Social Identity Theory and

Party Identification n

Steven Greene, North Carolina State University

Objective. Given that the group aspect of party identification forms a central, yet
largely unexplored element of American partisanship, social identity theory presents
a compelling social-psychological theory of group belonging through which to
reinterpret the contemporary understanding of partisanship. Methods. Using a
mail survey of 302 randomly selected Franklin County, Ohio residents, levels of
social identification with the Democratic Party, the Republican Party, and political
independents are measured using the Identification with a Psychological Group
(IDPG) scale. Scores on the IDPG are used to predict attitudes toward parties and
the consistency of partisan behavior. Results. Levels of partisan social identity
proved to be significant predictors of political party ratings, ideology, and party
activities, even when taking traditional measures of partisan strength into
account. Conclusions. Social identity is a fundamental aspect of partisanship,
which, when measured, can lead to superior prediction and understanding of related
political attitudes and behaviors.

Since The American Voter (Campbell et al., 1960) first described party
identification as a central component of political behavior, the concept has
been continually challenged, redefined, and modified. The most contested
areas of research on partisanship—the stability, dimensionality, and causal
ordering of partisanship—all leave aside the question of the psychological
investment in the political party as a group (see Niemi and Weisberg, 1993;
Weisberg and Greene, 2003). Yet, the original conception of party
identification in The American Voter is a precursor of social identity theory
years ahead of its time. The authors essentially state that just as people
identify with various racial, ethnic, and religious groups, so too do they
identify with political parties. Based on extensive experimentation and
development, social identity theory (Tajfel, 1978) is able to build on this
foundation and provide a rich theoretical framework for understanding the
perceptual patterns of modern partisanship. Consequently, the aim of this
study is not to argue for a new group-based conceptualization of
partisanship, but rather to build on this often-ignored foundation of party

n
Direct correspondence to Steven Greene, North Carolina State University, Department
of Political Science and Public Administration, Box 8102, Raleigh, NC 27695-8102
[email protected]. The author would like to acknowledge the valuable assistance of
Paul Allen Beck and the anonymous reviewers of the manuscript. The author will share data
for purposes of replication.

SOCIAL SCIENCE QUARTERLY, Volume 85, Number 1, March 2004

r2004 by the Southwestern Social Science Association
Social Identity Theory and Party Identification 137
identification with contemporary social-psychological theory. In this article,
I demonstrate that social identity theory yields important insights into the
nature of partisan-related attitudes and partisanship itself.
In short, I argue, (1) an individual’s attachment to a political party
as a meaningful psychological group is a fundamental aspect of partisanship;
(2) this group identification aspect of partisanship was properly addressed as
a fundamental aspect of partisanship in The American Voter, but has since
been largely ignored as scholars have focused on partisanship as an attitude
toward a political party; (3) substantial developments in social-psychological
theory, primarily social identity theory, allow us to return to the group-
belonging aspect of partisanship with an improved theoretical framework;
(4) social identity theory can help us to better understand the role of
partisanship in political attitudes and behavior as well as lead to more
complete measures of partisanship.

Explaining Partisan Attitudes and Behavior with Social Identity

Social identity theory attempts to explain how self-perceived membership

in a social group affects social perceptions and attitudes. Social identity is
defined as ‘‘that part of an individual’s self concept which derives from his
knowledge of his membership of a group (or groups) together with the value
and emotional significance attached to the membership’’ (Tajfel, 1978). The
theory holds that individuals attempt to maximize differences between the
in-group (the group to which one psychologically belongs) and the out-
group (psychologically relevant opposition group) and thus perceive greater
differences between one’s in-group and the relevant out-group than actually
exist and show favoritism toward in-group members (Tajfel and Turner,
1986). Importantly, these social identifications are not based on any formal
group membership, but rather self-perceived membership in a particular
group. Given the informal basis of membership in American political
parties, the already apparent perceptual biases discovered in political
partisanship, and evidence for social identity in partisanship in social-
psychological studies of Great Britain (Kelly, 1988), Scotland (Abrams,
1994), Australia (Duck, Hogg, and Terry, 1995), and American college
students (Greene, 1999), social identity theory presents a powerful paradigm
through which to examine party identification in the American context.
The study of political partisanship and group identification has a
substantial history and extends all the way back to the classic Columbia
voting studies (e.g., Berelson, Lazarsfeld, and McPhee, 1954), which finds
strong relationships between social groups and partisanship. More recently,
Stanley and Niemi (1995) have shown the recently attenuated, though still
substantial, relationship between various social groups and party identifica-
tion. Building on the reference group theory basis of the American Voter,
Wlezien and Miller (1997) and Miller, Wlezien, and Hildreth (1991)
138 Social Science Quarterly
demonstrate that partisanship is substantially influenced by perceptions of
the various groups associated with parties. Although these group studies have
much to offer, a common feature is that they fail to explicitly consider the
group-identification aspect of party membership. Consequently, it makes
sense to turn to social identity theory to fully address this area of concern.
So, what should we expect partisanship based on a social identification to
look like? Not all that different than the traditional interpretation, as again,
Campbell et al. (1960) implicitly describe partisanship as a social
identification. Following from the group-differentiation aspect of social
identity theory, the group nature of partisanship should naturally create a
bipolar partisanship where individuals characterize the political parties into
us and them and exaggerate perceived differences to favor their own group.
In social identity theory, bipolar partisan attitudes are a natural
psychological outgrowth of self-perceived membership in a political party.
In social identity theory, intergroup differentiation occurs in two primary
ways: in-group favoritism and out-group derogation (Brewer and Brown,
1998). In-group favoritism simply refers to the tendency for group members
to mentally exaggerate and enhance the favorable qualities of the relevant in-
group to which they see themselves belonging. Out-group derogation, in
contrast, is exaggeration of the negative characteristics of relevant out-
groups, thereby also making one’s in-group seem superior. The net result of
either process is enhanced group differentiation. Interestingly, in-group
favoritism and out-group derogation need not co-occur (Brewer and Brown,
1998). Consequently, in the case of partisanship, we may expect either or
both processes to be a prime causal factor behind bipolar partisan attitudes.
In addition to an impact on partisan attitudes, we might also expect
individuals with stronger partisan social identity to exhibit increased partisan
behavioral consistency. A stronger partisan social identity should lead to
greater differentiation between the parties, thus making defection from one’s
preferred party less likely. We might therefore expect that as a person’s
partisan social identity increases, he or she should be more likely to vote a
straight ticket, vote for his or her party’s presidential candidate, and vote for
his or her party more consistently across elections. Along with behavior at
the ballot box, we might also expect higher levels of partisan political
participation as social identity increases. Many partisan activities are group
activities, for example, attending meetings, rallies, and other forms of
volunteering. Certainly the commitment to the party as a group should
foster participation in such activities. Furthermore, even in noncommunal
activities, for example, contributing money or trying to persuade others how
to vote, the increased differentiation between the two parties provides more
incentive to work for one’s own party.
At least in this cross-sectional design, the above predictions of social
identity theory do not look all that different than what we might expect
from traditional predictions based on strength of partisanship in the
American Voter model. What is the advantage then? Explicitly measuring
Social Identity Theory and Party Identification 139
partisan group identification as an independent concept offers not only
enhanced predictive power in political models by finally providing an
accurate measurement of this fundamental basis of partisanship, but also
allows for a greater understanding of the related roles of partisan group
identification and partisan attitudes in explaining political behavior. Social
identity theory does, in fact, lead to testable propositions about specific
perceptions of in-group and out-group members (e.g., Kelly, 1990) and
dynamics in intergroup competition, but these are peripheral to the central
concerns of mass behavior. The existence of nontrivial levels of partisan
social identification on a proper scale and, more significantly, a clear
explanatory impact on major political variables above and beyond that of
traditional measures provides facial validity for the social identity approach.
As discussed further in the Conclusion, the greatest benefit may come in
incorporating social identity measures over time in order to address long-
controversial questions about the nature of partisan stability.

The Dimensionality of Partisanship and the Party Identification Scale

In addition to explaining basic attitudes toward political parties, social

identity also offers the potential to explain both partisanship and political
independence within a single coherent framework. Weisberg (1980) and
others (e.g., Alvarez, 1990; Dennis, 1988) have argued that the traditional
seven-point categorization of partisanship forces two separate attitudinal
dimensions—partisanship and independence—onto a single bipolar scale.
They argue that, to a considerable degree, attitudes toward parties and
attitudes toward political independence are separate and distinct. Social
identity offers a new perspective on what appears to be two quasi-distinct
dimensions of partisanship and independence. Certainly we should expect
that many persons will have a clear partisan social identity. However, given
the strong civic virtue and social norms placed on political independence in
America, it may be that either in addition to, or in place of, a social
identification with a political party, some citizens may also socially identify
with other political independents. Given the pervasive media treatment of
independents as a homogenous entity and the fact that in many states
persons can register as ‘‘independent,’’ just as they would as a Republican or
Democrat, it may not be too implausible for some citizens to conceive of
independents as just another political group to which they do or do not
belong.
Social identity in combination with a multidimensional view of
partisanship in which independence and partisanship occupy quasi-
independent dimensions (e.g., Alvarez, 1990; Dennis, 1988) offers a
potential explanation for the seemingly incongruous behavior and self-
identification of partisan leaners—independents who express a partisan
preference. These ‘‘leaners’’ are often just as partisan, if not more so, than
140 Social Science Quarterly
weak partisans, despite identifying themselves as independents (Greene,
2000; Keith et al., 1992; Petrocik, 1974). An independent social
identification may exist somewhat independently of a partisan social
identification and cause an individual to identify as an independent, yet not
prevent the person from behaving in a partisan fashion. In the multi-
dimensional perspective, this does not entail that the person has no partisan
social identification, but rather that independent social identification may
prove more personally relevant in the realm of politics. We might expect
that leaners are marked by a greater independent social identification than
true partisans and most likely less partisan social identification.

Data and Methods

The respondents for the study consisted of 302 randomly selected

registered voters in Franklin County, Ohio who completed a mail survey in
February and March 1998.1 Although not ideally representative, due to the
high levels of respondent motivation necessary in a mail survey, the
respondents to the survey proved to be a fairly diverse and heterogeneous
group.2 It is only fair to mention that the strength of social identity and the
relationships between variables may be overstated among those so motivated
to reply to the survey, yet any such findings should surely be quite suggestive
and nonetheless demonstrate the benefits of this theoretical approach.
Consequently, the focus here is not on generalizing absolute levels of social
identity to the American public, but rather on demonstrating the existence
of social identity in partisanship and how it influences one’s political
attitudes and behaviors in generalizeable patterns.
The key feature of the survey was an assessment of partisan and
independent social identification. Social identity was measured using Mael
and Tetrick’s (1992) Identification with a Psychological Group (IDPG)
scale. Brewer and Silver (2000) have found the scale to be widely applicable
to a variety of social groups, including political parties. The IDPG scale
measures social identity with a designated group through a series of 10
1
The response rate was 24 percent. This is an admittedly problematic response rate
although every effort was made: reward incentives, reminder postcards, and so forth, to
encourage response rate following Dillman’s (1978) Total Design Method.
2
The median age was 45. Modal education level was a college degree, 29 percent; with
considerable minorities having high school, 16 percent; some college, 26 percent; or post-
college education, 28 percent. Household income was divided fairly evenly between
categories based on $20,000 increments, with $40,000 to $60,000 being the median
category. Fifty-three percent of the respondents were female and 89 percent were white.
Twenty-nine percent were Democrats, 43 percent Republicans, and 27 percent independents.
Franklin County itself is quite representative of the nation as a whole, though a little better
educated. Following are several key statistics with the county percentages followed by U.S.
percentages: white: 75.5 percent to 75.1 percent; bachelor’s degree: 31.8 percent to 24.4
percent; median home value $116,200 to $119,600; persons under 18: 25.1 percent to 25.7
percent; median household income: $42,734 to $41,994.
Social Identity Theory and Party Identification 141
questions that assess the respondents’ perceptions of shared identity and
shared experiences with the group (Mael and Tetrick, 1992). The scale
consists of the following items.
1. When someone criticizes this group, it feels like a personal insult.
2. I don’t act like the typical person of this group (reversed).
3. I’m very interested in what others think about this group.
4. The limitations associated with this group apply to me also.
5. When I talk about this group, I usually say ‘‘we’’ rather than ‘‘they.’’
6. I have a number of qualities typical of members of this group.
7. This group’s successes are my successes.
8. If a story in the media criticized this group, I would feel embarrassed.
9. When someone praises this group, it feels like a personal compliment.
10. I act like a person of this group to a great extent.
Respondents completed this scale only for their preferred political party.
Additionally, all respondents completed the IDPG scale with ‘‘people who
call themselves political independents’’ as the reference group, in order to
measure any independent social identification. Respondents indicated how
much they agreed with the statements, for example, ‘‘This group’s successes
are my successes,’’ on a scale from 0, strongly disagree, to 3, strongly agree.
Two straightforward summary social identification measures were created.
For both partisan and independent social identity, respondents received the
mean score, from 0 to 3, of their 10 responses to the IDPG scale. Higher
scores indicated greater social identity. The Cronbach’s alphas for partisan
and independent social identity were 0.85 and 0.86, respectively, indicating
highly reliable instruments.
The survey included a variety of additional items. The measures emulated
standard National Election Study question wordings, modified where
necessary for a self-administered questionnaire. These items included
attitudinal, behavioral, and demographic measures to better explore aspects
of social identity and partisanship. Feeling thermometers assessed overall
party evaluations on the standard 0–100 scale. In-party or preferred-party
feeling thermometer refers to the score for the party with which the
respondent identifies or leans, out-party and nonpreferred-party feeling
thermometer refer to the opposition party. Party strength is simply the
standard seven-point scale folded in the middle, to range from 0, pure
independent, to 3, strong partisan. The ideology measure uses a seven-point
scale running from 1, liberal, to 7, conservative, for Republican; and
1, conservative, to 7 liberal, for Democrats. Consequently, higher values
indicate greater agreement with the preferred party’s ideology. The partisan
activity index is a six-point additive scale of how many activities the
respondent engaged in: trying to persuade others, displaying campaign
paraphernalia, attending rallies, contributing money, voting in the primary,
and other campaign work. Over-time party support is based on a question
asking how frequently one has voted for the presidential candidate of their
142 Social Science Quarterly
preferred party, ranging from 1, never, to 5, always. Political knowledge is a
five-point additive scale indicating the number of factual knowledge items
correctly answered. The items, adopted from Delli Carpini and Keeter
(1996), are the job of Al Gore, branch of government responsible for judicial
review, percent necessary for a veto override, party in control of Congress,
and the more conservative political party. Political interest is measured using
a weighted average of a three-category measure of interest in the 1996
presidential campaign and a four-category measure of general interest in
government, higher values indicating greater interest.
Measures of demographic data included gender, age, education, income,
and race. Gender was a dummy variable coded 0 for males, 1 for females.
Age was simply coded as age of respondent in years. Education was
measured with a four-point scale ranging from 1, HS diploma, to 4, post-
college work. Income was based on a four-category scale of household
income in $20,000 increments, with higher values indicating greater
income. Race was included using a dummy variable coded 1 for nonwhites
and 0 for whites.

Analyses
Social Identity and Political Attitudes

Since social identity should operate in a psychologically comparable

manner for both Democrats and Republicans, they were grouped together
throughout the analyses in order to provide a larger N and thus greater
analytical power. This should not adversely impact the conclusions as
Republicans and Democrats had statistically indistinguishable levels of
partisan identity and independent identity for every level of partisanship.
The most basic analysis is simply the descriptive statistics and frequency
distributions of independent and partisan social identifications (Table 1).

TABLE 1
Descriptive Statistics and Frequency Distribution of Social Identity

Partisan Social Identity Independent Social Identity

Mean 1.56 0.96
Standard deviation 0.52 0.50
Median 1.50 1.00
Minimum 0 0
Maximum 3.0 2.30
Percent in 0–1.0 range 13.4 52.2
Percent in 1.1–2.0 range 72.4 46.8
Percent in 2.1–3.0 range 14.2 1.0
Number 246 291
Social Identity Theory and Party Identification 143
Across the sample of respondents, partisan social identity presents a nearly
classic normal distribution. The results in Table 1 demonstrate a wide
distribution of and meaningful variation in partisan social identity across the
public. Independent social identification, in contrast, is likely not very
meaningful to much of the public. The mean and median responses were
both near 1, thus fully half the sample falls in the bottom possible third of
independent social identification.
It should not be too surprising that independent social identification
scores were so much lower than partisan social identity. Political parties
represent real and meaningful groups that any person could potentially
readily identify with in a way not true for political independents.
Furthermore, except for pure independents (consistently less than 15
percent of the public), the vast majority of Americans still have clear partisan
preferences. Nonetheless, half the sample fell in the range of at least
moderate independent social identity.
A key test for the role of social identity in partisanship is that as partisan
social identity increases, respondents should be more likely to evidence
perceptual exaggerations toward both parties. To test these predictions of
social identity theory, I ran two regressions using partisan social identity
with one’s preferred party as a key independent variable. To examine the
effects of social identity on in-group favoritism, I used preferred-party
feeling thermometer as the dependent variable. Conversely, to assess out-
group derogation, the nonpreferred-party feeling thermometer served as the
dependent variable.
Not surprisingly, strength of partisan social identity was considerably
correlated with partisan strength traditionally measured (r 5 0.48,
po0.001). It is clear, both theoretically and now empirically, that the
traditional measure is already to some degree assessing the group belonging
involved in partisanship. Nonetheless, in order to convincingly demonstrate
that social identity represents a meaningful aspect of partisan perceptions
that is not fully measured by the traditional items, partisan strength,
measured with the folded seven-point party identification scale, serves as an
important control variable. Similarly, ideology serves as a control in response
to the competing realistic group conflict perspective (e.g., Sherif, 1966). We
should expect conservatives to favor Republicans more and liberals to like
Democrats more, regardless of their social identity. A significant coefficient
for social identity in the presence of these controls should indicate that the
perceptual phenomena evoked by a partisan social identification account for
bipolar partisan attitudes above and beyond what can be explained by simple
partisan strength and ideology. Although the multidimensional perspective
would hold independent and partisan social identification to be largely
independent of each other, I included independent social identification to
assess whether this orientation might decrease differentiation between the
two parties. Demographic controls included education, income, gender,
race, and age. Finally, since the sample was more politically interested and
144 Social Science Quarterly
TABLE 2
Social Identity and Political Party Perceptions

In-Party FT Out-Party FT
Social identity 8.870 nn

2.698
(2.944) (10.897)
Ideology 3.298 n n
7.500 n n
(1.106) (1.185)
Independent identity
6.938 n n 0.548
(2.528) (2.707)
Partisan strength 4.154 n
1.594
(1.887) (2.016)
Interest 1.629
5.287
(5.565) (5.930)
Knowledge
0.218 3.690 n n
(1.205) (1.288)
Education
5.723 n n 0.983
(1.274) (1.361)
Income
1.355 0.0171
(1.072) (1.149)
Gender
6.304 n n 0.514
(2.384) (2.545)
Minority 3.804
9.175 n
(4.256) (4.533)
Age
0.023 0.046
(0.080) (0.084)
Constant 69.354 n n 57.469
(10.205) (10.897)
Number 191 192
R2 0.346 0.252
n
p  0.05; p  0.01, one-tailed test.
nn

knowledgeable than the general public, the inclusion of these variables

prevented these factors from serving as confounds.
The results of the first regression model clearly bore out the expectations
from social identity theory with regard to in-group favoritism (Table 2).
Partisan social identity had considerable statistical and substantive
significance. The coefficient, 8.87, was significant at the p 5 0.001 level
and indicated that for each one-point move on the 0–3 social identity scale,
the feeling thermometer toward one’s preferred party increased by nearly
nine points. Importantly, this was controlling for partisan strength and
ideology, which both proved to be statistically significant as well. In fact,
both measured on 0–3 scales, the impact of social identity was significantly
greater (po0.01) and nearly twice that of partisan strength. The social
identity tenet of in-group favoritism was clearly borne out within this
model. Interestingly, independent social identification had a significant
downward impact on in-party favoritism.
Social Identity Theory and Party Identification 145
The second model examined the impact of social identity on out-
group derogation. This model was identical to the previous, except that the
dependent variable was the feeling thermometer for the opposition party,
rather than one’s preferred party. The results here clearly demonstrated
that in-group favoritism and out-group derogation need not be recipro-
cally related (Table 2). Although partisan social identity proved to be a
dominant factor in how one perceives one’s preferred political party, it did
not have a statistically significant impact on how one perceives the
opposition party. Interestingly, partisan strength did not significantly
impact the dependent variable, either. Ideology was left as the sole political
variable driving down ratings of the opposition party. In short, the models
in Table 2 demonstrate that partisan social identity in the form of
intergroup differentiation is indeed a dominant cause of bipolar partisan
attitudes and that this is accomplished predominantly through in-
group favoritism. However, one caveat is in order: as pure independents
did not have a measured partisan social identity, they were excluded from
all these models. This truncates the range of the party strength scale, and
may partially account for the lower size of its coefficient, but given the
substantial difference in the size of coefficients, the lack of pure
independents in the model is unlikely to change the substantive
conclusions.
In addition to social identity’s effect on party ratings, there is good reason
to expect it to influence ideology as well. According to optimal
distinctiveness theory, persons try to perceive the in-group as clearly distinct
from (and presumably better than) the out-group on key characteristics
(Brewer and Brown, 1998). A primary way for partisans with greater social
identity to differentiate themselves from members of the other party is to
become more ideological. The causal arrow should run from social identity
to ideological extremity for the reasons above, whereas there is little
theoretical reason to expect that a more liberal (conservative) ideology
would imply an increased emotional group identification with the
Democratic (Republican) party. The basic psychological mechanisms of a
group social identification can likewise also be seen as most likely causally
prior to the more considered changes in belief required for ideological
change. Thus, a regression model nearly identical to that in Table 2, but
with ideology as the dependent variable, confirms the importance of partisan
social identity in shaping political ideology (Table 3). Partisan social identity
was a statistically and substantively significant predictor of ideological
extremity. As with the party thermometer models, partisan strength,
although statistically significant as well, had a considerably lesser substantive
impact. In this model, and the remaining models in Table 3, independent
identity and ideology are excluded, as they both fail to have a strong
theoretical rationale for predicting the behavioral variables, and when
included fail to provide significant coefficients and attenuate the fit of the
model.
146 Social Science Quarterly
TABLE 3
Social Identity, Ideology, and Partisan Political Behavior

Ideological Partisan Over-Time

Self-Placement Activity Index Party Voting
Social identity 0.418 n n 0.751 n n 0.281 n
(0.163) (0.189) (0.169)
Partisan strength 0.192 n
0.119 0.232 n
(0.109) (0.126) (0.113)
Interest 0.320 1.746 n n 0.314
(0.329) (0.382) (0.343)
Knowledge 0.212 n n 0.017 0.021
(0.070) (0.080) (0.071)
Education 0.076
0.022 0.025
(0.076) (0.088) (0.079)
Income
0.002 0.087 0.009
(0.065) (0.075) (0.067)
Gender
0.140 0.058
0.133
(0.146) (0.169) (0.152)
Minority
0.156 0.312 0.267
(0.257) (0.299) (0.265)
Age 0.001 0.001 0.004
(0.005) (0.005) (0.005)
Constant 2.866 n n
0.771 2.315 n n
(0.583) (0.676) (0.602)
Number 218 220 215
R2 0.189 0.218 0.093
n
p  0.05; p  0.01, one-tailed test.
nn

Social Identity and Partisan Political Activity

Of course, the impact of partisan social identity need not be restricted to

an impact on attitudes. To assess the impact of partisan social identification
on political behavior, regression models were run on two measures of
partisan political behavior. The first was the index of partisan activity. The
second measure was voting support for that party across elections. Partisan
social identity again served as the key independent variable. As with previous
models, partisan intensity and demographics provided controls.
The results from these two models demonstrated that not only does social
identity affect political perceptions, but it also affects important partisan
behaviors (Table 3). Partisan social identity was statistically and substantively
significant in each model. In the case of partisan political activities, a person at
the top of the social identity scale would participate in fully three more
activities than someone at the bottom. The results from over-time voting are
not as strong, but nonetheless statistically significant. As before, these findings
persist controlling for partisan strength, thus implicating the group-identity
aspect of partisanship as a key element in behavior.
Social Identity Theory and Party Identification 147
Social Identity and the Party Identification Scale

As hypothesized earlier, a social identity perspective including partisan

and independent social identifications can potentially bring further
understanding to the traditional seven-point party identification scale,
especially regarding intransitivities (Greene, 2000; Keith et al., 1992;
Petrocik, 1974). To examine the relationship between social identity and the
traditional party identification scale, I conducted simple comparisons of
means and t-tests for the four categories of partisan strength (pure
independent, leaner, weak partisan, strong partisan).
The basic comparisons of social identity across different categories of
partisanship presented several interesting findings (Table 4). There were two
clear transitive patterns in the data: as partisan strength increased, partisan
social identity increased; and as partisan strength increased, independent
social identity decreased. The sole apparent exception was the independent
social identification of pure independents and leaners, but this was not a
statistically distinct difference, unlike all the other differences between
categories. Social identity, then, is a meaningful part of the standard party
identification index, which indeed has the transitive properties we should
expect. Additionally, independent identity appears to be a meaningful aspect
of political independence. Although independent identity decreases as
partisan identity increases along the strength of partisanship scale,
independent identity and partisan social identity are not significantly
correlated (r 5 0.031, p 5 0.632).

Discussion

On the whole, the findings here clearly indicate that we can learn much by
applying social identity theory to American partisanship. Perhaps, most
importantly, when we think about what partisanship is, we need to consider

TABLE 4
Social Identity by Strength of Party Identification

Social Identity with Social Identity with

Preferred Party Independents
Pure independent (14) — 1.09w
Independent leaner (45) 1.24 1.15w
Weak partisan (104) 1.41 0.97
Strong partisan (96) 1.88 0.81

NOTE: Differences between all items within same column are statistically significant at po0.05,
except those denoted by w. Items in parentheses are number of respondents.
148 Social Science Quarterly
not just attitudes toward, but group belonging with, a political party. Social
identity theory provides a compelling theoretical perspective from which to
reinterpret the meaning and implications of party identification. I have
attempted to demonstrate here that by considering partisan social
identification in addition to our standard measure of partisanship, we can
not only better understand political perceptions and behaviors, but can
model them better as well.
The first series of regression models demonstrated that partisan social
identification can substantially explain why most citizens tend to view their
own party positively and the other negatively. The traditional strength of
one’s partisanship and ideological intensity are certainly contributing
factors, but as social identity theory predicts, greater social identification
leads to a perception of greater differences between the relevant in-group
and out-group. Interestingly, in the case of political parties, it appears that
intergroup differentiation is caused essentially by in-group favoritism, the
heightened preference for one’s own party.
Not only does partisan social identity influence political perceptions, it
appears to influence partisan behavior as well. As partisan social identity
increased, respondents were more likely to engage in partisan-oriented
activities, such as attending rallies and wearing buttons, as well as to vote for
that party across elections. Thus the group-identity aspect of partisanship
likewise contributed to partisan activity above and beyond partisan strength.
Social identity theory does not offer clear explanations for why this should
be the case. Possible explanations for this relationship are that the perceptual
biases cause individuals to behave in a more partisan manner. Additionally,
as partisan social identification increases, defection from the party may
become more psychologically difficult, if indeed partisan group belonging
does contribute to one’s self esteem, as is implied by social identity theory.
For the most part, the application of social identity theory does not lead
to dramatically new and different predictions of partisan attitudes and
behaviors—at least when confined to a cross-sectional study. What it does is
add a much richer theoretical structure and a change of emphasis for the
understanding of partisan influence. On the theoretical level, social identity
theory suggests that by focusing on partisanship explicitly as a group
membership, we can better understand individual partisan identifications
and how they structure political thought. On an empirical level, the
inclusion of partisan social identity scales in addition to traditional measures
can clearly bolster our ability to explain and predict political attitudes and
behaviors.
Importantly, the insights of social identity on partisanship are not just
fodder for students of partisanship; the findings that partisan social
identification affects political perceptions and behavior have real con-
sequences for political parties. Tajfel (1978) has shown in the minimal
groups paradigm that it takes only the slightest categorization to invoke a
group social identity. Consequently, it would seem that political parties need
Social Identity Theory and Party Identification 149
not go to great lengths to induce social identification in their supporters.
Any actions that political parties can take to heighten social identification
with the party and to stimulate the use of partisan categories among their
rank and file can create a more partisan, more active group of supporters.
Existing research suggests that emphasizing group distinctiveness and
emphasizing similarity and interdependence of group members should lead
to increased group categorization (Brewer and Brown, 1998).
Aldrich and McGinnis (1989) argue that, contrary to Downs’s (1957)
well-known median voter theorem, the need to attract party activists and
their resources gives candidates incentive to diverge from the median voter.
Simple ideological proximity aside, though, social identity theory suggests
even more compelling reasons for parties to take divergent stands. Following
optimal distinctiveness theory (Brewer and Brown, 1998), which holds that
persons try to perceive their in-groups as clearly distinct from out-groups,
political parties have much to gain by focusing on their own distinctiveness.
Since in America both parties represent broad cross-sections of the public,
the only area in which the parties can truly set themselves apart as distinct is
ideology. Even if the majority of voters are moderates, social identity theory
suggests that parties may nonetheless benefit from divergent stands if the
increased partisan attitudes, partisan consistency, and activities in support of
the party among those with meaningful levels of partisan social identity
outweigh the costs of losing favor with the median voter. Furthermore, since
persons with greater partisan social identity are more likely to both
participate in the primaries and to have more extreme ideologies, they likely
contribute toward moving the parties away from the political center. Thus
social identity theory serves to theoretically augment Aldrich and
McGinnis’s theory of party positions and supplement, as well, as the
related work of Merrill and Grofman (1999) and Alvarez (1998).
The existence of partisan and independent social identifications as
separate, uncorrelated identities would seem to support the multidimen-
sional perspective advocated by Weisberg (1980) and Alvarez (1990). Yet
with regard to group identity, these two dimensions are far from completely
distinct. An independent social identification significantly attenuated the
differences that persons perceived between groups, as one would expect if
independence were simply the opposite of partisanship. It did this primarily
through reducing the favoritism toward one’s own party—again, something
we should expect in a unidimensional framework. On balance, then, the
exploration of group belonging in partisanship and independence suggests
the two as quasi-independent dimensions.
The examination of separate partisan and independent social identities
yields important insight into the problem of intransitivities in the traditional
measure of partisanship (see also Greene, 2000). The patterns for both
partisan and independent social identity help to explain why leaners choose
to identify as independents. The answer is that in terms of independent
social identity, they are independents, no different from pure independents.
150 Social Science Quarterly
Additionally, they have significantly less social identification with their
chosen party than do true partisans. Mean values in both the partisan social
identification measure and independent social identification measure are
transitive across the traditional partisan categories. In the matter of social
identity, we see a scale as we would expect with no intransitivities.
Furthermore, although the average partisan identity was higher than the
average independent identity for leaners, over 60 percent of leaners had an
independent identity greater than or equal to their partisan social identity.
For weak and strong partisans, in contrast, the figures are 20 percent and 4
percent, respectively. For the most part, leaners seem to identify with both
their preferred party and other independents. Following self-categorization
theory (Turner et al., 1987), leaners may find either the independent or
partisan social identity more salient, depending on the particular political
cues.

Future Explorations of Partisanship with Social Identity Theory

Given the considerable group aspect inherent in partisanship, proper

understanding and modeling implies that we augment the still essential
traditional measure with measures of partisan and independent social
identity. The results here are intriguing and suggestive, but are based on an
admittedly suboptimal sample. With the low response rate, politically
interested citizens are probably overrepresented, which may overstate the
impact of social identity. Nonetheless, there are clear and intriguing patterns
worthy of further study. Obviously, a more thorough exploration of social
identity and partisanship demands a national telephone survey with greater
respondent participation. Greene (2002) argues for more precise psycho-
logical measures of all aspects of partisanship, including social identity, in
order to properly address many of the recurring questions of political
behavior.
Given that this study has established a basis for the role of social
identification in partisanship, there are a variety of extensions from social
identity theory that can be used to further explore politics and partisanship.
Self-categorization theory, which states that social identifications will be
more or less salient depending on features in one’s environment, suggests
that partisan social identifications may have a very dynamic aspect well
worth exploring. The nature of partisan social identification during an
election campaign, during times of political scandal, when a party is a
majority or a minority, when parties are more or less polarized, and so forth,
are all issues that could further enlighten our understanding of citizen
response to the political world.
Additionally, the inclusion of partisan group identity measures in
longitudinal studies has tremendous potential to address the seemingly
endless controversy about the stability and exogeneity/endogeneity of
Social Identity Theory and Party Identification 151
partisanship (e.g., Niemi and Weisberg, 1993, 2001). If partisanship is
ultimately more a group identification than a constellation of related
attitudes, this supports the traditional interpretation of stable long-term
identification, since a group identity is surely less malleable than attitudes
based on dynamic political objects. Assessing the degree to which partisan
attitudes, as opposed to partisan group identification, are responsive to
external political stimuli should go a long way toward more definitively
settling these debates on partisanship.
Importantly, there are also significant implications beyond the confines of
partisanship studies. Kelly’s (1990) study of party factions in Britain shows
a considerable impact of social identity depending on membership in
minority or majority factions. In the American context, then, we might be
able to use social identity theory to understand the behavior and attitudes of
convention delegates of winning and losing presidential nomination
contenders at national conventions. Even the seemingly unrelated topic of
media bias offers potential for further exploration from this perspective. In
the Australian context, Duck, Hogg, and Terry (1998) suggest that
perceptions of media bias are significantly shaped by political in-group
identification. Clearly, with the ongoing concerns about perceptions of
media bias in this country, similar explorations in the American political
context are warranted.
Taken as a whole, the findings discussed here clearly demonstrate that
including the social identity element of partisanship, both conceptually and
analytically, in research designs has much to offer for our understanding of
partisanship and its role in the American political system. In terms of
understanding the sources of partisan attitudes, partisan activities, and the
party identification index, social identity theory enhances, rather than
replaces, traditional conceptualizations and measures of partisanship.
Further research that incorporates partisan social identity measures into
more representative and comprehensive surveys offers considerable potential
for furthering our understanding of how Americans relate to political parties
and the political system more broadly.

REFERENCES

Abrams, Dominic. 1994. ‘‘Political Distinctiveness: An Identity Optimizing Approach.’’

European Journal of Social Psychology 24:357–65.
Aldrich, John, and Michael McGinnis. 1989. ‘‘A Model of Party Constraints on Optimal
Candidate Positions.’’ Mathematical and Computer Modeling 12:437–50.
Alvarez, R. Michael. 1990. ‘‘The Puzzle of Party Identification.’’ American Politics Quarterly
18:476–91.
———. 1998. Information and Elections, rev. ed. Ann Arbor, Mich.: University of Michigan
Press.
Berelson, Bernard, Paul Lazarsfeld, and William McPhee. 1954. Voting. Chicago, Ill.:
University of Chicago Press.
152 Social Science Quarterly
Brewer, Marilynn B., and Rupert J. Brown. 1998. ‘‘Intergroup Relations.’’ Pp. 554–94 in
Daniel T. Gilbert, Susan T. Fiske, and Gardner Lindzey, eds., Handbook of Social Psychology,
4th ed., Vol. II. Boston, Mass.: McGraw-Hill.
Brewer, Marilyn B., and Michael D. Silver. 2000. ‘‘Group Distinctiveness, Social Identity,
and Collective Mobilization.’’ In Sonia Stryker, T. J. Owens, and R. W. White, eds., Self,
Identity, and Social Movement. Minneapolis, Minn.: University of Minnesota.
Campbell, Angus, Philip Converse, Warren Miller, and Donald Stokes. 1960. The American
Voter. Chicago, Ill.: University of Chicago Press.
Delli Carpini, Michael X., and Scott Keeter. 1996. What Americans Know about Politics and
Why it Matters. New Haven, Conn.: Yale University Press.
Dennis, Jack. 1988. ‘‘Political Independence in America.’’ British Journal of Political Science
18:77–109.
Dillman, Don. 1978. Mail and Telephone Surveys: The Total Design Method. New York:
Wiley and Sons.
Downs, Anthony. 1957. An Economic Theory of Democracy. New York: Harper and Row.
Duck, Julie, Michael Hogg, and Deborah Terry. 1995. ‘‘Me, Us and Them: Political
Identification and the Third-Person Effect in the 1993 Australian Federal Election.’’
European Journal of Social Psychology 25:195–215.
———. 1998. ‘‘Perceptions of a Media Campaign: The Role of Social Identity and the
Changing Intergroup Context.’’ Personality and Social Psychology Bulletin 24:3–16.
Greene, Steven. 1999. ‘‘Understanding Party Identification: A Social Identity Approach.’’
Political Psychology 20:393–403.
———. 2000. ‘‘The Psychological Sources of Partisan-Leaning Independence.’’ American
Politics Quarterly 28:511–37.
———. 2002. ‘‘The Social-Psychological Measurement of Partisanship.’’ Political Behavior
24:171–97.
Keith, Bruce, David Magleby, Candice Nelson, Elizabeth Orr, Mark Westyle, and Raymond
Wolfinger. 1992. The Myth of the Independent Voter. Berkeley, Cal.: University of California
Press.
Kelly, Caroline. 1988. ‘‘Intergroup Differentiation in a Political Context.’’ British Journal of
Social Psychology 27:319–32.
———. 1990. ‘‘Social Identity and Intergroup Perceptions in Minority-Majority Contexts.’’
Human Relations 43:583–99.
Mael, Fred, and Lois Tetrick. 1992. ‘‘Identifying Organizational Identification.’’ Educational
and Psychological Measurement 54:813–24.
Merrill, Samuel, and Bernard Grofman. 1999. A Unified Theory of Voting. New York:
Cambridge University Press.
Miller, Arthur H., Christopher Wlezien, and Anne Hildreth. 1991. ‘‘A Reference Group
Theory of Partisan Coalitions.’’ Journal of Politics 53:1134–49.
Niemi, Richard G., and Herbert F. Weisberg. 1993. ‘‘Is Party Identification Stable?’’ In
Richard G. Niemi and Herbert F. Weisberg, eds., Controversies in Voting Behavior, 3rd ed.
Washington, D.C.: Congressional Quarterly Press.
Social Identity Theory and Party Identification 153
———. 2001. ‘‘How Much Does Politics Affect Party Identification?’’ In Richard G. Niemi
and Herbert F. Weisberg, eds., Controversies in Voting Behavior, 4th ed. Washington, D.C.:
Congressional Quarterly Press.
Petrocik, John. 1974. ‘‘An Analysis of Intransitivities in the Index of Party Identification.’’
Political Methodology 1:31–47.
Sherif, Muzafer. 1966. Group Conflict and Cooperation: Their Social Psychology. London:
Routledge.
Stanley, Harold W., and Richard G. Niemi. 1995. ‘‘The Demise of the New Deal Coalition:
Partisanship and Group Support, 1952–92.’’ In Herbert F. Weisberg, ed., Democracy’s Feast.
Chatham, N.J.: Chatham House Publishers.
Tajfel, Henri. 1978. ‘‘Social Categorization, Social Identity, and Social Comparisons.’’
In Henri Tajfel, ed., Differentiation Between Social Groups. London: Academic Press.
Tajfel, Henri, and John C. Turner. 1986. ‘‘The Social Identity Theory of Intergroup
Behavior.’’ In Stephen Worchel and William G. Austin, eds., Psychology of Intergroup
Relations. Chicago, Ill.: Nelson-Hall.
Turner, John C., Michael A. Hogg, P. J. Oakes, S. D. Reicher, and M. Wetherell. 1987.
Rediscovering the Social Group: A Self Categorization Theory. New York: Blackwell.
Weisberg, Herbert F. 1980. ‘‘A Multidimensional Conceptualization of Party Identification.’’
Political Behavior 2:33–60.
Weisberg, Herbert F., and Steven Greene. 2003. ‘‘The Political Psychology of Party
Identification.’’ In Michael MacKuen and George Rabinowitz, eds., Electoral Democracy. Ann
Arbor, Mich.: University of Michigan Press.
Wlezien, Christopher, and Arthur H. Miller. 1997. ‘‘Social Groups and Political
Judgments.’’ Social Science Quarterly 78:625–40.