Political ideology and voice pitch preference for political leaders : an experimental approach of ideology and voice pitch preference in Dutch political context

(1)

Political ideology and voice pitch

preference for political leaders

An experimental approach of ideology and voice pitch

preference in a Dutch political context

Master thesis political science: International Relations Frank Boels (10088423)

Political Elites and the Economy Supervisor: dr. Gijs Schumacher Second reader: dr. Daphne van der Pas June 21, 2017

(2)

2

Abstract

Psychological differences between people define how they perceive the social world. These worldviews determine political preferences and thus political ideology. These worldviews are often operationalized in terms of Right Wing Authoritarianism (RWA) and Social Dominance Orientation (SDO). According to Laustsen et al. (2015) people high in RWA and SDO are more likely to identify with republicans and conservatives as opposed to democrats and liberals. Since republicans and conservatives experience more uncertainty and threat they are more likely to vote for political leaders with lower voice pitches. Lower voice-pitches are after all associated with strength and social dominance, characteristics that are considered valuable when it comes to prevalence in conflicts. This research is aimed at applying this relationship between ideology and voice pitch preference for political leaders to a different political context. Therefore this research focuses on the Netherlands, that on the elite level is structured by a cultural and an economic dimension. It is demonstrated that authoritarianism (in this study measured on a SCA instead of a RWA scale) and SDO are associated with these two dimensions of political ideology. Besides, it is argued that measuring one’s ideology by a few political issue stances is a more appropriate way of measuring political ideology than is self-identification. By using an experiment, it is surprisingly demonstrated that political ideology does not predict voice pitch preference for political leaders in a Dutch political context. Moreover, in contrast to current theoretical assumptions this research shows that cultural conservatives who prefer lower voices in leaders do not reward dominance more than less cultural conservatives. Economic conservatives even give lower dominance evaluations than less economic conservatives. This study raises questions about the current ideas on ideology and voice pitch preference and therefore strongly encourages more research on this topic.

(3)

3

1. Introduction

People refer to political leadership in vast terms. Leaders are often called ‘charismatic’, but this is a rather vague term encompassing many traits, which makes it a comprehensive but shallow concept at the same time. Sometimes people are more specific: a leader is perceived as ‘very charming’, ‘extraordinary intelligent’ or is praised for his/her rhetorical skills. This study focuses on a more subtle and often under addressed characteristic of political leadership: voice pitch, or the perceived ‘highness or lowness’ of one’s voice.

In the animal kingdom vocal signals reveal much information about the caller of these signals. They inform receivers about fighting ability or social dominance, because acoustic properties reliably predict traits such as size or strength (Klofstad et al., 2012). Because vocalizations directly relate to physiological and physical properties of the caller, they cannot be faked or exaggerated (Maynard-Smith & Harper, 2003; Searcy & Nowicki, 2005). Vocal signals do not only play an important role in the animal kingdom; they also influence human behaviour. Recent research shows that (and how) voice pitch relates to the way speakers are perceived. Research shows that men with lower voice pitches are perceived as ‘more attractive’ (Feinberg et al., 2005; Oguchi & Kikuchi, 1997; Puts, 2005; Tigue et al., 2012; Zuckerman & Miyake, 1993) physically stronger (Tigue et al., 2012; Collins, 2000; Sell et al., 2010) and more socially dominant (Tigue et al., 2012; Puts et al., 2006; Wolff & Puts, 2010). Because of these associations with strength and dominance researchers have suggested that a preference for lower voices derives from dispositional abilities in leaders to protect and prevail in conflict (Tigue et al., 2012).

For women on the other hand, higher voice pitches are perceived as more attractive (Collins & Missing, 2003; Feinberg et al., 2008; Jones et al., 2008), while lower voice pitches are, just as for men, associated with social dominance (Borkowska & Pawlowski, 2011; Jones et al., 2010). Because voice pitch is highly sexually dimorphic (a female’s voice is on average twice as high as a male’s voice) (Klofstad et al., 2012), it can be argued that voice pitch plays a significant role for people electing their leaders.

Political scientists and psychologists are therefore increasingly interested in the relationship between a leader’s voice pitch and leadership evaluations by the public. The idea that lower voices are positively related to leadership evaluations is supported by recent research. Electoral and experimental data alike show that candidates with lower voices win more votes (Anderson & Klofstad, 2012; Klofstad, 2013; Klofstad et al., 2012; Tigue et al., 2012). This is presumably the result of the previous mentioned evolutionary factors: deeper voices are associated with strength and masculinity. But also because for both sexes, higher pitched voices

(5)

5

are associated with negative emotions like panic, fear and stress (Weeks et al., 2012). These findings suggest that on average politicians with lower pitched voices have a great advantage when it comes to being elected, compared to candidates with higher pitched voices. Although voice pitch is mainly determined by throat autonomy (and thus quite stable), a speaker can modulate his/her voice to a certain extent through voice training (website American Scientist).

Some researchers have argued that there are individual differences between people with regards to voice pitch preference. Laustsen et al. (2015) find that republicans are more likely to vote for a candidate with a lower voice than are democrats. This is due to the fact that republicans and conservatives are more likely than democrats and liberals to see the world as a dangerous place (Duckitt & Sibley, 2010). In theoretical terms: psychological differences cause people to adhere certain political ideologies. This thesis is written to critically evaluate the research by Laustsen et al. (2015) and to apply the presumed theoretical relationship between ideology and voice pitch preference to a two-dimensional political context i.e. the Netherlands.

2. Literature Review

2.1. Worldviews: RWA & SDO

The conclusion that republicans and conservatives are more likely to see the world as a dangerous place is based on the idea that political preferences are formed by attitudinal predispositions. These attitudinal predispositions are an expression of one’s worldview, which results from a person’s personality and environment (Laustsen et al., 2015: 3). The notion that socio-political attitudes result from psychological differences between people, started with research by Adorno and colleagues (1950) (Duckitt & Sibley, 2010: 1861). They assumed that a single left-right ideological dimension structured socio-political attitudes and beliefs. The ideological extremes of this dimension are respectively pro-democratic, egalitarian, tolerant, liberal attitudes as opposed to antidemocratic, pro-fascist, prejudiced and political conservative attitudes (ibid: 1862). However, many empirical findings did not support the assertion about the one-dimensional structure of socio-political attitudes (ibid: 1863). This eventually led to the development of two constructs, Right wing Authoritarianism (RWA) and Social Dominance Orientation (SDO) (ibid: 1863-1864). Research based on Schwartz universal motivationally based values (1992) found that RWA is strongly correlated with the value dimension of conservation (security, conformity and tradition) versus openness to change (stimulation and self-direction), whereas SDO is strongly correlated with the value dimension of self-enhancement (hedonism,

(6)

6

power and achievement) versus self-transcendence (benevolence, universalism) (ibid: 1866; Duriez et al., 2005: 300). RWA thus expresses the motivational goal of maintaining collective or societal security, order, cohesion and stability (which is at odds with autonomy, self-expression and individual freedom) reinforced by the social worldview belief that the world is inherently dangerous, unpredictable and threatening (Duckitt & Sibley, 2010: 1867). SDO on the other hand expresses the motivational goal of power, dominance and superiority over others, induced by the Machiavellian worldview belief that life is a struggle for power, whereby the strong win and the weak lose, as opposed to a cooperative place where people care for and help each other (ibid: 1868-1869). Research demonstrates that RWA tends to predict socio-cultural conservatism better than economic conservatism, whereas SDO tends to predict economic conservatism better than socio-cultural conservatism. However, RWA and SDO are also positively correlated (Duriez et al., 2005; Jost et al., 2009).

2.2 RWA, SDO and one dimensionality of political ideology

The idea that RWA and SDO represent the psychological roots of political conservatism also logically derives from work by Jost (2007). According to Jost (2007) political conservatism is a belief system that consists of two core components: resistance to change and opposition to equality. These components, as illustrated in the previous paragraph, form the base of RWA and SDO respectively. From a theoretical perspective, ‘resistance to change’ represents socio-cultural conservatism, while ‘opposition to equality’ represents economic conservatism (Feldman & Johnston, 2014: 339). Both elements serve to reduce uncertainty and threat. Resistance to change and opposition to equality are traits that preserve the status quo because they favour what is familiar and known over the risky project of social change. The broader argument is that ideological differences between left and right have psychological roots: stability and hierarchy provide reassurance and structure while change and equality lead to chaos and unpredictability (Jost, 2007: 990). So while socio-cultural conservatism and economic conservatism are conceptually distinct, they tend to be related in practice due to their common psychological roots (Feldman & Johnston, 2014: 339). Consequently, Laustsen et al. (2015) argue that these psychological differences explain differences between democrats and liberals versus republicans and conservatives. Given that socio-political attitudes not only derive from environmental factors but even more from genetic predispositions (Alford, Funk & Hibbing, 2005) it is assumed that genetics reasonably explain the universal left-right spectrum of political ideology, which remains applicable cross-culturally and over time (ibid: 153). In this thesis political

(7)

7

ideology is defined as a belief system or a “configuration of ideas and attitudes in which the elements are bound together by some form of constraint or functional interdependence” (Converse, 2006: 3).

In short, this perspective theorizes that universal differences between left-right (and liberal-conservative) derive from psychological differences between people in terms of experiencing uncertainty and threat. People high in RWA or SDO experience more uncertainty and threat than people lower on these worldview dimensions. A high score on RWA results in conservative cultural political preferences, while a high score on SDO results in conservative economic political preferences. Because both personalities seek to reduce uncertainty and threat ideological differences between individuals tend to be value coherent. Value coherence means “the degree to which leftist or progressive values one issue coherently go together with leftist or progressive ones on others, and to which rightist or conservative values on one issue coherently go together with rightist or conservative values on others” (Achterberg & Houtman, 2009: 1649-1650).

2.3 RWA, SDO and two dimensionality of political ideology

However many political scientists argue that political preferences of the masses are structured by two separate ideological dimensions instead of one singular dimension, i.e. a socio-cultural and an economic dimension. (Converse, 2006 [1964]; Van der Brug & Van Spanje, 2009; Achterberg & Houtman, 2009; Feldman & Johnston, 2014; Heath et al., 1994). The two underlying values of the socio-cultural and economic dimension are freedom and equality, respectively (Achterberg & Houtman: 1650). The economic dimension ranges from egalitarianism to laissez faire, while the cultural dimension ranges from authoritarianism to libertarianism. The term terms authoritarianism and libertarianism are slightly confusing in this context, because ‘authoritarianism’ also refers to the personal value dimension of authoritarianism. Besides that it is more common in the Netherlands to refer the cultural dimension in the Netherlands in terms of ‘progressive’ and ‘conservative’. Therefore in this research the terms ‘progressive’ and ‘conservative’ in the Dutch context refer to the extremes of the cultural dimension of political ideology. This cultural and economic ideological dimension are presumed to be clearly distinct and unrelated, which means that an egalitarian stance on economic issues is just as likely (or unlikely) to be combined with an progressive or conservative stance on cultural issues (ibid: 1650). Kitschelt (2004) argues that the cultural dimension of political preferences is about individual autonomy of citizens to govern their lifestyles and tolerance and respect for

(8)

social-8

cultural difference (Kitschelt, 2004: 2). Based on Kitschelt (2004), De Lange (2007) translates these to two components: citizenship and ethnocultural relations & individual freedom (De Lange, 2007: 420). The elements citizenship and ethnocultural relations can be specified to two components, ‘immigration’ and ‘integration of cultural minorities’, whereas individual freedom is about ‘diversity of lifestyles’ and ‘ethical legislation’ (De Lange, 2007: 420). These four components are therefore considered as the building blocks of the cultural dimension of political ideology.

The economic dimension is about the full political redistribution of income and material assets at one extreme (egalitarianism) versus the complete market allocation of economic resources at the other (laissez-faire) (Kitschelt, 2004: 2). There is strong theoretical support for the multidimensionality of political ideology. The dual process model by Duckitt (2010) suggests that the social and economic dimensions of political ideology derive from distinct personal worldviews. These different worldviews, often operationalized in terms of RWA and SDO, greatly correspond with the social and economic dimension of ideology, respectively (Feldman & Johnston, 2014: 339). According to Altemeyer (1998) these personality dimensions represent empirically independent strands of conservative attitudes. They are un- or at most weakly correlated (Altemeyer, 1998).

In short, it is demonstrated that there are two perspectives on political ideology. Both perspectives subscribe to the idea that political preferences derive from personality differences between people. However the latter perspective does not endorse the value coherence of the first perspective, but to the contrary assumes that two unrelated dimensions structure political ideology of the masses.

2.4 RWA versus Social Conformity – Autonomy dimension

Throughout the years, RWA has not been an undisputable concept. It has been criticized for several reasons. To start with, RWA has been criticized for the ideological bias it contains. Speaking of Right Wing Authoritarianism presumes that there is no such thing as Left Wing Authoritarianism. Altemeyer (1996) proposed a Left Wing Authoritarianism scale, but since there has not been empirical evidence for the existence of Left Wing Authoritarianism, it has been argued that it would be better to isolate a common ‘authoritarianism’ dimension (Feldman, 2003: 44). Another problem is that issues that should reflect authoritarianism often reflect social conservatism, the issues it aims to predict (ibid). The same has been argued with regards to intolerance to deviant groups (ibid: 44-45). Feldman (2003) therefore constructs a new

(9)

9

‘authoritarianism’ dimension, which is called the social conformity – autonomy (SCA) dimension (ibid: 49). This dimension is basically about the tension between the goals of personal autonomy versus social cohesion. More specifically it is about the question how much social rules and restrictions are needed to guarantee personal liberty (ibid: 46). It is hereby important to note that because of this tension between social conformity and personal autonomy, this dimension is defined by the relative priority people attach to these two values (ibid: 48). For theoretical reasons outlined above, the SCA dimension instead of the RWA dimension will be used to predict ‘conservativeness’ on the cultural dimension. However before addressing the research questions and the corresponding hypotheses, it is necessary to outline the American political context (in which the research of Laustsen et al. (2015) took place) and the Dutch political context. Not least because this contributes to the understanding of the methodological choices of this research.

2.5 The liberal-conservative dimension and partisanship in the United States

Since the late eighteen-century, political issues have often been discussed in terms of a ‘left-right’- and in the United States- a ‘liberal-conservative’ dimension (Malka & Lelkes, 2010: 158). Liberal historically means freedom from governmental interference in social and economic life, whereas conservative means support for preservation of monarchical, religious and aristocratic institutions (ibid: 158). However since then, political views that are consistent with liberalism and conservatism have changed. For example, since the 1930s opposition to and support of redistributive policies in the United States became associated with respectively conservatives and liberals, while liberals were historically in favour of minimalist government intervention in the economy (ibid: 158). Since the 1970s, contrasting cultural stances, such as those on abortion and homosexuality, have been conceptualized on the ‘liberal-conservative’ dimension (Baldassarri & Gelman, 2008). Besides that, on this one-dimensional dimension prevailing intuitive combinations of attitudes are not naturally consistent with the labels ‘conservative’ and ‘liberal’. For example, one might be in favour of abortion, based on the value principle of ‘freedom of choice’. This person might similarly reject governmental redistribution policies with regards to assisting the needy, based on the same principle of ‘freedom of choice’ (ibid: 157). In this case, the first issue stance would be regarded as liberal, while the latter is regarded as conservative. But would this person regard him/herself a liberal or a conservative? The examples described above show that there are inconsistencies with regards to the meanings of the terms ‘conservative’ and

(10)

10

‘liberal’, both at the elite (discursive) level (top down) as well as on the individual level (bottom-up). This raises the question whether identification with either liberals or conservatives actually represents one’s political ideology, if the terms ‘liberal’ and ‘conservative’ are a) subject to change on the elite level, and b) sometimes inconsistent with personal values.

Ellis (2012) addresses this question by making a distinction between ‘operational’ ideology, or actual political issue stance and ‘symbolic’ ideology, or self-identification. His research demonstrates that the American public is on average symbolically conservative and operationally liberal (Ellis, 2012). This implies operational ideology cannot be reliably measured by self-identification. In the United States this difference on the ideological liberal-conservative dimension manifests in partisanship i.e. identification with the democratic or the republican party (Huddy et al., 2015: 1; Abramowitz & Saunders, 2008). Partisanship in the United States has increased in strength in recent years and continues to have a great influence on political behaviour including vote choice and voter turnout (Bankert et al., 2). There are broadly two perspectives of partisanship: instrumental partisanship and expressive partisanship (Huddy et al., 2015: 1; Bankert et al., 2-3). The instrumental perspective assumes that partisanship is a manifestation of issue preferences, ideological beliefs and party performance. According to the expressive perspective on the other hand, partisanship is a social identity strengthened by social affiliations to gender, religious and ethnic groups (ibid: 2-3; Huddy et al., 2015: 1). Therefore the same logic as with ideology applies to partisanship: the instrumental perspective represents ideology, whereas the expressive perspective represents ideological identity.

There is another problem with the singular liberal-conservative dimension i.e. the meaning people attach to this dimension. Some people see it primarily as a conflict dimension of social issues, while others see it in terms of economics (Zumbrunnen & Gangl, 2008). This heterogeneity of meanings between people implies that a single liberal-conservative dimension is inadequate to be truly indicative for one’s political ideology.

From examining the American context it is concluded that with regards to both ideology and partisanship a distinction should be made between ideological identity and ideology. Otherwise it is impossible to determine whether identification with democrats/liberals or republicans/conservatives is motivated by ideological beliefs or a need for identity. The second problem relates to the issue that a single liberal-conservative dimension cannot be truly indicative of one’s political ideology.

Laustsen et al. (2015) however claim that self-declared party affiliation (on a democrat-republican scale) and ideological self-placement (on a liberal-conservative scale) are valid criteria for measuring individual differences in perceptions of the social world (Laustsen et al., 2015:

(11)

2-11

6). With this claim they do not take into account the difference between ideological identity and ideology. To accurately determine how personality differences lead to political ideology and subsequently voice pitch preference, political ideology needs to be measured differently. In this study political ideology will therefore be measured through ‘political issue stance’ (this is more in depth explained in the methodology section of this thesis). Besides that, to examine the effect of the two dimensions of political ideology on voice pitch preference, political ideology should be measured in a two-dimensional political context.

2.6 Two dimensionality of political ideology in the Netherlands

The Dutch political context is different from the American political context in the sense that political preferences of voters are structured by a cultural and economic dimension instead of one dimension (Middendorp, 1989; Van Der Brug & Van Spanje, 2009). While there is much debate on the question whether party competition takes place on two dimensions (e.g. Warwick, 2002; Sniderman & Bullock, 2004; Kriesi et al., 2006) or one dimension (e.g. Van Der Brug & Van Spanje, 2009), this debate implies that two ideological dimensions are prominent at the elite (political party/media) level. This is important because political elites (party leaders, media representatives) strongly influence the specific contents of political ideology (Jost et al., 2009: 315). In fact this means that people form an opinion about certain topics because these topics are addressed by political elites. But this also means that the two dimensional structure of political ideology is reinforced by political elites. An example of the latter is the Dutch Kieskompas (Election compass) developed by political scientist Andre Krouwel. Kieskompas provides voting advice based on a political space that is structured by two dimensions (Otjes & Louwerse, 2014: 263-264). These voting advice applications play a substantial role in shaping voter’s perceptions of the positions of political parties (ibid: 263). But these voting advice applications therefore also ‘shape’ voters’ political ideology in the sense that they create political consciousness: if they use a two dimensional structure, this structure is adopted by the public. As a consequence it is likely that the Dutch public is better capable than the American public to relate to political issues in a two dimensional way (since they are not constrained by partisanship or liberal/conservative ideological identification).

With this last theoretical addition, all the building blocks of the theoretical model of this research have been discussed. Figure 1 shows the theoretical model used in this research as opposed to the theoretical model used by Laustsen et al (2015).

(12)

12

Figure 1. Schematic representation of the research by Laustsen et al. (2015) (the upper model) and the research presented in this paper (the lower model). The concepts above the rectangular blocks represent the theoretical concepts, while the concepts within the rectangular blocks represent the operationalized concepts of both studies.

2.7 Dominance and leadership

Lautsen et al. (2015) also claim that the tendency of conservatives and republicans see the world as a hostile place will lead them to put a high premium on dominance in leaders (Laustsen et al., 2015: 3). In other words, according to this assertion ‘conservatives’ would reward dominance in leaders. However, disagreement exists with regards to the relationship between dominance and leadership, which makes this a questionable assertion. Van Vugt (2006) claims that dominance is not related to leadership for a number of reasons. In the first place, while it is attractive to replicate behaviour of some social species to human behaviour, this comparison is not applicable. One important reason is that human hierarchies are much flatter than those of most other social species, including non-human primates (Van Vugt, 2006: 358). Secondly, human hierarchies are built on prestige rather than on dominance (ibid). A third reason is that many resources only become available through cooperation, and once they are available they cannot be controlled by one individual (ibid). Finally, group members have alternatives for following a leader. They can follow another leader or leave the group entirely. This greatly restricts the power of one individual leader (ibid). Since research has shown that individuals who tend to see the world as a dangerous place show greater preference for ‘dominant’ faces (Spisak et al., 2012a; 2012b), similar patterns in voice pitch preference are expected. This means that voting for a

(13)

13

candidate with a lower voice would imply voting for a ‘dominant’ voice. This research will contribute to the debate about dominance and leadership by reviewing dominance evaluations with regards to voice pitch preference in an experimental political context. By taking dominance evaluations into account with regards to the relationship between ideology and voice pitch preference, it becomes evident whether ‘conservatives’ (in this study ‘conservatives’ on the cultural and economic dimension) really ‘put a premium on dominance’ when they opt for a leader with a lower voice pitch.

3. Research questions and hypotheses

The most general assumption with regard to voice pitch and voting behaviour derives from the introduction of this study. Because several studies indicate that male leaders with lower voice pitches are perceived as more attractive, socially more dominant and physically stronger and because higher voice pitches are associated with panic, fear and stress it is hypothesized that:

H1: People generally tend to prefer male political leaders with lower voice pitches compared to male political leaders with higher voice pitches.

As discussed previously, political ideology derives from socialized worldview beliefs. These worldviews are often operationalized in terms of RWA and SDO. For reasons outlined in the theoretical section, in this thesis the SCA scale of authoritarianism will be used instead of RWA. Based on these theories we would therefore expect a correlation between these socialized worldview beliefs and political ideology. The first sub question will therefore be: To what extent do

SCA and SDO predict the cultural and economic dimension of political ideology? This question is especially

relevant since this research measures political ideology differently (i.e. based on ‘political issue stance’). If SCA and SDO predict political ideology that would therefore also verify my operationalisation of political ideology. Based on the theoretical assumptions the following hypotheses are formulated:

H2a: SCA is better at predicting the cultural dimension of political ideology than the economic dimension.

(14)

14

H2b: SDO is better at predicting the economic dimension of political ideology than the cultural dimension.

H2c: Compared to SDO, SCA is better at predicting the cultural dimension of political ideology, whereas compared to SCA, SDO is better at predicting the economic dimension of political ideology.

To examine how political ideology is related to voice pitch preference in political leaders in a two dimensional context, this research is limited to the Dutch political context, that on the elite (political party/media) level is structured by a socio-cultural dimension and an economic dimension. Based on the theories it is hypothesized that these two dimensions of political ideology are clearly distinct and unrelated. The third hypothesis is therefore:

H3: The cultural dimension is unrelated or at most weakly correlated with the economic dimension.

This study also seeks to clarify the relationship between leadership and dominance. Laustsen et al. (2015) expect a positive relationship among conservatives between these two variables whereas Van Vugt (2006) argues that leadership is not based on dominance. This research will therefore also provide an answer to the question whether people who opt for leaders with lower voice pitches actually ‘put a premium on dominance’. This results in the sub question to what

extent do people who score ‘authoritarian’ on the cultural dimension and ‘laissez faire’ on the economic dimension give higher dominance evaluations when they vote for a leader with a lower voice pitch? Based on the theory,

the following hypothesis is formulated:

H4a: In people who tend to vote for leaders with lower voice pitches, the higher the dominance evaluations they give, the more ‘conservative’ on the cultural dimension they are.

H4b: In people who tend to vote for leaders with lower voice pitches, the higher the dominance evaluations they give, the more ‘laissez faire’ on the economic dimension they are.

(15)

15

Since the American political discourse is dominated by one dimension, i.e. a ‘liberal-conservative’ dimension, it is not clear how these separate ideological dimensions in the Netherlands predict voice pitch preference in political leaders. The central research question in this thesis will therefore be: To what extent does political ideology predict voice pitch preference for political leaders in a Dutch

political context? In the Netherlands, a similar relationship between ideology and voice pitch is

expected as in the United States, i.e. that a ‘conservative’ score on both dimensions of ideology leads to a preference for lower voices. The hypotheses with regards to the central research question are therefore:

H5a: A ‘conservative’ score on the cultural dimension leads to a preference for leaders with lower voice pitches.

H5b: A ‘laissez faire’ score on the economic dimension leads to a preference for leaders with lower voice pitches.

4. Material and methods

To test the proposed hypotheses a survey has been conducted, using the online survey software programme Qualtrics. The survey was conducted in Dutch, since it examines the relationship between worldviews, ideology and voice pitch preference in a Dutch political context. The survey was meant to measure the respondents’ score on SCA, SDO, ideology and voice pitch preference. The first part of the survey measures the respondents’ score on SCA, SDO and political ideology. The second part of the survey measures voice pitch preference for imaginary political leaders through an experiment. Six men were recorded speaking the politically relevant and partisan-neutral sentence ‘Ik verzoek u op mij te stemmen, de aankomende verkiezingen’. This statement is similar to the original ‘I urge you to vote for me this November’ used in the original study by Laustsen et al. (2015). It was explicitly mentioned that the political leaders were

imaginary (not real) politicians. The men ranged in age from 49 to 63 years (𝑥̅ ̅= 56 years). The

average voice pitches of the unaltered voice fragments ranged from 116 Hz to 169 Hz (𝑥̅ ̅= 139 Hz). Subsequently through the phonetic analysis and modulation program Praat (version 6.0.26) the voice fragments were manipulated to obtain a lower and higher version of each voice. The lower versions were altered to 20 Hz lower than the original ones, whereas the higher versions were altered to 20 Hz higher than the original ones. Previous studies have demonstrated that

(16)

16

voters can perceive which voice of each pair is higher in pitch, if the voice pitches are altered to a difference of 30-40 Hz between the voices (Anderson & Klofstad, 2012; Klofstad et al., 2012). Finally, the manipulated audio files were resynthesized so that they could be uploaded at the online audioplatform SoundCloud.

The response options for all questions in the survey were all based on forced choice. This means in this study that the respondents were forced to express an value/attitude or to vote for a political leader. One important reason for eliminating a ‘don’t know’ option is that this increases the number of survey records that are usable for analysis (website Sage research methods). Some have argued that eliminating a ‘don’t know’ option in a survey risks forcing people to express an opinion they actually do not hold (Converse & Presser, 1986: 35-38). However several experiments in the United States have demonstrated that people who express a lack of opinion on a certain topic in fact do have an opinion (Krosnick et al, 2002). They also show that lower educated are more inclined to select a ‘don’t know’ option (ibid: 382). These forced choice answer options were primarily used to get enough data on the two answer questions and the votes for the candidates, since the questions on ideology included a ‘neutral’ option.

The answer options of the questions that were used to measure ideology were based on a five point Likert scale. A Likert type scale is frequently used to measure attitudes or opinions (Likert, 1967). A Likert type scale is particularly useful to measure opinions that are too complex to answer with ‘yes’ or ‘no’ but rather allow for degrees of opinions. The five point Likert scale used in this research is therefore considered highly appropriate for measuring political ideology. Regarding all the statements that measure political ideology, terms that reflect partisan positions have intentionally been avoided to guarantee partisan neutrality. The answer options (whether ‘1’ represents a conservative/laissez faire answer as opposed to a progressive/egalitarianism answer) for this dimension of political ideology were randomized to avoid a ‘first answer option’ response bias.

The voice pitch experiment consisted of two parts. In the first part, respondents listened to three pairs of different voices. Subsequently the respondents were asked to evaluate the political leaders on ‘dominance’ and they were asked who they would vote for. It is naturally important that the respondents do not (assume to) know the ‘political leaders’ who speak the sentence, because that would bias their choice. For this reason the men whose voices were recorded live in a different city than where my social network is concentrated. The voices were paired based on the average voice pitch of the original voice fragments. The voice fragment of the first candidate had a mean voice pitch of 116 Hz, while the opposing candidate had an

(17)

17

average voice pitch of 125 Hz. The third candidate had an average voice pitch of 147 Hz, while his opposing candidate had an average voice pitch of 150 Hz. The fifth candidate had an average voice pitch of 169 Hz, while his opposing candidate had an average voice pitch of 125 Hz. Because of this great difference between the average pitch of the fifth and sixth candidate, in group 1 the original pitches were used so that the voices were clearly distinguishable with regards to their voice pitch (a difference of 44 Hz). In group 2 the original voice of candidate five was manipulated 40 Hz higher, whereas the original voice of candidate six was manipulated 40 Hz lower. In this first part of the experiment, respondents were randomly assigned to one of two groups. The first group listened to the lower version candidate one’s voice, and to the higher version of candidate two’s voice. The second group on the other hand listened to the higher version of candidate one’s voice, and to the lower version of candidate two’s voice etc. The lower voice was alternately presented first to control for a ‘first answer option’ response bias. All separate audio fragments were uploaded to SoundCloud and were only accessible via a private link in the survey. To give an overview, the two groups (and the voice pitches of the ‘political leaders’ in Hz) are presented in table 1.

Table 1. Schematic presentation of the candidates and their voice pitches (in Hz) for the two groups (part one of the experiment).

Candidate Group 1 Group 2

Candidate 1 96 Hz 145 Hz 136 Hz Candidate 2 105 Hz Candidate 3 167 Hz 127 Hz Candidate 4 130 Hz 170 Hz Candidate 5 125 Hz 165 Hz Candidate 6 169 Hz 129 Hz

An advantage of this kind of research design is that it is possible to compare the probability of choosing candidate 1 in group 1 with the probability of choosing candidate 1 in group 2 (provided that ceteris paribus). In that way it is convenient to discern the effect of voice pitch on the likelihood to vote for one political leader or the other. Another advantage is that the respondents do not know the goal of the research because they hear two different voices. Therefore they will not be able to give what they might perceive as social desirable answers.

In the second part of the experiment the respondents listened to the manipulated (higher and lower) versions of five of the same voices. In that respect, this research replicates the research design by Laustsen et al. (2015). One advantage of this design is that respondents cannot be biased by knowing (or assuming to know) a candidate (because they hear the same

(18)

18

voice twice at a different pitch). However the advantage of the first part does not apply to this part: when the respondents become aware they hear the same voice twice, they might get an idea where the research is about, which could lead to a less reliable response. Combining these two parts in the experiment therefore increases the reliability of the research. The voice pitches of the candidates (A to J) are presented in table 2. Again, the order with regards to which voice was presented first (the higher or lower version) was randomized.

Table 2. Schematic presentation of the candidates and their voice pitches (part two of the experiment).

Candidate Voice pitch

Candidate A 96 Hz Candidate B 136 Hz Candidate C 189 Hz Candidate D 150 Hz Candidate E 145 Hz Candidate F 105 Hz Candidate G 130 Hz Candidate H 170 Hz Candidate I 103 Hz Candidate J 143 Hz

The evident advantage this experimental research design is that it is easy to isolate and manipulate the independent variable to see the effect of the independent variable (in this case voice pitch) on the dependent variable (in this case voice pitch preference). (Creswell, 2012: 295). The internal validity of this research is therefore high. Figure 2 gives an schematic overview of the conducted survey.

(19)

19

Abbreviations: SCA = Social Conformity-Autonomy (measures authoritarianism); SDO = Social Dominance Orientation (measures social dominance orientation.

4.1 Operationalization

The first four survey questions were about child rearing values and are indicative for the SCA scale of authoritarianism. While Feldman (2003) presents a list of 17 items that reliably measure authoritarianism, he also argues that child rearing values are a good indicator of authoritarian values more generally. If a person does not believe that people are naturally inclined to conform, this person will be very likely to think that children should be trained to be obedient, not to challenge authority and to accept the way society is (Feldman, 2003; 490). Another advantage of measuring authoritarian values rather than directly socio-political attitudes, is that it avoids predicting ‘political conservatism’ by measuring ‘political conservatism’. The questions with regards to child rearing values were about the relative importance of obedience versus self-reliance, curiosity versus good manners, independence versus respect for elders and being considerate versus well behaved. All answers that reflect the relative importance of social conformity/authoritarianism were given the value ‘1’ and all answers that reflect personal autonomy/non-authoritarianism were given the value ‘0’.

The next set of items measures social dominance and was measured by using a standardized SDO set of items (Feldman, 2003: 70). This set consisted of eight agree/disagree items and was used to measure the respondents’ attitudes towards social equality. Like the questions with regards to authoritarianism, attitudes that reflect a preference for social inequality were given the value ‘1’, whereas attitudes that reflect a preference for social equality were given the value ‘0’.

Subsequently political ideology has been measured by ‘political issue stance’ on two ideological dimensions. In total 8 questions were used that measure a person’s political ideology. Five of these questions were extracted from the politics and values waves of the LISS (Longitudinal Internet Studies for Social Science) panel data archive (LISS panel data archive, politics and values; wave 9). The other three questions have been determined after thorough discussions with an expert on political ideology in the Netherlands. The first set of four statements was meant to determine the respondents’ score on the cultural dimension of political ideology. Based on the four components current and political relevant statements were collected that together represent the cultural dimension of political ideology. Three of these statements were extracted from the LISS panel data achieve (LISS panel data archive, politics and values; wave 9). The first statement was about asylum in the Netherlands and said that it should be easier to get asylum in the Netherlands (1 = totally agree, 5 = totally disagree). The question was

(20)

20

recoded so that ‘0’ means ‘progressive’ and ‘1’ or ‘conservative’. With regards to this question the answers were recoded so that 1 = ‘0’ and 5 = ‘1’. The second statement was about cultural diversity and said: ‘In the Netherlands, some people think that immigrants should be able to live here with the conservation of their own culture. Others think that they should fully adapt to the Dutch culture. Where would you place yourself on a scale from 1 to 5, whereby 1 means that should fully adapt to the Dutch culture and 5 that immigrants could fully conserve their own culture?’ These questions were recoded whereby 1 = ‘1’ and 5 = ‘0’. The third statement was related to ‘diversity of lifestyles’ and said: ‘Homosexuals should be able to kiss on the streets’ (1 = totally agree, 5 = totally disagree; recoded to 1 = ‘0’ and 5 = ‘1’). The fourth statement, also extracted from the LISS panel data archive, was about whether euthanasia should be allowed or prohibited (1 = prohibited, 5 = allowed; recoded to 1 = ‘1’ and 5 = ‘0’). The answer options (whether a 1 or a 5 represents a conservative answer) were intentionally randomized, to avoid an ‘first answer option’ response bias. The aggregate answer presents a score from 0-1 and was labeled ‘cultural dimension’.

The second set of four statements was meant to determine the respondents’ score on the economic dimension of political ideology. Two statements regarding this dimension were extracted from the LISS panel data archive (politics and values; wave 9). The answers were recoded so that 0 means ‘egalitarianism’ and 1 means ‘laissez faire’. The first statement was a general one about income differences in the Netherlands (1 = income differences should increase, 5 = income differences should decrease; recoded to 1 = ‘1’ and 5 = ‘0’). The second statement extracted from the LISS panel said: ‘Unemployment benefits in the Netherlands should be cut’ (1 = totally agree, 5 = totally disagree; recoded 1 = ‘1’ and 5 = ‘0’). The third and fourth statement were about salient political issues during the last parliamentary elections of March 15 2017. The third statement is about the privatization of the Dutch health care system versus one national health insurer (1 = preference for one national health insurer, 5 = preference for multiple private health insurers; recoded to 1 = ‘0’ and 5 = ‘1’). The fourth and last statement said: ‘Multinational corporations in the Netherlands should pay more taxes’ (1 = totally agree, 5 = totally disagree; recoded to 1 = ‘0’ and 5 = ‘1’). The aggregate answer presents a score from 0-1 and was labeled ‘economic dimension’.

In part two of the survey the participants listened to the paired voice fragments of the imaginative political leaders. Subsequently there were asked to evaluate both leaders on a five point scale on dominance (1 = ‘very dominant’ to 5= ‘not dominant at all’; recoded to 1 = ‘1’ and 5 = ‘0’). Finally there were asked who they would vote for. The answers were recoded so

(21)

21

that the higher pitched voice was given the value ‘0’ and the lower pitched voice the value ‘1’. All the survey questions can be found in appendix I.

In the results section the respondents are categorized according to their educational level. These categories are based on the Standaard onderwijsindeling 2016 from the Centraal Bureau voor de

Statistiek (Central Agency for Statistics). Based on this division the categories ‘geen opleiding’ (no

education) up to and including ‘MAVO, VMBO-t, MBO-kort’ were included in the category lower educated. ‘Anders’ (other) could not be categorized and has therefore been left out of this division. The categories ‘MBO, MTS, MEAO’ and ‘HAVO, VWO, Gymnasium’ were accordingly included in the category middle educated. The categories ‘HBO, HEAO, PABO, HTS’ and

‘Universiteit’ (University) were included in the category higher educated (Standaard onderwijsindeling

2016, Centraal Bureau voor de Statistiek). The distribution of respondents between the Dutch educational levels can be found in appendix II.

4.2 Procedures & sample

People were invited to participate through my online social network (Facebook & Facebook Messenger), through email and through the practice of a general practitioner. The latter method of recruiting participants was aimed at making the sample more representative of the population. In that way, the external validity of this research increases. People have been actively approached until the pre-specified target of ± 200 participants was met. The number of 203 participants in this study is considered appropriate for this kind of research, since similar studies have been conducted with either less (Andersen & Klofstad, 2012 [N=75]; Tigue et al., 2012 [N=125]) or more participants (Laustsen et al., 2015 [N=403]).

4.3 Data analysis plan

All the statistical analyses will be executed in IBM SPSS Statistics (version 22). To test hypothesis 1 the two parts of the voice experiment will be analysed separately, to get an overview of voice pitch preference for all the candidate pairs. In part one of the voice pitch experiment the voting differences between group 1 and group 2 (compared to the average distribution between the two groups) will be analysed. This difference will be analysed through a Chi-square test. Part two of the voice pitch experiment will be analysed through a one sample t test. The voice pitch

(22)

22

preference value for the respective candidates will be compared to the test value 0.5 (that assumes no difference between lower and higher pitched voices).

To test hypotheses 2a, 2b and 2c and hypothesis 3 a Pearson’s correlation test will be executed, whereby all the relationships between the variables SCA, SDO, the cultural and the economic dimension will be examined. If SCA and SDO are indeed correlated to the two dimensions of political ideology two univariate and two multivariate regression analyses will be executed to determine to what extent SCA and SDO predict political ideology.

To examine the relationship between leadership and dominance (and thus test the hypothesis about the relationship between the two dimensions of political ideology and dominance evaluations among the people who tend to vote for leaders with lower voice pitches) 141 people who score higher than 0.5 on average on voice pitch preference will be selected (by using the filter option in SPSS). Subsequently a Pearson’s correlation test will be executed to examine the relation between dominance evaluations and the cultural dimension and the economic dimensions of political ideology respectively.

Finally the association between political ideology and voice pitch preference will be examined. Several models will be used to test the associations between the two dimensions of ideology and voice pitch preference. First two univariate regression models will be used to examine the association of the cultural and economic dimension of political ideology with voice pitch preference. Subsequently a multivariate regression analysis will be executed to examine the independent association of the two dimensions of political ideology with voice pitch preference. At last, a multivariate regression analysis will be executed including all the examined variables in this research. To include all these variables in this multivariate regression model, a dummy variable for sex will be created (whereby 0 means female and 1 means male). For the educational categories, three new dummy variables will be created (whereby 1 means lower, middle and higher educated and 0 means the other educational categories).

5. Results

5.1 Descriptive statistics

The survey was completed by 203 participants, of which 106 were men and 97 were woman (52.2 and 47.8 percent respectively). The participants ranged in age from 17 to 73 years (𝑥̅̅ = 31.75 years, SD = 13.33 years). Table 3 gives an overview of all the variables used in this study.

(23)

23 Table 3. Overview of the variables used in this study.

Variable Mean ± SD N (%) Age 31.75 ± 13.33 Sex: male 106 (52.2%) Educational level - Low 5 (2.5%) - Middle 40 (19.8%) - High 157 (77.7%) SCA 0.298 ± 0.278 SDO 0.124 ± 0.175 Ideology 0.321 ± 0.134 - Cultural dimension 0.299 ± 0.126 - Economic dimension 0.344 ± 0.183 Voice pitch preference 0.607 ± 0.196 N = 203

5.2 Statistical results

5.2.1 Voice pitch preference

Table 4 shows the distribution of votes between the two groups with regards to the different voice pairs (part one of the voice pitch experiment) The N indicates the number of votes each candidate got and the (%) shows the percentage of the votes. Based on the average distribution between the two groups for each voice pair the difference between the groups has been calculated. The only difference that was not significant was the difference between group one and two with regards to candidate pair one (candidate 1 and candidate 2). For the other two pairs of candidates (candidate 3 versus candidate 4 & candidate 5 versus candidate 6) there is a significant difference between the two groups.

Table 4. Part 1 of the experiment. Differences between group 1 and 2 with regards to voting for the lower pitched candidate. Votes lower voice

Group 1 N (%)

Votes lower voice Group 2

N (%) Candidate 1 versus Candidate 2 44 (43.1%) 60 (59.4%) Candidate 3 versus Candidate 4 88 (86.3%)** 55 (54.5%)** Candidate 5 versus Candidate 6 85 (83.3%)* 31 (30.7%)* * p<0.02; ** p<0.001; group 1: N = 102; group 2: N = 101

(24)

24

Table 5 shows the distribution between the votes for the lower and higher pitched candidates (part two of the experiment). The table shows that in all cases the candidate with the lower voice pitch received more votes. The table also demonstrates that all observed differences are statistically significant except for candidate pair ‘Candidate E versus Candidate F’. The exact p values with regards to the candidate pairs AB, CD, GH and IJ are 0.021, 0.002, <0.001 and <0.001 respectively.

Table 5. Part 2 of the experiment. The (proportion of) votes on the lower pitched candidate as compared to the test value 0.5 (indicating no difference between the lower and higher pitched candidate).

Candidate Votes lower voice (% of total) Candidate A versus Candidate B 116 (57.1%)*

Candidate C versus Candidate D 122 (60.1%)** Candidate E versus Candidate F 104 (51.2%) Candidate G versus Candidate H 154 (75.9%)*** Candidate I versus Candidate J 131 (64.5%)*** * p<0.05; ** p<0.01; *** p<0.001; N = 203.

The average voice pitch preference value of both parts of the experiment combined is 0.6069. This result is statistically significant (p<0.001).

5.2.2 SCA & SDO and political ideology

Table 6 shows the correlations between SCA, SDO, the cultural and the economic dimension of political ideology. The results are presented in table 6.

Table 6. Correlation between SCA, SDO, the cultural and the economic dimension of political ideology.

Variable 1. 2. 3. 4. 1. SCA 1.00 2. SDO 0.246* 1.00 3. Cultural dimension 0.308* 0.401* 1.00 4. Economic dimension 0.282* 0.499* 0.492* 1.00 * p<0.01, N = 203

The results show that SCA more strongly correlates with the cultural dimension of political ideology (0.308) than the economic dimension of political ideology (0.282). They also demonstrate that SDO more strongly correlates with the economic dimension (0.499) than the

(25)

25

cultural dimension (0.401). However SDO is thus stronger correlated to both the cultural dimension and the economic dimension of political ideology compared to SCA. Table 6 also shows that the economic and the cultural dimension of political ideology are correlated (0.492). According to a ‘rule of thumb’ a correlation of 0.1-0.3 is considered a weak correlation, 0.40-0.60 a moderate correlation and 0.70-0.90 indicates a strong correlation (Dancey & Ready, 2007). The cultural and economic dimension of political ideology are thus moderately correlated.

Table 7 and table 8 show to what extent the variance on the cultural and economic dimension of political ideology is explained by SCA and SDO. The univariate parts of tables 7 and 8 show the extent to which these worldview dimensions explain the variance on the cultural and economic dimension of political ideology. The multivariate parts of the tables show the independent association between SCA and SDO and the cultural and economic dimension.

Table 7. Univariate and multivariate regression analysis with the cultural dimension as dependent variable and SCA and SDO as independent variables. Univariate Multivariate Beta R2 _Beta _R2 SCA 0.140* 0.095 0.101* 0.208 SDO 0.289* 0.161 0.249* * p<0.001

Table 8. Univariate and multivariate regression analysis with the economic dimension as dependent variable and SCA and SDO as independent variables. Univariate Multivariate Beta R2 _Beta _R2 SCA 0.186* 0.080 0.112* 0.269 SDO 0.521* 0.249 0.478* * p<0.001

Table 7 and 8 show that SCA alone explains 9.5 percent of the variance in cultural dimension and 8.0 percent of the variance in economic dimension. The tables also show that SDO alone explains 16.1 percent of the variance in economic dimension and 24.9 percent of the variance on the economic dimension. The multivariate regression models indicate that SCA and SDO combined explain 20.8 percent of the variance in cultural dimension and 26.9 percent of the variance in economic dimension. All these results are statistically significant (p<0.001).

(26)

26 5.2.3 Leadership and dominance

The mean dominance evaluation of the selected participants that have a preference for lower voices 0.504. Table 9 shows the correlation between the cultural dimension of political ideology and dominance evaluations for this group. Table 10 shows the correlation between the economic dimension of ideology and dominance evaluations.

Table 9. Correlation between the cultural dimension of political ideology and dominance evaluations by people who tend to vote on candidates with lower voice pitches.

Variable 1. 2.

1. Cultural dimension 1.00 -0.054 2. Dominance -0.054 1.00 N = 141

Table 10. Correlation between the economic dimension of political ideology and dominance evaluations by people who tend to vote on candidates with lower voice pitches.

Variable 1. 2.

1. Economic dimension 1.00 -0.171* 2. Dominance -0.171* 1.00 * p<0.05, N = 141

Table 9 shows that there is no significant correlation between the cultural dimension and dominance evaluations among the people who tend to vote for the candidate with the lower voice (p=0.527). Table 10 shows a significant negative correlation between the economic dimension and dominance evaluations among the people who tend to vote for the candidate with the lower voice pitch (p = 0.043).

5.2.4 Political ideology and voice pitch preference

Table 11 shows the correlation between the cultural dimension and the economic dimension of political ideology respectively and voice pitch preference. The table shows that neither the cultural dimension nor the economic dimension of political ideology predicts voice pitch preference significantly. Model 1A shows how strongly the cultural dimension of political ideology is associated with voice pitch preference. Model 1B shows how strongly the economic dimension of political ideology is associated with voice pitch preference. In model 2 the cultural and the economic dimension are taken together to examine their independent association with

(27)

27

voice pitch preference. In model 3, all the measured variables in this research have been included. The p values have been included in the table to demonstrate that none of the measured examined variables in this research are associated with the dependent variable voice pitch preference (since all the p values indicate insignificant associations).

Table 11. Factors associated with voice pitch preference.

Model 1A Model 1B Model 2 Model 3

beta p-value beta p-value beta p-value beta p-value Cultural 0.001 0.990 -0.009 0.945 0.011 0.899 Economic 0.011 0.883 0.014 0.872 0.005 0.959 Age -0.001 0.990 Sex: male 0.016 0.835 Education - Lower - Middle - Higher -0.131 -0.412 -0.357 0.456 0.329 0.419 SCA -0.110 0.163 SDO 0.052 0.547

Model 1A: association between voice pitch preference and cultural score Model 1B: association between voice pitch preference and economic score

Model 2: association between voice pitch preference, and cultural and economic score

Model 3: association between voice pitch preference, and cultural score, economic score, age, sex, education, SCA and SDO.

6. Discussion

6.1 Findings

6.2.1 Voice pitch preference

The data in the previous section clearly indicate that people generally prefer leaders with lower voice pitches, despite the fact that in two of the six cases in part one of the voice pitch experiment people chose the candidate with the higher instead of the lower voice. This is obviously due to the fact that voice pitch is not the only factor that plays a role with the evaluation of one’s voice. Some voices simply sound more appealing than others regardless of their voice pitch. The general preference for lower pitched voices is also illustrated by the second part of the experiment: in this part all cases indicate a preference for the lower pitched candidate. Because the voices in this part were the same voices on a different pitch, this clearly indicates a

(28)

28

general preference for lower pitched voices. The aggregate results (part one and two of the voice pitch experiment combined) demonstrate that in six of the eight cases the lower pitched candidate received more votes. Only in two of the eight cases the higher voice received more votes and in one case the lower pitched candidate received more votes, although this difference was not significant. The overall voice pitch preference value also showed a statistically significant preference for lower pitched voices. Hypothesis 1, which stated that people generally tend to prefer male political leaders with lower voice pitches compared to male leaders with higher voice pitches, is therefore accepted.

6.2.2 SCA & SDO and political ideology

The statistical results show that SCA is more strongly correlated with the cultural dimension of political ideology than with the economic dimension. SCA also explains more of the variance in cultural dimension than in economic dimension. Hypothesis 2a which states that SCA is better at predicting the cultural dimension of political ideology rather than the economic dimension, is therefore accepted.

The results also indicate that SDO is more strongly correlated with the economic dimension of political ideology than with the cultural dimension: SDO explains more of the variance in economic dimension than in cultural dimension. Hypothesis 2b which states that SDO is better at predicting the economic dimension of political ideology rather than the cultural dimension, is therefore accepted.

However the results also indicate that SDO predicts both the economic and the cultural dimension better than does SCA. Hypothesis 2c, which states that compared to SDO, SCA is better at predicting the cultural dimension of political ideology, whereas compared to SCA, SDO is better at predicting the economic dimension of political ideology, is therefore rejected.

The results demonstrate that the cultural and the economic dimension are moderately correlated. Hypothesis 3, which states that the cultural dimension is unrelated or at most weakly correlated with the economic dimension, is therefore rejected. These ideological dimensions prove to be not fully independent of one another.

The results lead to the conclusion that SCA and SDO are indeed correlated to the cultural and the economic dimension of political ideology respectively. Despite the fact that some of these hypotheses with regards to SCA, SDO and ideology have been accepted, SCA and SDO only explain the two dimensions of political ideology to a small extent.

(29)

29 6.2.3 Leadership and dominance

To test hypothesis 4 only the people who tend to vote for leaders with the lower voice pitches were taken into consideration. After all, votes on leaders with lower voices are considered by Laustsen et al. (2015) as votes on leaders with dominant voices. Laustsen et al (2015) expect conservatives to put a premium on dominance in leaders. It was therefore hypothesized that people who tend to vote for leaders with lower voice pitches give higher dominance evaluations the more conservative on the cultural dimension and the more laissez faire on the economic dimension they are (H4). It turns out that the cultural dimension is not correlated to dominance evaluations. Surprisingly however, it turns out that the economic dimension is negatively correlated with dominance evaluations in leaders. This means that the more laissez faire people are, the lower dominance evaluations they tend to give. This difference between the cultural and economic dimension and this negative correlation with the economic dimension is striking though hard to explain. Further research on the economic dimension of political ideology and dominance perceptions in leaders is therefore strongly encouraged.

6.2.4 Political ideology and voice pitch preference

The main question of this study was: ‘to what extent does political ideology predict voice pitch preference for political leaders in a Dutch political context?’ Surprisingly, the answer to this question is simple: this study demonstrates that political ideology does not predict voice pitch preference in a Dutch political context. It turns out that neither the cultural nor the economic dimension of political ideology is associated with voice pitch preference. In fact none of the variables examined in this study are associated with voice pitch preference. This is remarkable since according to the discussed theories ideological differences within countries are universal, given that political preferences derive from psychological differences between people.

Although it should be emphasized that the results only indicate that relative conservatives did not show a greater preference for lower pitched voices (since on average the participants were more progressive/egalitarian than conservative/laissez faire). It could be reasoned that these relative conservatives were in fact not conservative enough to show a general preference for lower pitched voices. This is something that further research could demonstrate.

Political ideology and voice pitch preference for political leaders : an experimental approach of ideology and voice pitch preference in Dutch political context