potential voters who identify strongly with one of the political parties. Party contact is modeled as the percentage
of the electorate that reports being contacted by one of the two political
parties. Party work is modeled as the
percentage of the electorate who worked for one of the two political parties
during the campaign. Presidential
candidate salience is modeled as the number of total likes and dislikes
mentioned about both major party candidates.
Finally, do you care which party wins is based on the proportion of the
electorate who indicate they do care which party wins.
The final block of variables to be considered is the cynicism block. The first variable is the level the electorate trusts the federal government, averaged over all of the response categories. Interest in public affairs is the level of interest the electorate has in public affairs, also averaged over all of the response categories. Big interests run the government is the proportion who believes big interests do run the government. Government wastes tax money is the level of tax money the electorate believes the government wastes, averaged over all of the response categories.
Findings
To the greatest extent possible
using this modeling technique and the data available, as many competing
variables were included in a single model as possible. However, for the survey-era variables, the
limited number of observations available made including all latter 20th
Century variables together in a single model impossible. As such, effects for political party
variables and cynicism variables were estimated in separate models, both of
which controlled for variables in the core model.
The
Entire 20th Century: Core
Model, Election Effects and Social Capital
The core model was
estimated along with election effects and social capital for the entire 20th
Century and results are presented in table 2.
From the core model we find suffrage, world wars, midterm elections,
rising education, and a younger electorate all cause lower levels of
turnout. We find that while rising
television viewership does not cause lower electoral participation, greater
participation in chapter-based organizations does. As for election effects, more votes netted by third party
candidates are associated with lower levels of turnout. Finally, greater ideological polarization in
the House is associated with lower turnout.
Table 2:
20th Century Model of Voter Turnout
OLS
R2:
.93 N: 49
Durbin-Watson d-statistic: 1.63
|
Variable |
Coefficient |
Standard Error |
P value |
|
Suffrage |
-18.2 |
2.4 |
0.00 |
|
26th Amendment |
0.04 |
3.1 |
0.99 |
|
World War |
-6.4 |
2.0 |
0.00 |
|
Midterm |
-11.6 |
4.2 |
0.00 |
|
Education |
-0.3 |
0.08 |
0.00 |
|
Income |
-0.0 |
0.0 |
0.82 |
|
Electorate Youthfulness |
-1.0 |
0.2 |
0.00 |
|
Group membership |
-10.7 |
2.5 |
0.00 |
|
Television viewership |
0.7 |
0.6 |
0.21 |
|
Presidential election margin |
16.6 |
26.4 |
0.53 |
|
Congressional election margin |
-7.1 |
8.4 |
0.40 |
|
Third candidate vote |
-0.2 |
0.1 |
0.01 |
|
House ideological polarization |
-34.3 |
9.6 |
0.00 |
|
Constant |
126.5 |
12.7 |
0.00 |
note: Durbin-Watson d statistic indicates
residuals are white noise.
Some of these results are as expected. The precipitous drop in turnout caused by suffrage is captured in the model, as is the effect of two world wars. The midterm effect is robust as well, but all of these are the predictable effects of history or institutional structure. The effect of the youthfulness of the electorate is important as a sociological variable which has a large effect in the direction theory predicts. Also, ideological polarization also depresses turnout, another finding informed by theory. The remaining findings are not consistent with theory, however. Greater rates of group membership should not cause voter turnout to go down. Social capital may not be related to turnout, but it certainly should not cause turnout to decline. Also, theory does not necessarily suggest why third party candidates garnering more votes would depress turnout. One possibility is that third party candidates garner their most support when the electorate is the most disaffected. An electorate disaffected with the status quo and unhappy with its options on the ballot would be both less likely to turnout and also more likely to support a third party candidate.
The
Later 20th Century: Core
Model and Political Parties
The core model was estimated along with survey-era variables measuring attributes of political parties, and results are presented in table 3. We find that while the proportion of the electorate contacted by political parties does not predict turnout, the proportion who worked as part of a campaign does. We also find that stronger partisan identification increases turnout, as does how salient the presidential candidates are to the electorate. After controlling for these factors, the degree to which the electorate cares which party wins does not predict turnout.
Table 3:
Later 20th Century Political Parties
ARIMA (1,0,0)
Log likelihood:
-33.21 N: 21
Wald chi2 (11) = 602.17 P
value: 0.00
|
Variable |
Coefficient |
Standard Error |
P value |
|
26th Amendment |
2.0 |
1.6 |
.22 |
|
Midterm |
3.5 |
7.5 |
.64 |
|
Education |
0.2 |
0.1 |
.22 |
|
Income |
-0.0 |
0.0 |
.93 |
|
Electorate youthfulness |
-0.5 |
0.1 |
.00 |
|
Contacted by party |
-0.2 |
0.1 |
.12 |
|
Worked for party |
1.5 |
0.5 |
.00 |
|
Strong pid |
0.4 |
0.1 |
.00 |
|
Presidential candidate salience |
5.9 |
1.8 |
.00 |
|
Care which party wins |
-14.9 |
13.0 |
.25 |
|
Lagged value |
-0.6 |
0.2 |
.00 |
|
Constant |
31.1 |
12.9 |
.01 |
note: The statistical significance of the lagged
value of the dependent variable indicates estimating the AR(1) parameter was
necessary to produce white noise residuals.
Hessian
standard errors were calculated to improve efficiency
With the reduction in cases from the model for the entire century many of the core variables lose significance (except for the youthfulness of the electorate). As theory suggests, simply being contacted by a political party does not predict turnout, but the proportion of the electorate that works in campaigns does. The earlier era when political party mobilization was truly a mass, community-based affair succeeded in motivating turnout in a way that modern mobilization efforts do not. The effect for strong identification is crisp as well, with the growing strength of party identification clearly causing higher turnout. These results tell us that stronger political parties, with deeper roots in the mass electorate, cause greater participation. The salience of the Presidential candidates matters as well, with campaigns that engage voters and cause them to think more positive and negative thoughts motivating greater turnout. So we see long-term factors like party mobilization and identification, as well as the election-specific factor, candidate salience, motivating participation.
The
Later 20th Century: Core Model and Cynicism
The core model was included with cynicism variables to assess their impact on voter turnout and the results are presented in table 4. Neither of the cynicism variables included in this analysis predict turnout, and neither of the other two cynicism variables are significant when included (results not reported).
Table 4:
Later 20th Century Cynicism
OLS
R2:
.96 N: 19
Durbin-Watson d-statistic: 1.55
|
Variable |
Coefficient |
Standard Error |
P value |
|
26th Amendment |
-1.7 |
2.7 |
0.54 |
|
Midterm |
-15.7 |
1.3 |
0.00 |
|
Education |
.3 |
0.2 |
0.20 |
|
Income |
-0.0 |
0.0 |
0.10 |
|
Electorate youthfulness |
-.8 |
0.4 |
0.05 |
|
Trust the federal gov’t |
.3 |
6.9 |
0.97 |
|
Interest in public affairs |
10.6 |
13.8 |
0.46 |
|
Constant |
78.8 |
34.9 |
0.05 |
Notes go here
The
cynicism variables fail to predict turnout to a stunning degree. Clearly the rising distrust of government
and declining interest in public affairs is not causing the electorate to
withdraw from politics. Turnout has
declined somewhat since 1960, but the story of an increasingly cynical pubic
withdrawing from politics is a myth.
The public is in fact more cynical, but this is not damaging electoral
participation.
Conclusions
What
should someone concerned about the health of the American electorate take from
this analysis? First, trends such as
declining participation in groups like voluntary associations, as well as
increasing distrust of government, are not associated with declining voter
turnout. Such arguments that they
should be associated appear persuasive at first glance, but a closer
examination forces one to specify what the precise links should be between
these trends and voter turnout. While
it is plausible that these trends could be causing a more politically
disengaged electorate, one less willing to participate, it is just as plausible
that this would not be the case. Going
to the polls on election day is not like joining a bowling league. And while rising cynicism might depress participation,
these results indicate it does not depress it because how cynical (or
optimistic) the electorate feels is not what moves people to the polls on
election day.
There are a number of other factors that do move people to the polls. First, aging. Previous studies have demonstrated that turnout increases as people grow older, and so as the electorate moves through cycles in which there are an unusually high number of young people, turnout will be depressed. Second, the more parties mobilize voters with community-based precinct organizations, the more successful mobilization efforts will be. It is not the mere mechanical act of contacting a voter that spurs them to vote as much as the importance of that act taking place in the context of a pre-existing social relationship. As Huckfeldt and Sprague argue, this social relationship transforms the meaning of the political contact. Greater identification with political parties also drives more of the electorate to the polls. As Campbell et al. argued, party identification personalizes the political contest for the voter by increasing how much voters care about the outcome. Election-specific factors matter as well, and in some elections voters are more engaged with the candidates. It is in these contests, when the candidates are the most salient to the voters, that voters express their opinions at the polls more often.
What keeps the
electorate from the polls? First is the
rising stratification of voters by education level. As Kleppner notes, at the end of the 19th Century
turnout differences did not exist along the lines of education or income. People with less education participated at
the same rate as those with more education because parties had substantially
greater institutional incentives to get voters to the polls. Also, education rates were much lower then
than they are now, meaning the disparities in education within the electorate
were much less. But party’s
institutional incentives to get the electorate to the polls have decreased
dramatically, thus increasing the importance of factors specific to voters to
motivate them to turnout. As
demonstrated time and again by individual-level analyses, one of the most
important of these factors is education.
This effect indicates that who votes is stratified by education,
reflecting both that those with more education are inculcated with civic values
and that the appeals political parties make are more attractive to better
educated people. Political party
leaders have presumably given up on making appeals to those with less
education, perhaps because they consider attempts to mobilize these voters as
unlikely to yield results. Parties
instead make post-materialist appeals that are more attractive to those with
more education. Thus, as the electorate
has become better educated on average, turnout has actually decreased because
parties draw better educated people to the polls at the expense of the less
educated members of the electorate.
This stratification leads to an aggregate negative effect of education
on turnout. Second is the degree to
which the electorate is dissatisfied with the choices of candidates. When such a situation arises, more voters
are likely to stay home, and those in the electorate who do vote are more
likely to vote for a third party candidate.
Finally, ideological differences between the two parties in the House
depresses turnout. This may have
several affects that depress participation.
An ideologically divided House may be more gridlocked, producing
frustration in the electorate. An
ideologically divided House may also conduct itself in a more divisive manner,
again causing the electorate to be frustrated.
References
Abramson, Paul R.,
Aldrich, John H. 1982. “The Decline of Electoral Participation in
America,” American Political Science Review, 76:502-21
Brody, Richard. 1978.
“The Puzzle of Political Participation in America,” in The New American
Political System, ed. Anthony King, Washington DC: American Enterprise Institute
Calhoun-Brown, Allison. 1996. “African American Churches and Political Mobilization: The Psychological Impact of Organizational Resources,” Journal of Politics, 58:935-953
Campbell, Angus, Converse, Phillip, Miller, Warren E., Stokes, Donald. 1960. The American Voter, Wiley
Citrin, Jack. 1974. “Comment: The Political Relevance of Trust in Government,” American Political Science Review, 68:973-988
Cox, Gary W., Munger, Michael C., “Closeness, Expenditures, and Turnout in the 1982 U.S. House Elections,” American Political Science Review, 83:217-231
Downs, Anthony. 1957.
An Economic Theory of Democracy, Harper
Dugan, William E.,
Taggart, William A. 1995. “The Changing Shape of the American
Political Universe Revisited,” Journal of Politics, 57:469-82
Green, Donald P., Shapiro, Ian. 1994. Pathologies of Rational Choice Theory, Yale University Press
Hanks, C., Grofman, B. 1998. “Turnout in Gubernatorial and Senatorial Primary and General Elections in The South 1922-1990: A Rational Choice Model of the Effects of Short-run and Long-run Electoral Competition on Relative Turnout,” Public Choice 94:407-21
Highton, Benjamin, Wolfinger, Raymond E. 2001. “The First Seven Years of the Political Life Cycle,” American Journal of Political Science 45:202-09
Huckfeldt, Robert, Sprague, John. 1992. “Political Parties and Electoral Mobilization: Political Structure, Social Structure, and the Party Canvass,” American Political Science Review 86:70-86
Jennings, M.Kent, Stoker, Laura. 1999. “The Persistence of the Past: The Class of 1965 Turns Fifty,” presented at the Midwest Political Science Association
Kleppner, Paul. 1982. Who Voted? The Dynamics of Electoral Turnout, 1870-1980, Praeger Publishers
Leighley, Jan E., Nagler, Jonathan. 1992. “Individual and Systemic Influences on Turnout: Who Votes? 1984,” Journal of Politics, 54:718-40
MacDonald, Michael P.,
Popkin, Samuel. 2000. “The Myth of the Vanishing Voter,” presented
at the Midwest Political Science Association, 2000
Milbrath, Lester W.,
Goel, M.L. 1977. Political participation: How and why do people get involved in
politics? University Press of America
Putnam, Robert D. 1996.
“Tuning In, Tuning Out: The
Strange Disappearance of Social Capital in America,” PS 28
Putnam, Robert D. 2000.
Bowling Alone: The Collapse and Revival of American Community,
Simon and Schuster
Rosenstone, Steven J.,
Hansen, John Mark. 1993. Mobilization, Participation and Democracy
in America, Allyn and Bacon
Shaffer, Steven D. 1981. “A Multivariate Explanation of Decreasing Turnout in Presidential Elections, 1960-1976,” American Journal of Political Science 25:68-95
Strate, John M., Parrish, Charles J., Elder, Charles D., Ford, Coit. 1989. “Life Span Civic Development and Voting Participation,” American Political Science Review 83:443-464
Tate, Katherine. 1991. “Black Political Participation in the 1984 and 1988 Presidential Elections,” American Political Science Review 85:1159-1176
Teixeira, Ruy A. 1987.
Why Americans Don’t Vote:
Turnout Decline in the United States, 1960-1984, Greenwood Press
Timpone, Rich. 1998.
“Structure, Behavior and Voter Turnout in the United States,” American
Political Science Review 92:145-58
Verba, Sidney, Nie,
Norman H. 1972. Participation in America: Political Democracy and Social Equality,
Harper and Row
Verba, Sidney,
Schlozman, Kay Lehman, Brady, Henry E.
1995. Voice and Equality:
Civic Voluntarism in American Politics, Harvard
Welch, Susan, Gruhl,
John, Comer, John, Rigdon, Susan, Steinman, Michael. 1999. American
Government, West Wadsworth
Wielhouwer, Peter W., Lockerbie, Brad. 1994. “Party Contacting and Political Participation, 1952-90,” American Journal of Political Science 38:211-229
Wolfinger, Raymond, E.,
Rosenstone, Steven J. 1980. Who Votes, Yale University Press
Appendix
A
Note:
In some cases, observations were missing from the series and they were
calculated assuming a linear trend between the last observation and the next
observation. This was only done in
series in which the author noted a stable linear trend apparent across the
entire series. Further explanations are
provided in the notes for the individual series.
Variable: Voter turnout, 1900-2000
Source: 1900-1946, Walter Dean Burnham, Democracy in the Making:
American Government and Politics (1986); 1946-1998, MacDonald and Popkin
(2000); 2000, correspondence with
MacDonald
Variable: Suffrage, 1900-1962
Definition: 1900-1918, the proportion of the U.S. population living in a
state where women had full suffrage.
1920 is coded as 1. 1920-1962 is
a decay function, modeling the decaying impact of suffrage. The decay rate is 80% of the previous value.
Source: 1900-1918, http://women.eb.com/women/mapdocs/suffra002d4.html,
Women in American History Brittanica Online.
Variable: 26th Amendment
Definition: 1 in 1972, 0 otherwise
Source: The 26th Amendment was ratified in 1971; Welch et al.,
American Government: Seventh Edition (1999)
Variable: World war
Definition: 1 in 1918, 1942 and 1944; 0 otherwise
Source: Welch et al., American Government: Seventh Edition (1999)
Variable: Midterm
Definition: 1 for midterm elections, 0 otherwise
Variable: Education, 1900-1996
Definition: The percentage of high school graduates attending college
Sources: 1900-1970 education; Historical Statistics of the United
States: Colonial Times to 1970
1972-1990 education; 120 Years of
American Education: A Statistical Portrait
1990-1996 education; Statistical
Abstract of the United States (1999); Section 31 20th Century Statistics
Variable: Income, 1900-2000
Definition: Per capita gdp of workers
Source: J.Bradford DeLong, “Cornucopia: Increasing Wealth in the
Twentieth Century” (2000) working paper, Berkeley econ department
Variable: Youthfulness of electorate, 1900-1998
Definition: Proportion of electorate 15-34 years of age
Source: 1900-1970, Historical Statistics of the United States:
Colonial Times to 1970, Series A 29-42
1972-1998, Statistical Abstract of
the US
1973 no.35, SAUS
1975 no 35, SAUS
1984 no. 30, SAUS
1999 no. 14, SAUS
Variable: Presidential election margin, 1900-2000
Definition: The inverse of the natural log of the difference in the raw votes
received by the two major party Presidential candidates.
Source: Congressional quarterly, 1900-1992.
1996, 2000: CNN election website.
Variable: Congressional seat difference, 1902-1998
Definition: The inverse of the natural log of the difference in the number of
seats held by the Democrats and the Republicans after each Congressional
election.
Source: 1902-1990: CQ, 1994-98: Clerk
of the U.S. url:http://clerkweb.house.gov/elections/1994/94Recapi.htm
Variable: Third party candidate vote, 1900-2000
Definition: The number of votes received by the highest vote earning third
party candidate
Source: 1900-1984, Elections American Style (1987), The Brookings
Institution
1988-2000,
CNN.com website
Variable: Membership in 32 national chapter-based associations, 1900-1998
Definition: The average rate of membership in these groups.
Source: Correspondence with Robert Putnam.
Variable: TV viewership
Definition: The number of average hours of television watched multiplied by
the proportion of US homes with television
Source: For average number of hours of television, correspondence with
Robert Putnam. For proportion of US
homes with television, Statistical Abstract of the United States
Notes: Some missing observations for average number of hours of television were calculated assuming a linear trend between observed time points. There were observations for 1950,54,60,64,66,70,72,76,80-98. At no point did the observed values move in a non-linear way.
Variable: Party identification strength, 1952-2000
Definition: The percentage of respondents identifying strongly with either of
the political parties
Source: ANES
Variable: House ideological polarization, 1900-1996
Definition: Difference in DW-nominate scores for Republican and Democratic
parties in each Congress
Source: Keith Poole website
Note:
The series generated by taking this difference was compared to the
series for the difference between ADA scores for the two parties from the 1940s
to 1996. The series were very similar.
Variable: Party contact, 1956-2000
Definition: The percentage of respondents who had been contacted by a
political party
Source: ANES. For 1958, 1962;
Gallup poll retrieved from a Roper search.
Note:
1970 value calculated assuming a linear trend from 1968 to 1972.
Variable: Party work, 1952-2000
Definition: The proportion of respondents who report they worked for a
political party during the past election
Source: ANES. For 1954, 1958,
1966 from Roper search, all Gallup results.
Variable: Presidential candidate salience, 1952-1996.
Definition: Total number of likes/dislikes about both major party candidates
for President
Source: ANES
Variable: Care which party wins, 1952-2000
Definition: During Presidential election years, the proportion of respondents
who state they care which party wins
Source: ANES.
Variable: Trust federal government, 1958-2000
Definition: Average over four response categories for how often respondents
can trust the federal government in Washington to do what’s right
Source: ANES
Notes:
1960 and 1962 were missing values so I calculated them assuming a linear
relationship from 1958 to 1964. The
values for 1958 and 1964 were the same so I assumed that if the variable had
been measured in 1960 and 1962 those values would have been the same as
well. The decline in the variable
occurred later.
Variable: Interest in public affairs, 1960-2000
Definition: Average over four response categories for all respondents.
Source: ANES
Note:
1970 was missing, so I calculated it assuming a linear trend from 1968
to 1972.
Variable: Big interests run government, 1964-2000
Definition: The proportion of respondents who believe big interests run the government
Source : ANES. For 1986, from Roper search, NYT, Dec. 1985.
Variable: Government wastes tax money, 1958-2000
Definition: Average over three response categories for all respondents.
Source: ANES. For 1986, from
Roper search, NYT, Dec 1985.
Note: Values calculated for 1960, 1962 and 1966 assuming a linear trend between adjacent time points.