Todd Donovan Caroline Tolbert Competitive Elections and Voter Participation: Mobilizing Turnout of the Less Engaged Abstract Existing research largely focuses on reducing the costs of voting as a means to affect turnout. We provide theoretical and empirical support for the idea that multiple forms of competitive elections increase turnout, and that competitive elections have stronger mobilizing effects on a distinct set of citizens. By stimulating interest among people who are less engaged with politics, electoral competition has a greater propensity to mobilize the young, and those with less formal education. We demonstrate the differential mobilizing effects of exposure to presidential, congressional and issue elections, and suggest limited exposure to competitive elections may be one reason for lower levels of turnout recorded since the 1960s. Competitive elections may be a process that affects the existing bias in who votes in America.
39
Embed
Todd Donovan Caroline Tolbert Competitive Elections and ...
This document is posted to help you gain knowledge. Please leave a comment to let me know what you think about it! Share it to your friends and learn new things together.
Transcript
Todd Donovan
Caroline Tolbert
Competitive Elections and Voter Participation: Mobilizing Turnout of the Less Engaged
Abstract
Existing research largely focuses on reducing the costs of voting as a means to affect turnout. We
provide theoretical and empirical support for the idea that multiple forms of competitive
elections increase turnout, and that competitive elections have stronger mobilizing effects on a
distinct set of citizens. By stimulating interest among people who are less engaged with politics,
electoral competition has a greater propensity to mobilize the young, and those with less formal
education. We demonstrate the differential mobilizing effects of exposure to presidential,
congressional and issue elections, and suggest limited exposure to competitive elections may be
one reason for lower levels of turnout recorded since the 1960s. Competitive elections may be a
process that affects the existing bias in who votes in America.
1
Electoral Competition and Voter Participation
Introduction
A large body of theory and research has improved our understanding of demographic and
attitudinal characteristics that distinguish voters from non-voters (for reviews see Wolfinger and
Rosenstone 1980; Rosenstone and Hansen 2003). Likewise, scholars have identified the
important effects of institutions such as registration laws on turnout (e.g. Highton and Wolfinger
1998; Nagler 1991; Squire, Wolfinger and Glass 1987). Election reform efforts in the United
States have focused primarily on changing rules to ease voter registration and make voting more
convenient, with the explicit goal of increasing turnout (Highton 1997). Yet some literature
suggests convenience voting reforms, such as early voting, may fail to significantly increase
turnout or alter the demographic composition of an electorate (Berinsky 2005; Fitzgerald 2005;
Karp and Banducci 2000). Why? Beyond individual-level demographics, attitudinal factors, or
electoral rules, what contextual factors affect turnout and the composition of the electorate?
An important contextual factor that is often overlooked in analysis of turnout is
competitive elections. When elites spend more effort and resources to contest elections, more
information becomes available to voters. The resource-laden environment associated with
electoral competition may reduce information costs of voting for individuals (especially those
with less interest), leading to higher turnout. Conversely, if elections are uncompetitive or
uncontested, they generate little political information. Absent active campaigns, individuals may
have fewer opportunities to become interested in a contest, and may have less incentive to vote.
We propose that disinterest in politics may be a significant barrier to voting and that competitive
elections generate political interest. A number of studies have identified effects of electoral
competition on turnout (e.g. Patterson and Caldeira 1983; Cox and Munger, 1989; Jackson 2002,
2
1996, 1997; Holbrook and McClurg 2005). We build on such work and propose a theory to
explain how electoral competition can have a more pronounced mobilization effect on a distinct
category of citizens.
We argue that competitive elections may increase demand for participation, especially
among individuals who may otherwise be less likely to vote. We take the electoral context that a
citizen resides in seriously, and test if citizens mobilized by active political campaigns are more
likely to vote. We examine how state-level and congressional district-level electoral competition
interacts with individual-level characteristics to affect participation, and demonstrate that
variation in the competitiveness of several types of elections affects the composition of the
electorate. We propose that the mobilizing effects of a state's electoral context have different
effects on specific categories of citizens. We expect the mobilizing effects of competitive
campaigns are contingent on an individual's level of political interest. We propose that citizens
with higher education and older citizens, respectively, having greater political interest, are more
likely to vote regardless of the mobilizing effects of campaigns. Younger citizens, and the less
educated, having less political interest, may be more likely to be affected by the mobilizing
effects of campaigns.
One limitation of previous turnout research is that much is based on aggregate data, or
individual-level survey data, but seldom both.1 Studying one level of analysis in isolation may
mask the effects of variation in contextual factors on individual-level turnout decisions. We
contend that geographic context matters since there is a great deal of variation in campaign
activity across place. We examine the effects of individual characteristics and contextual factors
simultaneously and interactively, and thus move beyond previous research on multiple fronts. By
drawing on large sample Current Population Surveys (CPS) and the 2006 Cooperative
Comparative Election Study (CCES), rich measures of electoral context, and sophisticated
3
statistical modeling, our research provides a rigorous diagnosis of who turns out in American
elections.
Electoral Competition and the Composition of the Electorate
It is reasonable to expect that campaigns mobilize voters, and that this might affect the
composition of the electorate. Yet there is limited research on effects of electoral competition on
the composition of electorates (for exceptions, see Hill and Leighley 1994; Brians and Grofman
1999; Holbrook and McClurg 2005). Party mobilization efforts (Rosenstone and Hansen 2003;
Gerber and Green 2000) are known to be associated with higher voter turnout in the U.S. and
other democracies. A large body of cross-national research also demonstrates the consistent
effects that closely contested elections have on increasing voter turnout (e.g. Blais and
Dobrzynska 1998; Franklin 2004). Blais (2006: 60) finds that closeness predicted turnout in 27
of 32 studies testing for the effect, yet many individual-level models of turnout in the U.S. give
the mobilizing effects of elections limited attention (but see Jackson 2002, 1996; Holbrook and
McClurg 2005). Although we know electoral activity can increase turnout we know much less
about who is mobilized, what they are mobilized by, and why they are mobilized. We suggest
that the mobilizing forces of electoral competition affect the composition of electorates by
generating interest in elections among the young and less educated. This argument is somewhat
similar to Campbell's (1966) "surge and decline thesis" which proposes that highly salient
(Spending U.S. House Races) + β5 (Percent Uncontested U.S. House Races) + β6 (Number/Spending Ballot Initiatives) + β7 (Closing Date Voter Registration) + β8 (Educational Attainment) + β9 (Percent black) + β10
(Percent Latino) +ε
An advantage of multilevel data is the ability to investigate cross-level hypotheses or
multilevel interactions. In our case, we are interested in how exposure to competitive elections
affects voter turnout for people at different ages and at different levels of education. Our models
include two additional random effect components, denoted as εu1 and εu2 above, to model
interactive effects of age and education. We hypothesize the effects of age and education on the
probability of voting may vary, depending on levels of exposure to information associated with
competitive elections. We allow the covariates for individual-level age and education to vary
across the state contextual (level 2) variables. These random effects interact age and education
with all the state electoral competition variables simultaneously. We thus avoid collinearity
problems that can be induced by multiple interaction terms.20 We also estimate cross-level
interactions, directly interacting the education and age of the respondent with campaign spending.
Table 1 presents evidence that exposure to competitive elections of all sorts increases the
probability of voting at the individual level. The effects of closer gubernatorial and U.S. Senate
races, as well as having ballot initiatives, both appear most pronounced at increasing turnout in
midterm elections. This makes sense given that the stimulus of presidential elections may swap
the mobilizing effects of down-ticket races in presidential years. While presidential years have
higher aggregrate mobilizing effects, we also see state-level presidential electoral competition
affects turnout in the presidential contest. Residing in a state with more ballot initiatives
increased the probability of voting in all years - this result holds even when the analysis is
constrained to only those states having ballot initiatives. (not shown).
12
Table 1 and Figures 1-3 about here
State laws regulating voting (closing date) were found to predict turnout in the
presidential race (when a greater volume of previously less-interested voters may be mobilized),
but not in the midterm election. This suggests that when the overall mobilizing effects of
elections are comparatively weak (e.g. non-presidential years), regulatory barriers to voting
(closing date) may have muted consequence on turnout. Competitive elections, in contrast,
appear to induce turnout in both midterm and presidential election years. Individual-level
predictors of voting were also in the expected direction in each election: higher educated,
women, wealthier and older citizens were more likely to report voting. Other factors held
constant, African Americans voted more than non-Hispanic whites, while Latinos and Asian
Americans had a significantly lower probability of voting. Residential stability, government
employment, veteran status, and higher status occupations were each associated with an
increased the probability of voting.
Coefficients at the bottom of Table 1 report the random effect components for each
model in the three election years, including the level 1 intercept and error terms for individual
level age and education (allowed to randomly vary across the state contextual variables). We see
the Chi-Square test is statistically significant for each of these components, indicating that the
effects of age and education on the probability of voting do vary significantly with exposure to
competitive elections at the state level.21
We translate some of our MLM estimates from Table 1 into probability simulations of
reported turnout to demonstrate how the effects of competitive elections on individual turnout
varies by a respondent's education and age. Simulations are displayed as graphs that illustrate
the probability of an individual voting at different levels of electoral presidenial and senate
competitiveness, holding all other variables in the model constant at their mean/modal values.
13
Figure 1 illustrates the probability that individuals with different levels of education reported
voting in 2004, predicted across state levels of presidential competitiveness. The figure
illustrates that residence in a state where the presidential race was more competitive had a greater
effect on turnout for people with less education than people with a college degree. Another
simulation (not reported) shows a similar effect for age: living in a competitive presidential state
had a disproportionate effect on turnout among younger voters. The capacity for a competitive
election context to change a marginal non-voter into a marginal voter can also be seen in Figure
1. Someone with a 10th Grade education residing in the least competitive presidential state in
2004 had a .46 estimated probability of reporting she voted, compared to a .55 probability for an
identical respondent in the most competitive state; a .09 difference.
Figure 2 displays the disproportionate effects of increased spending in Senate races on
the probability of voting in 2002, by age. A 33 year old living in the state with the most
expensive Senate race had a .40 greater probability of reporting voting than an identical 33 year
old living in a state with the no spending on a Senate race. Compare this to the predicted effect
of Senate spending on a 64 year old; the difference between exposure to the most expensive
Senate race versus no exposure to a Senate race was a .25 increase in the probability of a 64 year
old reporting voting. Electoral competition mobilizes the older and younger voter, but the effect
is much more substantial for the 33 year old.
Figure 3 plots similar disproportionate substantive effects of the competitive Senate race
by levels of education. Low educated individuals residing in a state with the most competitive
Senate race had a significantly higher probability of voting in 2002 than low educated people in
the least competitive states, all else held constant. Again, highly competitive Senate races
stimulate turnout of the less educated and the more highly educated in a midterm election, but
the effect is more substantial for the less educated. These data suggest that exposure to
14
competitive presidential and congressional races may not only increase turnout, but may also
subtly alter the composition of the participating electorate by turning young or less educated
marginal non-voters into marginal voters. The effects of each form of electoral competitiveness
were estimated from additive models and thus may be cumulative. When considered on a
national scale, these effects may have the capacity to mobilize millions of voters.
Effects of Presidential Campaign Activity on Voting
Some of our results from Table 1 might be critiqued for failing to account for campaign
activity directly. After all, we assume that less interested voters are mobilized by the activity
associated with actual campaigns (rallies, commercials, direct mail, local media coverage of
candidate activities, etc.), not by perceptions of how close elections may be. We address this by
replicating our multilevel models, but in place of the presidential vote margin and number of
initiatives we substitute three direct measures of campaign activity that were available from the
2004 campaign: 1) number of presidential and vice presidential candidate visits to a state, 2)
number of presidential campaign television advertisements in a state, 3) presidential campaign
television spending per capita in a state, and 4) spending on ballot initiatives. Table 2 illustrates
that increased exposure to this campaign activity significantly increased the probability
respondents reported voting, particularly for the young and less educated. The Chi-Square tests
for the level 1 intercept and error terms for individual-level age and education again indicate that
the effects of age and education on the probability of voting vary significantly with exposure to
each type of state-level presidential campaign activity. Results are very similar whether actual
campaign activity (Table 2) or presidential vote margin (Table 1) are used to predict turnout.
Table 2 and Figure 4 about here
15
Figure 4 graphs the probability of voting simulated from Table 2 to illustrate the
disproportionate mobilization effects of presidential and vice presidential campaign visits on
turnout for voters of different ages. It demonstrates that an 32 year old living in a state with no
visits (and no corresponding media coverage of visits) had a .69 probability of reporting she
voted, whereas an identical respondent living in a state with the most visits (Ohio) had a .76
probability (an increase of .07). For a 58 year old, the increased probability of reporting voting
associated with residence in the most visited state was .05. For older voters, the mobilizing
effect of the presidential campaign is even less. Plots of the effect of the other campaign
measures from Table 2 (not reported) produce similar results.
The last column of Table 2 includes a measure of combined spending in all federal races
per capita (presidential, U.S. Senate and U.S. House) in the respondent’s state, and a measure of
per-capita campaign spending on ballot initiatives. These spending variables are interacted by the
education of the respondent as an additional test. We see that candidate spending and ballot
measure spending in the respondent’s state each significantly increase turnout of the low
educated (see base term for the two spending variables, which is the effect of increased federal
candidate and initiative spending when education is set to low— Jaccard et al 1990). This is
consistent with our hypothesis that electoral activity has a larger effect on turnout among people
who have less interest in politics.
Effects of Competitive Elections on Sub-samples of the Young and Less-Educated
To demonstrate the robustness of these MLM findings, we replicate our analysis with
logit models estimating turnout among CPS sub-samples of the youngest age quartile (18 to 32
years) and the less-educated (high school graduates and below). These are compared to sub-
samples of the well-educated (college graduates and above) and older voters (58 and older, top
16
quartile), respectively. Full results of these models are available from the authors, but key
coefficients for the age sub-samples are reported in Table 3a. Comparisons of the size of
coeffiecients across the first two columns of Table 3a demonstrate that spending on ballot
initiatives, Senate races, and competitive gubernatorial campaigns have larger estimated effects
on turnout of younger voters than older voters in the midterm election. We find similar patterns
when estimating logit models that compare sub samples of the less educated to the more highly
educated (not shown). Results in Table 3a also suggests that these forces operated differently in
a presidential election year. The coefficient for closeness of a state's presidential contest is larger
for younger than older voters, but close gubernatorial races and more ballot measures in the
presidential year appear to have mobilized older voters only.
Tables 3a, 3b and 3c about here
We translated some of the logit estimates from Table 3a (and from our estimates
comparing the less educated to the more highly educated) into the predicted probability that a
person voted in order to illustrate the differences across these subgroups. Estimates using this
method suggest that residence in a highly competitive presidential state increased the probability
of voting in 2004 for the less educated by 8% and the higher educated by just 2% (Table 3b).
Similarly, residence in a highly competitive presidential state increased turnout among the young
by 11%, while state presidential competitiveness had no relationship with the likelihood that an
older respondent reported voting in 2004 (Table 3c). Although the high predicted probability of
voting for older and higher educated voters across all levels of electoral competitiveness may
suggest a ceiling effect associated with the logit model (e.g. Nagler 1991: 1402), results from the
models disaggregating by sub samples (in Table 3a) demonstrate real substantive differences in
the effects of electoral competition across the sub-samples, as does the significant effect of the
cross-level interaction term for exposure to competition multiplied by education (in Table 2).
17
In short, multiple modes of estimation here demostrate that electoral competition has a
disporportionate effect on mobilizing the young and less educated, a result consistent with our
assumtion that competition stimultates interest, which stimulates participation. But there is a
rival hypothesis. Key (1949) proposes that close elections may increase the incentives that
parties have to appeal to society's "have nots;" this suggests that electoral competition may
mobilize less affluent voters. We replicated our MLM models and simulations to for this
(available from authors). Rather than allowing age and education to interact with the effects of
elections, we allowed the respondent's income to interact with exposure to electoral competition.
We found no evidence that competitive elections mobilized turnout of low-income citizens more
or less relative to higher income citizens. This result is consistent with our theory grounded in
political interest - a theory that offers a mechanism to explain how competitive elections alter the
composition of electorates by generating interest among the young and less educated; rather than
by mobilizing voters with class-based appeals. As noted above, the CPS lacks direct measures of
political interest. We now turn to CCES data to address these shortcomings.
Electoral Competition, Political Interest and Turnout: CCES Data
Our theory suggests a two-stage process, where electoral competition increases interest in
politics, which in turns affects participation. The 2006 CCES allows us to model this process, as
it included identifiers linking respondents to their House districts and it included questions
asking about political interest and turnout.22 We merged state and congressional district-level
measures of electoral competition onto these CCES data to estimate a two-stage model, where
interest is our primary dependent variable in the first stage, and turnout is the dependent variable
in the second stage. In the second stage estimates, turnout is predicted with an instrumental
variable formed with the component of interest predicted by district-level US House race
18
campaign spending and other measures of electoral activity included in the first stage prediction
of interest. Interest is measured with responses to the question, “How interested are you in
politics and current affairs?” Response are coded as 1 for those “very much interested” and 0 if
the respondent said they were somewhat or not interested in politics. Turnout was coded 1 if the
respondent reported voting and 0 for not voting.
In the first stage, political interest is predicted by exposure to competitive elections and
other control variables. We measure 2006 campaign spending in the respondent’s U.S. House
district with raw data from the Federal Election Commission (FEC),23 since congressional
districts have comparable populations. We also measure total spending per capita on ballot
initiatives and referendums in the respondent’s state in 2006.24 Spending in U.S. Senate races is
again measured per capita. The models include standard demographic and ideology controls
matching as best as possible the CPS models. Models are estimated with logistic regression with
standard errors by clustered by congressional district.25
Table 4
Results in Table 4 demonstrate that individuals residing in US House districts with
greater spending were significantly more likely to report being very interested in politics.
Residence in a states with competitive US Senate races, and residence in states with greater
spending on ballot initiatives were also associated with a greater likelihood of being interested in
politics in 2006. This provides direct evidence that electoral competition increases political
interest. Most importantly, when we replicate the estimates in Table 4 across sub-samples that
compare younger voters to older voters, and less educated voters to higher educated voters (not
shown), we find the effects of spending on interest more pronounced among the young, and the
less educated. None of the measures of electoral competition increased interest among the high-
19
educated subsample. Column 2 in Table 4 demonstrates that interest stimulated by campaign
activity (estimated from Column 1) increases the likelihood of voting.
Discussion
Taken together, the sum of our results offer strong evidence that competitive elections
affect the composition of electorates. Competition mobilizes younger and less educated voters
more than others by stimulating interest in politics. Although it is difficult to generalize across
time from the data assessed here, our results suggest electoral competition affects who votes
across place, and over time. In presidential years, presidential campaigns may change a younger
or less-educated voter from a marginal non-voter into a marginal voter in highly competitive
states. Although relatively high levels of interest associated with presidential elections would
suggest a limit to how much competition might shape the electorate, our theory and results
suggest that when interest in a presidential election rises, turnout may increase disproportionately
among those previously less interested – the younger and less educated.
Although we do find that other forms of electoral competition stimulate interest and
disproportionately boost turnout of the young and less educated in midterm years, these results
are noteworthy since relatively few US House races are competitive in any given year and only
one-third of US Senate races can be competitive. The lack of a high-stimulus presidential
contest at the midterm, and a dearth of competitive congressional contests correspond with less
interest in politics and less turnout of the young and less educated at midterm elections. We
suggest this explains something we observe in the CPS: In 2002 and 2006, the average age of a
voter was 52. In 2004 the average voter was 49. The electorates were different because the
midterm contests were less interesting, and this muted participation of younger voters.
This is consistent with Schattschneider’s (1960) concern that low political participation is
one of the primary failures of American politics because it may produce a bias between the
20
general population and those who are actually represented. We have offered a theory of turnout,
grounded in political interest, to explain how active campaigns increase turnout
disproportionately among people with lower levels of political interest. Our contribution to the
understanding of voter turnout is to provide theoretical and empirical support for the idea that
competitive elections (and non competitive elections) do not have neutral effects on the
composition of an electorate. Electoral competition tends to increase turnout generally, but it can
also have a greater propensity to mobilize voters in groups known to have lower levels of
political interest: the young, and the less-educated. Active political campaigns that accompany
competitive elections may thus alter the composition of the electorate and reduce age and
education biases.
We illustrate how the effects of the electoral context a citizen resides in - namely,
exposure to competitive forces that stimulate political interest and mobilize voters - should be
seen as variable across time and place. In any given place, at any point in time, a unique set of
elections may stimulate a voter’s interest: state legislative races, contests for statewide office,
ballot measures, congressional races, and presidential elections. The more competitive these are,
the more likely it is that people less-interested in politics will vote. We suggest that the effects of
simultaneous exposure to multiple competitive elections may be cumulative. The relatively large
mobilizing effects of presidential elections on the young and less educated may mute the effects
of other contests in presidential years, but our results suggest there is substantial room for other
contests to mobilize these voters during midterms.
This study has implications for broader discussions about turnout. Our results illustrate
that laws placing barriers on voting are only part of the story about why so many people fail to
vote. Increased convenience of voting may have modest effects on turnout and on mitigating
existing bias in who participates (Fitzgerald 2005; Berinsky 2005). Barriers to voting declined
21
substantially after 1960, yet turnout (with the exceptions of the 2004 and 2008 presidential
contests) was consistently lower after the 1960s. Incumbent advantages and the geographic
distributions of partisans may have reduced the scope of exposure to competitive elections in the
U.S. since then. There were fewer competitive U.S. House races by 2000, and the geography of
U.S. presidential elections also changed substantially. In 1960 and 1968, the eight most
populous states were highly competitive in presidential politics, with an average margin of 2.4%
(s.d. 2.2) and 3.4% (s.d. 1.7), respectively. By 2004 and 2008, the eight most populous U.S.
states had many of the least competitive in presidential contests, with an average margin of 8.7%
(6.6 s.d) and 15.2% (9.3 s.d), respectively. New York, California and Texas were among the
most competitive states in presidential elections in the 1960s; in 2004 and 2008 they were among
the least competitive.
We suggest election reforms that may increase competition such as legislative
redistricting, changes to campaign finance regimes, and even modified forms of proportional
representation, may increase turnout overall - especially among the young and low-educated.
We concur with Franklin's suggestion (2004:3) that "the idea that declining turnout is due largely
to 'something about citizens' runs counter to some very obvious facts." Low turnout may have as
much to do with the character of elections as with the character of citizens. People vote for
various reasons, with one important reason being interest generated by elections. Compared to 40
years ago, a greater proportion of Americans now reside in places where the electoral context is
likely to offer little mobilizing effects from presidential and congressional campaigns. We have
demonstrated the importance of such exposure, and suggest this may be one reason for lower
levels of turnout since the 1960s.
22
Table 1: Probability of Voting, Predicted by Electoral Competition in a Respondent’s State. Hierarchical Linear Models. 2002-2006 Elections 2002
Midterm 2004 Presidential
2006 Midterm
β (S.E.) β (S.E.) β (S.E.) Level 2 (State Context) MODEL 1 MODEL 2 MODEL 3 Presidential Vote Margin -- .764*** -- -- (.235) -- Senate Race Spending Per Capita .040***
(.009) -.002 (.004)
.007** (.003)
Governor's Race Vote Margin .221** .042 .286*** (.079) (.067) (.100) Spending U.S. House Races (State, per capita) .005 .017 .033 (.031) (.026) (.023) Number of Initiatives State Ballot .039**
(.017) .032** (.015)
.045*** (.013)
Closing Date Register Vote -.004 -.011*** .003 (.004) (.003) (.003) Percent High School Graduates .013 .004 .007 (.011) (.007) (.012) Percent Black -.001 -.002 -.009** (.005) (.003) (.004) Percent Latino -.009* -.005 -.013** (.005) (.004) (.005) Level-2 Intercept -5.24*** -4.491*** -5.841*** (.157) (.630) (1.034) Level 1 (Individual-level) Education .217*** .265*** .210*** (.006) (.007) (.005) Income .055*** .069*** .046*** (.003) (.004) (.004) Age .043*** .004 .042*** (.004) (.005) (.004) Age Squared -.003-2 .002-1*** -.081-3** (.004-2) (.005-2) (.036-3) Male -.047*** -.176*** -.058*** (.018) (.015) (.017) Married .385*** .379*** .323*** (.021) (.030) (.025) Child .049** -.013 -.013 (.024) (.028) (.031) Black .370*** .449*** .262*** (.039) (.074) (.069) Latino -.158*** -.233*** -.170*** (.042) (.053) (.051) Asian -.729*** -1.083*** -.842*** (.095) (.088) (.122) Urban .066 .147*** -.025 (.052) (.048) (.040) Suburban -.107*** .014 -.079* (.036) (.034) (.032) Residential Mobility (5 years at residence or more)
.594*** (.025)
.466*** (.019)
.802*** (.030)
Military Veteran .079*** .163*** .126***
23
(.022) (.032) (.028) Government Employee .456*** .446*** .456*** (.036) (.042) (.029) Occupation Management .251*** .455*** .226*** (.032) (.040) (.028) Professional .192*** .394*** .251*** (.033) (.039) (.028) Service .074** .132*** .073*** (.030) (.037) (.030) Sales .191*** .277*** .175*** (.041) (.030) (.032) Secretarial/Admin Support .171*** .346*** .221*** (.035) (.039) (.034) Farming .274*** .088 .077 (.064) (.121) (.135) Transportation -.057 -.139*** -.131*** (.050) (.043) (.050) Random Effects Variance component .315*** .376*** .204*** Age (u1) .00001*** .00003 *** .00002 *** Education (u2) .002 *** .002 *** .001 *** -Log likelihood function -109700.00 -104100.00 -106040.00 Level-1 N 77,619 74,044 75,188 Level-2 N 50 50 50 Note: The dependent variable is coded 1 if the respondent reported voting, and 0 otherwise. Hierarchical linear models estimated using HLM 6.0. Random coefficient models using a Bernoulli distribution and logit link function. Unstandardized logistic regression coefficients and robust standard errors in parentheses. Models were run to convergence. Reliability estimates for random effects (level 1 intercept, age and education) above critical threshold. * p<.1; ** p<.05; *** p<.01. Source: 2002, 2004, 2006 Current Population Surveys.
24
Table 2: Probability of Voting in 2004 Varying Presidential Campaign Activity in a Respondent’s State. Hierarchical Linear Models. β (S.E.) β (S.E.) β (S.E.) β (S.E.) Level 2 (State Context) MODEL 1 MODEL 2 MODEL 3 MODEL 4 Presidential Visits .004** -- -- -- (.002) -- -- -- Presidential TV Ads -- .004-3** -- -- -- (.002-3) -- -- Presidential TV Spending Per Capita -- -- .045** -- -- -- (.022) -- Senate Race Spending Per Capita -.002 -.002 -.002 -- (.004) (.004) (.004) -- Governor’s Race Vote Margin .065 .058 .054 .029 (.067) (.067) (.066) (.069) Spending in U.S. House Races .028 .025 .024 -- (.030) (.031) (.031) -- All Spending in Federal Races (Pres, Senate, U.S. House) Per Capita
-- -- -- .018*** (.005)
Spending on Ballot Initiatives Per Capita .040** (.018)
Urban .146*** .146*** .146*** .146*** (.047) (.048) (.048) (.048) Suburban .013 .014 .014 .014 (.034) (.034) (.034) (.034) Residential Mobility (5 years at residence or more)
.466*** (.019)
.466*** (.019)
.466*** (.019)
.466*** (.019)
Military Veteran .163*** .163*** .163*** .163*** (.032) (.031) (.031) (.031) Government Employee .445*** .445*** .445*** .445*** (.042) (.042) (.042) (.042) Occupation Management .454*** .455*** .455*** .455*** (.040) (.040) (.040) (.040) Professional .394*** .394*** .394*** .394*** (.039) (.039) (.039) (.039) Service .132*** .132*** .132*** .132*** (.037) (.037) (.037) (.037) Sales .277*** .277*** .277*** .277*** (.029) (.029) (.030) (.030) Secretarial .346*** .346*** .346*** .346*** (.039) (.039) (.039) (.039) Farming .087 .087 .088 .088 (.120) (.120) (.120) (.120) Transportation -.139*** -.139*** -.139*** -.139*** (.043) (.043) (.043) (.043) Random Effects Level 1 Intercept/Variance component .371*** .377*** .380*** .398*** Age (u1) .00003*** .00003*** .00003*** .00003*** Education (u2) .002 *** .002 *** .002 *** .002 *** -Log likelihood function -104100.00 -104100.00 -103981.90 -1.039850 Level-1 N 74,044 74,044 74,044 74,044 Level-2 N 50 50 50 50 Note: The dependent variable is whether the respondent voted in 2004, coded as 1 if yes and 0 otherwise. Hierarchical linear models estimated using HLM 6.0. Random coefficient models using a Bernoulli distribution and logit link function. Unstandardized logistic regression coefficients and robust standard errors in parentheses. Reliability estimates for random effects (level 1 intercept, age and education) above critical threshold. * p<.1; ** p<.05; *** p<.01. Source: 2004 Current Population Survey.
26
Table 3a: Probability of Voting, Sub-Samples of Young and Old Respondents, 2002 and 2004 2002 Midterm 2004 Presidential Young (32 years or
less/ bottom quartile)
Old (58 years or older/ top quartile)
Young (32 years or less/ bottom quartile)
Old (58 years or older/ top quartile)
Level 2 (State Context) β (S.E.) β (S.E.) β (S.E.) β (S.E.) Presidential Vote Margin -- -- .991*** .388 -- -- (.368) (.472) Senate Race Spending Per Capita .279*** .161* -.003 -.128 (.065) (.094) (.095) (.110) Governor's Race Vote Margin .229*** .134 -.020 .184* (.076) (.115) (.105) (.109) Spending U.S. House Races (State .024 .018 -.021 .044 Average) (.027) (.037) (.022) (.032) Number of Initiatives State Ballot .068*** .043** .023 .063*** (.013) (.020) (.016) (.022) N 19358 19226 17930 1913 Wald Chi2 3689.24 8076.63 3750.40 3116.52 The dependent variable is whether the respondent voted, coded as 1 if yes and 0 otherwise using the CPS. Unstandardized logistic regression coefficients, with robust standard errors in parentheses. Probabilities based on two-tailed tests. Standard errors adjusted by clustering by state. * p<.1; ** p<.05; *** p<.01. Control variables omitted from table are identical to Table 1. Similar finding for sub-samples of low and high educated. Table 3b: Sub-samples by Education--Probability of Voting in 2004, predicted by the Margin of the Presidential Race in a Respondent’s State. Competitiveness of Presidential Race (1-vote margin)
Low Educated Sub-sample (high school graduate or lower)
High Educated Sub-sample (college degree or higher)
Very Low=.545 .62 (.023) .92 (.009) Low=.70 .65 (.015) .93 (.006) Moderate=.80 .67 (.012) .94 (.004) High=.90 .68 (.012) .94 (.004) Very High=.99 .70 (.014) .94 (.004) First Difference (min to max)
+.08 +.02
Table 3c: Sub-samples by Age--Probability of Voting in 2004, Predicted by the Margin of the Presidential Race in a Respondent’s State. Competitiveness of Presidential Race (1-vote margin)
Young Sub-sample (32 years or younger/bottom quartile)
Older Sub-sample (58 years or older/top quartile)
Very Low=.545 .49 (.029) .87 (.008) Low=.70 .53 (.019) .87 (.008) Moderate=.80 .55 (.015) .87 (.008) High=.90 .58 (.016) .87 (.008) Very High=.99 .60 (.021) .87 (.008) First Difference (min to max)
+.11 0
27
Table 4: Impact of U.S. House District-level Electoral Competition on Interest in the Election and Turnout Very Interested in the Election Voted in
the Election Competitive Elections Β (S.E.) p Spending in U.S. House District 2.380-08 (0.020) (Dollars) (1.02-08) Competitive Governor Race 0.019 (0.768) (0.066) Competitive Senate Race 0.136 (0.027) (0.061) Spending Initiatives and 0.018 (0.005) Referendums Per Capita (0.006) Prob. of Interested in Politics (Estimated from Column 1) 5.988
(1.044) (0.000)
Control Variables Strong Democrat 0.414 (0.000) 0.436 (0.000) (0.065) (0.110) Strong Republican 0.444 (0.000) 0.150 (0.305) (0.074) (0.146) Weak Republican -0.754 (0.000) 0.940 (0.000) (0.068) (0.174) Weak Democrat -0.678 (0.000) 0.874 (0.000) (0.064) (0.150) Black -0.716 (0.000) 0.425 (0.035) (0.079) (0.202) Latino -0.387 (0.000) -0.003 (0.978) (0.083) (0.116) Asian -0.905 (0.000) 0.508 (0.067) (0.224) (0.277) Male 1.053 (0.000) -0.714 (0.004) (.047) (.244) Education 0.358 (0.000) 0.003 (0.964) (0.019) (0.069) Married -0.207 (0.000) 0.202 (0.010) (0.048) (0.078) Age 0.026 (0.000) 0.016 (0.009) (0.002) (0.006) Liberal Ideology 0.410 (0.000) -0.362 (0.001) (0.058) (0.105) Conservative Ideology 0.301 (0.000) -0.305 (0.000) (0.054) (0.080) Income 0.107 (0.000) -0.017 (0.503) (0.008) (0.026) Frequency of Church Attendance -0.007 (0.715) 0.188 (0.000) (0.018) (0.029) Military or Military Veteran -0.055 (0.377) 0.008 (0.934) (0.062) (0.195) Closing Date to register 0.003 (0.605) (0.005) Percent Black -.017 (.003) (.006) Percent Latino .0003 (.937) (.004)
28
Constant -2.900 (0.000) -2.396 (0.000) (0.131) (0.199) Pseudo R Square .16 .16 Wald χ2 2016.44 (.000) 2124,67 (.000) Number of Observations 16689 16689 Note: The dependent variable in column 1 is very interested in politics. Dependent variable in column 2 is reported turnout. Unstandardized logistic regression coefficients with robust standard errors in parentheses. Standard errors adjusted by clustering cases by US House districts. Probabilities based on two-tailed significance tests. Models estimated using Polimetrix survey weights. When a logged version of spending in one’s Congressional District is estimated instead, the b=.055 (p value=.033). When an additional covariate for media consumption is added to the model in column 1, the results are unchanged. Variable omitted to avoid a loss of cases. Source: Cooperative Comparative Election Study (CCES) 2006 conducted by Polimetrix. a Predicted probability of being very interested in politics from Table 4, Column 1.
29
Fig. 1. Predicted Probability of Voting by Education Category (2004); At various levels of presidential election competitiveness in respondent’s state.
Fig. 2. Predicted Probability of Voting by Age (2002); At various levels of per capita spending on U.S. Senate races in respondent’s state.
30
Fig. 3 Predicted Probability of Voting by Education (2002); At various levels of per capita spending on U.S. Senate Races in respondent’s state.
Fig. 4. Predicted Probability of Voting by Age (2004); At various numbers of presidential candidate visits to respondent’s state.
31
References
Aldrich, John. 1993. “Rational Choice and Turnout.” American Journal of Political Science 37 (1):
246–78.
Berinsky, Adam. 2005. “The Perverse Consequences of Electoral Reform in the United States.”
American Politics Research 33: 471-491.
Blais, Andre. 2006. "What Affects Voter Turnout?" Annual Review of Political Science. 9:111-25.
Blais, Andre and A. Dobrzynska. 1998. "Turnout in Electoral Democracies." European Journal of
Political Research. 18:167-81
Brians, Craig and Bernard Grofman. 1999. "When Registration Barriers Fall, Who Votes? An
Empirical Test of a Rational Choice Model." Public Choice. 99-161-76,
Brians, Craig and Bernard Grofman. 2001. “Election Day Registration’s Effect on U.S. Voter
Turnout.” Social Science Quarterly 82 (1): 170-183.
Bryk, AS and SW Raudenbush. 2002. Hierarchical Linear Models: Applications and Data Analysis
Methods. Sage Publications.
Caldeira, Gregory A., Patterson, Samuel C., and Markko, Gregory A. 1985. “The Mobilization of
Voters.” Journal of Politics 47: 490–509.
Campbell, Angus. 1966. “Surge and Decline: A Study of Electoral Change.” In A. Campbell, P.E.
Converse, W.E. Miller and D.E. Stokes (Eds.) Elections and the Political Order. New York: Wiley.
Cox, Gary and M. C. Munger. 1989. Closeness, Expenditures and Turnout in the 1982 U.S. House
Elections” American Political Science Review. 83(1): 217-231.
Downs, Anthony. 1957. An Economic Theory of Democracy. New York: Harper & Row.
Ferejohn, John A., and Morris P. Fiorina. 1974. "The Paradox of Not Voting: A Decision Theoretic
Analysis." American Political Science Review 68: 525-546.
32
Fitzgerald, Mary. 2005. “Greater Convenience but not Greater Turnout: The Impact of Alternative
Voting Methods on Electoral Participation in the United States.” American Political
Research 33: 842-867.
Franklin, Mark N. 2004. Voter Turnout and the Dynamics of Electoral Competition in Established
Democracies Since 1945. Cambridge University Press.
Gerber, Alan S. and Green, Donald P. 2000. “The Effects of Canvassing, Telephone Calls, and
Direct Mail on Voter Turnout: A Field Experiment.” American Political Science Review 94:
653–663.
Green, Donald P. and Ian Shapiro. 1994. Pathologies of Rational Choice Theory. New Haven: Yale
University Press.
Highton, Benjamin. 1997. “Easy Registration and Voter Turnout.” Journal of Politics 59 (2): 565-575.
Highton, Benjamin. 2004. “Voter Registration and Turnout in the United States.” Perspectives on
Politics. 2(2):507-505.
Highton, Benjamin and A. Burris 2002. "New Perspectives on Latino Voter Turnout in the United
States. American Politics Research. 30:3:285 -306.
Highton, Benjamin and Raymond Wolfinger 1998. Estimating the Effects of the National Voter
Registration Act of 1993. Political Behavior. 20(2): 79-104.
Hill, Kim Quaile and Jan Leighley. 1994. "Mobilizing Institutions and Class Representation in U.S.
State Electorates. Political Research Quarterly. 47(1): 137-50.
Hill, Kim and Jan Leighley. 1999. “Racial Diversity, Voter Turnout, and Moblizing Institutions in
the United States.” American Politics Quarterly 27: 275-295.
Holbrook, Thomas and Scott McClurg. 2005. “The Mobilization of Core Supporters: Campaigns,
Turnout, and Electoral Composition in United States Presidential Elections.” American
33
Journal of Political Science 49 (4): 689-703.
Jackson, Robert A. 1996. “A Reassessment of Voter Mobilization.” Political Research Quarterly
49: 331– 349.
Jackson, Robert A. 1997. “The Mobilization of U.S. State Electorates in the 1998 and 1990
Elections.” Journal of Politics 59: 520-37.
Jackson, Robert A. 2002. "Gubernatorial and Senatorial Campaign Mobilization of Voters."
Political Research Quarterly. 55(4): 825-844.
Jaccard, James, Robert Turrisi, and Choi Wan. 1990. "Interactive Effects in Multiple Regression."
Sage Series on Quantitative Applications in Social Sciences, # 72. Newbury Park, CA: Sage.
Kahn, K.F. and P.J. Kenny. 1999. The Spectacle of U.S. Senate Campaigns. Princeton, NJ:
Princeton University Press.
Karp, Jeffrey A. and Susan A. Banducci. 2000. “Going postal: How All-Mail Elections Influence
Turnout.” Political Behavior 22: 223-239.
Key, V.O. 1949. Southern Politics in State and Nation. New York, NY: Knopf.
Leighley, Jan and Jonathan Nagler. 1992. “Individual and Systemic Influences on Turnout: Who
Votes?” Journal of Politics 54: 718-741.
McDonald, Michael and John Samples, eds. 2006. The Marketplace of Democracy: Electoral
Competition and American Politics. Washington, DC: Brookings Institute Press.
Nagler, Jonathan. 1991. “The Effect of Registration Laws and Education on U.S. Voter Turnout."
American Political Science Review 85:1393-1405.
Oliver, J. Eric. 1996. “The Effects of Eligibility Restrictions and Party Activity on Absentee Voting
and Overall Turnout.” American Journal of Political Science 40: 498-514.
Patterson, Samuel C. and Gregory Caldeira. 1983. “Getting Out the Vote: Participation in
34
Gubernatorial Elections.” American Political Science Review 77(3): 675-89.
Primo, David M., Matthew L. Jacobsmeier, and Jeffrey Milyo. 2007. “Estimating the Impact of
State Policies and Institutions with Mixed-Level Data,” State Politics and Policy Quarterly
7: 446-59.
Rosenstone, Steven and John Mark Hansen. 2003. Mobilization, Participation, and Democracy in
America. New York: Macmillan.
Schattschneider, E. E. 1960. The Semi-sovereign People. New York: Harcourt Brace.
Schier, Steven. 2000. By Invitation Only: The Rise of Exclusive Politics in the United States.
Pittsburgh: University of Pittsburgh Press.
Smith, Mark. 2001. “The Contingent Effects of Ballot Initiatives and Candidate Races on
Turnout.” American Journal of Political Science 45: 700-06.
Squire, Peverill, Raymond E. Wolfinger, David P. Glass. 1987. “Residential Mobility and Voter
Turnout. The American Political Science Review 81 (1): 45-66.
Steenbergen, Maro, and Bradford Jones. 2002. “Modeling Multilevel Data Structures.” American
Journal of Political Science 41(1): 218–37.
Wolfinger, Raymond, and Steven J. Rosenstone. 1980. Who Votes? New Haven: Yale University
Press.
Wolfinger, Robert, Benjamin Highton and M. Mullin. 2005. “How Postregistration Laws affect the
Turnout of Citizens Registered to Vote.” State Politics and Policy Quarterly 5: 1-23.
35
Appendix: Variable coding for Current Population Surveys
Voted / not voted: Following the CPS published reports, we code respondents indicating they did
not vote (question pes1) as non voters, as well refused, don’t know and no response.
Respondents reporting “yes” on question pes1 were coded as voters. In 2002 of 97,684 valid
respondents, 47,377 (48.5%) reported voting. In 2004 of 95,408 respondents, 62,328 reported
they voted. In the 2006 survey had 93, 331 respondents. Occupational categories include: 1)
management, business, and financial, 2) professional and related, 3) service, 4) sales and related,
5) office and administrative support, 6) farming, fishing, and forestry, 7) construction and
extraction, 8) installation, maintenance, and repair, 9) production, 10) transportation and material
moving, and 11) armed forces. Minor variations in these categories over the eights years of the
study. Residential Mobility: Respondents living at the same address for fives years or longer are
coded 1, and those less than five years 0. Military veteran (or currently in the military), coded 1
with non-veterans coded as 0. Education: 1=Less than 1st; 2=1st-3rd grade; 3= 5th-6th grade;
4=7th-8th grade; 5=9th grade; 6=10th; 7= 11th; 8=12th grade, no diploma; 9= high school grad-
diploma or equivalent; 10=some college, no degree; 11=associate degree-occupational/vocational;