Content uploaded by Alan I. Abramowitz
Author content
All content in this area was uploaded by Alan I. Abramowitz on Oct 06, 2015
Content may be subject to copyright.
Exploring the Bases of Partisanship in the American Electorate: Social Identity vs.
Ideology
Alan I. Abramowitz
Department of Political Science
Emory University
Atlanta, Georgia 30322
polsaa@emory.edu
Kyle L. Saunders
Department of Political Science
Colorado State University
Fort Collins, Colorado 80523
kyle.saunders@colostate.edu
Prepared for delivery at the State of the Parties Conference, Crown Plaza Quaker Square,
Akron, Ohio, October 5-7, 2005.
Abstract
This paper uses data from the 1952-2004 American National Election Studies and the
2004 U.S. National Exit Poll to compare the influence of ideology and membership in
social groups on party identification. Contrary to the claim by Green, Palmquist, and
Schickler that party loyalties are rooted in voters’ social identities, we find that party
identification is much more strongly related to voters’ ideological preferences than to
their social identities as defined by their group memberships. Since the 1970s,
Republican identification has increased substantially among whites inside and outside of
the South with the most dramatic gains occurring among married voters, men and
Catholics. Within these subgroups, however, Republican gains have occurred mainly or
exclusively among self-identified conservatives. As a result, the relationship between
ideology and party identification has increased dramatically. This has important
implications for voting behavior. Increased consistency between ideology and party
identification has contributed to higher levels of party loyalty in presidential and
congressional elections.
In Partisan Hearts and Minds: Political Parties and the Social Identities of Voters,
Donald Green, Bradley Palmquist, and Eric Schickler (2002) argue that party
identification in the United States is based on voters’ social identities rather than on a
rational assessment of the parties’ policies or performance in office. Challenging many
of the conclusions of recent research on party identification in the American electorate,
Partisan Hearts and Minds has attracted the attention of pundits (Brooks 2004) as well as
scholars.
Green, Palmquist, and Schickler make four major claims about the nature of
contemporary party identification:
1. Party identification is more stable at both the aggregate and the individual level than
most recent scholarship has suggested. Outside of the South, there has been little change
in the distribution of party identification in the U.S. for several decades (pp. 52-84).
2. Voters’ party loyalties are largely insulated from the effects of current issues such as
the state of the economy and the performance of the incumbent president (pp. 85-108).
3. Party loyalties exert a powerful influence on citizens’ issue positions, evaluations of
political leaders, and voting decisions (pp. 204-229).
4. Most importantly, party identification is based mainly on identification with social
groups rather than a rational evaluation of the parties’ ideological orientations or policies
(pp. 25-51). According to Green et al., “people ask themselves two questions when
deciding which party to support: What kinds of social groups come to mind as I think
about Democrats, Republicans, and Independents? Which assemblage of groups (if any)
best describes me (p. 8)?”
2
In proposing this social identity theory, Green et al. explicitly challenge rational
choice explanations of party identification such as those proposed by Downs (1957) and
Fiorina (1981). Green et al. view party identification as an emotional attachment
grounded in enduring group loyalties rather than a deliberate choice based on a
preference for one set of policy positions over another—a choice that can be modified if
parties’ policy positions change or new issues arise (Page and Jones 1979; Franklin and
Jackson 1983; Carmines, McIver, and Stimson 1987; Luskin, McIver, and Carmines
1989; Franklin 1992).
Like Campbell, Converse, Miller and Stokes (1960), Green et al. downplay the role
of issues and ideology in the formation of party identification. While recognizing that
party loyalties can be influenced by dramatic changes in the parties’ policy stands or
ideological positions, Green et al. argue that such shifts are relatively rare and generally
confined to periods of major realignment such as the New Deal era in the United States.
In this regard, social identity theory stands in sharp contrast to ideological realignment
theory which claims that as a result of the growing ideological polarization of the two
major parties since the 1980s, Americans have increasingly been choosing a party
identification on the basis of their ideological preferences, leading to a gradual
realignment of party loyalties along ideological lines (Abramowitz and Saunders 1998).
According to Green et al., even the one exception to the rule of partisan stability in
recent American political history, the dramatic realignment of southern white voters’
party loyalties since the end of World War II, was based more on changing perceptions of
the parties’ ties to social groups than on issues or ideology. They argue that as
southerners began to assume leadership positions in the Republican Party during the
3
1980s and 1990s, Republicanism came to be seen as a respectable affiliation among white
southerners. According to Green et al., “As the Republican image improved, Republican
identification became increasingly prevalent among all segments of the ideological
continuum (p. 160).”
Green et al. argue that “the growing correlation between liberalism-conservatism
and party [among southern whites] reflects cohort replacement as older conservative
Democrats pass away (p. 161).” With this process of generational replacement largely
completed, they claim that, “the pace of partisan conversion [in the South] appears to
have slowed to the near-standstill characteristic of party identification in the non-South
(p. 163).”
We find much that is persuasive about the evidence presented in Partisan Hearts
and Minds. We agree with its conclusions that party identification is usually quite stable
at both the individual and the aggregate level, that party loyalties are relatively immune
from short-term fluctuations in economic conditions and presidential popularity, and that
party identification exerts a powerful influence on evaluations of political leaders and
voting decisions. However, we take issue with the claim that partisan change in the
American electorate in recent years has been limited exclusively to southern whites. We
present evidence in this paper that there has been a substantial increase in Republican
identification among white voters outside of the South over the past three decades and
that this shift has been quite dramatic among several major subgroups including men,
Catholics, and the religiously devout. We further show that the extent and direction of
these shifts are strongly related to ideology, that the correlation between ideology and
party identification has increased substantially over time across a wide variety of social
4
groups, and that this increase is not due simply to generational replacement. We also
present evidence that the increasing correlation between ideology and party identification
over time is due primarily to the influence of ideology on party identification.
Most fundamentally, we take issue with the claim that party identification in the
U.S. is based mainly on the social identities of citizens rather than their ideological
orientations or policy preferences. In fact, Green, Palmquist, and Schickler never
actually test this social identity theory, nor do they compare the influence of social
background characteristics with the influence of issues and ideology. They argue that,
“we lack the luxury of examining a broad range of social identities [because] social class,
ethnicity, religion, and party exhaust the list of social categories about which we have
adequate longitudinal data (p. 83).”
We find this argument unpersuasive. While measures of identification with social
groups may not be widely available, measures of objective membership in a large variety
of social groups are widely available and social identity theorists generally view objective
membership as a necessary condition for identification with a group and the development
of group political consciousness (Gurin, Miller and Gurin 1980; Miller, Gurin, Gurin, and
Malanchuk 1981; Tajfel 1981; Turner 1982; Tajfel and Turner 1986; Gurin 1985; Turner
1987). Tajfel (1981), for example, defines social identity as, “that part of an individual’s
self-concept which derives from his knowledge of his membership in a social group
together with the value and emotional significance attached to the membership (p. 255).”
In addition to social class, ethnicity, and religion, data on age, race, gender, region,
urban-rural residence, and union affiliation are available in the American National
Election Studies and many other surveys conducted over the past fifty years. Along with
5
social class, ethnicity, and religion, these are precisely the social characteristics most
commonly associated with support for the major parties in the United States. Social
identity theory clearly implies that voters who belong to groups generally associated with
one major party or the other—groups such as the poor, union members, single women,
and Jews in the case of the Democratic Party, or the wealthy, married men, and
evangelicals in the case of the Republican Party—should be more likely to identify with
that party. Based on this reasoning, we test the social identity theory by examining the
influence on party identification of membership in a wide variety of social groups that are
closely aligned with the two major parties. Our results indicate that while there is a
group basis to party loyalties, most of these social characteristics are only weakly related
to party identification. With the exception of the overwhelmingly Democratic
identification of African Americans, party identification is much more strongly related to
voters’ ideological orientations than to their social identities as defined by their group
memberships.
We believe that our evidence shows that there is a much larger rational component
to party identification that Green et al. acknowledge. Moreover, the increasing
consistency of voters’ ideological orientations and party loyalties has had important
political consequences. This trend helps to explain the extraordinary levels of partisan
voting seen in the 2000 and 2004 presidential contests and other recent elections.
Trends in Party Identification
One of the key claims made by Green et al. is that outside of the South there has
been little change in partisanship since the 1960s. However, this claim appears to be
contradicted by a considerable body of research that has documented changes in
6
partisanship based on such factors as gender (Wirls 1986; Kaufmann and Petrocik 1999),
marital status (Weisberg 1987), religiosity (Guth and Green 1990; Layman and Carmines
1997) and social class (Stonecash, Brewer, and Mariani 2003).
In order to test the claim that partisanship outside of the South has been stable,
Table 1 presents data from American National Election Study surveys on trends in party
identification in the U.S. since the 1960s. We have grouped the data by decade in order
to minimize fluctuations due to short-term forces or sampling variation. The measure
presented here is simply the difference between the percentage of Democratic identifiers
and leaners and the percentage of Republican identifiers and leaners in the overall
electorate.1
[Table 1 goes here]
The evidence presented in Table 1 does not support the conclusion that outside of
party identification outside of the South has been stable since the 1960s. Although the
most dramatic change has occurred among white southerners, there has also been a
substantial increase in Republican identification among whites outside of the South.
During the 1960s, the Democratic Party enjoyed an average advantage of 13 points over
the Republican Party among non-southern whites. Since 1980, however, this advantage
has disappeared: in the 2002-2004 NES surveys, non-southern whites favored the
Republican Party over the Democratic Party by a 5 point margin.
Data from national exit polls also show a substantial increase in Republican
identification among non-southern whites since the 1970s. Between 1976 and 2004, the
percentage of non-southern whites identifying with the Republican Party in national exit
1 Excluding leaning independents and calculating net party identification based on the difference between
the percentage of Democratic identifiers and the percentage of Republican identifiers produces almost
identical results.
7
polls increased from 28 percent to 43 percent. Moreover, contrary to the claim by Green
et al. that the party loyalties of white southerners have stabilized in recent years, the
national exit poll data show a continuing movement toward the Republican Party in this
group: the Republican advantage in party identification among white southerners
increased from 17 points in 2000 to 31 points in 2004.
Outside of the South, Republican gains have been much larger in certain white
subgroups than in the overall white electorate. Table 2 presents data from American
National Election Study surveys on trends in party identification since the 1960s among
various subgroups of non-southern whites. These results show that Republican
identification has increased dramatically among men, married voters, Catholics, and the
religiously observant. Among white Catholics, for example, the Democratic advantage,
which was 42 points in the 1960s and 36 points in the 1970s, has completely disappeared.
In the 2002-2004 NES surveys, Republican identifiers slightly outnumbered Democratic
identifiers among non-southern white Catholics. Similarly, in the 2004 national exit poll,
Republican identifiers outnumbered Democratic identifiers by 41 percent to 34 percent
among non-southern white Catholics. The Republican gains in this group are very
significant politically because, according to the national exit poll, Catholics comprised
more than 30 percent of the white electorate outside of the South in 2004.
[Table 2 goes here]
Ideology in the American Electorate: Meaning and Measurement
The evidence examined thus far indicates that since the 1970s there has been a
substantial increase in Republican identification among whites outside the South as well
as among those in the South and that this increase has been quite dramatic among certain
8
subgroups such as Catholics. But why has this shift occurred? Contrary to Green et al.,
we believe that ideology has played a major role in producing a secular realignment of
party loyalties in the United States since the 1970s. According to this ideological
realignment hypothesis, the increasing clarity of ideological differences between the
parties during the Reagan and post-Reagan eras has made it easier for citizens to choose a
party identification based on their ideological orientations.
Before examining the impact of ideological orientations on party identification,
however, we need to demonstrate that members of the public, or at least a substantial
proportion of them, have meaningful ideological orientations. While the concept of
ideology has been defined in many different ways (Gerring 1997), political scientists
generally view an ideology as a set of beliefs about the role of government that shapes
responses to a wide range of specific policy issues (Converse 1964; Peffley and Hurwitz
1985). Among political elites in the U.S., positions on a wide range of economic, social,
and foreign policy issues appear to be structured by a single liberal-conservative
dimension (Poole and Rosenthal 1991). However, the extent of ideological thinking in
the public has been a subject of debate since the publication of Converse’s (1964)
seminal study of belief systems in mass publics which suggested that awareness of
ideological concepts and use of such concepts by ordinary citizens were quite limited.
Although some subsequent studies have supported Converse’s conclusions about
the lack of ideological sophistication among the general public in the U.S. (Axelrod
1967; Bishop, Oldenick, Tuchfarber, and Bennett 1978; Sullivan, Piereson, and Marcus
1978; Conover and Feldman 1981; Knight 1985; Jennings 1992), other studies have
suggested that the ability of ordinary citizens to comprehend and employ ideological
9
concepts depends on the extent and clarity of ideological cues provided by political elites.
According to this view, the greater the prevalence and clarity of ideological cues in the
political environment, the higher the level of ideological comprehension and reasoning
should be among the electorate (Field and Anderson 1969; Nie and Anderson 1974; Nie,
Verba, and Petrocik 1979; Nie and Rabjohn 1979; Craig and Hurley 1984; Jacoby 1995).
From this standpoint, the increased ideological polarization of the parties in recent years
and the increased salience of ideological conflict in the media should have produced an
increase in ideological comprehension and reasoning among the American public.
While we do not claim that ordinary citizens in the U.S. now possess belief systems
as elaborate or constrained as those evident among political activists and elites (Jennings
1992; Saunders and Abramowitz 2004), our evidence does point to a substantial increase
in the ability of citizens to apply ideological labels to the political parties, an increase in
the coherence of citizens’ views across different issues, and a growing connection
between the ideological labels that citizens choose and their positions on a wide range of
domestic and foreign policy issues.
In 1972, when the NES began asking respondents to place themselves and the two
major parties on a 7-point liberal-conservative scale, only 48 percent of respondents were
able to place themselves on the scale and to place the Democratic Party to the left of the
Republican Party. By 1996 and 2004, however, 67 percent of respondents were able to
place themselves on the scale and to place the Democrats to the left of the Republicans.
These results indicate that public awareness of ideological differences between the parties
has increased substantially in the past three decades.
10
The NES data also indicate that there has also been an increase in the ideological
coherence of citizens’ policy preferences and in the correlation between ideological
identification and policy preferences. Table 3 displays a measure of the internal
consistency (Cronbach’s alpha) of responses to 8 items that were included in every
presidential election year survey between 1984 and 2004. The 8 items include liberal-
conservative identification and opinions on 7 policy issues: government responsibility for
jobs and living standards, government responsibility for health insurance, government
services and spending, defense spending, government aid to blacks, abortion, and
women’s equality. The table also shows the correlation between liberal-conservative
identification and responses to the seven policy issues.
[Table 3 goes here]
The increasing value of Cronbach’s alpha over time indicates that citizens’
responses to these 8 questions have become more internally consistent since 1984. In
addition, contrary to the claim that ideological labels have little policy content for most
Americans (Conover and Feldman 1981), the evidence in Table 3 shows that liberal-
conservative self-identification was strongly related to preferences on every policy issue
in every survey and that this relationship has grown stronger over time. These results
indicate that there is an ideological structure to Americans’ opinions on policy issues and
that ideological self-identification is a valid indicator of the liberalism or conservatism of
citizens’ policy orientations.
In testing the ideological realignment hypothesis, we use different measures of
ideology with different data sets. We use the 7-point ideological identification scale to
classify respondents in NES surveys as liberal (1-3), moderate (4), or conservative (5-7)
11
since this question is correlated with preferences on a wide range of policy issues and it
has been included in every survey since 1972. We use a similar ideological identification
question with three response categories—liberal, moderate, and conservative—in our
analysis of 2004 national exit poll data because the split-sample procedures used in the
exit poll make it impossible to create a multiple-item scale for the entire sample.
However, in our analyses of the 1992-1996 NES panel survey and the 2004 NES survey
we measure ideological orientations with multiple-item scales that include the 7-point
ideological identification question along with a number of questions about specific policy
issues.
The 1992-1996 ideology scale is based on 11 items included in both the 1992 and
1996 waves of the panel: liberal-conservative identification, abortion, government aid to
blacks, defense spending, the death penalty, laws barring discrimination against gays and
lesbians, allowing gays and lesbians to serve in the military, government vs. personal
responsibility for jobs and living standards, government vs. private responsibility for
health insurance, government spending and services, and the role of women in society.
Because of the small number of respondents interviewed in both waves of the panel, we
recoded all of the 7-point issue scales to place respondents with no opinion in the middle
position (4) in order to avoid losing cases due to missing data. The 1992 scale has a
reliability coefficient (Cronbach’s alpha) of .70. The 1996 scale has a reliability
coefficient of .74.
The 2004 ideology scale is based on 16 items: liberal-conservative identification,
abortion, abortion funding, partial birth abortion, gay marriage, government vs. private
responsibility for health insurance, government vs. personal responsibility for jobs and
12
living standards, government services and spending, gun control, the death penalty,
government aid to blacks, government aid to Hispanics, environmental protection vs. job
creation, defense spending, use of diplomacy vs. military force, and the role of women in
society. This scale has a reliability coefficient (Cronbach’s alpha) of .80.
Group Membership, Ideology, and Partisan Change
In order to test the ideological realignment hypothesis, we will first examine trends
in party identification among some of the white subgroups that have experienced the
largest Republican gains since the 1970s, while controlling for ideological identification.
If the ideological realignment hypothesis is correct, we should find that Republican gains
have been greatest among conservative identifiers and smallest among liberal identifiers.
[Table 4 goes here]
The evidence presented in Table 4 provides strong support for the ideological
realignment hypothesis. For every subgroup examined, the increase in Republican
identification was much larger among conservative identifiers than among moderate or
liberal identifiers. In fact, Republican identification declined among liberal identifiers in
every subgroup except Catholics. Among southern whites, for example, there was a 54
point increase in net Republican identification among conservatives and a 7 point
decrease in net Republican identification among liberals. Similarly, among religiously
observant whites, there was a 40 point increase in net Republican identification among
conservatives and an 18 point decrease in net Republican identification among liberals.
The end result of the process of ideological realignment has been a marked increase
in the correlation between ideology and party identification. Table 5 displays the trend in
the correlation between ideology and party identification between the 1970s and 2004 for
13
the entire electorate and for several major subgroups. The correlation between ideology
and party identification increases in all groups including southern and non-southern
whites. In fact, the increase in the correlation between ideology and party identification
is just as great for non-southern whites as it is for southern whites.
[Table 5 goes here]
The increasing correlation between ideology and party identification was not simply
a result of generational replacement. Table 6 presents the results of a cohort analysis of
the relationship between ideology and party identification among northern and southern
whites from the 1970s through the 1990s. Almost every 10-year age cohort shows an
increase in the correlation between ideology and party identification over time. For
example, among southern whites who were in their 20s during the 1970s, the correlation
between ideology and party identification was only .27. However, among members of
the same cohort during the 1990s, the correlation between ideology and party
identification was .54. Similarly, among northern whites who were in their 20s during
the 1970s, the correlation between ideology and party identification was only .34.
However, among members of the same cohort during the 1990s, the correlation between
ideology and party identification was .61.
[Table 6 goes here]
Ideological Realignment vs. Partisan Persuasion Among White Southerners and Catholics
The evidence examined thus far indicates that the relationship between ideology
and party identification became considerably stronger among both northern and southern
whites who remained in the electorate between the 1970s and the 1990s. However,
cohort analysis does not allow us to determine whether ideology was influencing party
14
identification, as the ideological realignment hypothesis suggests, or whether party
identification was influencing ideology, as Green et al. suggest.
It is possible that the increasing clarity of ideological differences between the
parties during the 1980s and 1990s caused Democratic and Republican partisans to adopt
ideological positions consistent with their existing party loyalties in a process that might
be termed partisan persuasion. However, partisan persuasion cannot explain increasing
Republican identification among major subgroups within the electorate. Moreover,
evidence from National Election Study surveys indicates that for white southerners and
Catholics, two subgroups within the white electorate that experienced substantial
increases in Republican identification between 1972 and 2004, ideological realignment
rather than partisan persuasion was the primary mechanism of change.
If partisan persuasion was at work, increases in Republican identification among
white southerners and Catholics between 1972 and 2004 should have led to substantial
increases in conservatism in these groups as the growing ranks of Republican identifiers
adopted the conservative ideology of their new party. But the data displayed in Figures 1
and 2 show that there was no increase in conservatism among either white southerners or
white Catholics. Throughout the period from 1972 though 2004, the mean conservatism
score of white southerners hovered around 4.5 while the mean conservatism score of
white Catholics remained in the vicinity of 4.2.
[Figures 1 and 2 go here]
Contrary to the partisan persuasion hypothesis, white southerners and Catholics did
not become much more conservative between 1972 and 2004; however, conservative
white southerners and Catholics did become much more Republican. Between 1972 and
15
2004, the proportion of conservative southern whites identifying with the Republican
Party increased from 39 percent to 80 percent and the mean score of conservative
southerners on the seven-point party identification scale rose from 3.7 to 5.5. During the
same period, the proportion of conservative Catholics identifying with the Republican
Party increased from 48 percent to 82 percent and the mean score of conservative
Catholics on the seven-point party identification scale rose from 4.1 to 5.7. This
evidence clearly indicates that ideological realignment rather than partisan persuasion
was responsible for the increasing correlation between ideology and party identification
among white southerners and Catholics.
Evidence from the 1992-1996 NES Panel Survey
Additional evidence concerning the relative importance of ideological realignment
and partisan persuasion can be obtained from the only major long-term panel study
conducted by the NES between 1976 and 2000—the 1992-96 panel survey. Although the
study covers only a four year period and the sample is fairly small (only 597 respondents
were interviewed in both 1992 and 1996), the panel design of the study allows us to
estimate the influence of ideology on party identification as well as the influence of party
identification on ideology.
[Figure 3 goes here]
Figure 3 presents the results of a path analysis of ideology and party identification
among white respondents in the 1992-96 NES panel survey. These results indicate that
there was a high degree of stability in both ideological orientations and party
identification among survey respondents. In fact, ideological orientations were even
more stable than party identification over the four years of the panel. Despite the high
16
degree of stability of party identification, however, the results of the path analysis
provide strong support for the ideological realignment hypothesis. Even after controlling
for 1992 party identification and a wide variety of social background characteristics,
1992 ideological orientations had a significant impact on 1996 party identification. Over
this four year period, conservatives tended to become more Republican while liberals
tended to become more Democratic. There is much less support for the partisan
persuasion hypothesis. Over the same four year period, Democrats did not become
significantly more liberal and Republicans did not become significantly more
conservative.
Social Identity, Ideology and Party Identification
According to the social identity theory, party identification is based largely on
membership in social groups—citizens choose a party identification based on their
perception of the fit between their own social characteristics and the social characteristics
of supporters of the two major parties. Since the New Deal, the Democrats have
generally been viewed as the party of the poor, the working class, union members, urban
dwellers, racial and ethnic minorities, Catholics, and Jews while the Republicans have
generally been viewed the party of the wealthy, business executives, small town and rural
residents, and white Protestants outside of the South. However, the social images of the
Democratic and Republican parties have undergone considerable change in recent years.
As Green et al. point out, southern whites, who were once a key component of the
Democratic coalition, have been moving into the Republican camp since the 1950s.
More recently, gender, marital status, sexual orientation, and religious beliefs have
emerged as important correlates of party affiliation: members of traditional families and
17
those with strong religious convictions tend to be Republicans while singles, gays, and
less religious voters tend to be Democrats.
The changing relationship between social groups and the parties raises the question
of whether membership in social groups has a direct impact on party identification, as the
social identity theory proposes, or whether partisan differences between social groups are
simply a result of the policy preferences of group members. According to this
ideological differences hypothesis, the reason that some groups such as white
evangelicals have become increasingly Republican in recent years while other groups
such as gays and lesbians have become increasingly Democratic is because of the policy
preferences of their members.
As a first test of the social identity and ideological differences hypotheses, Table 7
presents data from the 2004 national exit poll on net party identification among members
of a number of groups that are closely aligned with either the Democratic Party or the
Republican Party, controlling for ideological orientation. We measured the ideological
orientations of respondents in the exit poll with the three-point liberal-conservative
identification question because this question was included in all three versions of the exit
poll questionnaire.2
[Table 7 goes here]
The results in Table 7 show that, except for African-Americans, the differences
between liberals and conservatives within each social group were much larger than the
differences between social groups. African-Americans, regardless of their ideological
2 However, responses to this question were strongly related to opinions on three policy questions that were
included in different versions of the questionnaire: abortion (r = .40), gay marriage (r = .42) and the role of
the federal government (r = .31).
18
orientation, strongly favored the Democratic Party. Otherwise, across all social groups,
liberals strongly preferred the Democratic Party and conservatives strongly preferred the
Republican Party. While the large majority of Hispanics identified with the Democratic
Party, the large majority of conservative Hispanics identified with the Republican Party.
Jews overwhelmingly identified with the Democratic Party but conservative Jews
overwhelmingly identified with the Republican Party. Wealthy liberals favored the
Democrats while poor conservatives favored the Republicans; conservative gays and
lesbians preferred the Republican Party by a wide margin while liberal evangelicals (yes,
there were some) preferred the Democratic Party by a wide margin.
The results in Table 7 provide only limited supported for the social identity theory.
It is true that the partisan orientations of certain groups cannot be completely explained
by their policy preferences. For African-Americans, in particular, social identity and
party identification seem to be closely connected. African-Americans, regardless of
ideology, tend to be Democrats. For other groups, however, the connection between
social identity and party identification is much weaker or nonexistent. Even for members
of groups with very close ties to one party or the other, such as Jews or evangelical
Christians, ideology trumps social identity. The reason why most Jews identify with the
Democratic Party is because of their liberal policy preferences, not because of their social
identity; the reason why most evangelical Christians identify with the Republican Party is
because of their conservative policy preferences, not because of their social identity.
Social Identity, Ideology, and Party Identification in 2004
In order to directly compare the effects of ideology and group membership on
contemporary party identification, we performed a logistic regression analysis of party
19
identification. Our dependent variable in this analysis was a dichotomous measure of
party identification with strong, weak, and independent Democrats classified as
Democrats and strong, weak, and independent Republicans classified as Republicans.
Pure independents were excluded from the analysis.3 Independent variables in the
analysis included a 16-item liberal-conservative scale and a variety of social background
characteristics including age, education, income, gender, marital status, religion, church
attendance, and household union affiliation. The results of the logistic regression
analysis are presented in Table 8.
[Table 8 goes here]
To facilitate comparisons of the effects of the independent variables on party
identification, we converted each of the logistic regression coefficients into a change in
probability score. This score can be interpreted as the change in the probability of
identifying with the Republican Party associated with a change between categories of any
of the dichotomous independent variables such as gender, union membership, or martial
status, or the change in probability associated with a change between the 25th percentile
and the 75th percentile on any of the continuous independent variables such as age,
education, or ideology.
The major conclusion that emerges from Table 8 is that the impact of ideology on
party identification was much stronger than that of any of the social background
variables. Most of the estimated coefficients for the social background variables are not
statistically significant. In addition, the change in probability scores for most of the
social background variables are generally small, with the largest effect (.16) being for
3 An OLS regression analysis using the 7-point party identification scale as the dependent variable
produced almost identical results.
20
education. Many social characteristics including age, income, gender, martial status, and
church attendance had little or no impact on party identification after controlling for
ideology. In contrast, the estimated coefficient for the ideology scale is highly
statistically significant and the change in probability score is almost four times larger
than that for education. Even after controlling for social background characteristics, the
probability of identifying with the Republican Party was 63 percentage points higher for
a voter at the 75th percentile of the liberal conservative scale than it was for a voter at the
25th percentile of the liberal-conservative scale.
According to Green, Palmquist, and Schickler, the questions most voters ask
themselves in deciding which party to support are: “What kinds of social groups come to
mind as I think about Democrats, Republicans, and Independents?” and “Which
assemblage of groups (if any) best describes me?” Based on our evidence, however, it
appears that the questions most voters ask themselves in deciding which party to support
are actually: “What do Democrats and Republicans stand for?” and “Which party’s
positions are closer to mine?”
The Consequences of Ideological Realignment for Voting Behavior
The growing consistency of ideology and party identification has important
consequences for voting behavior because voters whose party identification and
ideological orientation are consistent are much more loyal to their party than voters
whose party identification and ideological orientation are inconsistent. In the 2004
presidential election, according to data from the national exit poll, 96 percent of liberal
white Democrats voted for John Kerry compared with only 62 percent of conservative
21
white Democrats. Similarly, 97 percent of conservative white Republicans voted for
George W. Bush compared with only 58 percent of liberal white Republicans.
Overall, according to both national exit polls and the NES post-election surveys,
over 90 percent of Republican identifiers and almost 90 percent of Democratic identifiers
voted for their own party’s presidential candidates in the 2000 and 2004 elections. These
two elections produced the highest levels of party voting in the history of the National
Election Studies. Party voting was also very prevalent in recent congressional elections,
especially in competitive races (Abramowitz and Alexander 2004).
The high level of partisan voting in recent presidential and congressional elections
is due largely to the fact that an ideological realignment has taken place among white
voters in both the South and the North since the 1970s. As a result of this realignment,
voters’ party affiliations are now more consistent with their ideological orientations than
in the past. According to data from the American National Election Studies, liberal
Democrats and conservative Republicans made up only 42 percent of all white party
identifiers in 1972, while conservative Democrats and liberal Republicans made up 20
percent. By 2004, liberal Democrats and conservative Republicans made up 59 percent
of all white party identifiers while conservative Democrats and liberal Republicans made
up only 9 percent. Because of this growing consistency, the outlook for the 2006 and
2008 elections is for a continuation of high levels of partisan voting.
22
References
Abramowitz, Alan I., and Brad Alexander. 2004. Incumbency, Redistricting, and the
Decline of Competition in Congressional Elections: Evidence from the 2002
Midterm Election. Paper presented at the Annual Meeting of the Western Political
Science Association, Portland, Oregon.
Abramowitz, Alan I., and Kyle L. Saunders. 1998. Ideological Realignment in the U.S.
Electorate. Journal of Politics, 60: 634-652.
Axelrod, Robert. 1967. The Structure of Public Opinion on Policy Issues. Public
Opinion Quarterly, 31: 51-60.
Bishop, George F., Robert W. Oldenick, Alfred J. Tuchfarber, and Stephen E. Bennett.
1978. The Changing Structure of Mass Belief Systems: Fact or Artifact? Journal
of Politics, 40: 781-787.
Brooks, David. 2004. Circling the Wagons. New York Times, June 5, Section A, p. 15.
Campbell, Angus, Philip E. Converse, Warren E. Miller, and Donald E. Stokes. The
American Voter. 1960. New York: John Wiley & Sons.
Carmines, Edward G., John P. McIver, and James A. Stimson. 1987. Unrealized
Partisanship: A Theory of Dealignment. Journal of Politics, 49: 376-400.
Conover, Pamela J., and Stanley Feldman. 1981. The Origins and Meaning of
Liberal/Conservative Self-Identification. American Political Science Review, 25:
617-645.
Converse, Philip E. 1964. The Nature of Belief Systems in Mass Publics. In Ideology
and Discontent, edited by David Apter. Glencoe, Illinois: The Free Press.
23
Craig, Stephen C., and Thomas L. Hurley. 1984. Political Rhetoric and the Structure of
Political Opinion: Some Experimental Findings. Western Political Quarterly, 37:
632-640.
Downs, Anthony. 1957. An Economic Theory of Democracy. New York: Harper and
Row.
Field, John O., and Ronald E. Anderson. 1969. Ideology in the Public’s
Conceptualization of the 1964 Election. Public Opinion Quarterly, 33: 380-398.
Fiorina, Morris P. 1981. Retrospective Voting in American National Elections. New
Haven, CT: Yale University Press.
Franklin, Charles H. 1992. Measurement and the Dynamics of Party Identification.
Political Behavior, 14: 297-309.
Franklin, Charles H., and John E. Jackson. 1983. The Dynamics of Party Identification.
American Political Science Review, 77: 957-973.
Gerring, John. 1997. Ideology: A Definitional Analysis. Political Research Quarterly,
50: 957-994.
Green, Donald, Bradley Palmquist, and Eric Schickler. 2002. Partisan Hearts and
Minds: Political Parties and the Social Identities of Voters. New Haven, CT: Yale
University Press.
Gurin, Patricia. 1985. Women’s Gender Consciousness. Public Opinion Quarterly, 49:
143-163.
Gurin, Patricia, Arthur H. Miller, and Gerald Gurin. 1980. Stratum Identification and
Consciousness. Social Psychology Quarterly, 43: 30-47.
24
Guth, James L., and John C. Green. 1990. Politics in a New Key: Religiosity and
Participation among Political Activists. Western Political Quarterly, 43: 153-179.
Jacoby, William G. 1995. The Structure of Ideological Thinking in the American
Electorate. American Journal of Political Science, 39: 314-335.
Jennings, M. Kent. 1992. Ideological Thinking Among Mass Publics and Political
Elites. Public Opinion Quarterly, 56: 419-441.
Kaufmann, Karen M., and John R. Petrocik. 1999. The Changing Politics of American
Men: Understanding the Sources of the Gender Gap. American Journal of Political
Science, 43: 864-887.
Knight, Kathleen. 1985. Ideology in the 1980 Election: Ideological Sophistication Does
Matter. Journal of Politics, 47: 828-853.
Layman, Geoffrey C., and Edward G. Carmines. 1997. Cultural Conflict in American
Politics: Religious Traditionalism, Postmaterialism, and U.S. Political Behavior.
Journal of Politics, 59: 751-777.
Luskin, Robert C., John P. McIver, and Edward G. Carmines. 1989. Issues and the
Transmission of Partisanship. American Journal of Political Science, 33: 440-458.
Miller, Arthur H., Patricia Gurin, Gerald Gurin, and Oksana Malanchuk. 1981. Group
Consciousness and Political Participation. American Journal of Political Science,
25: 494-511.
Nie, Norman H., and Kristi Anderson. 1974. Mass Belief Systems Revisited: Political
Change and Attitude Structure. Journal of Politics, 36: 540-591.
25
Nie, Norman H., and James N. Rabjohn. 1979. Revisiting Mass Beliefs Systems
Revisited: Or, Doing Research is Like Watching a Tennis Match. American
Journal of Political Science, 23: 139-175.
Nie, Norman H., Sidney Verba, and John R. Petrocik. 1979. The Changing American
Voter. Cambridge, MA: Harvard University Press.
Page, Benjamin I., and Calvin C. Jones. 1979. Reciprocal Effects of Policy Preferences,
Party Loyalties, and the Vote. American Political Science Review, 73: 1071-1089.
Peffley, Mark A., and Jon Hurwitz. 1985. A Hierarchical Model of Attitude Constraint.
American Journal of Political Science, 29: 871-890.
Poole, Keith T., and Howard Rosenthal. 1991. Patterns of Congressional Voting.
American Journal of Political Science, 35: 228-278.
Saunders, Kyle L., and Alan I. Abramowitz. 2004. Ideological Realignment and Active
Partisans in the American Electorate. American Politics Research, 32: 285-309.
Stonecash, Jeffrey M., Mark D. Brewer, and Mack D. Mariani. 2003. Diverging Parties:
Social Change, Realignment, and Party Polarization. Boulder, CO: Westview
Press.
Sullivan, John L., James E. Piereson, George E. Marcus, and Stanley Feldman. 1978.
The More Things Change, the More They Stay the Same: The Stability of Mass
Belief Systems. American Journal of Political Science, 23: 176-186.
Tajfel, Henri. 1981. Human Groups and Social Categories: Studies in Social
Psychology. Cambridge, UK: Cambridge University Press.
26
Tajfel, Henri, and John C. Turner. 1986. The Social Identity Theory of Intergroup
Behavior. In Stephen Worchel and William G. Austin, ed., Psychology of
Intergroup Relations. Chicago: Nelson-Hall.
Turner, John C. 1982. Toward a Cognitive Redefinition of the Social Group. In Henri
Tajfel, ed., Social Identity and Intergroup Relations. Cambridge, UK: Cambridge
University Press.
Turner, John C. 1987. Rediscovering the Social Group: A Self Categorization Theory.
Oxford, UK: Basil Blackwell.
Wirls, Daniel. 1986. Reinterpreting the Gender Gap. Public Opinion Quarterly, 50:
316-330.
Weisberg, Herbert F. 1987. The Demographics of a New Voting Gap: Marital
Differences in American Voting. Public Opinion Quarterly, 51: 335-343.
27
Table 1. Net Party Identification in the United States by Race and Region
_______________________________________________________________________
Group 1962-1970 1972-1980 1982-1990 1992-2000 2002-2004 Change
_______________________________________________________________________
ALL + 24 + 20 + 13 + 11 + 7 - 17
African-
Americans + 72 + 74 + 72 + 72 + 75 + 3
Whites + 18 + 13 + 1 0 - 7 - 25
South + 36 + 25 + 9 - 7 - 17 - 53
North + 13 + 9 - 2 + 3 - 5 - 18
________________________________________________________________________
Source: American National Election Studies
Note: Net party identification = percentage of Democratic identifiers and leaners minus
percentage of Republican identifiers and leaners.
28
Table 2. Net Party Identification of Northern White Subgroups
_______________________________________________________________________
Group 1962-1970 1972-1980 1982-1990 1992-2000 2002-2004 Change
_______________________________________________________________________
Males + 14 + 8 - 6 - 6 - 10 - 24
Females + 12 + 9 + 2 + 11 - 1 - 13
Married +13 + 6 - 7 - 5 - 14 - 27
Unmarried + 12 + 16 + 6 + 14 + 10 - 2
Protestant - 5 - 10 - 18 - 14 - 20 - 15
Catholic + 42 + 36 + 17 + 13 - 2 - 44
Jewish + 72 + 58 + 46 + 72 + 67 - 5
Other, None + 29 + 28 + 16 + 23 + 15 - 14
Observant + 6 0 - 9 - 16 - 16 - 22
Nonobservant + 27 + 21 + 9 + 18 + 15 - 12
________________________________________________________________________
Source: American National Election Studies
Note: Net party identification = percentage of Democratic identifiers and leaners minus
percentage of Republican identifiers and leaners.
29
Table 3. Analysis of 8-Item Liberal-Conservative Scale, 1984-2004
________________________________________________________________________
1984 1988 1992 1996 2000 2004
________________________________________________________________________
Cronbach’s Alpha .65 .66 .70 .75 .74 .77
Correlation of
Lib/Con Id with
Jobs/Living Standards .36 .23 .28 .36 .35 .43
Health Insurance .19 .23 .30 .39 .31 .43
Spending/Services .24 .33 .32 .38 .46 .42
Aid to Blacks .31 .26 .27 .37 .32 .37
Defense Spending .27 .30 .28 .35 .35 .34
Abortion .14 .21 .32 .37 .39 .33
Women’s Role .19 .23 .28 .31 .32 .32
7 Policy Issues .44 .46 .51 .60 .60 .62
________________________________________________________________________
Source: American National Election Studies
Note: Correlations are Pearson’s r. All coefficients are highly statistically significant (p <
.001).
30
Table 4. Net Party Identification of Selected White Subgroups by Ideological
Self-Identification
_______________________________________________________________________
Group 1972-1980 1982-1990 1992-2000 2002-2004 Change
_______________________________________________________________________
ALL WHITES
Liberals + 53 + 52 + 66 + 66 + 13
Moderates + 20 + 10 + 11 + 12 - 8
Conservatives - 29 - 44 - 54 - 67 - 38
SOUTH
Liberals + 55 + 49 + 60 + 62 + 7
Moderates + 33 + 27 + 6 + 3 - 30
Conservatives - 6 - 31 - 59 - 60 - 54
NORTH
Liberals + 52 + 53 + 68 + 68 + 16
Moderates + 16 + 7 + 13 + 11 - 5
Conservatives - 37 - 49 - 52 - 70 - 33
MALES
Liberals + 53 + 49 + 63 + 62 + 9
Moderates + 22 + 12 + 10 + 14 - 8
Conservatives - 30 - 49 - 59 - 70 - 40
MARRIED
Liberals + 51 + 49 + 68 + 63 + 12
Moderates + 20 + 11 + 8 + 7 - 13
Conservatives - 31 - 48 - 56 - 69 - 38
CATHOLICS
Liberals + 70 + 54 + 71 + 59 - 11
Moderates + 43 + 28 + 21 + 6 - 37
Conservatives - 7 - 30 - 42 - 59 - 52
OBSERVANT
Liberals + 42 + 54 + 65 + 60 + 18
Moderates + 18 + 13 + 9 + 13 - 5
Conservatives - 29 - 46 - 60 - 69 - 40
________________________________________________________________________
Source: American National Election Studies
Note: Net party identification = percentage of Democratic identifiers and leaners minus
percentage of Republican identifiers and leaners.
31
Table 5. Correlation between Party Identification and Ideological
Self-Identification by Decade
_______________________________________________________________________
Group 1972-1980 1982-1990 1992-2000 2004 Change
_______________________________________________________________________
All Respondents .35 .39 .49 .58 + .23
African-Americans .14 .14 .24 .23 + .09
Whites .34 .42 .53 .64 + .30
South .26 .33 .50 .55 + .29
North .38 .45 .54 .66 + .28
No College .26 .27 .34 .49 + .23
College .47 .54 .65 .71 + .24
________________________________________________________________________
Source: American National Election Studies
Note: Product-moment correlations between 7-point party identification scale and 7-point
liberal-conservative scale. All coefficients are highly statistically significant (p < .001).
32
Table 6. Cohort Analysis of Correlation between Ideological Self-Identification and
Party Identification for White Respondents by Decade
________________________________________________________________________
Region Age Group 1972-1980 1982-1990 1992-2000
________________________________________________________________________
South 20-29 .27 .28 .54
30-39 .36 .39 .58
40-49 .22 .44 .54
50-59 .21 .40 .56
60 + .32 .32 .45
North 20-29 .34 .44 .59
30-39 .37 .52 .57
40-49 .39 .48 .61
50-59 .38 .49 .57
60 + .38 .39 .46
________________________________________________________________________
Source: American National Election Studies.
Note: Pearson product-moment correlations between 7-point liberal-conservative scale
and 7-point party identification scale. Are coefficients are highly statistically significant
(p <.001).
33
Year
2004
2002
2000
1998
1996
1994
1992
1990
1988
1986
1984
1982
1980
1978
1976
1974
1972
Mean
4.8
4.6
4.4
4.2
4.0
3.8
3.6
3.4
3.2
3.0
Libcon
Partyid
Figure 1. Trends in Ideological Self-Identification and Party Identification for
White Southerners
34
Year
2004
2002
2000
1998
1996
1994
1992
1990
1988
1986
1984
1982
1980
1978
1976
1974
1972
Mean
4.6
4.4
4.2
4.0
3.8
3.6
3.4
3.2
3.0
Libc on
Partyid
Figure 2. Trends in Ideological Self-Identification and Party Identification for
White Catholics
35
Source: 1992-1996 American National Election Study Panel Survey.
Note: Party Identification measured by standard 7-point party id scale. Ideological
orientation measured by 11-item scale. Coefficients shown are standardized regression
coefficients. Control variables included in regression analyses are age, church
attendance, religion, region, gender, marital status, education, family income, and
household union membership. Coefficients marked with asterisk are highly statistically
significant (p < .001).
Figure 3. Path Analysis of Ideology and Party Identification for White Respondents
in 1992-1996 Panel Survey
1992
Libcon
Scale
1996
Libcon
Scale
1992
Party
Id
1996
Party
Id
36
Table 7. Net Party Identification by Social Identity and Ideological Identification in
2004 National Exit Poll
_______________________________________________________________________
Group Liberal Moderate Conservative
________________________________________________________________________
African-Americans + 76 + 71 + 44
Hispanics + 44 + 16 - 42
Whites + 52 - 4 - 70
Income LT $15,000 + 50 + 33 - 47
Income $200,000 + + 73 - 21 - 88
Union Members + 65 + 24 - 53
Northeast +53 + 1 - 61
South + 38 - 6 - 75
Big City Dwellers + 61 + 8 - 83
Small Town, Rural Residents + 30 - 2 - 62
Protestant + 46 - 6 - 74
Catholic + 49 - 4 - 61
Jewish + 87 + 27 - 50
Most Observant + 39 - 9 - 75
Least Observant + 53 - 2 - 55
Born Again/Evangelical + 22 - 11 - 73
Male + 29 - 19 - 57
Female + 45 - 1 - 46
Married + 53 - 6 - 73
Single + 52 + 2 - 60
Gun Owners + 43 - 8 - 71
Gay, Lesbian, Bisexual + 45 + 24 - 28
________________________________________________________________________
Note: Net Party Identification = Percentage of Democratic Identifiers – Percentage of
Republican Identifiers.
37
Table 8. Logistic Regression Analysis of Party Identification of White Respondents with
Liberal-Conservative Policy Scale and Social Background Characteristics in 2004 NES
Survey
________________________________________________________________________
Independent Change in
Variable B (S.E.) z Probability Sig.
________________________________________________________________________
Lib-Con Scale .313 (.025) 12.35 .63 .001
Age - .010 (.006) - 1.57 - .06 N.S.
Education .254 (.074) 3.43 .16 .001
Income .005 (.021) .26 .01 N.S.
Female - .230 (.208) - 1.10 - .05 N.S.
Married .342 (.222) 1.54 .08 N.S.
Non-Union .116 (.064) 1.72 .10 .05
Religion
Catholic - .584 (.242) - 2.41 - .14 .01
Jewish - .672 (.621) - 1.08 - .16 N.S.
No Religion - .024 (.323) - .07 - .01 N.S.
Church Frequency .031 (.076) .41 .02 N.S.
Constant - 1.068 (.745) - 1.43 N.S.
Pseudo R2 = .38
N = 706
_______________________________________________________________________
Source: 2004 American National Election Study
Note: Dependent variable is dichotomous party identification (strong, weak, and
independent Republican vs. strong, weak, and independent Democrat). Change in
probability is estimated difference in probability of Republican identification between
categories of dichotomous variables (non-union vs. union household; female vs. male;
Catholic, Jewish, or no religion vs. Protestant; married vs. not married) or between 25th
and 75th percentiles of continuous variables and scales (age, education, church frequency,
lib-con).
