What is the True Cost of Mass Polarization?
A Study of the Relationship Between Political Polarization and Trust in Political Institutions in the United States
Sama Serena Dean Fadji
Political Science, bachelor's level 2020
Luleå University of Technology
Department of Business Administration, Technology and Social Sciences
What is the true cost of Mass Partisan Polarization?
A Study of the Relationship Between Political Polarization and Trust in Political Institutions in the United States
Democracy is defined by the element of competition. Elite party competition has become one of the most discussed contemporary developments in the United States. Elected representatives from the main parties
have become internally homogeneous, deepening the divide of ideologies between one another. This thesis seeks to establish the relationship between mass partisan polarization and the level of trust in political institutions across the United States. What happens when the public trusts the Elites more than Congress? Elite polarization has divided the masses so deeply in the U.S by electing representatives from
the two major parties whom carry ideologies so distinct from another that the public begin change their ways of forming opinions. This thesis acknowledges that there is high elite and mass political polarization
in the U.S., which is attributed to the heterogeneity in ideologies across the three main political parties (Democrats, Republicans and Independents) and intra-party homogeneity. The elite partisan theoretical
framework expounds the relationship such that the public tends to hold a low level of trust towards the U.S. congress because majority of voters’ partisan motivated decision making is influenced by political
endorsements. The implication is that the public is more likely to hold a considerable level of trust towards their political parties as opposed to the U.S. congress.
Sama Serena Dean Fadji
Bachelor Programme in Political Science
Search Words: Political Polarization, Mass Partisan Polarization, Elite Polarization, Political Trust
TABLE OF CONTENTS
1. INTRODUCTION. . . .3
1.1 Overview. . . 3
1.2 Purpose and Aim. . . 4
1.3 Problem: Numbers into Practice. . . 6
2. THEORY. . . .8
2.1 Theoretical Framework. . . .8
2.2 Previous Empirical Research . . . .10
3. METHOD. . . .11
3.1 Research Designs. . . .12
3.2 Data . . . .13
3.3 Variable Measurements. . . .13
3.4 Data Analysis Design. . . .14
4. ANALYSIS . . . .15
4.1 Descriptive Statistics . . . 15
4.2 Correlation Analysis. . . 18
4.3 Validity and Reliability of the Data . . . 20
4.3.1 Summary on the Validity and Reliability of the OLS Statistical Test. . . .21
4.4 Multivariate Regression Analysis . . . .21
4.4.1 Evaluation of Type I Error in the Multivariate Regression Analysis. . . 24
4.4.2 Summary of the Multivariate Regression Analysis . . . 25
5. Discussion. . . 25
References. . . 28
Appendix. . . 31
1. INTRODUCTION 1.1 Overview
America’s main political parties have grown further apart in their ideologies, deepening a divide that has today come to be known as polarization (Epstein & Graham, 2007). Political polarization comes in many forms, as it is a multifaceted concept defining different levels and degree of segregations between and amongst the public and government.
The state of polarization defines the degree of which opinions are opposed, whereas the process of polarization defines the increase of the opposition during the course of time. Its meaning refers to the opposing of political opinions towards each end of an extreme spectrum and whilst democracy is defined by the element of competition, it is in a natural occurrence, that polarization is the divergence within and amongst political parties and its democratic system of governance. When the political attitudes of the elites are divided between the party-in- government and party-in-opposition, it is defined as elite polarization, and when the masses who vote before them are divided in accordance, it refers to the process of mass partisan polarization (Epstein & Graham, 2007). This thesis will focus on one of the most noteworthy and the most discussed contemporary developments in U.S. politics, of where the split has been apparent amongst the political elites, by which means members of Congress, party activists and other influential players in the process of politics (Druckman et al. 2013) and measure the causes whilst dissecting its consequences. Who is polarized? And what are its consequences for a democracy?
Mass Partisan Polarization is measured by taking into account the ideological distance between voters and parties as well as the level of public support and rejection of the main political parties in the U.S., which include the Democratic, Republican and the Independent parties. This deep division creates a shift in the nature of citizen decision-making and political trust.
Although many studies look at the relationship between the polarization of the public and government, there is a yet to be a great deal of empirical literature dedicated to evaluating the nature of the relationship, specifically between political polarization and the level of trust in political institutions. This study uses SPSS output to measure and evaluate the survey conducted
by the Cooperative Congressional Election Study (CCES) in 2018 querying directly of trust in Congress.
The insight based on previous studies depict findings based on the studies that was conducted, including that by Druckman et al. (2013) depict that there is a negative relationship between elite partisan polarization and political trust in legislature. According to Druckman et al. (2013), partisan polarization adversely affects the level of trust in political institutions because the opinion of voters that operate in a highly polarized environment is apparent to be determined by the party-affiliated framing and endorsements. The study that was conducted by Linde (2018) also confirmed that there is a significant relationship between political communication and cooperative decisions in political parties. Specifically, the insight from Linde (2018) depict that effective communication enhances the strength of cooperative decisions in political parties. This study seeks to extend the research by Linde (2018) by addressing the existing gap with respect to how political polarization determines trust in political institutions.
1.2 Purpose and Aim
The aim of this study is to evaluate the extent of the relationship that exists between mass partisan polarization and the level of trust that exists in the U.S. political institutions. A purely quantitative statistical research design is conducted to measure the gradient of the relationship that exists between elite partisan polarization and trust in U.S. congress. The purpose of this research to calculate the degree of a polarized environment and opinion-formation towards mistrust in political institutions through statistical research and furthermore, gain knowledge on the causes of its direction that leads towards mistrust. When the masses develop mistrust towards their government, it questions the political legitimacy of a nation as political trust indicates towards a legitimate political system, the consequences will be looked at in the chapter portraying the Theoretical Framework as well as direct consequences in the realities of today in the following chapter, Problem. However to understand the level of of mistrust in political institutions, previous studies have shown to result in citizens developing behaviours often challenging the political system, less likely to comply with the laws, engage less in institutionalized forms of political participation (Hooghe & Marien 2013).
While the common understanding of the importance of maintaining a healthy democracy, is to understand the ‘build-up’ through statistical research aimed directly at measuring the negative effects that leads its public towards mistrust, this study aims to gain an understanding that it can be dire and critical within forming the strategy of political communication and party endorsements, and further take the matter and consideration of the masses whom live under the governance of the state.
This study will begin by aiming to grasp the pervasive concepts of Polarization, study its different levels, forms and the degree in regards to its political theory and furthermore, test and measure the cost of the effects that can be consequential on both the citizens and government.
Within Political Science, a healthy democratic state has within itself a competitive political system, this means that there should be a platform where leaders compete with organizations and institutions that provide alternative policies for its public (Schattschneider 1960, 138), the public before it, can then participate in the decision-making process through part taking in large-scale collective action (i.e., voting, petitions and protests). However this theory, as peaceful as it may sound, can be counter productive by resulting in an influence that leads to the installment of a gridlock on important policies. This has been seen as recently as in the United States. The political parties in the U.S have grown further apart drastically and these changes do not go unpunished, ideologies that grow apart in polar opposite directions cause challenges that diminish the efforts that compromise towards tackling serious problems within the country (Epstein & Graham, 2007).
This thesis will aim to relay the theories of political polarization and its consequences into practice, by studying the statistics from the survey, and measuring the division that exists between the trust attitudes of the public, and the U.S legislature on a national level. The elite polarization and partisan polarization theoretical framework are tested at the national level rather than the state level, this is because the state offers a considerable level of variance in legislative polarization compared to the existing trust attitudes that mainly apply in the U.S. legislature (Congress) (Banda & Kirkland, 2018).
The study draws on published data from the individual opinion data included in the Cooperative Congressional Election Study (CCES). The CCES is a large-scale survey that incorporates the political opinion of U.S. adult citizens in the United States. Specifically, the data on mass and
elite partisan polarization as well as trust in political institutions is drawn from the 2018 Cooperative Congressional Election Survey (CCES). Since this study carries more than one explanatory variable, a multivariate linear regression model will be incorporated in the analysis to evaluate the effect of elite partisan polarization on the level of trust in the U.S. congress.
Within the theories of statistics, is the understanding that: when observing the population subjects at a period of time, a cross- sectional data is to be published, and since this thesis relies on the population from the 2018 national political opinion surveys, this data will seek to evaluate mass polarization and the level of trust that individuals hold over the political institutions in the U.S. To edit the gathered data and prepare it for further analysis, a preliminary data analysis will entail the summary of the numbers, that will summarize the given data and represent the population from the survey within the descriptive statistics, in terms of the mean, median, and standard deviation. The intent of conducting the descriptive statistics is to provide a broad picture on the extent of political polarization and trust in political institutions across the U.S.
The multivariate regression analysis is used as a method to measure the degree of the relationship and will be based on the technique for modeling continuous data, the four OLS assumptions (linearity, normality, no multicollinearity and constant variance/ homoscedasticity) as proposed by Franzco and Farmer (2014). The four OLS assumption tests are important because they enhance validity of the estimated regression model (Franzco & Farmer, 2014).
Therefore, all the four OLS assumptions will be tested to establish the validity of the data before conducting multivariate linear regression analysis.
Within the analysis section, it will propose the outcome of the descriptive statistics and use the statistical method when evaluating the strength of the relationship between two variables, a correlation analysis, to evaluate the strength and extent of the relationship between mass partisan polarization and the level of trust in political institutions in the U.S.
1.3 Problem: Numbers into Practice
The problems that arise from public mistrust in political institutions in theory, is dire. This subsection will match the theory with the realities of today, by publishing the consequences
partisan polarization, is seen to have demonstrated a carry of consequences across recent U.S news.
The decisions citizens make that is influenced by the elites who assume the role of ideologies that argue for better representation of its people, creates a polarized environment (Mansbridge 1983, 25). A polarized environment between the elites and the legislative has been observed to be costly by complicating long-term policy challenges, challenges that could be avoided should it have been under circumstances of true bipartisan collaboration (Epstein & Graham, 2007). Yet, while studies mainly focus on the consequences of mistrust on public policies, the existing gap remains, that goes beyond policy gridlock, but rather further into a concerning but yet dire, threat to democracy; How can there be trust in the political institutions, when the elite is making counter arguments?
An example of the elite split of political opinion and its consequences occurred in the United States, when the major parties failed to compromise on a spending bill, known in U.S political parlance as the appropriations bill, that required to be passed by Congress and signed by the President by midnight of the 19th of January in 2018. Since this failed to happen, the federal government closed its doors, resulting in the longest government shutdown in history. According to the BBC World News (2018) Republicans put blame on the minority leader Senate, Chuck Schumer who led the Democrats, whilst the Democrats say they stood by the immigration deal but it was rather brought to a haul by the President. This put a gridlock on a long-term spending bill which would have funded the U.S government for the budget year of 2018, until October, but more so, it saw several unpopular taxes on healthcare remain, it brought the Children’s Health Insurance Programme (CHIP) to a halt, which would have for six years, seen a re-authorized coverage of low-income families to providing healthcare for their children... A halt crafted by the Republican leadership in the House of Representatives and the Senate.
According to Lauka et al. (2018), the level of rejection towards the opposing political party is expounded by the distance in political ideological orientation. To build on this stance a study observed what is called “Partisan Identity Strength”, this endorses the idea of intergroup violence (Gerber et al. 2011). Since the election of 2016, hate crimes have witness a rise and more Americans who identify as a Democrat or Republican within the ideology, strengthen their identity, intensifying the tension between one and another, resulting in the resonation and belief
that by taking the side of their partisan party, comes an approval in the act of violence (Connolly, 2018). When the masses lose trust in their government, it can be a threat to democracy, this occurs as partisan polarization paves a path towards mistrust in the legislative, which causes the violation of state orders. A case of this threat occurred recently, known as the “American Patriot Rally” of May of 2020, when armed protesters entered the Michigan Statehouse in Lansing, going against the orders by not wearing masks nor social distancing, in protest against the stay- at-home order. The Governor, a Democrat, had extended the stay-at-home mandate as a result of more than 41,000 infected and 3,788 deaths due to Covid-19 reported by Al Jazeera News (2020). Republican-led state legislature sued the Governor to put a gridlock on the policies made as emergency measures to prevent the spreading of Covid-19. Trump allies took it as far as to organize demonstrations called “Operation Gridlock” that saw thousands of people, some armed and dressed in militia uniforms, taking to blocking the streets of Lansing, to protest Governor of Michigan, Gretchen Whitmer’s order to stay at home. This escalated into an anti-lockdown protest across the US waving signs in support of Trump and Confederate flags. This example is brought up as, during these partisan polarized times of the public with firearms in demonstration against the Governor’s emergency measures, the President, Donald Trump tweeted in their support and calling them “very good people”.
2. THEORY
2.1 Theoretical Framework
Polarization, like two opposing magnets, in its definition refers to the division in very contrasting sets of beliefs, the opposing of opinions in relation to a theoretical maximum. According to DiMaggio et al. (1996), the state of polarization in political science defines the degree of which opinions are opposed, whereas the process of polarization defines the increase of the opposition during the course of time. This thesis will fall under the Partisan theoretical framework of both Elite and Mass Polarization and the cost of its process in the United States. This chapter covers the side-effects of the divide and further, the fundamental reasons for its persistence though before it, the understanding of the different forms of polarization, from the polarization of the
masses and upwards towards polarization within the government, all within the field of politics, will be mentioned throughout this thesis, but first defined in this subsection to understand the dynamics accordingly.
Political Polarization
The sharp divisions caused by strong opinion in society come in many forms, as sectors do not share the same ideologies, the multifaceted concept of polarizations in politics has different nouns for its different verb of segregations. However within Political Science, it can be simplified by understanding that the big complex picture can be understood further in binary: the Masses and the Elite. The political elites are the elected officials together with their party organizers, whilst the polarization of the masses is the voter opinion by the general public, the electorate’s (McCarty et al. 2006).
Elite Partisan Polarization: This thesis will measure the effect of partisan elite polarization between the party in government and the party in opposition as well as the polarizing effects it has on the masses that further may lead towards political mistrust. Elite Polarization sees its party as internally unified and carry ideologies unique to itself. Scholars mostly use elite polarization to focus on the legislative and deliberative bodies, and in a two-party system, the legislature itself can be polarized. A polarized legislature is the result of when there is a conflict of ideologies between members of the two-parties over legislation and policies, collapsing the notion of an ideological center. To measure the degree of polarity in the legislature, between the elites, political scientists analyze the recorded votes, known as roll call, it is the votes by party members voting for and against each member of the assembly, recorded and later published by interest groups, contrary to the ballot in which the voting of the assembly member’s choice is carried out in secrecy The roll call is measured by looking at its pattern to be able to identify the party line voting and party unity. Speech patterns between parties can be analyzed by looking at the differences to measure polarization (Garand, 2010). One of the most noteworthy and the most discussed contemporary developments in U.S. politics is Elite Polarization, as American politics has been reshaped by the distance between the Democratic and Republican parties in growing more distant in their ideologies from another, they have polarized (McCarty et al., 2006). Elite polarization is the process of which the distance occurs within a state and federal legislatures, the members of Congress, party activists and other influential actors in any political process by the
party-in-government and the party-in-opposition and although party competition is the integral feature of a successful democracy (Dahl, 1971), in the past four decades the polarization has resulted in consequences.
Mass Partisan Polarization: The polarization of the public occurs when the electorate’s beliefs towards policies, political issues and the actors representing them are divided. The differences grow further apart towards polar sides of a spectrum in towards an extreme approach; that the belief in which they carry the moral legitimacy, whilst the ‘other’, the opposing group carries policies that are a threat to the nation and their way of life. According to the insight based on the study that was conducted by Druckman et al. (2018), elite and mass partisan polarization tends to intensity the impact of political party endorsements on public opinions. The fact that partisan identity and partisan motivated decision making is influenced by party endorsements, the voters are unlikely to be swayed by substantive information and the opinion of political institutions such as the U.S. congress. This is because their overall conscious decision- making is determined by party endorsements (Davis & Dunaway, 2016). Taking this into consideration, Partisan Polarization will be measured and included as the independent variable and based on the Lauka, McCoy and Firat (2018) approach of measuring mass partisan polarization. Specifically, the mass partisan polarization will be measured by taking into account the ideological distance between voters and parties as well as the level of public support and rejection of the main political parties in the U.S., which include the Democratic, Republican and the independent parties.
Political Trust: The theory of political trust relies heavily as an indicator of political legitimacy.
The legitimacy of a political system is seen to decrease when it loses the trust of its citizens due to the dire consequences it has on a democracy. The consequences of mistrust has been studied to result in citizens developing behaviours often challenging the political system, less likely to comply with the laws and engage less in institutionalized forms of political participation (Hooghe & Marien 2013). Political trust plays a central role in understanding how citizens relate to political authorities as the behaviour of citizens directly affects a healthy democracy. Without trust in political institutions the political system can see its worst-case scenario of a descent into anarchy, therefore it is important that a government actively ensures the trust of their public. The conventional way in which to measure political trust is via surveying the citizens. This thesis
measures political trust by the survey participants’ level of trust in the U.S. congress. In the CCES survey the level of trust in congress was assessed based on the following question:
“Do you trust and approve in the way the U.S. congress is doing” (CCES.Gov. Harvard, 2020) The responses were restricted based on the 5-point Likert scale, with 1 Strongly Approve, 2 Somewhat Approve, 3 Somewhat Disapprove, 4 Strongly Disapprove, 5 Not Sure. Therefore, dependent endogenous variable (trust in congress) was measured as numerical interval variable.
The measuring of political trust followed an estimated multivariate linear regression model where political trust is the dependent variable determined by the answers from the CCES survey and the value of the dependent variable (political trust) is endogenously determined in the model based on the variations in the exogenous independent variables. The formulated multivariate equation also depicts that the main independent explanatory variable that will determine the extent of political trust is partisan polarization. The other exogenous variables (age and income) are incorporated in the multivariate regression model as control variables to mediate the hypothesized relationship between political trusts and partisan polarization in the U.S.
2.2 Previous Empirical Research
There is yet a great deal of empirical literature that has been conducted to evaluate the nature of the relationship between political polarization (elite and mass partisan polarization) and the level of trust in political institutions (Druckman et al., 2013; Davis & Dunaway, 2016; Lauka et al., 2018). The findings based on the study that was conducted by Druckman et al. (2013) depict that there is a negative relationship between elite partisan polarization and political trust in legislature. According to Druckman et al. (2013), partisan polarization adversely effects the level of trust in political institutions because the opinion of voters that operate in a highly polarized environment is determined by the party-affiliated framing and endorsements.
The study that was conducted by Lauka et al. (2018) also observed that mass partisan polarization has an adverse effect on the extent of political trust towards democratically established political institutions. According to the study, the level of support or rejection of a political party is expounded by the distance in political ideological orientation that exists between the public (voters) and the political parties. However, Davis and Dunaway (2016)
explain that media fragmentation plays an important role in determining partisan-ideological sorting, which in turns influences the level of trust in political institutions. Druckman et al, (2013) suggested that such media framing of social issues is likely to be superseded or counterbalanced by political endorsements especially if the general public and voters operate in a highly polarized political climate where there is considerable heterogeneity in political party ideologies. Finally, the study that was conducted by Linde (2018) also confirmed that there is a significant relationship between political communication and cooperative decisions in political parties. Specifically, the insight from Linde (2018) depict that effective communication enhances the strength of cooperative decisions in political parties. This study seeks to extend the research by Linde (2018) by addressing the existing gap with respect to how political polarization determines trust in political institutions.
Based on the empirical foundation provided by Druckman et al. (2013), which examined how elite polarization influences public opinion, this study extends the stated research by addressing the research gap associated with elite and mass partisan polarization. Specifically, this study is different from Druckman et al. (2013) because it extends the stated research by examining how mass political polarization influences trust in the U.S. congress rather than public opinion formation. However, Druckman et al. (2013) explored how elite polarization affects the formation of public opinion given the mediating effect of the mainstream media. The study is also unique compared to previous empirical research because it focuses specifically on the U.S.
congress and incorporates the mediating effect of individual age and income on the extent to which mass partisan polarization affects the level of trust in the U.S. congress.
3. METHOD
3.0 Methods
The thesis seeks to establish the relationship between partisan polarization and the level of trust in political institutions with a specific reference to the U.S. congress. This subsection of the data analysis describes the research design that was integrated in the study to analyze the data and
accomplish the primary aim of the research. The variable definition, variable measurement and statistical analysis are also briefly described in the methods section.
3.1 Research Design
A purely quantitative statistical research design is employed to measure the relationship that exists between elite partisan polarization and trust in political institutions (U.S. congress). The elite polarization and partisan polarization theoretical framework are tested at the national level rather than the state level as proposed by Banda & Kirkland, 2018). The main justification for testing the elite and partisan polarization at the national level is because states offer a considerable level of variance in legislative polarization compared to the existing trust attitudes that mainly apply in the U.S. legislature (Congress) (Banda & Kirkland, 2018). The study relies on published data from the individual opinion survey included in the year 2018 Cooperative Congressional Election Study (CCES), which is published by Harvard University (CCES.Gov.
Harvard, 2020). The CCES is a large-scale survey that incorporates the political opinion of U.S.
adult citizens in the United States (Banda & Kirkland, 2018).
The following multivariate linear regression model is incorporated in the analysis to evaluate the effect of elite partisan polarization on the level of trust in the U.S. congress.
Political Trust = α + β1Partisan Polarization + β2Age + β3Income + ε (eq. 1)
The endogenous dependent variable in the estimated model is represented by the trust in political institutions, which in this case is measured by the survey participants’ level of trust in the U.S.
congress. The value of the dependent variable (political trust) is endogenously determined in the model based on the variations in the exogenous independent variables. The formulated multivariate equation 1 also depicts that the main independent explanatory variable that will determine the extent of political trust is partisan polarization. The other exogenous variables (age and income) are incorporated in the multivariate regression model as control variables to mediate the hypothesized relationship between political trusts and partisan polarization in the U.S. The insight based on previous studies depict that there is a negative association between age and trust in political institutions such that as individuals get older, their overall trust, which is influenced
by the experience with the country’s democratic political processes and partisan politics declines (Marien & Hooghe, 2011). On the other hand, income was found to be positively correlated to trust in political institutions such that as household income levels increases, the public trust in political institutions also increases (Catterberg & Moreno, 2006).
3.2 Data
The study relies on published cross-sectional secondary data based on national political opinion surveys, which seek to evaluate mass polarization and the level of trust that individuals hold over the political institutions in the U.S. Specifically, the data on mass and elite partisan polarization as well as trust in political institutions is drawn from the 2018 Cooperative Congressional Election Survey (CCES). The stated national political opinion survey, which was published by Harvard University incorporates a sample of 60,000 respondents drawn from various U.S. states (CCES.Gov. Harvard, 2020). All the data was retrieved from the Harvard database, which maintains the 2018 CCES survey. Besides information on the participants’ demographic attributes, the respondents were also asked to present their views, opinion and perspective with respect to the level of trust and approval ratings for the various political institutions in the U.S.
The individuals’ level of political orientation as either Democrats, Republicans and Independents coupled with the subjects’ degree of liberalism and conservatism is also presented in the 2018 CCES survey.
3.3 Variable Measurements
Partisan Polarization: The independent variable (partisan polarization) is measured based on the Lauka, McCoy and Firat (2018) approach of measuring mass partisan polarization.
Specifically, the mass partisan polarization was measured by taking into account the ideological distance between voters and parties as well as the level of public support and rejection of the main political parties in the U.S., which include the Democratic, Republican and the independent parties. Therefore, using the Lauka et al. (2018) approach, I aggregated the individual subjects’
level of support for the three main political parties (Democrats, Republicans and Independents) and the individuals’ level of ideological orientation in the U.S. political system using the CCES
survey data, which was retrieved from CCES.Gov. Harvard (2020). In summary, mass partisan polarization was measured as follows;
Mass partisan polarization = Strength of Political Party Support + Ideological Distance
The strength of political party support is incorporated as an 8-level Likert scale variable with 1 being ‘strong Democrat’ and ‘7 being strong Republican’ while 8 captures the participants’ level of uncertainty (‘Not sure’). In this case, polarization is depicted as the variation in the continuum (1 to 7) between an individual being ascribed to the political party ideals of the Democrats, Independents and Republicans. The component variable, ideological distance is a six-point Likert scale variable on the individuals’ level of political divergent viewpoints with 1 being ‘very liberal’, 5 being ‘very conservative’ and 6 being ‘not sure’. Therefore, the final composite variable, mass polarization is a scale variable (2 – 16) with values less than 9 indicating higher mass polarization and higher extreme values (≥ 15) depicting low mass polarization.
Political Trust: The endogenous factor variable (trust in political institutions) was measured as the level of trust in the U.S. legislature (congress). In the CCES survey (CCES.Gov. Harvard, 2020), the level of trust in congress was assessed based on the following question;
“Do you trust and approve in the way the U.S. congress is doing”. The responses were restricted based on the 5-point Likert scale, with 1 Strongly Approve, 2 Somewhat Approve, 3 Somewhat Disapprove, 4 Strongly Disapprove, 5 Not Sure. Therefore, endogenous variable (trust in congress) was measured as numerical interval variable.
3.4 Data Analysis Design
The preliminary data analysis entailed the assessment of the descriptive statistics in terms of the mean, median, and standard deviation. The primary aim of conducting the descriptive statistics is to provide a broad picture on the extent of political polarization and trust in political institutions across the U.S. The data analysis also entailed estimating the Pearson correlation coefficient to establish the level of association between mass partisan polarization and trust in political institutions in the country. The justification for conducting Pearson correlation analysis is to
provide a rough insight on the level of association between political polarization and trust in political institutions (Lumley et al., 2002). The multivariate regression analysis is based on the four OLS assumptions of linearity, normality, no multicollinearity and constant variance (homoscedasticity) as proposed by Franzco and Farmer (2014). The four OLS assumption tests are important because they enhance validity of the estimated regression model (Franzco &
Farmer, 2014). Therefore, all the four OLS assumptions were tested to establish the validity of the data before conducting multivariate linear regression analysis.
4. ANALYSIS
4.0 Analysis
The analysis section presents the outcome of the descriptive statistics and correlation analysis to assess the extent of association between mass partisan polarization and the level of trust in political institutions in the U.S. The assumptions based on which the multivariate OLS regression model was estimated are assessed at the 5% significance level. The significance level of 5% is selected as the tolerable error threshold such that the accuracy (validity) of the estimated coefficients can vary up to the allowable error rate of 5%. The implication is that the analysis requires all the estimated coefficients to have a 95% confidence of capturing the true population and a tolerable error rate of only 5% based on standard statistical tests (Lumley et al., 2002).
Specifically, the four OLS assumptions that were evaluated include linearity, normality, constant variance and no multicollinearity among the exogenously determined independent factor variables. The statistical evidence based on the outcome of the multivariate OLS regression analysis is also presented in the analysis section.
4.1 Descriptive Statistics
Descriptive statistics are important metrics in exploratory research. Descriptive statistics such as the measures of central tendency including mean and median provide a broad picture on the average value of the endogenous and the exogenous variables (Freedman, 2017). On the other
extent of variation in the factor variables that have been incorporated in the study. Table 1 presents a summary of the descriptives for all the four factor variables that were incorporated in the study including the level of political trust and mass partisan polarization in the U.S. during the year 2018 when the CCES survey was conducted.
Table 1.
Descriptive Statistics
Variable Mean Median Standard Deviation
Political Trust 3.3 3 1.0
Mass Partisan Polarization 7.1 7 3.3
Age 48 48 18
Income 15 6 27
Sample Size (N) 60,000 60,000 60,000
Source: SPSS Output Based on Analysis of the CCES Survey Data (Appendix)
The descriptives depict that the average public rating on trust in the U.S. congress (Mean = 3.3;
Median = 3; STD = 1.0) is moderate. This is because, based on the 5-point Likert scale, the average rating on the level of political trust can be construed to capture the participants’ assertion that they somewhat disapprove (mistrust) the U.S. congress. The lower standard deviation of 1 implies that the extent of political trust does not vary substantially from the mean. The implication is that on average, a considerable proportion of the CCES survey respondents had a moderate level of mistrust in the political institutions and especially, the U.S. congress.
The outcome based on the descriptive statistics also depict that the average level of mass partisan polarization (Mean = 7.7; Median = 7; STD = 3.3) is very high. Therefore, based on the descriptives, there is a very high level of mass partisan polarization in the U.S. as far as the results of the 2018 CCES survey are concerned. Specifically, there is a very strong variation in terms of the support for the three main political parties in the country (Democrats, Republicans and Independents). The analysis can broadly conclude that the two variables are negatively
associated with each other. The average value of political trust and political polarization are presented in Figure 1 below.
Source: SPSS Output Based on Analysis of the CCES Survey Data
Figure 1. Mean Partisan Polarization and Trust in Congress
The mean value of age (Mean = 48; Median = 48; STD = 18) indicates that majority of the survey respondents who participated in the year 2018 CCES survey are 48 years old. The results are consistent with the outcome based on the study that was conducted by Linde (2018), which depicts that most of the participants in national surveys are within 40-50 years of age. However, age has a considerable level of standard variation given that the participants’ age can vary below and above 18 years from the mean age.
The mean value of income (Mean = 15; Median = 6; STD = 27) indicates that majority of the CCES survey respondents reported earning higher annual income levels. Specifically, based on the survey, the mean annual household income represented by the average value of 15 is
$350,000 - $499,999. It means that on average, a greater proportion of the survey respondents are high income earners. However, the data for annual household income appears to be highly
3.3
7.1
0 1 2 3 4 5 6 7 8
Poli1cal Trust Poli1cal Polariza1on
RATINGS
POLARIZATION/TRUST
skewed to the right given that the median value of 6 ($50,000 - $59,999) implies that the survey respondents earn a low average income.
4.2 Correlation Analysis
Correlation analysis is an important metric in statistics, which supplements the insight from the descriptives by presenting the extent of relationship that exists between two or more research variables that have been incorporated in the study (Freedman, 2017). The correlation between the endogenous (dependent) and the exogenous (independent) variables is measured by the Pearson correlation coefficient, which is assessed at the 5% significance level. The significance level of 5% is selected as the allowable threshold for error rate. It means that analysis of the association between political polarization and trust in congress will be assessed with 95% confidence and 5% tolerable error (Franzco & Farmer, 2014). In most studies, the statistical threshold for correlation analysis is set at the Pearson correlation coefficient of r > 0.5 (Franzco & Farmer, 2014). A correlation coefficient greater than 0.5 indicates a fair level of association between the endogenous and the exogenous factor variable (Freedman, 2017).
The outcome based on the Pearson correlation analysis is presented in Table 2 below. The correlation analysis is conducted to establish the level of association between political trust, mass partisan polarization and the two control variables (respondents’ age and income level).
Table 2.
Correlation Analysis
Variable Political Trust Partisan Polarization Age Income
Political Trust 1.00
Partisan Polarization -0.26** 1.00
Age -0.03** 0.10** 1.00
Income 0.03** 0.06** 0.08** 1.00
Sample Size (N) 60,000 60,000 60,000 60,000
ρ**<0.05
Source: SPSS Output Based on Analysis of the CCES Survey Data (Appendix)
The correlation coefficient between political trust and mass partisan polarization (r = -0.26; ρ <
0.05) indicates that there is a fairly low level of negative association between the two factor variables. However, given that the significance (ρ < 0.05) is less than 5%, the insinuation based on the analysis is that there is a fairly low but significant negative relationship between mass partisan polarization and the level of public trust in the U.S. congress. The stated results of correlation analysis can be explained using the elite polarization and the partisan polarization theoretical framework. The mass partisan theoretical framework depicts that the when the political parties are highly polarized as characterized by interparty heterogeneity and intra-party homogeneity, the public trust in the political institutions tend to decline. The partisan motivated reasoning hypothesis that was tested by Druckman et al. (2013) indicated that partisans are more likely to support a frame that has been endorsed by their parties as opposed to a frame that has been endorsed by a different political party. The stated insight depicts that decision making among individuals is dictated by partisan-based motivation and endorsements (Linde, 2018). The results therefore, imply that the public have a higher level of mistrust in political institutions that are not dominated by their preferred political parties. However, the low correlation coefficient between mass partisan polarization and trust in political institution depicts that there are other factor variables that might account for the variation in the level of public trust.
The correlation analysis results reveal that there is a very low negative but significant level of association between the respondents’ age and the trust in the U.S. congress (r = -0.03; ρ < 0.05).
This means that the participants’ age and the level of public trust in political institutions are negatively associated but with a very low level of magnitude on the extent of the stated relationship. The low negative correlation indicates that as the respondents’ get older, their overall approval rating (trust) in political institutions such as the U.S. congress declines marginally at the 5% significance level. Lauka et al. (2018) explains the stated negative relationship between age and trust in political institutions as being explained by the level of experience with the country’s political system. Therefore, older respondents tend to have a low level of trust and approval rating towards political institutions compared to the younger voters due to their wide exposure and experience with the political system in the country. The level of experience with a political system determines the political attitudes that individuals hold against
There is a low positive correlation between the annual family income and the respondents’ trust in the political institutions (r = 0.03; ρ < 0.05). The weak relationship between the respondents’
annual income and trust in political institutions is not statistically significant because the ρ-value of the coefficient is less than 5%. The implication based on the stated relationship is that trust in political institutions is high among the high-income earners and low among the low-income earners. According to Poole (2009), the roots of political polarization can mainly be attributed to the enactment of less popular tax legislations, unemployment and fiscal policies, which have a considerable negative effect on household income. For instance, high income taxes tend to reduce the level of individuals’ disposal income; a situation that is considered a catalyst for political polarization as the public holds divergent views on how the tax system should be formulated. The high income-earners are likely to maintain a fair level of trust in political institutions because such unpopular tax legislations tend to have marginal effects on their wealth.
However, the low-income earners who bear the greatest burden on such unfavorable income tax legislations are likely to be highly polarized and therefore maintain a low approval rating and/or trust in political institutions (Davis & Dunaway, 2016).
4.3 Validity and Reliability of the Data
The validity and reliability of any cross-sectional, longitudinal and time series data, which is subjected to OLS regression is determined by the degree to which the data meets the four (4) OLS assumptions. These include the assumption of linearity, normality, no multicollinearity and homogeneity in the variance of the residuals (Freedman, 2017). This subsection of the analysis presents the outcome of the linearity, normality, multicollinearity and homogeneity tests. The failure to attain at least two of the stated four OLS assumptions is likely to render the formulated multivariate linear regression model ineffective in estimating the multivariate OLS regression model. However, Moore (2017) argues that the violation of any statistical OLS assumption is likely to have a significant adverse effect on the estimated OLS regression model.
The linearity assumption is assessed at the 5% significance level based on the following null and alternative hypothesis;
Ho: There is a linear relationship between political trust and mass partisan polarization.
Ha: There is no linear relationship between political trust and mass partisan polarization.
The normality test is evaluated using the K-S test of normality based on the following null and alternative hypothesis.
Ho: Mass partisan polarization and political trust are normally distributed.
Ha: Mass partisan polarization and political trust are not normally distributed.
The test of multicollinearity is assessed using the following null and alternative hypothesis formulation:
Ho: There is no association between mass partisan polarization, respondents’ age and income.
Ha: Mass partisan polarization, respondents’ age and income are associated.
The following null and alternative hypothesis was used as the basis for conducting the homogeneity test based on the Levene’s equality of variance test.
Ho: The residuals of the political trust multivariate regression model have a constant . variance.
Ha: The residuals of the political trust multivariate regression model have no constant .. . . variance.
4.3.1 Summary on the Validity and Reliability of the OLS Statistical Tests
The data validity and reliability test depict that only two of the four main OLS multivariate assumptions were met when assessed at the 5% significance level (Exhibit 8, Appendix).
Specifically, the normality and the no multicollinearity assumptions were met at the 5%
significance level, which meant that the data was normally distributed and there was not significant association among the independent explanatory factor variables. However, the linearity and constant variance assumption were not met when assessed at the 5% significance level. Therefore, one of the measures that were integrated to address the lack of homogeneity in the variance of the residuals was to incorporate two control variables (age and income) in order to moderate the hypothesized relationship between the level of political trust and the mass partisan polarization in the U.S. The analysis did not consider incorporating data or model transformation into logarithmic, quadratic or other non-linear form given that the extent of non-
multivariate regression model (Franzco & Farmer, 2014). This is because, in most cases, when the linearity assumption is violated, the model has to be transformed using logarithmic (changing values of dependent and independent variables into log form) and quadratic form to enhance its validity as suggested by Franzco and Farmer (2014). Therefore, as suggested by Moore (2017), the fact that two (2) of the four (4) OLS assumptions are met imply that the multivariate regression results would be fairly valid.
4.4 Multivariate Regression Analysis
This subsection of the analysis presents the results of the multivariate regression analysis to estimate the effect of mass partisan polarization on political trust in the U.S. congress. The respondents’ age and income level were integrated as the control variables to moderate the hypothesized relationship between partisan polarization and trust in political institutions. The results of the political trust multivariate regression analysis are presented in Table 3 with evaluation of the statistical coefficients being assessed at the 5% significance level. The following multivariate OLS regression model is formulated based on the estimated coefficients;
Political Trust = 3.89 – 0.078(Partisan Polarization) – 0.001 (Age) + 0.002 (Income)
Table 3.
Regression Analysis: Political Trust
Variable Unstandardized Beta Coefficient Standardized Beta Coefficient Sig. (ρ-value)
Intercept 3.89 0.000
Mass Partisan Polarization -0.078 -0.264 0.000
Age -0.001 -0.012 0.003
Income 0.002 0.045 0.000
Adjusted R2 0.071
F-value (3, 59,999) 1523.655 0.000
Sample Size (N) 60,000
Source: SPSS Output Based on Analysis of the CCES Survey Data (Appendix)
Partisan Polarization: The coefficient of mass partisan polarization (β1 = -0.078; ρ <0.05) is statistically significant when assessed at 5% level of significance. This means that mass partisan polarization among the voters has a significant negative effect on the level of public trust in the congress. Specifically, evaluation of the stated independent variable coefficient depicts that when mass partisan polarization rises by 1 percentage point, the level of public trust in the congress declines by 7.8%. This means that the extent of mass partisan polarization, which is used as the basis for generating partisan motivated decisions has a substantial negative effect on the individual voters’ level of public trust in the U.S. congress. According to Druckman et al. (2013), the negative influence of partisan and elite polarization is attributed to the fact that in a highly polarized political environment, the opinion of voters are public at large is influenced to a considerable extent by the party endorsements at the expense of substantive information related to factual issues such as abortions and same sex marriages. The implication is that given the divergent opinion across the interparty, the voters seem to have less trust in the political institutions (Hetherington, 2009).
Respondents’ Age: The coefficient representing the survey respondents’ age polarization (β2 = - 0.001; ρ <0.05) also depicts that age has a low negative effect on the level of trust in the political institutions. However, in contrast to the mass partisan polarization, age appears to have a relatively low impact on political trust. The coefficient of age depicts that when the respondents’
age increases by 1 year, the individual voters’ level of trust in congress declines by 0.1 percentage point. The findings are consistent with the insight based on Theiss-Morse and Wagner (2015) who noted that age has a substantial negative effect on the public trust in the political institutions. This means that younger respondents tend to exhibit a considerably higher level of trust in congress when compared to the older participants due to the variation associated with intergenerational experience.
Respondents’ Family Income: The estimated coefficient of the respondents’ family income (β3
= 0.002; ρ <0.05) has a statistically significant positive effect on the level of political trust in the U.S. congress. The results are explained by the fact that the annual family income has a significantly weak positive impact on the extent of trust in the U.S. congress. This means that as
fact that a greater majority of the voters would have a considerable level of confidence and ability in the legislature to enact favorable tax and wealth distribution legislations (Poole, 2009).
According to Lauka et el. (2018), the level of political trust in the U.S. political institutions is relatively low among the low-income earners compared to the high-income earners.
Predictive Ability of the Regression Model: The estimated political trust multivariate regression model is significant. This is because, the F-statistic (F = 1523.66; ρ <0.05) meets the statistical significance at the 5% level. However, the estimated political trust multivariate model has a low predictive (explanatory) ability as depicted by its relatively low adjusted R2 (0.071).
The estimated coefficient of determination implies that the model explains only 7.1% of the variation in the level of political trust in the country. This means that when mass partisan polarization, age and family income are incorporated as independent and control factor variables, they explain only 7.1% of the fluctuations in the level of trust that the public holds over the political institutions in the country.
4.4.1 Evaluation of Type I Error in the Multivariate Regression Analysis
The analysis acknowledges that there is a risk of type I error in the estimated regression model due to the fact that the results are based on a large sample of 60,000 respondents. Type I error occurs when the analysis falsely rejects a true null hypothesis formulated in the model.
Therefore, in order to test for the presence of type I error, the formulated multivariate regression model is rerun using a small sample of 970 respondents (n = 970). Table 4 presents a summary of the rerun regression results based on a small sample of 970 respondents.
Table 4.
Regression Analysis: Political Trust
Variable Unstandardized Beta Coefficient Standardized Beta Coefficients Sig. (ρ-value)
Intercept 0.004 0.831
Mass Partisan Polarization -0.008 -0.154 0.000
Age -0.001 -0.053 0.097
Income 0.000 -0.037 0.242
Adjusted R2 0.022
F-value (3, 59,999) 8.376 0.000
Sample Size (N) 970
Source: SPSS Output Based on Analysis of the CCES Survey Data (Appendix
The results based on a small sample of n = 970 respondents depicts that the possibility of type I error is very remote. The justification is that the model’s F-statistics (F = 8.376; ρ < 0.05) is significant at the 5% level. This means that despite the small sample of 970 respondents, the estimated multivariate regression model is still robust. However, based on the small sample, the coefficient of age (β2 = -0.001; ρ = 0.097 > 0.05) is not significant when evaluated at the 5%
significance level.
4.4.2 Summary of the Multivariate Regression Analysis
The outcome based on the multivariate regression analysis is consistent with the insight from the prior studies, which have examined the relationship between elite and mass partisan polarization and the trust in political institutions with respect to the U.S. political system (Catterberg &
Moreno, 2005; Marien & Hooghe, 2011; Linde, 2018). Specifically, the results show that mass partisan polarization has an adverse effect in terms of the trust and the level of approval rating of the public towards the various political institutions in the country. The analysis also found that
