The Effect of Rising Income Inequality on Partisan Voting in
Majoritarian Electoral Systems: A Comparative Case Study of the
United States and Ghana
Maaike van Woerden Leiden University
Political Science: International Relations and Organizations Bachelor Project 6: Inequality in Political Perspective
Dr. B.K.S. van Coppenolle Word Count: 8399
1. Introduction
Income dispersion has been significantly growing within several developing and developed countries since the 1970s, such as Ghana and the United States (Boix, 2010, p. 511). Implementing policies such as higher taxation for upper-income citizens and redistribution to lower-income citizens within these countries can help reduce this income gap. Many articles on the topic of income inequality, such as Esping-Andersen and Myles (2011) have concluded that these redistributive policies are more likely to emerge in governments led by economically left-wing parties than ones led by right-left-wing parties (p. 643). Voting for a specific party is one of the primary channels through which a large share of citizens can influence government and policy and can therefore be a potential indicator of the public’s demand for redistribution.
However, the frequently cited theories on voter demand for redistribution (such as Meltzer & Richard, 1981, Iversen & Soskice, 2006, and Moene & Wallerstein, 2001) have contradicting explanations and predictions for how, and if, rising inequality changes the electorate’s preference for redistribution, and how this relates to the percentage of votes received by right-wing and left-wing parties. A commonality of these theories is that their causal mechanisms can theoretically all take place in a majoritarian, two-party electoral system. Thus, by examining the different expected effects of income inequality over time on partisan voting in only majoritarian two-party systems, the assumptions and predictive values of these theories can best be compared. The main question of this paper is therefore: Do changes in income inequality over time affect the share of votes received by left-wing and right-wing parties in countries with two-party majoritarian systems? Why or why not?
To answer this question, I will first discuss the relevant existing theories and extract the varying hypotheses from this literature. Then, using the cases of the United States and Ghana from 1992 to 2012, I will examine whether the assumptions, mechanisms, and expected results of these models can be empirically observed. Partly through determining the empirical value of these theories, this paper will analyze why a relationship between income inequality and the percentage of votes received by left- or right-wing parties in elections over time can or cannot be established. This will help provide new insights on whether partisan voting is a means through which income inequality impacts politics.
This paper will show that, despite certain theoretical predictions, a consistent relationship between rising income inequality and partisan voting over time cannot be identified in the two countries. This is partly due to the theories’ assumptions differing from empirical conditions, as well as factors unrelated to income inequality having a significant effect on voter
behavior in Ghana and the United States. These findings create a starting point for future studies and contribute to the academic efforts to understand the effect of rising income inequality on different aspects of politics.
2. Theoretical Framework
In this section, I will examine and compare several theories that provide potential answers to my research question. To help interpret the relevant literature relating to this question, reoccurring concepts, such as majoritarian electoral systems, right- and left- wing parties, and income inequality, will first be clearly defined.
To understand how the division between left and right in two-party democracies occurs, the relevant characteristics of a majoritarian electoral system must be identified. In majoritarian elections, voters are distributed in single-member voting districts. The electoral formula in each district requires candidates to gain a plurality of the votes to be elected. Put simply, each district can vote for one seat in the national legislature, which is gained by the party with the highest vote share in the district (Norris, 1997, p. 300). It has widely been argued that electoral systems shape the “nature of political parties and the composition of governing coalitions” (Iversen and Soskice, 2006, p. 165). Persson, Roland, and Tabellini (2007) further elaborate on the distinct effects of majoritarian elections and proportional elections on party formation. They find that plurality rule, as part of majoritarian elections, creates a greater incentive for politicians to merge into larger parties than under proportional rule (p. 30). This explains their empirical observation that majoritarian rule is correlated with a higher incidence of single-party governments than coalition governments (Persson et al., 2007, p. 28). The single party in government acts as a uniform actor, making it difficult for voters to distinguish between different factions of a single-party government (Persson et al., 2007, p. 6). According to Duverger’s Law, electoral systems with plurality-rule elections within single-winner districts tend to only sustain two parties in equilibrium, since voters are disinclined to vote for parties with little chance of winning (Duverger, 1963, p. 239). Persson et al. argue that the strategic and opportunistic choices of politicians in majoritarian systems ultimately lead to an equilibrium of two major parties, rather than strategic voting (2007, p. 7). Either way, the expected outcome remains similar: a two-party democracy. Iversen and Soskice (2006) create a model where these parties tend to represent classes, which is paired with the assumption that, within two-party systems, one of the parties will be left and the other will be center-right.
This introduces the second concept that must be further clarified- what is meant in the remainder of this thesis by left/center-left parties and right/center-right parties. Keeping in mind the general preferences of both types of parties on the supply of redistribution helps clarify theories relating to the main question- whether changes in inequality can potentially shift the electorate’s preference in favor of a right- or left -wing party in a majoritarian electoral system. Not all theories relating to voter demand for redistribution directly refer to parties or candidates as being left- or right-wing. Although there are several ideological dimensions in which parties can be categorized as right or left, I will focus primarily on a more one-dimensional economic categorization, due to its relevance for income inequality. This corresponds to the “partisan theory” of political competition. An assumption of this theory is that left-wing parties tend to represent voters with a relatively lower income and right-wing parties represent voters with a relatively higher income in a country (Boix, 1998, p. 40). While these parties do not have identical ideologies across countries, substantial evidence has been found to support that, generally, if left-wing parties gain sufficient legislative and political power, they will tax and redistribute in favor of the less well-off more than right-wing parties will (Esping-Andersen & Myles, 2011, p. 643).
Lastly, the concept of income inequality must be clarified. Most articles provide only an operationalization of income inequality, which is then also used to conceptualize economic inequality. However, I will attempt to provide a more universal definition of income inequality that can be applied in most of this literature, regardless of their specific operationalization. Income inequality is, in the broadest sense, the disparity of income between individuals, households, specific groups, social classes, or countries. The wider the income dispersion is, the greater the economic inequality is (Charles-Coll, 2011, p. 17). For my research question, income inequality within a country is most relevant. Since voter preference is relevant for party voting, the type of income inequality can be narrowed even further to be most closely related to voter preference. Kenworthy and Pontusson (2005) argue that looking at inequality as the distribution of household income is most effective for understanding voter behavior, because voters tend to think about their household as a reference point for their relative position in income distribution rather than their individual income (p. 456).
With these concepts defined, the different theories and hypotheses on the effects of income inequality on voting within countries can be examined. It should be noted that the political economy theories relevant to my study will be primarily focused on how income inequality interacts with the “demand” side of redistribution, relating to the preferences of voters for redistribution, as opposed to the “supply” side, or to the provision of policy (McCarty
& Pontusson, 2011, p. 688). Not all these “demand” models directly refer to right-/left-wing voting. However, the ideas and assumptions presented in these theories all provide insights on voting behavior determinants related to income inequality, which I will then link to my question by incorporating partisan voting. Thus, the theories and literature discussed below both directly and indirectly provide hypotheses on the effect of rising income inequality on partisan voting. Noticeably, the expected effects range from an electoral advantage for the left (Meltzer and Richard, 1981) or the right (Iversen and Soskice, 2006 and the campaign finance mechanism) to no effect at all (Moene and Wallerstein, 2001).
In their 1981 article, “A Rational Theory on the Size of Government”, Meltzer and Richard hypothesize that the size of government depends on the relation of a country’s mean income to the income of the decisive voter. In majoritarian electoral systems, the voter with the median income among enfranchised citizens is identified as the decisive voter, meaning that only they can guarantee a majority (Meltzer & Richard, 1981, p. 916). Their explanation of the size of government emphasizes the critical role of the decisive voter’s demand for income redistribution, as expected by median voter theorem. If all enfranchised citizens vote and the only activities of a government are redistribution and taxation, then the size of the government is determined by the “welfare-maximizing choice” of the decisive, median-income voter. Voters with an income above the decisive voter desire lower taxes and less redistribution, while voters below the decisive voter favor higher taxes and more redistribution. In this model, the decisive voter determines the actual tax share. Therefore, if the mean income of society is higher than median voters’ income, the median voters’ preference for redistributive policy increases (Meltzer & Richard, 1981, p. 924).
Meltzer and Richard only briefly refer to the effect of inequality on the median voters’ demand for redistribution, arguing that “economic growth can lead to rising inequality and, if our hypothesis is correct, to votes for redistribution” (1981, p. 925). This is because rising inequality is expected to widen the gap between median voter income and mean income. If Meltzer and Richard are correct, a rise in income inequality would be predicted to increase demand for redistribution by the median voter given that income distribution remains right-skewed and the mean income stays constant (McCarty & Pontusson, 2011, p. 671).
Meltzer and Richard do not directly refer to parties or left-right ideological dimensions. To make their model more applicable to my research question, I will combine it with partisan voting, rather than voting for specific redistributive policies. To do so, I will use the ideas of partisanship presented by Pontusson and Rueda (2010), who directly connect the demand for redistribution to partisan voting. They claim that rising inequality shapes policy preferences of
lower-income groups for redistribution and left parties, while high-income voters prefer less redistribution and right-wing parties (Pontusson & Rueda, 2010, p. 680). An important assumption of their argument is that political parties are not solely motivated by their desire to win elections. Unlike Meltzer and Richard, who argue that rising inequality would shift the ideologies of existing parties to cater to the policies preferred by the decisive voter, Pontusson and Rueda argue that parties are often still ideologically tied to their core constituencies. Essentially, this means that even though a right-party tries to appeal to the median voter, it will still be responsive to the demands of its high-income voter base, and vice versa. The ties to core constituencies thus cause left and right parties to refrain from completely moving to the ideological center. This allows the two parties to retain their distinctive economic platforms to the extent that their prospects of reelection are not severely threatened (Pontusson & Rueda, 2010, p. 679).
As Kenworthy and Pontusson (2005) explain, “median-voter logic and partisanship need not be construed as mutually exclusive”, because ideological commitments can be combined with wanting to also win elections (p. 457). Therefore, partisanship, as described by Pontusson and Rueda, can in fact be combined with Meltzer and Richard’s median-voter theory. This combination leads to the prediction that a rise in income inequality would likely increase the share of votes for the party arguing in favor of redistribution. In a majoritarian two-party system, this puts the center-left party at an advantage, since their economic ideology caters more to growing redistributive preferences of the decisive voter than the center-right party, which remains to some extent loyal to the preferences of their higher-income core constituents.
In short, the main hypothesis derived from combining Meltzer and Richard with partisanship is that rising income inequality likely increases the share of votes received by the center-left party. However, this hypothesis is based on several assumptions that may not be empirically observable, such as 100 percent voter turnout (Meltzer & Richard, 1981, p. 924). The empirical absence of these conditions could lower the predictive value of the hypothesis, or at least weaken the expected effect. According to Barnes (2012), lower turnout results in a higher median voter income than general median income, because higher-income citizens are more likely to vote (p. 85). Due to this smaller difference between mean income and median income than under complete voter turnout, the decisive voter’s demand for redistribution, and thus the advantage for the center-left party, is lower. Lower turnout therefore weakens Meltzer and Richard’s expected effects of rising income inequality on (partisan) voting. Furthermore, Meltzer and Richard (1981) assume in their model that the only functions of the government are taxation and redistribution, meaning that voters base their vote on only their economic
preferences and income level (p. 925). If non-economic dimensions or cleavages strongly influence voting behavior and partisan identification within a country, then the impact of changes in income inequality on partisan voting is expected to be, in the least, weaker than the original hypothesis predicts. Lastly, partisanship may not be universally observed in the polarized manner described by Pontusson and Rueda. If the two parties within a majoritarian system are closer in ideology and have weaker ties to core constituents than expected by Pontusson and Rueda, then they could both be willing to argue in favor of redistribution to cater to the demands of the median voter (McCarty & Pontusson, 2011, p. 679). Therefore, if there is little to no polarization in a country, rising income inequality does not necessarily lead to a significant electoral advantage for the center-left as was originally predicted by combining partisanship with Meltzer and Richard’s theory.
An alternative hypothesis, predicting an advantage for the center-left party as a result of rising income inequality, is provided by Iversen and Soskice in their 2006 article “Electoral Institutions and the Politics of Coalitions: Why Some Democracies Redistribute More Than Others”. The main objective of this article is to explain why democracies with proportional representation electoral systems redistribute more than countries with majoritarian electoral systems. I will focus specifically on their model and assumptions relating to partisan voting in majoritarian systems here. The authors assume that voters are evenly split between three groups, low-, middle-, or high-income (Iversen & Soskice, 2006, p. 166). The claim that economic interests of individuals are divided nationally along class is further established in the authors’ later work. They argue that in single-member district majoritarian systems, income is the “only salient economic dimension” (Cusack, Iversen, & Soskice, 2007, p. 377).
Assuming that Duverger’s Law holds, there are two major parties in this system, which the authors categorize as a center-left and center-right party. The policy differences between these two parties are assumed to be constant across countries and over time, so the variance analyzed by the authors is how often each type of party wins elections (McCarty & Pontusson, 2011, p. 680). The low-income individuals will always vote for the center-left party, while the high-income individuals will always vote for the center-right party (Iversen & Soskice, 2006, p. 168). Therefore, the middle-class vote ends up being decisive.
As Iversen and Soskice empirically prove, center-right parties win elections more often than center-left parties in majoritarian systems. They explain this bias by arguing that if the middle-income votes for center-left parties, they cannot be sure that the low-income group will set up policies that could potentially financially exploit the middle-class. The middle-class faces the risk that redistributive taxes will also be imposed on them, instead of only the rich.
Therefore, middle-income voters have a greater incentive to vote for center-right parties, in order to avoid being “soaked” by the lower-income citizens (Iversen & Soskice, 2006, p. 178).
Whether Iversen and Soskice’s model can be expanded to explain changes over time, is a question they pose for further research, since it has not been empirically tested. However, they do hypothesize that rising income inequality would likely further increase the fears of the middle-class of being financially exploited by the lower-class. Given that this model can indeed be used to explain how partisan advantage changes over time, the share of votes for the center-right party are hypothesized to increase as income inequality rises. This is primarily facilitated by creating a greater financial incentive for the middle-class to vote for the center-right party (Iversen & Soskice, 2006, p. 179).
When the central assumptions made by Iversen and Soskice change, so does the likelihood that their hypothesis can be observed. First, similar to Meltzer-Richard, their model assumes that voting occurs along income lines. If income is not the only salient voting dimension, then the impact of changes in income inequality on partisan voting is weaker than expected in this model. Secondly, the assumption that policy differences between center-left and center-right are constant across countries may not be empirically observable. Specific parties are more likely to diverge from the preferences of the median-voter than others, due in part to varying degrees of party polarization across countries (McCarty & Pontusson, 2011, p. 679). Therefore, the hypothesis of Iversen and Soskice may be found to be more applicable to countries with greater polarization and clearer policy differences between the left- and right-wing parties.
Another explanation for why rising economic inequality could theoretically lead to a larger share of votes for the center-right party is related to the fact that rising income inequality tends to increase the relative economic advantage of high-income citizens. Income inequality therefore allows the upper-class to have significantly more financial resources compared to the rest of the population. Several authors (such as Ferguson, 1995 and Bonica et al., 2013) have explored the claim that this potentially allows the wealthy to skew the political discourse to their advantage, partially through exerting influence on voter behavior. This would occur primarily through economic means, specifically campaign contributions (Gelman, Kenworthy & Su, 2010, p. 1213). If there is a two-party majoritarian system where the center-right party acts in line with the economic interests of high-income voters (as is assumed by Boix, 1998 and Iversen & Soskice, 2006), then campaign contributions by high-income citizens would generally be expected to be in favor of the right. By this logic, it is predicted that as income inequality increases, the electoral system would be expected to be continually biased in favor
of the right-wing parties that help these high-income groups retain their wealth (Gelman, Kenworthy & Su, 2010, p. 1214). However, according to Campante (2011), there are some limitations to the extent that campaign donations translate into political influence. First off, laws on campaign finance place restrictions on the amount of money that can be donated to one candidate or party. Secondly, empirical research has found that campaign contributions do not directly influence election outcomes, because they have little to no effect on the voting behavior of individuals (Campante, 2011, p. 646). Therefore, campaign donations by high-income individuals, generally expected to increase with rising income inequality, are unlikely to be a main causal mechanism of how income inequality influences partisan voting.
Moene and Wallerstein (2001) reach more ambiguous conclusions on the demand for redistribution created by inequality in advanced democracies than Meltzer-Richard and Iversen-Soskice. Their model is similar to these theories in the sense that it also observes political processes that tend to only occur in majoritarian systems and operates in the same median-voter framework as Meltzer and Richard. The main findings of their article are that rising inequality creates a two-fold, counteracting effect. The first effect is similar to the Meltzer-Richard model, since Moene and Wallerstein also hypothesize that rising inequality increases voter demand for redistribution (2001, p. 859). By looking at this effect alone, it would be expected that, given the arguments of Pontusson and Rueda (2010) on partisan ideologies, the center-left/ left-wing party would gain votes. However, the second effect predicted by Moene and Wallerstein is that greater inequality can also raise the median voter’s demand for non-redistributive insurance (2001, p. 860). This risk-averse median voting behavior is closer to the main voting motivation of the middle-class described by Iversen and Soskice (2006), where the center-right/ right-wing party is at an advantage. The two effects described by Moene and Wallerstein essentially cancel each other out, which means that rising income inequality is not predicted to cause the median voter to strongly favor one type of party over another. Therefore, the main hypothesis derived from Moene and Wallerstein’s model is that little to no significant effect of rising income inequality on partisan voting will likely be observed (Kenworthy & Pontusson, 2005, p. 457). The theories and literature discussed in this section provide, directly or indirectly, predictions of the effect of rising income inequality on partisan voting. The variance in these hypotheses can best be explained by the fact that each theory emphasizes a different causal mechanism of income inequality that could influence partisan voting. There has been little to no empirical research on whether the theories’ projected voting effects and causal mechanisms can be observed in specific countries in specific time periods. The purpose of this thesis is to contribute further to this debate by analyzing why the different expected effects of income
inequality on votes received by left- and right-wing parties can or cannot be observed in two-party states, particularly in the United States and Ghana.
3. Research Methodology
In this paper, a comparable cases design will be used to compare the effects of changes in inequality on the share of votes received by left- and right-wing parties in majoritarian democracies with two major parties. I will be analyzing only two countries, the United States and Ghana, over a period of time from 1992 to 2016, which allows me to look more in depth at qualitative and quantitative factors within each country that can explain the result. This also enables me to analyze whether elements of the theories described in the previous section are present in each country, and whether the theories have any predictive value in these countries.
A “Most Different Systems Design” (MDSD) will be used to conduct this longitudinal comparative political research. Here, units of research which are most different with regard to extraneous variables but have similarities in a certain result and explanatory variable will be compared (Anckar, 2008, p. 390). In short, cases with different characteristics are chosen, so the effect of a key similarity can be tracked. The logic behind this is that similar results in cases cannot be explained by other variables, since all variables, apart from the independent and dependent variable, differ across the cases. However, if the cases have different results, then it is likely that an alternative variable has a stronger effect on the dependent variable than the observed independent variable does (Faure, 1994, p. 316). In this research, the units of research will be two countries over a specific period of time. Ghana and the United States were specifically selected for this comparative research, due to their differences in many variables, but key similarities in electoral and party-system and general trend in income inequality (table 1).
Table 1. Key similarities and differences between Ghana and the United States. United States Ghana
Difference: Length (in
years) of consolidated democracy
Long lasting democracy in place (since 1787)
Relatively new consolidated democracy (since 1992)
Difference: Geography North America, 9.834 million km2
West Africa, 285,535 km2
Difference: Population 325.7 million 28.21 million
Difference:
Income-classification of country
High income/ OECD-member country
Lower middle income/ High average developing country
Difference: GDP growth
1992-2016
184.8% growth 565.6 % growth
Difference: Years since
independence from colonial rule
Independent since 1776 Independent since 1957
Similarity/ Conditioning variable: Electoral system
for legislature
Majoritarian electoral system with first-past-the-post voting and single-member districts, resulting in a dominant two-party system with a center-left and center-right two-party.
Similarity/ Independent Variable: Changes in
income inequality (Gini) from 1992- year of most recent data
Similar rises in Gini coefficients, U.S. Gini score rose about .33 points from 1992-2016 and Ghana about .4 points from 1992-2012. In both countries, the median income has been consistently estimated to be less than the mean income from 1994-2013, as is expected with rising income inequality.
Similarity / Dependent variable: The share of
votes received by left-/ right-wing parties in legislative elections from 1992-2016
My research will test whether left- or right-wing parties in Ghana and the U.S. appear to gain or lose votes in a way consistent with the rise of income inequality. If one of the hypotheses of the theoretical framework holds true for these specific majoritarian systems, then the results in the two countries would be expected to be similar, either seeing a rise in right-wing voting (Iversen-Soskice), left-wing voting (Meltzer-Richard), or no effect at all (Moene-Wallerstein).
Source Data: World Bank, 2016.
The conditioning variable in this research, which provides the context in which the relationship between the independent and dependent variable will be studied, is the electoral system in place. In this research, this is a majoritarian voting system, because the different models in the theoretical framework operating within this system all base their models on different aspects of majoritarian voting. As part of the MDSD, this variable must be most similar between the two countries. Therefore, the party system resulting from this system must also be most similar. As expected by Duverger’s Law, The United States and Ghana both have the close to two effective legislative parties- 2.1 and 1.9 respectively (Kollman et al., 2017).
In my research, I will examine how and if a rise in income inequality over time, the independent variable, affects the share of votes received by each party in majoritarian two-party systems. As mentioned in the conceptualization of income inequality, ideally the distribution of household income should be measured. This measurement most strongly influences voter behavior, and thus election results (Kenworthy and Pontusson, 2005, p. 456). The Gini coefficient, which is widely used in literature and based on household data, gives an indication of relative wealth and will be the most useful tool for measuring changes in inequality over time. The closer to 1 the coefficient is, the more unequal the income distribution in a country is (Dodge, 2010). For the Gini measurement in the United States, I will use data from the United States Census Bureau (2016), based off household surveys, as it provides detailed annual data. The Gini index data for Ghana, however, is less widely available for the time period I will examine. I will be using World Bank’s (2016) Gini estimates for Ghana which are derived from household data from the Ghana Living Standards Survey (GLSS). Unfortunately, this survey has only been held 4 times since 1991. While this data will show general patterns in income inequality in Ghana, it is less specific than the United States data.
To measure the dependent variable, I will graph the changes in percentage of votes received by each party in every legislative election from 1992 to 2016. The specific starting year is 1992 since Ghana had their first free and fair majoritarian election in 1992, meaning that before then, the United States and Ghana did not share a common conditioning variable (Adams & Agomor, 2015, p. 366). The ending year, 2016, is the last year of available election data for both countries. I will only analyze the election results for the Parliament of Ghana and the House of Representatives in the United States, due to the likeness of their electoral system (Table 1).
By graphing the changes in income inequality along with the changes in votes received by each party, it will become clear whether a correlation between the two variables is present
in both countries. To determine whether the theories of Meltzer-Richard and Iversen-Soskice can explain the observed correlations, or lack thereof, I will first explore to what extent their assumptions of 100 percent voter turnout, a clear difference between the demand for redistribution of the left- and right-wing parties, and voting based on economic considerations and income are present in Ghana and the United States from 1992 to 2016. I will also determine whether the causal mechanisms described by these theories, as well as the campaign finance theory, can be observed in these cases. Lastly, I will identify alternative explanations for the observed correlations between income inequality and election results in Ghana and/or the United States.
4. Comparative Case Study: Income Inequality and Partisan
Voting in the United States and Ghana
4.1 The legislative electoral systems and party systems in the United States and
Ghana
The United States has a majoritarian voting system that can best be described as first-past-the-post (FPTP), meaning that a plurality of the vote is required for an electoral victory. The 435 members of the House of Representatives are elected in single-seat voting districts and serve two-year terms. The United States is generally classified as a two-party system, with the Democratic and Republican parties holding dominant positions within the federal government since 1854 (Hague & Harrop, 2013, pp. 206-207). The Republican Party is the more economically-right party of the two. During the 20th and 21st centuries, this party came to be associated with laissez-faire capitalism, a push for low taxes, and conservative social policies (Huber and Stanig, 2007, p. 2). These underlying ideological principles continue to be a common denominator in most Republican candidates’ platforms today (Hague & Harrop, 2013, p. 207). The Democratic Party, on the other hand, is classified as a leftist, social democratic party. Democrats generally advocate for greater social and economic equality (Hague & Harrop, 2013, p. 206). Republicans are traditionally viewed as the party of business, while Democrats represent labor interests (Gelman, Kenworthy, & Su, 2010, p. 1203).
Ghana has consistently been holding peaceful democratic elections since the enactment of its 1992 constitution. The 275 members of the Parliament of Ghana are elected in a similar fashion to U.S. representatives: in single-seat constituencies with FPTP voting. Since the election in 1994, the NPP and NDC have been dominant parties in Ghana’s two-party system
(Fobih, 2010, p. 27). The NPP often compares itself to the Republican party of the United States. Its platform is economically conservative and is considered to be the more right-wing of the two parties (Fobih, 2010, p. 28). Its ideology is based on pro-business and free-market policies, with a focus on growing the private sector. The NDC shares ideological positions with the Democratic party of the United States. It categorizes itself as a social-democratic party with more center-left views than the NPP (Abdulai & Crawford, 2009, p. 44). According to its 2004 manifesto, the NDC seeks to ‘marry the efficiency of the market and private initiative with the compassion of state intervention to protect the disadvantaged and the marginalized and to ensure optimum production and distributive justice’ (p. xiv).
4.2 Income inequality trends in the United States and Ghana
Despite its status as a high-income country, income inequality has been rising in the United States since the 1970s (Gelman et al., 2010, p. 1204). This rise is characterized by a decline in the total share of income received by the bottom three household income quintiles and a rise in the share received by the top two quintiles (Karoly, 1992, p. 97).
Ghana’s economy, in terms of GNI, has been steadily growing and progressing over the last decades, which contributed to its status as a high average developing country (World Bank, 2016). However, this growth is paired with widening gap between the highest and lowest income quintiles, similar to the United States. This rise in household income inequality has been attributed to the inability of the economic growth to reach the lower-income groups (Novignon, 2017, p. 515).
4.3 The trends in income inequality and votes received by the right- and left-wing
parties in the United States and Ghana from 1992 to 2016
Figure 1 compares the general trend in income inequality from 1992 to 2016 to changes in the share of votes received by Democrats and Republicans in elections for the House of Representatives during that same period. The value of the correlation coefficient (r) between income inequality and votes received by the Republican candidates is 0.24, indicating a very weak positive linear relationship between the two variables (Field, 2013, p. 295). An even weaker negative correlation coefficient, -0.18, exists between income inequality and votes for Democratic candidates.
Figure 1. Changes in U.S. national income inequality and the percentage of votes received by the Republican
and Democratic Party in elections for the House of Representatives from 1992 to 2016.
Source Gini data: U.S. Census Bureau, 2016.
Source election data: Office of the U.S. House of Representatives Clerk, 2017.
Figure 2 compares the general trend in income inequality from 1992 to 2016 to changes in the share of votes received by NPP and NDC in elections for the Parliament of Ghana during that same period. It should be noted that the NPP did not participate in the parliamentary elections of 1992, which likely contributed to the large share of votes received by the NDC that year. Generally, as income inequality rises, so does the percentage of votes for the NPP, and its slight drop in inequality from 2005 to 2012 is met with a slight drop in the share of votes received by the right-wing party (figure 2). This indicates a strong positive correlation between the two variables, calculated at 0.93 using the limited available data. A strong negative correlation between the share of votes for the NDC and income inequality., r= -0.91, further demonstrates that income inequality varies in a similar manner to partisan voting in Ghana.
42 44 46 48 50 52 54
Figure 2. Changes in Ghana’s national income inequality and the percentage of votes received by the NPP and
NDC in elections for the Parliament of Ghana from 1992 to 2016.
Source Gini data: World Bank, 2016.
Source election data: Electoral Commission of Ghana, 2016.
It cannot be concluded based on solely a correlation that rising income inequality is the causal mechanism behind the gains and losses of votes for each party in Ghana. It is, however, more likely that a causal relationship exists between the two variables in Ghana than in the United States, where no strong correlation exists. To determine if this is so, I will analyze whether assumptions and mechanisms of the theories predicting an advantage for the right-wing are more applicable to Ghana than to the United States. I will also determine why the results predicted by the partisan version of Meltzer-Richard are not observable in either country and lastly, explore alternative explanations for the correlation in Ghana.
4.4 Testing assumptions
4.4.1 Income-based voting
An important assumption shared by Iversen and Soskice’s and Meltzer and Richard’s models is that citizens base their voting choices solely on economic considerations, specifically
30 35 40 45 50 55 60 65 70 75 80
their own relative income. The causal mechanisms that lead to their respective hypotheses are largely based on this notion.
McCarty et al. (2006) find evidence that, in the United States, partisan identification is increasingly linked to relative economic status. High-income voters increasingly vote Republican, while low-income voters prefer Democratic candidates, which matches the broad expectations for partisan voting of both models (p. 71). Gelman et al. (2010) also find that the rich-poor gap in partisan voting has increased along with income inequality since the 1970s (p. 1205). Despite this increase, structural factors such as religion and education still influence partisan voting in the United States (Gelman et al., 2010, p. 1214). Therefore, it cannot be concluded that relative income is the United States’ only salient partisan voting determinant.
Income-based voting is, however, still more present in the United States than in Ghana. According to Lindberg and Morrison (2008), partisan voting in Ghana since 1992 has been based on evaluation of party policies, campaign messages, and achievements in government (p. 121). Economic platforms are therefore taken into consideration by voters, but the most critical factor in determining vote choice remains each party’s proposed policies on education and health (Youde, 2005, p. 13). An individual’s income has not been proven to be one of the structural factors that affects voting rationale in Ghana. Adams and Agomor (2015) studied voting behavior from 1992-2012 and found that a voter’s age, sex, and ethnicity significantly affected which party they voted for, while level of household income (low, middle, or high) did not (p. 376).
Since voting behavior in both countries is influenced by dimensions other than income, the impact of changes in income inequality on voting is already less than originally expected in these models. Especially in Ghana, where income has not been proven to influence voting, income inequality is unlikely to influence voter behavior through median- or middle-income voter mechanisms described by Meltzer-Richard and Iversen-Soskice.
4.4.2 A distinction between redistributive demand of the left- and right- wing party
The assumption that there exists a clear ideological distinction between the preferred redistribution rate of the left- and right-wing parties is shared by Pontusson-Rueda (and therefore my partisan interpretation of Meltzer-Richard) and Iversen-Soskice.
In the United States, the two dominant parties have different approaches to the problem of rising income inequality. Democrats generally favor creating more income equality through issuing progressive taxes and transfers, while Republicans argue against strong government involvement in redistribution (Krugman, 2007, p. 11). McCarty et al. (2006) suggest that
ideologies of the Democratic and Republican parties have grown even more polarized with the rise of income inequality (p. 5). Therefore, it can be confidently concluded that there have been clear differences between the economic ideologies of the left- and right-wing parties from 1992 to 2016 in the United States.
The extent of fiscal redistribution in Ghana is smaller than in Western countries, but, according to Younger et al. (2017), “the government of Ghana regularly commits itself to reducing poverty and inequality and increasingly adopts policies explicitly intended to alter the distribution of income” (p. 47). In their platforms, the NDC generally emphasize the need to reduce income inequality more than the NPP. However, it has been commented that during its years in government, the NDC pushed pro-market policies as much as the NPP, which demonstrates that these parties are less economically polarized than the American parties (Abdulai & Crawford, 2009, p. 46). Retrospective voting, which influences voter choice for about 8.5 percent of Ghanaians, further weakens the role of ideology of each party, since these voters evaluate performance rather than campaign message (Adams & Agomor, 2015, p. 371). Due to less polarization than in the United States, the NPP and NDC can be more responsive to general needs and therefore both cater more to the needs of the swing-voters, as expected in Meltzer and Richard’s original median-voter framework (Smidt, 2015, p. 365). This further weakens the applicability and explanatory value of Iversen-Soskice and the partisan Meltzer-Richard for Ghana, as these assume that the differences between left- and right-wing parties remain more constant due to core constituencies.
4.4.3 Voter turnout
The assumption of 100 percent voter turnout is primarily important for Meltzer and Richard’s model. However, in both the United States and Ghana, turnout has been significantly lower during the observed years of this research. In the United States, turnout for House of Representatives elections from 1992 to 2016 has ranged from 36,4 to 55,3 percent of the voting age population (Brookings Institute, 2014). In Ghana, this has ranged from 28 to 85 percent of the voting age population (IDEA, 2016). As was discussed in the theoretical framework, this would likely reduce the gap between the median voter income and the mean income, and thus weaken the likelihood that Meltzer and Richard’s main causal mechanism is in effect during this period. Due to a lack of longitudinal data on median-voting behavior, it cannot be determined to what extent Ghana and the United States deviate from Meltzer and Richard’s causal mechanism. However, the deviance from the assumption of complete voter turnout could, in part, explain why Meltzer and Richard’s hypothesized effects of income inequality on
partisan voting, an increase of votes for the left, cannot be consistently observed in either country during this time.
4.5 Testing causal mechanisms
4.5.1 The causal mechanism of Iversen-Soskice
In the United States, the empirical conditions are generally similar to the assumptions made by Iversen and Soskice, meaning that the theory could have some predictive value. However, their hypothesis, that rising income inequality is generally expected to increase the share of votes received by the center-right party, cannot be strongly observed in figure 1. This likely indicates that their main causal mechanism- that within a majoritarian system, rising income inequality increases the incentive for the risk-averse middle-income voters to vote for the party that is less likely to increase taxes- does not appear to be consistently at work in the United States. This can partly be attributed to the fact that a partisan split, unrelated to income inequality, already exists in the American middle-income group. Professionals and white-collar workers have shifted more towards the Democrats, while managers and self-employed workers are more likely to identify as Republicans (Evans, 2000, p. 409). This demonstrates that voting behavior in the middle-class is not solely influenced by an increasing fear of being financially exploited by the lower-class. Although income-based voting increasingly occurs in the upper- and lower-income groups, the middle-class does not only vote based on income-related considerations. The existence of structural voting determinants within the middle-class, such as occupation and religion, that are unrelated to rising income inequality can help account for the absence of Iversen and Soskice’s expected outcome in the United States.
The correlation between income inequality and partisan voting in Ghana over time seemingly reflects the expectations of Iversen and Soskice. However, Ghana’s empirical conditions significantly divert from their assumptions. As discussed earlier, if income does not influence voting behavior, then the usefulness of the income-based model of Iversen-Soskice for understanding Ghana’s election results is significantly reduced. Therefore, it seems unlikely that this theory can help establish causality between income inequality and partisan voting in Ghana.
4.5.2 The campaign finance causal mechanism
As was predicted in the theoretical framework, there is no evidence that campaign finance has shifted the popular vote in favor of one party in the U.S. House of Representatives.
According to Jezer and Miller (2012), no correlation between campaign donations and legislative voting in the United States has been established thus far (p. 474). Furthermore, the influence of donations has been limited further due to the tightening of campaign finance laws, starting with the Bipartisan Campaign Reform Act in 2002 (Powell, 2010, p. 20).
The party financing system in Ghana, however, is almost completely unregulated. As opposed to the United States, there are no restrictions on how much a Ghanaian citizen can donate to a party or campaign (Arthur, 2017, p. 1130). Rich party individuals were identified as the most predominant funding source for political parties in Ghana according to a survey held in 2014 (Sakyi, Agomor & Appiah, 2015, p. 27). Although the exact data is unknown, there is no evidence to suggest that campaign contributions have grown with income inequality, nor that this benefits the NPP more than the NDC, or vice versa. This can be partly attributed to the fact that higher-income individuals in Ghana do not necessarily identify with one party over the other (Adams & Agomor, 2015, p. 376).
4.6 Alternative explanations
A strong causal relationship between rising income inequality and partisan voting in the United States is unlikely, as is primarily illustrated by the lack of a notable correlation. In Ghana, however, causality cannot be ruled out yet. Since the theories analyzed earlier do not appear to have any significant explanatory value for Ghana, potential alternative explanations for the correlation will be briefly discussed.
4.6.1. Retrospective voting
Income inequality can still influence voter behavior for Ghanaians that vote retrospectively. If the incumbent government fails to reduce the income gap, it contributes to lower satisfaction with the governing party and could thus influence the party choice of citizens for the following election (Youde, 2005, p. 8). If this type of voting is based on trends in income inequality, it could explain why the opposition party, the NPP, gained votes with the rise of inequality from 1992 to 2000 (figure 2). However, it cannot explain why a drop in inequality from 2006 to 2012 resulted in a loss of votes for the NPP, which was ruling at the time, since retrospective voters would reward the party for bettering conditions. This may be explained by the fact that retrospective voting in Ghana is not solely based on government performance regarding income inequality, but also strongly on educational policy and overall economic
growth and development (Youde, 2005, p. 9). Therefore, looking at changes in income inequality as the only indicator for retrospective voting in the following election is insufficient.
4.6.2. Swing-voters
About 20 to 25 percent of the population of Ghana are swing-voters, frequently changing who they vote for, and these voters are critical in determining election outcomes. These voters do not all vote retrospectively, since this only accounts for 8.5 percent of the population. Rather, the ability of each party to hold effective and campaigns that convince these voters that they understand their needs is most likely to influence their vote (Adams & Agmor, 2015, pp. 378-379). Unfortunately, no literature has identified the demographic makeup of these swing-voters, which makes it difficult to link their needs and voting behavior over time to income inequality.
5. Discussion
For the United States and Ghana, I was unable to prove that changes in income inequality strongly affected the share of votes received by left- or right-wing parties from 1992-2016. Rising income inequality in the United States and Ghana during this period did not consistently cause left- or right-wing parties to gain votes through mechanisms hypothesized by Iversen-Soskice and Meltzer-Richard. This was in part due to the empirical conditions differing from the theories’ main assumptions, which made the models less applicable to these countries and significantly reduced their usefulness in explaining changes in these election results. Campaign contributions were also not a causal mechanism through which income inequality influenced election results in either country. By comparing Ghana and the United States, it became evident that the two cases did not work against the hypotheses in similar ways. This means that the presence of a majoritarian two-party system with rising income inequality is not a sufficient condition for guaranteeing a consistent empirical result. On the contrary, country-specific voting determinants reduced the influence of partisan voting on income inequality in varying ways.
This research can, however, be improved upon. The theory of Moene and Wallerstein, discussed in the theoretical framework, may have provided an explanation for the lack of a significant correlation between rising income inequality and partisan voting in the United States. However, due to a lack of data on the voter population’s demand for redistribution and redistributive insurance, I was unable to empirically evaluate the applicability of their theory for the United States and Ghana. A lack of extensive research on what factors drive
middle-income voting behavior also made it difficult to fully test the model of Iversen and Soskice for both countries.
The positive correlation between income inequality and the share of votes received by the center-right party in Ghana was based on less extensive information than that of the United States, due to less frequent elections for the Parliament of Ghana and sparser Gini data. More complete data for Ghana would have helped more strongly establish or invalidate this correlation. None of the models addressed in the theoretical framework were able to help establish causality in Ghana. These models have generally only been applied to advanced democracies, which are more focused on redistribution, while Ghana is a relatively new democracy. This may have resulted in a lower relevance of these models for Ghana. However, alternate explanations, such as retrospective voting and swing-voters, could not fully explain the correlation either, meaning that I was unable to prove a causal relationship between income inequality and right-wing voting, given the available literature and data. Still, the possibility of an indirect causal relationship between the two variables in Ghana cannot be completely eliminated based on my research.
6. Conclusion
This thesis aimed to answer the question: do changes in income inequality over time affect the share of votes received by left-wing and right-wing parties in countries with two-party majoritarian systems? Why or why not? Much of the existing literature on the effects of income inequality on partisan voting was ultimately found to have little predictive and explanatory value for Ghana and the United States. This was largely due to the theories’ assumptions significantly differing from the countries’ empirical conditions. If the rise in income inequality did lead to more votes for a right-wing party, which is still a possibility due to the positive correlation between the two in Ghana, it would not likely be through campaign contributions or through the causal mechanisms described by Iversen-Soskice and Meltzer-Richard. Future research could be devoted to further analyzing whether any causality can be established in Ghana.
The main contribution of this paper to the academic efforts to understand the effect of rising income inequality on different aspects of politics is the conclusion that simply the presence of a majoritarian two-party system does not ensure that a positive correlation between income inequality and share of votes received by a right- or left-wing party will exist. This is primarily illustrated in the election results for the United States from 1992 to 2016. Future
research can be conducted to determine whether this lack of a correlation is primarily due to other factors having more influence on voting behavior or if the counteracting effects hypothesized by Moene-Wallerstein can be observed in the United States.
References
Abdulai, A. & Crawford, G. (2010). Consolidating democracy in Ghana: Progress and prospects? Democratization, 17(1), 26-67.
Adams, S. & Agomor, K. S. (2015). Democratic politics and voting behaviour in Ghana. International Area Studies Review, 18(4), 365-481.
Anckar, C. (2008). On the applicability of the Most Similar Systems Design and the Most Different Systems Design in comparative research. International Journal for Social Research Methodology, 11(5), 289-401.
Arhur, P. (2017). Political parties’ campaign financing in Ghana’s Fourth Republic: A
contribution to the discourse. Journal of Asian and African Studies, 52(8), 1124-1140. Barnes, L. (2012). Does median voter income matter? The effects of inequality and turnout on
government spending. Political Studies, 61(1), 82-100.
Boix, C. (1998). Political parties, growth and equality: Conservative and social democratic economic strategies in the world economy. Cambridge: Cambridge University Press. Boix, C. (2010) Origins and persistence of economic inequality. Annual Review of Political
Science, 13, 489-516.
Bonica, A., McCarty, N., Poole, K.T. & Rosenthal, H. (2013). Why hasn’t democracy slowed rising inequality? Journal of Economic Perspectives, 27(3), 103-124.
Borck, R. (2007) Voting, inequality, and redistribution. Journal of Economic Surveys, 21(1), 90-109.
Brookings Institute. (2014). Vital statistics on Congress [Data file]. Retrieved from: http://www.brookings.edu/vitalstats
Campanante, F.R. (2011). Redistribution in a model of voting and campaign contributions. Journal of Public Economics, 95, 646-656.
Charles-Coll, J.A. (2011). Understanding income inequality: Concept, causes and
measurement. International Journal of Economics and Management Sciences, 1(3), 17-28.
Clerk of the House of Representatives. (2016). Election statistics, 1920 to present [Data file]. Retrieved from: http://history.house.gov/Institution/Election-Statistics/
Cusack, T.R., Iversen, I. & Soskice, D. (2007). Economic interests and the origins of electoral systems. American Political Science Review, 101(3), 373-391.
Dodge, Y. (2010). The concise encyclopedia of statistics. New York: Springer-Verlag. Duverger, M. (1963). Political parties: Their organization and activity in the modern state.
New Jersey: Wiley.
Electoral Commission of Ghana. (2016). Election results [Data file]. Retrieved from: http://www.ec.gov.gh/resources/downloads/election-results.html
Esping-Andersen, G. & Myles, J. (2011). Economic inequality and the welfare state. In B. Nolan, W. Salverda, & T.M. Smeeding (Eds.), The Oxford handbook of economic inequality (640-665). Oxford: Oxford University Press.
Evans, G. (2000). The continued significance of class voting. Annual Review of Political Science, 3, 401-417.
Faure, A.M. (1994). Some methodological problems in comparative politics. Journal of Theoretical Politics, 6(3), 307-322.
Ferguson, T. (1995). Golden rule: The investment theory of party competition and the logic of money-drive political systems. Chicago: University of Chicago Press.
Field, A. (2013). Discovering statistics using IBM SPSS statistics. London: SAGE Publications.
Fohib, N. (2010). Dynamics of political parties’ administration in Ghana. Africa Today, 1, 25-41.
Gelman, A., Kenworthy, L. & Su, Y. (2010). Income inequality and partisan voting in the United States. Social Science Quarterly, 91(5), 1204-1219.
Hague, R. & Harrop, M. (2013). Comparative government and politics: An introduction. Basingstoke: Palgrave Macmillan.
Huber, J. D.& Stanig, P. (2007). Voting Polarization on Redistribution across Democracies. Working Paper, Department of Political Science, Columbia University.
International Institute for Democracy and Electoral Assistance. (2017). Electoral System Design [Data file]. Retrieved from: https://www.idea.int/data-tools/data/electoral-system-design
Iversen, T. & Soskice, D. (2006). Electoral institutions and the politics of coalitions: Why some democracies redistribute more than others. American Political Science Review, 100(2), 165-181.
Jezer, M. & Miller, E. (2012). Money politics: Campaign finance and the subversion of American democracy. Notre Dame Journal of Law, Ethics & Public Policy, 8(2), 464-498.
Karoly, L.A. (1992). Trends in income inequality: the impact of, and implications for, tax policy. Santa Monica: RAND Corporation.
Kenworthy, L. & Pontusson, J. (2005). Rising inequality and the politics of redistribution in affluent countries. Perspectives on Politics, 3(3), 449-471.
Kollman, K., Hicken, A., Caramani, D., Backer, D., & Lublin, D. (2017). Constituency-level elections archive [data file and codebook]. Retrieved from:
http://www.electiondataarchive.org/datacenter.html
Krugman, P. (2007). The conscience of a liberal. New York: W.W. Norton & Company. Laver, M. & Budge, I. (1992). Party policy and government coalitions. Basingstoke,
Lindberg, S.I. & Morrison, M.K.C. (2008). Are African voters really ethnic or clientelistic? Survey evidence from Ghana. Political Science Quarterly, 123(1), 95-122.
Lyons, T. (1997). A major step forward. Journal of Democracy, 8(2), 65-76.
McCarty, N. & Pontusson, J. (2011). The political economy of inequality and redistribution. In B. Nolan, W. Salverda, & T.M. Smeeding (Eds.), The Oxford handbook of
economic inequality (pp. 666-692). Oxford: Oxford University Press.
McCarty, N., Poole, K.T. & Rosenthal, H. (2006). Polarized America: The dance of ideology and unequal riches. Boston: MIT Press.
Meltzer, A.H. & Richard, S.F. (1981). A rational theory on the size of government. Journal of Political Economy, 89(5), 914-927.
Moene, K.O. & Wallerstein, M. (2001). Inequality, social insurance, and redistribution. The American Political Science Review, 95(4), 859-874.
Moene, K.O. & Wallerstein, M. (2003). Earnings inequality and welfare spending: A disagrregated analysis. World Politics, 55(4), 485-516.
National Democratic Congress. (2004). A better Ghana: NDC manifesto. Retrieved from: www.ndcghanaonline.com/index.php/manifesto
Norris, P. (1997). Choosing electoral systems: Proportional, majoritarian and mixed systems. International Political Science Review, 18(3), 297-312.
Novignon, J. (2017). Household income inequality in Ghana: A decomposition analysis. African Journal of Economic and Management Studies, 8(4), 515-526.
Persson, T., Roland, G., & Tabellini, G. (2007). Electoral rules and government spending in parliamentary democracies. Quarterly Journal of Political Science, 2(2), 1-34. Pontusson, J. & Rueda, D. (2010). The politics of inequality: Voter mobilization and left
parties in advanced industrial states. Comparative Political Studies, 43(6), 675-705. Powell, L. (2010). Political parties and the finance law. In M.M. Smith (Ed.), Campaign
finance reform: The shell game (pp. 20-30). Lanham: Lexington Books.
Sakyi, E.K., Agomor, K.S. & Appiah, D. (2015). Funding political parties in Ghana: Nature, challenges and implications (International Growth Centre No. 6812). London:
International Growth Centre.
Smidt, C.D. (2015). Polarization and the decline of the American floating voter. American Journal of Political Science, 61(2), 365-381.
United States Census Bureau. (2016). Income and poverty in the United States [Data file]. Retrieved from
https://www.census.gov/content/dam/Census/library/publications/2017/demo/P60-259.pdf
Wissenschaftszentrum Berlin für Sozialforschung. (2016). Manifesto project database [Data file]. Retrieved from: https://manifestoproject.wzb.eu/
World Bank. (2016). Ghana [Data file]. Retrieved from: https://data.worldbank.org/country/ghana
World Bank. (2016). United States [Data file]. Retrieved from: https://data.worldbank.org/country/united-states
Youde, J. (2005). Economics and government popularity in Ghana. Electoral Studies, 24, 1-16.
Younger, S.D., Osei-Assibey, E. & Oppong, F. (2017). Fiscal incidence in Ghana. Review of Development Economics, 21(4), 47-66.
