Values and Votes: The Indirect Effect of Personal Values on Voting Behavior

Authors


Philipp Leimgruber is a Ph.D. candidate at the Institute of Political Science, University of Bern. His research interests include political behavior, political attitudes and values, and human values. Address for correspondence: Philipp Leimbruger, Institute of Political Science, University of Bern, Lerchenweg 36, CH-3000 Bern 9. Phone: +41 31 631 51 57; e-mail: philipp.leimgruber@gmail.com

Abstract

Recent research from social psychology suggests that personal values predict political behavior, such as vote choice. In contrast to previous studies, it is hypothesized in this article that personal values influence voting behavior only indirectly through political value orientation. Drawing on the personal value concept of Shalom Schwartz, structural equation models based on Swiss electoral data (SELECTS 2007) are applied to test the hypothesis of indirect effects. The empirical analysis shows that the effects of personal values are indeed mediated by political values, but that their indirect impact on vote choice remains substantial. It is argued on a theoretical level that personal values need first to be translated (or transformed) into political values to become effective on voting behavior.

Introduction

A decade ago, Kuklinski (2001: 355) noted that the study of values was cyclical in political science: “It reached its nadir in the early 1970s, fell from prominence following a barrage of criticisms, and now once again is gaining momentum”. Ten years after Kuklinski introduced this notion, it is safe to say that the study of values has kept its momentum. Numerous studies have demonstrated the great and arguably growing importance of political values in the process of political reasoning and decision-making (see Knutsen and Kumlin 2005, Goren 2005, Ansolabehere et al. 2008, to name a few). It is now widely accepted that core political values are important determinants of political orientations and behavior (Jacoby 2006).

While the significance of political values is rather undisputed, one important question remains: What are the origins of political value orientation itself? Why does one person give high importance to social egalitarianism while the other, with a similar social background, accepts inequality? What is, in the end, the “backstop” of a political belief system (Tetlock et al. 1996: 27)? It seems that there must be something besides social and political predispositions that citizens can use to inform and organize their political preferences (Goren 2007). As Kuklinski (2001) rightly notes, few of us have reasoned and distinct attitudes on every political issue, but most of us have ideas about right and wrong, good and bad, from which ideas we can reach political judgments.

Early political psychologists believed that the backstop of a political value system lies within one’s identity (Lasswell 1930, Lane 1972). According to this view, political opinions are no more and no less than transformed private needs and personal motives, and are thereby closely related to the structure of personality. Thus a citizen does not merely hold conservative values, but is, as such, a conservative personality (Wilson 1973). More recently, social psychologists have revived this idea of the political man. According to them, political reasoning and behavior is mainly determined by abstract principles or personal values which reflect basic social and human needs (cf. Rohan and Zanna 1994, Braithwaite 1997).

One potentially valuable approach to studying the relationship between personal values and political reasoning is the values construct proposed by Shalom Schwartz (Feldman 2003). Schwartz and his colleagues not only demonstrated cross-culturally that people possess by and large a dozen basic personal values which are organized along two central dimensions (Schwartz 1992). Furthermore, Schwartz et al. (Schwartz 1996, Barnea and Schwartz 1998, Caprara et al. 2006) argue that these basic dimensions underlie political value orientations. One of these basic value dimensions runs from openness to change to conservation and is assumed to influence the political value orientation libertarianism vs. authoritarianism. The second personal value dimension runs from self-enhancement to self-transcendence and is assumed to be associated with the political conflict line left vs. right.

At first sight, Schwartz’ theory seems to be confirmed by empirical research. Several studies show that, as hypothesized, personal values are associated not only with political opinion but also with political behavior (Schwartz 1996, Barnea and Schwartz 1998, Caprara et al. 2006). However, this proposed relationship between personal values and political behavior needs to be scrutinized. Virtually all empirical studies presume – and test – direct relationships between personal values and voting behavior exclusively. Such research designs entirely neglect the well-known impact of political values on voting and as a result are not only prone to yielding biased findings in general, but to overestimating the direct impact of personal values on political behavior in particular.

Keeping this criticism in mind but still drawing on the main theory of Schwartz, I will argue in this paper that citizens are indeed affected by their abstract principles when they vote – but only indirectly. I hypothesize that personal values are first translated into political meaning, i.e. transformed into political values, to become subsequently effective on political behavior. I will thoroughly discuss the theoretical foundation of my hypothesis in the next section and will subsequently support it empirically by employing structural equation models on Swiss electoral data.

Theory

Values are defined as general and enduring beliefs or ideals of an individual about what is good or desirable and what is not. Values are therefore essentially a conception of “the desirable” (Kluckhohn 1951: 395). As guiding principles, they are limited in number and serve as a basis for numerous specific evaluations and subsequently constrain behavior. Schwartz (1992: 4) summarizes the properties of values as follows “[They] (1) are concepts or beliefs, (2) pertain to desirable end states or behaviors, (3) transcend specific situations, (4) guide selection or evaluation of behavior and events, and (5) are ordered by relative importance.”

Personal Values

In psychology, the starting point of value research has essentially been the purpose that values actually serve. Like Rokeach (1973), Schwartz deduces values from fundamental human and societal needs. Schwartz (1992) forms a unique typology of human values – or personal values – which distinguishes three universal requirements of human existence: needs of individuals as biological organisms, requisites of coordinated social interactions, and survival needs of groups. In tying values to human and social needs, Schwartz identified ten value types: power, achievement, hedonism, stimulation, self-direction, universalism, benevolence, tradition, conformity, and security (Schwartz 1996).

One strength of Schwartz’s theory is the integration of these basic value types into a broader value system (Davidov et al. 2008). By understanding value types in terms of basic human and social needs, he is able to specify relationships among these types, as some types are more compatible with each other and some are more opposed to each other. For instance, universalism and benevolence values are compatible because actions that express both values promote the welfare of others. Universalism and power, on the other hand, are opposite value domains, as power involves dominance and control seeking while universalism reflects appreciation and tolerance of all people. The interrelationships among all the value types can be displayed in a two-dimensional value structure with four higher order value types.

One dimension opposes openness to change (combining the value types self-direction and stimulation) to conservation (combining security, conformity, and tradition). “This dimension reflects a conflict between emphases on own independent thought and action and favoring change versus submissive self-restriction, preservation of traditional practices, and protection” (Schwartz 1996: 5). The second dimension opposes self-transcendence (combining benevolence and universalism) to self-enhancement (combining power and achievement). “This dimension reflects a conflict between acceptance of others as equals and concern for their welfare versus pursuit of one’s own relative success and dominance over others” (ibid, p. 5).

Political Values and Value Orientation

Similar to the concept of personal values, empirical research on political values indicates that at least two political value dimensions can be isolated from each other (cf. Feldman 2003). The first dimension is the traditional left-right continuum (or liberal-conservative in the US), which reflects the economic conflict within a modern democracy. Contestation on this dimension has predominated in most Western nations in the postwar period (Bartolini and Mair 1990). Specific issues within this struggle are, among others, taxation, wealth redistribution, social security, size of government, and free economic enterprise. More simply put, this is the conflict between socialist and capitalist ideology (Kitschelt 1994).

With the rise of new challenges to modern democracies a new political dimension has emerged (Flanagan 1987). As Kitschelt (1994: 11) argues, “citizens’ experiences in their work and social life generate political predispositions that can be mapped onto the two-dimensional space created by the socialist-capitalist and the libertarian-authoritarian dimensions, and party strategies situate themselves in their field of competing programs”. This second dimension, which Kitschelt termed “authoritarian-libertarian”, reflects political issues such as freedom of thought and conscience, authority, law and order, civic protests, minority rights, and tradition. Convincing empirical evidence has been presented showing that this dimension, albeit bearing many names, measures a discrete value dimension.1

The two-dimensional structure described above is well suited to assess ideologies of Western democracies and the political parties representing them. Classic liberalism, for instance, stresses both (economic) rightist and libertarian values. It encompasses libertarianism insofar as the moral core of liberalism contains an affirmation of basic human rights – freedom, dignity, and life – and the political core includes political rights – particularly the right to vote and to participate. It includes economic rightism insofar as liberalism postulates economic individualism and free enterprise system, i.e. a market economy that is free from state control (Macridis and Hulliung 1997, Bellamy 1993).2 This, for instance, is the ideological foundation of the Swiss Liberal Party (FDP). Democratic Socialism, on the other hand, can be classified as libertarian and economic leftist. Like liberals, socialists have always claimed civic and political rights for all people. In contrast to liberalism, however, democratic socialism also stresses social rights in terms of wealth redistribution and state intervention in order to remove social inequality (Wright 1993). This can be regarded as the current ideological position of the Swiss Social Democratic Party (SP). Conservatism, lastly, may be classified as economic rightist and authoritarian. While pursuing a similar economic agenda to liberalism, conservatives particularly value principles of hierarchy and status, as well as the legitimacy of authority (Nisbet 1952). This political orientation is broadly represented by the Swiss Conservative Party (SVP).

From Personal Values to Political Values to Political Behavior

Having laid out the concepts of personal and political values, it is necessary to discuss their relationship. Steenbergen and Leimgruber (2010) assert that personal and political values differ in their levels of abstraction. Political values are beliefs which citizens seek to see implemented in the political realm, while personal values are prescriptive beliefs that are preferable to other end states or mode of conducts. Because personal values speak to fundamental human and societal needs, they do not only relate to the social sphere, but also, as a consequence of modern times, to the political world as well, yet from a greater “distance”. So, rather than being equal subclasses, personal values can be conceptualized as enduring beliefs which precede political values. As Tetlock et al. (1996) have argued, underlying all political belief systems are core values that specify what the ultimate goals of public policy should be. These basic values are the ultimate justification for political preferences, as citizens must be using something besides political predispositions to build their preferences (Goren 2007).

This view has elements in common with the image of political personalities that early scholars of political psychology (or pathology) portrayed. The study of “politics within” (Lasswell 1930) maintains that “[a] person’s political belief system […] will be heavily influenced by his identity” (Lane 1972: 175), that private motives and personal needs come to be transformed into political opinions, and that personal values are crucial for the framing of a political belief system. A prominent showcase for illustrating the interdependence of identity, personal needs and political ideology has been the idea of the conservative personality. Several empirical studies have concluded that the “conservatism syndrome” (Wilson 1973: 261) is mainly a response to the feeling of insecurity, inferiority, and a generalized fear of uncertainty (see Jost et al. 2003).3 It has even been averred that “conservatism is less a political doctrine than a habit of mind” (White 1964: 1).

The theory of value-based political reasoning proposed by Schwartz and his collaborators shares the same theoretical foundation. Caprara et al. (2006: 2), for instance, point out that personal values “are the crucial grounding of ideology.” Schwartz and his colleagues theorize essentially that the two dimensions of personal values underlie the two dimensions of political value orientation (Caprara et al. 2006, Barnea and Schwartz 1998, Schwartz 1996, Schwartz 1994): the superordinate value openness to change is assumed to underlie political values that endorse individual autonomy, tolerance, and openness, that is: libertarianism. On the other side of the spectrum, conservation underlies political values favoring social order, certainty, hierarchy, and national values, i.e. authoritarianism. Analogously, self-transcendence underpins political values that endorse social welfare and equality, i.e. (economic) leftism, whereas the opposite pole self-enhancement corresponds to political values giving primacy to differential merit and unequal distribution of income and wealth, i.e. (economic) rightism.

Empirical research supports the hypothesis that personal values are related to political values, as predicted by Schwartz and his colleagues. Authoritarianism, for instance, has been found to be positively correlated with personal values representing conservation (Davidov et al. 2008), but negatively correlated with openness to change (Rohan and Zanna 1996). In the same vein, Steenbergen and Leimgruber (2010) showed that conservation not only corresponded with diverse political values, but was also associated with left-right self-placements of Swiss citizens. In addition, Davidov et al. (2008) found that self-transcendence and conservation explain political attitudes towards immigration in Switzerland and other countries.

Moreover, personal values have been found to predict voting behavior. Barnea and Schwartz (1998) and Schwartz (1996) showed that the conflict of openness to change with conservation discriminated to a large extent between party supporters in Israel. Caprara et al. (2006), studying Italian general elections in 2001, found that “universalism” predicted a preference for the center-left coalition, whereas the superordinate value conservation predicted a preference for the center-right. Devos et al. (2002) found that Swiss right-wing supporters attributed more importance to “power,”“security,” and “conformity,” whereas left-wing supporters cherished “universalism” more.

Although these studies clearly demonstrate the significant influence of personal values on voting behavior, they inherit a major shortcoming regarding the operationalization of the assumed relationship. In theory, Schwartz and his colleagues argue that personal values correspond primarily to political value orientations (Schwartz 1994, Schwartz 1996, Barnea and Schwartz 1998). They even acknowledge that personal values influence voting behavior through political ideology (Barnea and Schwartz 1998: 30). Yet on other occasions they claim that this mediating effect will only play a minor role: “[Political] values may mediate the effects of basic values on political choice but basic values are more fundamental” (Caprara et al. 2006: 5). In light of this inconsistence it is unclear how Schwartz et al. conceptualize the role of political values in the causal chain between personal values and party choice. What is evident, however, is that in previous research no empirical models of voting behavior have been developed which incorporate both personal and political values. Virtually all studies test only the direct effect of personal values on voting behavior (see Schwartz 1996, Barnea and Schwartz 1998, Verplanken and Holland 2002, Devos 2002, Caprara et al. 2006). In doing so, they entirely neglect the role of political values in the process of voting. This is a strong assumption, as research has consistently demonstrated the importance of political values in the process of voting.4 By neglecting them entirely, one not only runs the risk of missing a crucial link in the causal chain but also of overestimating the impact of personal values in general.

Hypothesis: The Indirect Effect of Personal Values on Voting Behavior

In light of this, it is necessary to bring political values back into the causal chain, as preliminary findings indicate a mediating function of political attitudes or values. For instance, one study shows that personality traits, besides personal values defined as a major constituent of an individual’s identity, have an impact on vote choice that is mediated by political predispositions (Schoen and Schumann 2007). In the same vein, Braithwaite (1997) presumes that basic values influence voting behavior only indirectly though political attitudes. These findings should not come as a surprise if one takes the theoretical conception of personal values seriously. If basic values are considered as general principles and used as a basis for numerous specific evaluations across situations (Feldman 2003), then there must be an entity – such as political values – that makes these abstract values applicable to a specific situation – such as voting in an election.

This argument can be made from two perspectives. From a value-based perspective one can theorize that abstract beliefs like personal values need to be translated into political meaning in order to have any behavioral consequences. As van Deth (1995: 6) points out, basic “[…] values of individuals can be transformed into political orientations which have some impact on their behavioral intentions.” Thus an individual’s endorsement of, say, “understanding, appreciation, and tolerance of others” (i.e. “universalism”) needs to be translated, even if subconsciously, into actual political stances. Thus, only when personal universalism is transformed into political universalism, which is commonly endorsed by Social Democratic parties, can it have a behavioral impact such as inducing a vote for the Social Democrats. From an action-based perspective, on the other hand, it can be argued that “values need to be activated to affect […] behavior” (Verplanken and Holland, 2002: 434). Behavioral intentions need to be connected with basic values through attitudes which serve to evaluate a specific object. Considered this way, political attitudes (or political values, more generally) constitute a value-expressive function (Maio and Olson 1995). Only when vote intentions connote political values can very basic motives be expressed through them. Both perspectives equally imply that political values function as a mediator between personal values and political behavior.

Considering this, I hypothesize that personal values as conceptualized by Schwartz (1992) do not affect voting behavior in a direct fashion. Rather they are first translated (or transformed) into political values, which in turn influence vote choice. The impact of personal values on voting behavior is thus assumed to be indirect, as they are mediated by political orientation. By assuming a deductive concept of political reasoning (Peffley and Hurwitz 1985) from abstract beliefs to more specific political preferences to actual voting, I model a more complex causal chain. More precisely, it is hypothesized, following Schwartz’s theory, that the higher value type conservation is negatively correlated with libertarianism but uncorrelated with leftism. Libertarianism, in turn, is expected to discriminate between voters of the conservative Swiss People’s Party as well as the supporters of the left-libertarian Social Democratic Party from voters of the centrist Christian Democratic Party. Accordingly, self-transcendence is expected to be positively correlated with leftism but uncorrelated with libertarianism. Leftism, in turn, is expected to discriminate between voters of the left, i.e. the Social Democratic Party, and supporters of the economic right, i.e. of the Liberal Party and the People’s Party, from the party supporters of the centrist Christian Democratic Party. The higher-order personal values self-transcendence and conservation are assumed to have no direct effect on party choice.

Method

Data and Measurement

The guiding research question of this study will be analyzed with data from the Swiss Electoral Study (SELECTS) of 2007. The SELECTS 2007 survey contains information not only about vote choice and the demographics of Swiss voters, but also items measuring personal and political values5. In order to avoid cultural variation in regard to political values, the analysis will be restricted to voters in the Swiss German cantons.

Concerning our dependent variable, vote choice, a four category variable is defined. It includes the Social Democratic Party (SP), the Christian Democratic Party (CVP), the Liberal Democratic Party (FDP), and the Swiss People’s Party (SVP). These four parties represent the current government (Bundesrat), and together they secured roughly 80% of the vote share in the 2007 elections6. Concerning the ideological positions of the party voters (represented by the libertarian-authoritarian and left-right dimensions), the following is known (see Leimgruber et al. 2010): the median voter of the Social Democratic Party is ideologically left-libertarian, while the median voter of the People’s Party, his ideological counterpart, is right-authoritarian. The average voter of the Liberal Party most distinctively stresses economic right values but is moderate in libertarian values. The median supporter of the Christian Democratic Party, lastly, is located in the center of the ideological space. Given the centrist ideological position of CVP supporters, this party is defined as the reference category of the dependent variable in subsequent empirical analyses (cf. Leemann 2008).

In order to measure the two political value orientations described above, ten items covering attitudes towards diverse political issues7 are selected based on preliminary exploratory factor analysis8: items addressing “Swiss tradition,”“immigrant customs,”“law and order,”“equal chances for foreigners,” and “support for the army” measure libertarianism (see Table I for the question wording and item format). Likewise, questions addressing “social expenditure,”“security,”“redistribution,” and “taxation” measure leftism. Note that, for efficiency reasons, the two constructs are described by their poles leftism and libertarianism, although they in fact reflect the whole spectrum of each dimension (left-right and libertarianism-authoritarianism).

Table I.   Variable Description
  1. Note. In the analysis, all human value items have been inverted. The same applies for the political value items “trad”, “law1”, and “taxinc”.

PERSONAL VALUES
Conservation
 SECurityIt is important to him to live in secure surroundings. He avoids anything that might endanger his safety.
1 = “very much like me”, 2 = “like me”, 3 = “somewhat like me”, 4 = “a little like me”, 5 = “not like me at all”.
 COnformity1He believes that people should do what they’re told. He thinks people should follow rules at all times, even when no one is watching.
 COnformity2It is important to him to always behave properly. He wants to avoid doing anything people would say is wrong.
 TRaditionTradition is important to him. He tries to follow the customs handed down by his religion or his family.
Self-Transcendence
 UNiversalism1He thinks it is important that every person in the world should be treated equally. He believes everyone should have equal opportunities in life.
 UNiversalism2It is important to him to listen to people who are different from him. Even when he disagrees with them, he still wants to understand them.
 UNiversalism3He strongly believes that people should care for nature. Looking after the environment is important to him.
 BEnevolenceIt’s very important to him to help the people around him. He wants to care for their well-being.
POLITICAL VALUES
Libertarianism
 ARMYAre you against or in favor of a strong army?
1 = “strong army”, 2 “Rather strong” 3 “neither nor” 4 “rather no army” 5 “No army”.
 Equal CHANcesDo you think foreigners should have equal opportunities or should Swiss citizens have better chances?
1 = “equal opportunities”, (…) 5 = “better opportunities for Swiss”
 Swiss TRADitionDo you think that Swiss traditions should be defended or rather questioned?
1 = “defend Swiss traditions”, (…) 5 = “question traditions”.
 LAW and order1Do you think “law and order” should be stressed in Switzerland?
1 = “not stress law and order”, (…) 5 = “stress”.
 Immigrant CUSTomsImmigrants should be required to adjust to the customs of Switzerland.
1 = “strongly agree”, (…) 5 = “strongly disagree”.
 LAW and order2People who break the law should be given tougher sentences.
1 = “strongly agree”, (…) 5 = “strongly disagree”.
Leftism
 Social EXPenditureAre you in favor of expansion of social expenditure or cutting of social expenditure?
1 = “Cut social expenditure”, (…) 5 = “increase”.
 TAXing high incomeDo you think taxes on high income should be increased?
1 = “for higher taxes”, (…) 5 = “lower taxes”.
 Social SECUirityProviding a stable network of social security should be the primary goal of government.
1 = “strongly agree”, (…) 5 = “strongly disagree”.
 REDIstributionIncome and wealth should be distributed towards ordinary people.
1 = “strongly agree”, (…) 5 = “strongly disagree”.
DEPENDENT VARIABLE (DUMMY PARTY CHOICE)
 FDP1 = Voted for Swiss Liberal Party FDP
 SVP1 = Voted for Swiss People’s Party SVP
 SP1 = Voted for Social Democratic Party SP
 CVP1 = Voted for Christian Democratic Party CVP. (Reference category)
SOCIO-ECONOMIC COVARIATES
 Education1 = High education (High School or higher), 0 = Low education
 Gender1 = Female, 0 = Male
 Income Household income; 11 point-scale (1 = less than 2′000 CHF, 2 = 2’001 - 3’000 CHF, …, 10 = 10’001- 11’000 CHF, 11 = more than 11’000)
 AgeYears (18–95)
 Religion1 = Catholic, 0 = Non-Catholic

Personal values are measured by the Portrait Values Questionnaire (PVQ) developed by Schwartz et al. (2001), which has been validated in various cultures (see Table I for the question wording and item format). Unfortunately, only a subset of Schwartz’ PVQ has been included in the SELECTS survey. These 12 items do not allow us to fully reproduce the two dimensions self-transcendence vs. self-enhancement and openness to change vs. conservation. Rather, the higher-order values self-transcendence (measured directly by three indicators of the value type “universalism” and by one of “benevolence”) and conservation (measured by two indicators of “conformity” and one each of “security” and “tradition”) are generated with the available data (see Davidov et al. 2008 for a similar procedure).9 Although this represents a limitation of the present analysis, it does not undermine the theoretical assumptions proposed by Schwartz et al., which are adapted here: self-transcendence is predicted to correspond with leftism, and conservation with libertarianism (Barnea and Schwartz 1998: 29). It is also not assumed that the reduction to two poles interferes with the general hypothesis of indirect effects. If these two higher order values can be shown to affect vote choice only indirectly, then it is likely that this also holds for the full dimensions.

As the measurement model shows (see Figure 1), the items selected measure both the political and the personal value constructs very well: All factor loadings are greater than .5 (note that the indicators are classified as ordered categorical variables in order not to underestimate the standard errors [DiStefano 2002], thus the factor loadings are standardized probit regression coefficients); the political value dimensions are uncorrelated; only the moderately though significant correlation (.336) between the personal factors is not assumed from the theoretical model.10 All in all, the measurement model of personal and political values fits well to the data and produces an acceptable model fit.11

Figure 1.

 The Measurement Model of Personal and Political Values.
Note. Estimator is MLR. Factor loadings are standardized probit regression coefficients; measurement errors are proportions of unexplained variance. All estimates are significant at the .05 except for those designated “ns” or “nt” (not tested). N = 626; AIC = 30191; BIC = 30652 (See footnote 7 for more details on model fit.)

As Feldman (2003: 488) has pointed out, one cannot correlate personal values with political values without controlling for variables that could predict one or the other. Thus, in the structural equation models applied, socio-economic covariates will be included, since they are known to correlate with personal values, political values, and vote choice alike. In regard to personal values, it has been found that gender, age, and religious affiliation discriminate on some values (cf. Lindeman and Verkasalo 2005, Lyons et al. 2005, Devos et al. 2002). In addition to education and income, the same SES variables have been found to correlate with political values and value orientations (cf. Andersen and Heath 2003, Brunner and Sciarini 2002, Howell and Day 2000, Alwin and Krosnick 1991). Finally, with the exception of gender, all of these socio-economic indicators are correlated with party choice in the Swiss elections of 2007 (Lutz 2008). Therefore, the following SES covariates will be included as control variables in the structural equation models: gender, age, education, income (household), and religious (catholic vs. non-catholic) denomination (see Table I).

Models

To model causal paths and to test complex relationships, in our case direct and indirect effects on vote choice, structural equation modeling (SEM) is the most appropriate statistical tool. The goal is to demonstrate that personal values influence voting behavior only indirectly, through political values. For this endeavor it is advisable to test alternative models and compare them with each other in order to demonstrate the supremacy of one theoretical model (MacCallum and Austin 2000, Homer and Kahle 1988). The three alternative models tested in this analysis are illustrated in Figure 2 (the corresponding path estimates are displayed in Table II). Note that in these figures, the circle “personal values” represents both higher values self-transcendence and conservation, and likewise the circle “political values” stands for both libertarianism and leftism.

Figure 2.

 Three Theoretical Models of Value-Based Voting

Table II.   The Simplistic, Complex, and Parsimonious Model of Value-Based Voting
 Simplistic ModelComplex ModelParsimonious Model
Est. SEEst. SEEst. SE
  1. Note. Estimator is MLR. Directs effects on vote choice (FDP, SVP, SP) are standardized probit regression coefficients with party CVP as the reference category, all other path estimates are standardized linear regression coefficients. Thus the estimates of the indirect effects are in brackets, as they are not directly interpretable. The significance of the indirect effects is estimated with Sobel’s (1982) test based on the SE of the unstandardized coefficients (not shown).

  2. **p < 0.01; *p < 0.05.

Direct effects
FDP ON
 Conservation.379 .201.709 .590Fixed to 0
 Self-Transcendence−.565**.163−.653 .379Fixed to 0
 LibertarianismFixed to 0.457 .637−.226 .200
 LeftismFixed to 0−.159 .260−.512**.140
 Education.260*.120.178 .127.153 .122
 Age.166 .138.201 .147.117 .137
 Gender.173 .121.178 .121.089 .113
 Income (Household).317*.128.247 .140.176 .139
 Religious Denomination−.798**.094−.801**.092−.764**.098
SVP ON
 Conservation.825**.090−.033 .384Fixed to 0
 Self-Transcendence−.694**.088−.185 .256Fixed to 0
 LibertarianismFixed to 0−.917**.343−.887**.058
 LeftismFixed to 0−.090 .127−.239**.084
 Education−.103 .088−.028 .073−.065 .069
 Age−.144 .083−.226**.068−.251**.065
 Gender.042 .083−.006 .074−.034 .063
 Income (Household).072 .087.070 .081.045 .078
 Religious Denomination−.584**.081−.400**.089−.423**.069
SP ON
 Conservation−.689**.147.598 .405Fixed to 0
 Self-Transcendence.482**.150−.494 .257Fixed to 0
 LibertarianismFixed to 0.994**.365.602**.099
 LeftismFixed to 0.770**.127.608**.094
 Education.029 .125.043 .092.031 .091
 Age.130 .119.215*.093.172 .092
 Gender−.046 .119.060 .085−.016 .078
 Income (Household).055 .122.225*.092.192*.094
 Religious Denomination−.659**.119−.439**.092−.444**.088
Libertarianism ON
 ConversationFixed to 0−.872**.054−.858**.055
 Self-TranscendenceFixed to 0.433**.058.444**.058
 Education.354**.043.120**.042.121**.042
 Age−.243**.047−.097*.045−.099*.045
 Gender.165**.043−.032 .040−.033 .040
 Income (Household).072 .047.009 .042.013 .041
 Religious Denomination−.039 .042.074 .038.071 .038
Leftism ON
 ConservationFixed to 0−.292**.070−.287**.070
 Self-TranscendenceFixed to 0.595**.061.596**.060
 Education−.060 .048−.188**.049−.187**.049
 Age−.006 .057.000 .048−.002 .048
 Gender.097*.046−.071 .048−.071 .048
 Income (Household)−.327**.051−.294**.052−.293**.052
 Religious Denomination.004 .044.011 .045.010 .044
Conservation ON
 Education−.211**.046−.213**.045−.213**.045
 Age.242**.047.238**.048.239**.048
 Gender−.125**.047−.110*.047−.110*.047
 Income (Household)−.122*.053−.129*.053−.128*.053
 Religious Denomination.157**.044.159**.043.159**.043
Self-Transcendence ON
 Education.116*.048.115*.048.114*.048
 Age.150**.050.144**.050.145**.050
 Gender.241**.046.236**.047.236**.047
 Income (Household)−.102 .053−.102*.052−.103*.052
 Religious Denomination.066 .047.065 .047.066 .047
Indirect effects
FDP ON
 Conserv. VIA Libertarianism--- ---(−.399) .556(.194) .172
 Conserv. VIA Leftism--- ---(.046) .077(.147)*.054
 Self-Tran. VIA Libertarianism--- ---(.198) .277(−.100) .090
 Self-Tran. VIA Leftism--- ---(-.095) .155(−.305)**.089
SVP ON
 Conserv. VIA Libertarianism--- ---(.800)*.303(.761)**.070
 Conserv. VIA Leftism--- ---(.026) .038(.069)*.029
 Self-Tran. VIA Libertarianism--- ---(−.397)*.158(−.394)**.058
 Self-Tran. VIA Leftism--- ---(−.054) .076(−.142)*.052
SP ON
 Conserv. VIA Libertarianism--- ---(−.954)*.324(−.517)**.091
 Conserv. VIA Leftism--- ---(−.225)*.065(−.174)**.050
 Self-Tran. VIA Libertarianism--- ---(.474)*.170(.267)**.056
 Self-Tran. VIA Leftism--- ---(.458)**.089(.362)**.067
Factor covariances
 Libertarianism WITH Leftism.242**.066−.150 .140−.149 .129
 Conserv. WITH Self-Tran..376**.064.357**.065.359**.065
 N626626626
 Degrees of freedom209199205
 Loglikelihood−15711.961−15467.347−15472.109
 AIC31711.92331242.69431240.217

In the first model, the simplistic model, the research designs applied by Schwartz and his colleagues are translated into the SEM framework. It is assumed that the effect of political values on vote choice can be neglected and that the latter is sufficiently explained by personal values and socio-economic factors (Caprara et al. 2006). Hence, as the paths in model (a) illustrate, vote choice is determined exclusively by personal values and socio-economic factors.12 As argued in the theory section, ignoring political values in the process of voting might not be advisable. Therefore, in the second (complex) model political values are introduced, generating additional paths: vote choice is now also dependent on political values, and the latter on personal values. This means that, in addition to the direct effects of personal values on vote choice, the indirect effects of personal values on vote choice through political values are also estimated. In the final model, the parsimonious model, the direct effects of personal values on vote choice are set to zero, imposing an assumption that the impact of personal values on vote choice is fully mediated by political values. The supremacy of the parsimonious model is successfully demonstrated if (a) in the simplistic model the direct effects of personal values on vote choice are significant but (b) become insignificant in the complex model (when indirect effects are controlled for), and if (c) the model fit of the parsimonious model is better than or equals the model fit of the complex model (see Baron and Kenny 1986).

Results

Let us turn to the results yielded by the structural equation models. The simplistic model, i.e. the model of Schwartz et al. put into structural equation form, corroborates their original findings (Schwartz 1996, Barnea and Schwartz 1998, Caprara et al. 2006). In this model the paths leading from personal to political values and from political values to vote choice are fixed to zero. As reported in Table II, Column 1, citizens valuing self-transcendence are less likely to vote for rightist parties (the standardized probit coefficient for FDP is −.565, for SVP = −.694) but more likely to vote for the leftist Social Democratic Party (.482). Similarly, conservation is positively correlated with voting for the conservative SVP (.825) but negatively with voting for the left-libertarian SP (.-689). Since the variation in personal values accounts for variation in vote choice, personal values fulfill the necessary condition of being possibly mediated by political values.

In the complex model, the paths from personal values to political values and from the latter to vote choice are now freed (refer to Table II, Column 2). As expected, all of the direct effects of personal values on party choice become insignificant. In other words, they are completely mediated by political values. Still, the main finding of the Schwartz theory is confirmed, as conservation substantially correlates with libertarianism (−.872) and likewise self-transcendence with leftism (.595). Yet the variation of vote choice is now fully explained by the direct effects of libertarianism and leftism (and those of economic status). Libertarianism discriminates social democratic voters (SP) and conservative voters (SVP) from centrist voters (CVP), the reference category; Leftism discriminates SP voters from the reference category. Also note that the complex model is superior to the simplistic model, as both log-likelihood and AIC indicate a better model fit.

To make the case that vote choice is only indirectly affected by personal values even stronger, we now fix the direct effects of personal values on vote choice to zero in the parsimonious model. As can be seen in Table II, bottom of Column 3, fixing these parameters decreases the log-likelihood only marginally (from −.15467 to −15472) and, taking the increase in degrees of freedom into account, the difference of the model fit is equal: the chi-square difference test on the log-likelihood yields a value of 9.273, which is insignificant (p = .158) given the increase of 6 degrees of freedom.13 That the parsimonious model might fit the data even better than the complex model is indicated by the lower (and thus favorable) Akaike information criteria (AIC) value, which penalizes overfitting.

Having demonstrated the supremacy of the parsimonious model, we can now discuss its results, which are also illustrated in Figure 3, in more detail. First, political values determine vote choice, as predicted. The more (economic) leftist a voter, the more likely it is he votes for the Social Democratic Party (the standardized probit coefficient equals .608) and the less likely it is he votes for both the Liberal Party (FDP) and the Swiss People’s Party (SVP) (.−512 and −.239, respectively). Similarly, stressing libertarianism increases the likelihood of voting for the Social Democratic Party (SP) and decreases the likelihood of voting for the SVP, but does not discriminate between FDP and CVP (the reference category) voters. The relationships between political value orientations and vote preferences found here parallel previous findings (Leimgruber et al. 2010).

Figure 3.

 The Parsimonious Model of Value-Based Voting.
Note. All estimates are statistically significant at the .05 level except for those designated “ns”. The estimated effects of socio-economic covariates on personal values, political values, and vote choice are not shown. Refer to Table II for these estimates and more details on that model. (N = 626)

Second, the relationships among values detected here broadly support the theory of Schwartz et al. (Schwartz 1996, Barnea and Schwartz 1998, Caprara et al. 2006), though they are not as clear-cut as predicted. While the effect of conservation on political value orientations is fairly consistent with the theory (−.872 on libertarianism but only −.292 on leftism), the effect of self-transcendence on both value orientations is substantial, with the predicted effect on leftism (.596) being only marginally larger than the unpredicted effect on libertarianism (.444).

Third and most consequently, this finding also affects the indirect effects of basic values on vote choice. It is not the case, as the theory implies, that conservation influences voting behavior only through libertarianism, nor does self-transcendence influence vote choice exclusively through leftism. In fact, whether personal values do affect voting behavior at all is predominantly determined by how strongly political value orientation discriminates between party voters (again in reference to the centrist CVP voter) – the two indirect effects that are not significant (p < 0.5) are those via libertarianism on voting for FDP, which themselves are not significant.14 But for all that, it is fairly impressive to see how big the impact of personal value orientation is on vote choice overall: ten out of twelve indirect paths are statistically significant (p < .05). It is important to note, however, that the estimates of the indirect effects are not directly interpretable. Because the direct effects of personal on political values are standardized linear regression coefficients, while the direct effects of political values on vote choice are standardized probit coefficients, it is not appropriate to multiply the former with the latter as it is commonly done. (The multiplied estimates depicted in Table II are therefore put in brackets).15 Although they are not clearly interpretable, we can still acknowledge that the largest indirect effects are those in line with the theory. So conservation does strongly predict vote choice through the assumed mediator libertarianism, as its indirect effect on voting for SVP is .761 and for SP −.517.

Conclusion

In contrast to the value based voting model proposed by Schwartz and his collaborators (Schwartz 1996, Barnea and Schwartz 1998, Caprara et al. 2006), this article posits that personal values affect voting behavior only indirectly through political value orientations. Empirical analysis supports this modified hypothesis. The parsimonious model, which fits best to the Swiss electoral data, shows that self-transcendence and conservation are completely mediated by leftism and libertarianism and thus exert only an indirect influence on vote choice. This means, for instance, that a citizen attributing high importance to self-transcendence is likely to vote for the Social Democratic Party, because he, given his personal values, also endorses political leftist values. Similarly, a voter prioritizing conservation tends to prefer the Swiss People’s Party because his personal values are expressed through his authoritarian values.

However, the results of the applied structural equation models also support the fundamentals of Schwartz’ and his colleagues’ theory. First, personal values are strongly associated with political values. Conservation, for instance, has an enormous effect on libertarianism, and the high correlation between self-transcendence and leftism is similarly considerable. It must be noted, however, that also the not assumed relationship between self-transcendence and libertarianism is substantial. This, of course, weakens Schwartz’, and equally my, hypotheses to some extent, as the pattern is not as clear-cut as theorized. Yet it does not undermine the general argument entirely, nor is it totally unreasonable. As Barnea and Schwartz (1998) argue, giving high importance to “universalism” (which essentially means understanding, tolerance, and caring for others) does not necessarily have to affect political stances exclusively on the (economic) left-right continuum, rather it may also affect where one stands culturally (or morally). The second finding supporting Schwartz’ reasoning is the enormous total – even if mediated – impact of personal values on voting behavior, as ten out of twelve indirect paths on vote choice have been found to be significant.

This result brings us back to the main argument of this paper: that of the indirect effects. To be fair, the criticism raised against Schwartz et al. rather concerns the research designs they employed than the theoretical argument they made. The heart of their theory is the connection between personal and political value orientation (Barnea and Schwartz 1998), even if the empirical focus lies on the direct effects of personal values on voting behavior. It seems that the authors are rather unaware of this apparent inconsistency between theory and empirics. For instance, they inappropriately concluded that “[...] this analysis supports the association of values with voting through their ties [...] of political ideology” in a study where only direct effects had been tested (Barnea and Schwartz 1998: 30; my emphasis). This conclusion, as it has been shown here, is not unfounded, but it was premature based on their empirics. Considered that way, the study undertaken here only confirms, now in a methodologically adequate way, what has been taken as theoretically correct ever since.

Nonetheless, if it is empirically demonstrated that the relationship is in fact more complex, this must consequently lead to a broader conclusion and a refinement of the theory of value-based voting. My point is that personal values need to be translated (or transformed) first into political meaning in order to have an effect on political behavior (van Deth 1995). Or, from another perspective, personal values need to get activated by political values in a specific situation such as voting (cf. Maio and Olson 1995). Thus, a citizen must be aware of what his private needs mean in the political realm, which political program conforms to his personal guidelines, or which party serves his private motives. It is as if social psychologists have gone a little too far in demonstrating the importance of basic values in politics and have thereby lost sight of what politics is really about. Even more, they have lost sight of what elections are about. In elections, parties and candidates compete against each other, they advertise and sell programs (Downs 1957), and they make their case through campaigning and communication. And the language of campaigns and communication is first and foremost political. These days, politicians do not promote an abstract, desirable end-state of society. Rather, they much more mundanely “defend their old policies, sell new policies, and justify their rule” (Popkins 1991: 8).

Thus this analysis specifies more clearly how and when personal values are related to political reasoning and behavior (Feldman 2003, Verplanken and Holland, 2002). It also reconfirms the large impact of political values on vote choice, particularly when controlling for measurement error (Ansolabehere et al. 2008). Apparently, the electorate of the main Swiss parties – save the Christian Democrats and their still predominant Catholic electorate – orient themselves nowadays by shared values rather than by socio-economic status (cf. Kitschelt 1994). These insights fit nicely with the idea of the “political man” portrayed by early political psychologists, as “[…] personal values are important for the framing of a political belief system” (Lane 1972: 181).

It is up to future research to determine whether the hypothesis of indirect effects also holds in different electoral settings and, in particular, when data allow for generating all of the entire dimensions of Schwartz’ (1992) personal values. It could also prove promising to model issue voting in a similar framework as the one that is employed here, with attitudes towards a specific political issue regressed on personal values. For instance, it may well be that attitudes towards salient issues (Niemi and Bartels 1985), and thus decisive issues, in an election may be backed up by specific personal values. Or similarly, it may be that the effect of personal values on vote choice depends upon whether the choice is over easy or hard issues (Carmines and Stimson 1980). And lastly, it would be interesting to know more about how different styles of political communication affect personal values. It can be inferred from our findings that a campaign slogan such as “Change,” successfully heralded by Barack Obama and his campaign, may not have only attracted political progressives but, even more fundamentally, may have enticed people endorsing openness to change in their daily life.

Footnotes

  • 1

     See for instance Kerlinger (1984), Evans et al. (1996), Marks et al. (2006); for the Swiss context in particular see Leimgruber et al. (2010) and Kriesi et al. (2006).

  • 2

     Note that Kitschelt (1994) considers liberalism to be different from libertarianism.

  • 3

     See Altemeyer (1981) for a very similar conclusion concerning the authoritarian personality.

  • 4

     See, for instance, Ansolabehere et al. (2008), Jacoby (2006), Goren (2005), Knutsen and Kumlin (2005).

  • 5

     Note that vote choice and some of the political items were interviewed in the post-electoral telephone survey, while the personal value items and some additional political items were measured in a follow-up mail-back questionnaire.

  • 6

     The Green party is not included in the analysis because of the limited number of Green voters in the data set.

  • 7

     Note that some of the political items are formulated fairly generally and are therefore not far from tapping political values. However, whether these items measure attitudes or values is not of a serious concern here, as the underlying constructs measured in the structural equation models are political value dimensions either way (see van Deth 1995).

  • 8

     Exploratory factor analysis based on these ten items clearly suggests a two-factor model. The eigenvalues for the first five factors are: 2.802; 1.972; 0.832; 0.771; 0.742. Hence, Kaiser’s (1960) rule of retaining only factors having eigenvalues greater than one as well as Cattell’s (1966) scree test suggest a two-factor solution.

  • 9

     The higher order value openness to change is only covered with one item of the “stimulation” domain (“He looks for adventures and likes to take risks. He wants to have an exciting life.”). The superordinate value self-enhancement is covered with one item of each the “achievement” and the “power” domain (“Being very successful is important to him. He hopes people will recognize his achievements.” And “It is important to him to be rich. He wants to have a lot of money and expensive things”, respectively). Preliminary inclusion of these three items neither lead to a satisfactory two-dimensional factor solution suggested by Schwartz’ theory, nor did the latter two items satisfactorily reproduce self-enhancement. Hence, these three items are excluded from the analysis together with one item of the domain “security”, as it contains an undesired political reference (“It is important to him that the government ensures him safety against all threats. He wants the state to be strong so it can defend citizens.”). All other personal value items are considered not containing political reference.

  • 10

    Schwartz (1994) argues that the response tendency to rate personal values in general as more or less important might generate (“artificially”) positive correlations between factors.

  • 11

     The statistical software package used in this analysis is Mplus 5.1. The structural equation models will include categorical indicators (the personal and political value items) as well as a nominal outcome variable (vote choice). The only estimation procedure than can be used in Mplus 5.1 for such a model is MLR (maximum likelihood parameter estimates with standard errors and a chi-square test statistic that are robust to non-normality and non-independence of observations; see Muthén and Muthén [2007]). For consistency reasons, MLR has been applied to the measurement model as well, even though Mplus 5.1 does not report the common model fit statistics. Yet, classifying the indicators as continuous and estimating the measurement model with maximum likelihood (ML), the test statistics indicate, according to Hu and Bentler (1998), an acceptable model fit (CFI = .898; RMSEA = .050; SRMR = .060).

  • 12

     Note that in each model employed, socio-economic factors are modeled as independent determinants of personal values, political values, and vote choice. Hence, it is first assumed that the causal direction runs from socio-economic status to values. This can be questioned, as both directions of causality between values and some socio-economic categories are possible. For instance, some personal values might be influential on religiosity (Roccas 2005). Secondly, SES covariates are modeled neither as moderating nor as mediating factors. This is again an assumption that can be questioned. For instance, it could be argued that political values have a stronger impact on vote choice for well educated voters than for less educated voters (cf. Zaller 1992), i.e. education is a moderator between political values and voting. However, this moderating effect has not been found in preliminary analyses here.

  • 13

     The log-likelihood difference test applied here takes a correction factor due to MLR estimation into account. For more details see Muthén and Muthén on http://www.statmodel.com/chidiff.shtml.

  • 14

     The significance of the indirect effects is estimated with Sobel’s (1982) test based on the SE of the unstandardized coefficients: z-value = a*b/SQRT(b2*sa2+a2*sb2sa2*sb2); where a = unstandardized coefficient between independent variable and mediator, sa standard error of a; b = unstandardized coefficient between the mediator and the dependent variable, sb = standard error of b.

  • 15

     For the same reason, and also because it has not been a primary goal of this study, total effects of personal values on vote choice are not estimated.

Ancillary