Take my notes?

Latin America has the highest inequality of any region in the world. Uneven distribution of income fundamentally threatens the continu- ance and quality of democracy in the region. Unequal democracies face threat from both the rich, who fear redistribution to the poor majority, and the poor, who see existing institutions as unresponsive to their demands (Acemoglu and Robinson 2006). Related to this, inequality may have dis- torting effects on the party system and encourage clientelism. The quality of democratic institutions, including voter turnout especially among the poor, is also known to be worse where income is more disparate. These dynamics are troubling in any nation, but are particularly so where nascent democratic institutions and young governments may not be able to manage these con- flicts.

This chapter addresses the role of inequality in Latin America from a different approach than nearly all previous work. I focus on the threats to democratic quality that may come from interregional inequality—diver- gence in income level and economic productivity across geographic regions within a nation. I compare these effects to those expected from interpersonal inequality and examine how democratic political institutions structure attention on these distinct types of inequality. My modest goal in this chapter is to show a systematic difference between interpersonal inequality, the unequal distribution across individuals, and interregional inequality. Latin America has long been discussed as the region of the world with the highest interpersonal income inequality. The data below show that Latin America also has the highest level of interregional inequality of any global region.

Using a new dataset on interregional inequality, interpersonal inequal- ity, and government spending in Latin America, I show that these types of inequality display different incidence and effects. Not only are these types of inequality conceptually distinct, they also drive divergent government behavior. I address the classic political economy concern with the effect

4 Inequality and

Democratic Representation

Melissa Ziegler Rogers

51

C o p y r i g h t 2 0 1 6 . L y n n e R i e n n e r P u b l i s h e r s .

A l l r i g h t s r e s e r v e d . M a y n o t b e r e p r o d u c e d i n a n y f o r m w i t h o u t p e r m i s s i o n f r o m t h e p u b l i s h e r , e x c e p t f a i r u s e s p e r m i t t e d u n d e r U . S . o r a p p l i c a b l e c o p y r i g h t l a w .

EBSCO Publishing : eBook Academic Collection (EBSCOhost) - printed on 9/7/2021 10:30 AM via UNIVERSITY OF SOUTH ALABAMA AN: 1353743 ; Foweraker, Joe, Trevizo, Dolores.; Democracy and Its Discontents in Latin America Account: s4595122.main.eds

52 Melissa Ziegler Rogers

of inequality on government spending, applying the existing theoretical work to interregional as well as interpersonal inequality. I found that while interpersonal inequality tends to prompt higher government spend- ing to address social ills in the region, interregional inequality has an opposite and robust effect to drive down government spending as regions become more lopsided. This finding alone has implications for the quality of democracy if indeed the regional demand for redistribution is discon- nected from its supply.

In this study, I took a cautious further step to address whether certain political institutions may condition government responses to inequality. I suggest that electoral rules vary in their attentiveness and reactions to regional versus interpersonal concerns. Some nations have electoral sys- tems more attuned to the challenges of interpersonal inequality: namely, centralized strong party systems with broad social constituencies. Other countries’ national institutions are more clearly geographic in orientation, with decentralized political and administrative systems and strong person- al vote connections to local zones. How and whether nations address these challenges of inequality will depend fundamentally on how political bar- gains are struck in nations and the incentives of politicians to represent constituents.

The chapter is structured as follows. First, I address the concept and incidence of interpersonal and interregional inequality within Latin America and in comparative perspective with the rest of the world. Second, I suggest theoretical reasons why interpersonal and interregional inequality may drive conflict in governments and different ways these conflicts are managed within democratic political institutions. Third, I show empirical results on the relationship between inequality and government spending and the inter- action between inequality and political institutions on government spend- ing. Finally, I link the findings in this chapter to the quality of representa- tion for the less privileged in Latin America’s democracies.

Background: Inequality and Democracy in Latin America

The historic focus of inequality and democracy in Latin America has been on the uneven distribution of income and resources (especially land) among individuals within a nation (Acemoglu and Robinson 2006; Boix 2003). Interpersonal inequality is the concept that motivates nearly all studies (and quantitative measures) of inequality in academic and policy research. This focus on distribution among individuals is appropriate and important, but does not paint a complete picture of politically relevant inequality. Most countries, especially in the developing world, have wide variance in admin-

EBSCOhost - printed on 9/7/2021 10:30 AM via UNIVERSITY OF SOUTH ALABAMA. All use subject to https://www.ebsco.com/terms-of-use

Inequality and Democratic Representation 53

istrative and political authority within their territory and weak mechanisms for government distribution across individuals and regions. Moreover, dem- ocratic politics in most nations of the world are organized spatially, around geographic jurisdictions, rather than individuals. Geography, and distribu- tion across regions, is thus important to politics in many nations, and inequality of individuals and regions is irrevocably intertwined. To advance our understanding of distributive conflict in Latin America, we need to con- sider in more detail how inequality across regions influences political out- comes and thus democratic quality.

Latin America, like much of the developing (but also the developed) world, is characterized by uneven development and agglomeration around a central city. In most nations, the population of the largest city dwarfs the size of any other city. In Latin America, the examples of population concen- tration come easily to mind—the megacities of Mexico City, Buenos Aires, and São Paolo dominate the population and economic activity of their respective nations. Latin America is a highly urbanized region of the world, second only to the wealthiest Organisation for Economic Co-operation and Development (OECD) nations in urban population, and much of this popu- lation is concentrated in one city center. These cities in Latin America are modern, economically diverse, and reminiscent of their counterparts in the developed world. However, much of the area outside of the capitals of Latin American countries is set apart by economic stagnation and depressed development (Sawers 1996).

The history of any country is structured by distributive conflict not only between social groups (rich and poor classes), but also between regions and economic sectors. The wealthy in agricultural regions, for example, may have fundamentally different economic interests from the wealthy in indus- trial regions. Their attitudes toward redistribution, trade openness, and gov- ernment investment may be more strictly opposed than those of rich and poor within the same sectors and regions (Cusack, Iversen, and Soskice 2007). National policies strongly affect interpersonal distributions of income, of course, through labor and social welfare policies. Central poli- cies are perhaps even more critical, however, to adjudicating between regions and economic sectors through laws addressing trade policy, taxation and subsidization, and fiscal federalism. Import-substitution industrializa- tion provides a cogent example of political adjudication of regional interests in Latin America.

Regional inequality is largely absent from politics research for both theoretical and empirical reasons. Historical accounts of inequality in most nations have typically centered on class interests, in narratives motivated by the experience of class conflict in Western European nations. Additionally, the measurement of inequality has focused nearly exclusively on individual

EBSCOhost - printed on 9/7/2021 10:30 AM via UNIVERSITY OF SOUTH ALABAMA. All use subject to https://www.ebsco.com/terms-of-use

54 Melissa Ziegler Rogers

or household distributions of income quintiles, through indicators such as the Gini coefficient and the ratio of low to high incomes (e.g., 90/10 quin- tiles, 50/10 quintiles). This chapter is meant to address both gaps in this lit- erature, by raising theoretical reasons to focus on regional distributive con- flict and its role in democratic quality, and by adding quantitative evidence to begin to explore these questions.

In Table 4.1, the incidence of interregional inequality is shown for available Latin American countries for the period 2000−2010. The measure listed is the population-weighted coefficient of variance in regional GDP per capita, discussed in more detail below. It is apparent in the table that the level of interregional inequality in the region varies quite widely, from the low in Bolivia to the high in Mexico. The large federations of Latin America have strikingly high levels of interregional inequality. Not surpris- ingly, differences in regional income are a pervasive political concern in Argentina, Brazil, and Mexico. Even the countries with relatively low levels in comparison to their neighbors, such as Bolivia or Colombia, are nonethe- less on the upper side of the global scale.

Table 4.2 features average values of global regions to compare Latin America to other parts of the world, again in the period 2000−2010. Table 4.2 reveals that the Latin American region has very high levels of interre- gional inequality. This level is particularly striking in comparison with Western Europe. Each Latin American country has interregional inequality levels higher than each Western European nation. In fact, if you exclude Indonesia (the highest value in the world) from the Southeast Asian sample, Latin America has the highest inequality of any global region. These simple charts reveal the potential relevance of the question of interregional inequality to the political environment of Latin America today.

Argentina 0.621

Brazil 0.485

Bolivia 0.306

Chile 0.421

Colombia 0.388

Ecuador 0.610

Mexico 0.719

Panama 0.483

Peru 0.495

Table 4.1 Regional Inequality in Latin American Countries, 2000−2010

Country Regional Inequality Score

EBSCOhost - printed on 9/7/2021 10:30 AM via UNIVERSITY OF SOUTH ALABAMA. All use subject to https://www.ebsco.com/terms-of-use

Inequality and Democratic Representation 55

The interregional measures of inequality shown above are calculations of coefficients of variance in regional GDP per capita. The regional unit used in this study is the state, province, or department as defined by the nation itself. The conceptual analog is to the US state, although the powers designated to those units clearly vary across nations. The coefficient of variance aggregates the values of regional GDP to one value for a nation that is comparable across nations. The interregional inequality measure shown above is weighted according to regional population, on the assumption that regional wealth or poverty is more politically meaningful when attributed to a large segment of the population. The calculation formula is shown in Figure 4.1 below and the summary statistics for nations in the sample can be found in Appendix 1.

A few examples provide some context for the regional gaps in Latin America that are apparent in these indicators. In Argentina, the city and province of Buenos Aires together produce nearly 70 percent of national GDP. Thus, two provinces out of twenty-four generate the vast majority of economic output, and nineteen of twenty-four produce less than 1 percent of GDP each. Of course, the measures use regional GDP per capita, rather than GDP, meaning certain provinces with high productivity relative to sparse population (e.g., Santa Cruz or Neuquén) dampen the statistical effect of what is highly lopsided economic output. Similarly, São Paolo in Brazil and the federal district of Mexico alone produce 33 percent and 16 percent of GDP, respectively. Bogotá produces 25 percent of GDP in Colombia. These figures indicate a geographic concentration of economic might within Latin American countries.

These simple summary statistics reveal a notable difference between inequality in Latin America and the rest of the world. In the sections below, I apply these new data to much-studied relationships between inequality and redistributive politics in the political economy literature. As the most com- mon entry point into this literature, I examine the differences between the

Latin America 0.477

Western Europe, North America, Australia, and New Zealand 0.206

Eastern Europe and the former Soviet Union 0.373

East Asia 0.342

Southeast Asia 0.626

South Asia 0.402

Table 4.2 Regional Inequality in Global Perspective, 2000−2010

Global Region Regional Inequality Score

EBSCOhost - printed on 9/7/2021 10:30 AM via UNIVERSITY OF SOUTH ALABAMA. All use subject to https://www.ebsco.com/terms-of-use

56 Melissa Ziegler Rogers

two types of inequality for their effect on total government spending. Next, I address the role that political institutions play in this distribution—what incentives do politicians under different systems of representation have to deliver goods to social groups or regions?

Inequality and Government Spending in Latin America

Inequality encourages political conflict as rich and poor groups articulate different preferences for government spending. In the most simplistic for- mulation of economic reasoning, rich groups should press for lower govern- ment spending under increasing inequality. In a progressive tax system, government spending is inherently redistributive—those who benefit from government spend less to consume it. Poor groups should have opposite preferences, pressing for more redistributive government spending as inequality grows. These dynamics have been almost exclusively attributed to rich and poor individuals, most famously through the Romer-Meltzer- Richard model (RMR model; Romer 1975; Meltzer and Richard 1981).1 They can also be reasonably attributed to regional actors who face similar incentives to block or advocate redistribution. Rich regions, like rich indi- viduals, should hope to limit government spending and, under certain insti- tutional conditions, have the means to do it (Giuranno 2009; Rogers 2016).

Where politics is structured and resources are administered along geographic lines, politicians and voters are incentivized to evaluate distri- butional concerns in terms of their region (Beramendi 2012). Certain political institutions, notably geographic constituencies, federalism, per- sonal vote elections, and, to some extent, presidentialism territorialize political competition (Rogers 2016). In contrast, parliamentarism, closed list proportional representation, and unitarism have centripetal effects— they encourage focus on national politics (Gerring, Thacker, and Moreno 2005).

Where resources are divided to provinces or districts, politicians push to maximize resources going to their district, regardless of whether they prefer fiscal constraint or expansion at the national level (Rehfeld 2005). Federalism, bicameralism, electoral rules, and even the presidential- parliamentary distinction not present in Latin America inform politicians and voters about whether they should think primarily in terms of their dis- tricts or their social group, or both. All political systems have some concept of both geography and social group representation, of course, but most sys- tems fall closer to one or the other. Chile, for example, has very nationally oriented politics while Colombia is a strongly local political system. These institutional features shape the distributive debate in a nation to be either

EBSCOhost - printed on 9/7/2021 10:30 AM via UNIVERSITY OF SOUTH ALABAMA. All use subject to https://www.ebsco.com/terms-of-use

Inequality and Democratic Representation 57

more concerned with local public goods or with broad social goods (Milesi- Ferretti, Perotti, and Rostagno 2002).

The primary reason politicians in geographically oriented systems focus on local allocation is that politicians’ careers depend on pleasing their districts. This takes different forms depending on the structure of political institutions. For US politicians, this means bringing home pork to their districts to win reelection (Grimmer 2013). For Argentina and Brazil, federalized career paths incentivize attracting central transfers to the cof- fers of governors. In Chile, politicians’ career paths depend more on national parties that aim to please social constituencies that span regions (Harbers 2009). Political institutions fundamentally shape whether politi- cians (and voters) evaluate the national distributive game in regional terms.

The examples above of the United States, Argentina, and Brazil pro- vide clear examples of institutions that influence distributive structures in national politics. First, the US localized credit claiming is attributed pri- marily to the personal vote, which stems from single-member district sim- ple plurality systems for members of Congress (Cain, Ferejohn, and Fiorina 1987). All legislators in the United States (and some in Latin America) are elected according to these rules, which motivate members to bring resources back home. In Argentina, there is not an equivalent personal vote because members are elected through closed list proportional represen- tation. However, list access is determined by local party leaders, typically the governor. Accordingly, the nomination process for national legislators makes politicians think about national policy in provincial terms (Jones et al. 2002). Open list proportional representation in Brazil incentivizes credit claiming for politicians who must compete within their own party lists. Bringing resources to their home territory is one way to distinguish them- selves, although the specific geographic pull is less clear than in single- member districts (Samuels 2002).

In this analysis, I focused on the effects of electoral rules, as opposed to the more obvious choice of federalism, on the territorialization of politics. Importantly for both the theoretical implications and the empirics, the design of electoral rules is not obviously endogenous to regional disparity. The case has been made that the design of federalism is endogenous to regional inequality (Bolton and Roland 1997; Beramendi 2012). The design of electoral rules has been argued to be endogenously driven by class con- flict, but not regional inequality (Boix 1999; Cusack, Iversen, and Soskice 2007). Variation in electoral system design in Western Europe, for example, has been attributed to concerns with income redistribution and social insur- ance (Boix 1999; Iversen and Soskice 2006). Class conflict may have regional implications, of course, but the primary motivation for electoral

EBSCOhost - printed on 9/7/2021 10:30 AM via UNIVERSITY OF SOUTH ALABAMA. All use subject to https://www.ebsco.com/terms-of-use

58 Melissa Ziegler Rogers

rule design appears to be whether incumbent groups can win a majority, or will need to share power with rising socialist parties. In this regard, elec- toral rules that territorialize politics may be plausibly exogenous to con- cerns of regional inequality per se. Of course, this question remains for future research.

The dynamics of regional preferences are not as straightforward as those of individuals.2 While rich regions may want to keep their resources within their borders, there is individual heterogeneity within those borders. More specifically, rich regions have both rich and poor individuals with dif- ferent preferences for government spending. Under different institutional and economic conditions, poor or rich groups can form coalitions across regional borders to maximize their gains as social groups, rather than regional units (Beramendi, Rogers, and Diaz-Cayeros, forthcoming). Just as with regional interests, therefore, the institutional structure of national poli- tics should be crucial to how and whether groups of individuals or regions are able to collectively act to attain their policy preferences.

This introductory empirical investigation of regional and interpersonal dynamics tested two primary questions: What is the effect of regional inequality on government spending? And how do political institutions con- dition the effect of inequality on government spending? These questions are complex and the empirical analysis is necessarily preliminary, so I keep the theoretical explanation short and concise, relying on existing theories with specific application to regional dynamics in the Latin American context in particular.

Theoretical Expectations

The RMR model of interpersonal inequality shows that increasing inequali- ty should drive higher government spending. In this theory and its exten- sions, the decisive median voter will push for higher government spending as inequality grows. The median voter, who grows poorer relative to the mean voter with rising inequality, should advocate more government spend- ing that is increasingly redistributive as incomes become more uneven in a progressive tax system. The simplified model of this theory has been largely dismissed for empirical and theoretical reasons in comparative political economy. The empirical predictions are not born out in the data; in fact, the opposite is often shown to be more accurate—inequality drives down gov- ernment spending rather than increases it (cf. Alesina and Glaeser 2004; Benabou 2000; Gouveia and Masia 1998).

One reason for these weak empirical findings is the inadequate por- trayal of the political process in the RMR model. That the median voter is

EBSCOhost - printed on 9/7/2021 10:30 AM via UNIVERSITY OF SOUTH ALABAMA. All use subject to https://www.ebsco.com/terms-of-use

Inequality and Democratic Representation 59

decisive (or, more cogently, that the full spectrum of individuals votes) has been seriously refuted in formal and theoretical models (Benabou 2000; Gerber and Lewis 2004; Beramendi 2007). More importantly for the purposes here, RMR assumes a form of political representation—a single district with direct democracy—that is far removed from politics in any nation. Rather, nations divide their territory into voting districts, most often in geographic units, and establish a range of actors with veto author- ity able to halt the interests of the aggregate median voter. Regional inequality will press distributive conflict that discourages cooperation and coordination to boost government spending. For a large sample of coun- tries, Rogers (2016) has shown that political institutions empower regions to constrain government spending through veto authority and enable poor individuals to increase government spending through majority power. As interregional inequality grows, therefore, rich regions should have greater incentive to block spending, leading to my expectation in Hypothesis 1.

Hypothesis 1: Higher interregional inequality drives lower govern- ment spending.

Electoral systems structure the bargaining dynamics of politicians, whether on the grounds of interpersonal or interregional inequality. By dividing nations into regional electoral units, the interests of regions are highlighted and voter incentives to express preferences on a regional basis become more likely. Electoral systems affect whether politicians think more in local or national terms, increasing or decreasing the relevance of the region-specific preferences. In this manner, incentives created by elec- toral systems interact with inequality to influence government distribu- tion.

Hypothesis 2: Electoral institutions interact with regional inequali- ty to shape government spending.

Research Design

The basic statistical model for Hypotheses 1 and 2 is

Total Government Spending/GDP = ßRegional Inequality + ßInterpersonal Inequality + ßRegional Inequality * Personal Vote + ßInterpersonal Inequality * Personal Vote +ßPersonal Vote + ßPopulation(log) + ßTrade Openness + ßCapital Openness + ßGDP per capita (log) +

EBSCOhost - printed on 9/7/2021 10:30 AM via UNIVERSITY OF SOUTH ALABAMA. All use subject to https://www.ebsco.com/terms-of-use

60 Melissa Ziegler Rogers

ßGDP Growth + ßEthnic Fractionalization + ßPopulation over 65 + ßFederalism + ε

Dependent Variables

The large literature on the effects of income inequality on government poli- cy typically begins with an analysis of government spending overall and redistributive spending (typically, social transfers) in particular. This analy- sis is a comparison of interregional measures of inequality to interpersonal measures of inequality, so I employed this most common measure, govern- ment spending, as the primary dependent variable. Of course, other dependent variables are highly relevant and will be explored in the future. For this preliminary research, I preferred to evaluate the effects of inequal- ity as they have been traditionally measured in the field so that scholars could have a direct apples-to-apples comparison.

Within the government spending categories, I used two common measures—general government and central government expenditures.3 General government expenditure refers to spending at all levels of govern- ment (central, state, local). Central government expenditure is restricted to resources distributed by the national government. In theory, the central government variable may seem the most appropriate for evaluating the effects of regional inequality on national policymaking. The framework I have provided is one in which representatives of regions and social groups come to the national bargaining table to divide up the spoils of the central government. However, I think it is appropriate to consider general govern- ment expenditures for at least two reasons. First, one primary effect of fis- cal federalism (which would be captured only in general government measures) is for subnational regions to tax and spend to their own prefer- ences. Expenditures in jurisdictions below the national level could be strongly affected by regional inequality as rich territories spend at high levels to their preferences and tax bases, and poorer areas have fewer resources to distribute. Moreover, poor regions are subsidized with (some- what) progressive national transfers that plump their spending.

Second, regional inequality may be endogenously related to the fiscal structure of a nation (Beramendi 2012). That is, rich regions may prefer to decentralize many government functions so that they can consume to their preferences without subsidizing poorer regions. Regional inequality in this case would not necessarily reduce government spending, but shift its geo- graphic incidence. In that case, central government spending would still be the best theoretical indicator of the effect of regional inequality on shared resources, but the comparison with general government spending would provide a more complete picture of its overall impact.

EBSCOhost - printed on 9/7/2021 10:30 AM via UNIVERSITY OF SOUTH ALABAMA. All use subject to https://www.ebsco.com/terms-of-use

Inequality and Democratic Representation 61

Independent Variables

Regional development has long been a topic of interest in political geogra- phy and economics but not much explored in political science due, in large part, to a dearth of data and theoretical models devoid of geography. In this study, I utilized a large dataset of interregional inequality collected from fifty nations around the world for the period 1980−2010 (Rogers 2016).4 The focus of this examination was the nine available Latin American coun- tries (Argentina, Brazil, Bolivia, Chile, Colombia, Ecuador, Mexico, Panama, Peru), with comparison to results for the entire sample. Although not all Latin American countries have available data, those countries includ- ed account for the vast majority of citizens and economic productivity in the region.

The interregional inequality measures used in this analysis are country- year observations synthesizing region-level GDP per capita data for each country. I employed the population-weighted coefficient of variance of regional GDP per capita.5 The formula for this measure is shown in Figure 4.1, with y representing regional GDP per capita, n the number of regional units, and p the population. This indicator is calculated independent of the number of regions considered, is not sensitive to shifts in average GDP level, and satisfies the Pigou-Dalton principle.6

Importantly, regional inequality is not fixed in Latin America. Several of the nations included have experienced significant changes in regional inequality in the period under examination. For example, in Bolivia, the regional inequality value fluctuated between .2 and .39, a nearly 100 per- cent difference between the minimum and maximum value between 1988 and 2011. Notable changes are also observable in regional inequality in Mexico and Chile, with fluctuations of 50 percent between the minimum and maximum observed value. Other countries’ values, such as Peru and

Figure 4.1 Calculation of Regional Inequality Measure

[∑ ]1—ȳ pi ( ȳ–yi )2 i=1

1/2n

Note: y-bar represents the country’s average GDP per capita, yi is the GDP per capita of region i, n is the number of regional units, and pi is the share of the country’s total population in region i.

EBSCOhost - printed on 9/7/2021 10:30 AM via UNIVERSITY OF SOUTH ALABAMA. All use subject to https://www.ebsco.com/terms-of-use

62 Melissa Ziegler Rogers

Panama, were relatively static. Appendix 3 and Appendix 4 show changes over time in the Latin American sample nations, and descriptive statistics by nation. Overall regional inequality fluctuated somewhat less in the Latin American (standard deviation = .14) than in the global sample (standard deviation = .20), but the changes are nonetheless notable. In comparison, the Gini coefficient of income inequality had a standard deviation of .04 in Latin America and .09 in the global sample. This variation over time makes the regional inequality data plausibly suitable for time series cross-sectional regression analysis.7

I compared the effects of interregional inequality to available measures of interpersonal inequality, namely, the Gini coefficient. The Gini coeffi- cient is the most widely used measure of interpersonal inequality and it is based on a nationally aggregated concept of inequality between quintiles of income.8 Of several available cross-national datasets of interpersonal inequality, I used Frederick Solt’s data, the Standardized World Income Inequality Database (SWIID) (see Solt 2009). These data have advantages both in coverage (which tends to be spotty in Latin American countries) and in distinguishing income before (Gini market) and after (Gini net) govern- ment transfers. For theoretical reasons, I used the Gini market value in the regression analysis. I was looking for the effect of inequality on government spending, therefore the Gini market value provided a cleaner indicator because it excludes government taxation and spending in its calculations. Intraregional inequality is also a relevant distributive concern to national politics but data are sparse, even in OECD countries, and are not commonly collected by Latin American countries (see Beramendi, Rogers, and Diaz- Cayeros, forthcoming). Accordingly, I could not include these values in the analysis.

The theoretical focus of this chapter is on the interaction between inequality and political institutions. I focused in this preliminary research on the effect of electoral rules, although I recognize many other institutions can influence these dynamics (Rogers 2016). Accordingly, I measured polit- ical institutions to capture the incentives of politicians to deliver resources to social or geographic constituencies with a measure of the personal vote (Carey and Shugart 1995). The personal vote is not, by definition, a geo- graphic concept but tends to relate highly to the spatial orientation of a country’s political institutions.

The personal vote is the relative value of an individual politician’s rep- utation to his or her party’s reputation in the electoral fate of that politician. The personal vote is high when politicians must distinguish themselves on personal characteristics rather than party characteristics. This highlights the intraparty conflicts as well as the localized incentives that are the focus of my theoretical development. The personal vote measure is an additive index

EBSCOhost - printed on 9/7/2021 10:30 AM via UNIVERSITY OF SOUTH ALABAMA. All use subject to https://www.ebsco.com/terms-of-use

Inequality and Democratic Representation 63

of ballot structure (takes a value of 2 if parties do not control access or ordering of candidates, 1 if parties control access or order, 0 if parties con- trol both access and ordering), vote pooling (2 if votes are not pooled, 1 if votes pooled across some members in a district, 0 if votes pooled across all members), vote type (2 if voters cast a vote for one individual candidate, 1 if voters cast a vote for a party, 0 if voters cast one vote for a party), and district magnitude (Johnson and Wallack 2006).9 For each component of the index except the district magnitude, a higher value implies a higher personal vote.10

For my purposes, I was concerned with the extent to which politicians must think about their local jurisdiction more than the nation as a whole. I argue the same electoral system structures that tend to cultivate personal reputations are also those most likely to encourage geographic-focused rep- resentation. Although this is not a perfect measure of the concept, it does provide a reasonable proxy for the geographic orientation of an electoral system. The average values of the personal vote scores, by Latin American country, are shown in Appendix 2.

Control Variables

When measuring government spending, several economic and demographic characteristics are necessary to isolate specific political effects. The first is level of development, measured with the log value of gross domestic prod- uct per capita (GDPPC) corrected for purchasing power parity. This variable helps to control for Wagner’s law, which predicts higher government spend- ing as countries grow richer. Second is population, again logged, because larger populations might offer returns to scale in delivery of public services (the numerator) or increase productive capacity (the denominator). The third is an age ratio, the percentage of the population aged sixty-five years and older. In developed countries, this is an important variable to capture the size of the population dependent on government health and income sub- sidies. This variable may not be quite as relevant in certain Latin American countries, but should play a role in places such as Argentina, Chile, or Mexico, with notable pension systems.

I included ethnic fractionalization because some studies have found social expenditure is lower where ethnic heterogeneity (and, presumably, ethnic tensions) is high (Alesina, Baqir, and Easterly 1999). This may be important, in particular, in the countries with considerable indigenous popu- lations such as Bolivia and Peru, and racial diversity such as Brazil. Globalization, measured as trade openness and capital openness, also con- trols for the likelihood that countries with open borders are constrained from taxing at high levels to provide government services. With very open

EBSCOhost - printed on 9/7/2021 10:30 AM via UNIVERSITY OF SOUTH ALABAMA. All use subject to https://www.ebsco.com/terms-of-use

64 Melissa Ziegler Rogers

markets, this variable could be even more important in Latin American countries (Wibbels and Arce 2003).

Political institutions other than electoral rules may have similar effects on the territorial scope of national politics. Most importantly, federalism should orient budgets and politician behavior more toward the regions than unitary systems. Federalism is particularly important toward explaining central government expenditure, which should be low relative to unitary systems because significant fiscal activity occurs at the subnational level. I also controlled for institutions that may influence the speed and veto author- ity in the budget process, including the presidential budget authority (Alesina et al. 1999) and territorial bicameralism in alternative models. The budget authority was similar (and high) for the president in the included countries, with the possible (relative) exception of Peru. Bicameralism is frequently linked to territorial politics and was present in the bigger (and more decentralized) countries in the sample. Only Chile featured a unitary system and bicameralism in the sample. I excluded country fixed effects because I was concerned primarily with the variance across countries on regional inequality and political institutions. The political institutions were fixed in the sample, with the exception of Bolivia, prohibiting over time within country comparisons. Regional inequality did change over time in the sample, but it was a slow-moving variable. Accordingly, cross-country comparisons offered more theoretical and empirical value in this initial examination.

Estimation Techniques

Predicting government spending requires several adaptations to the standard ordinary least squares model to correct for violations of its assumptions. Throughout, I employed panel corrected standard errors (PCSE). I con- trolled for autoregression in spending with the lagged dependent variable, or alternative, an autoregressive (AR1) process, to reduce the considerable variance absorbed by the lagged dependent variable.11 In the models pre- sented in the body of the text, I focused on PCSE models with AR1 autocor- relation.

Results

Effects

Interregional inequality had a consistent effect of reducing government spending relative to GDP in a cross-national sample. This result, and its comparison with interpersonal inequality, is shown in Table 4.3. This effect

EBSCOhost - printed on 9/7/2021 10:30 AM via UNIVERSITY OF SOUTH ALABAMA. All use subject to https://www.ebsco.com/terms-of-use

Inequality and Democratic Representation 65

Regional −6.266** −21.514*** −3.370*** −4.594** inequality (3.105) (4.555) (1.210) (2.281)

Gini 0.430*** 0.658*** 0.019 0.181** coefficient (0.099) (0.116) (0.028) (0.077)

GDP −0.100*** −0.098** −0.108*** −0.203*** growth (0.039) (0.044) (0.020) (0.042)

Population 0.983* 3.122*** 0.525*** 1.619*** over age 65 (0.588) (0.434) (0.101) (0.231)

GDP per capita 0.403 0.756 0.000 −2.754*** (logged) (2.317) (2.563) (0.539) (0.996)

Population 0.411 0.645 −0.445* −1.958*** (logged) (0.644) (0.691) (0.247) (0.457)

Capital 0.599* −0.124 0.119 −0.734*** openness (0.326) (0.292) (0.163) (0.285)

Trade −0.011 0.056*** −0.009 0.026 openness (0.017) (0.017) (0.008) (0.017)

Personal vote 0.261** 0.606*** 0.044 0.1 rank (0.119) (0.153) (0.070) (0.126)

Ethnic 11.687** 24.462*** −0.391 −4.185 fractionalization (5.509) (6.301) (1.089) (2.772)

Federalism 1.492** 0.746 (0.650) (0.602)

Constant −25.187 −49.890* 13.918*** 38.252*** (22.126) (25.643) (5.060) (9.639)

R2 0.518 0.842 0.586 0.614

Observations 127 83 875 725

Countries 9 8 45 42

Country fixed No No No No effects

χ2, F [Prob > F] 58.773 (0.00) 200.935 (0.00) 269.423 (0.00) 332.485 (0.00)

Note: GDP is gross domestic product. *p <0.1, **p<0.05, ***p<0.01, two-tailed test.

Table 4.3 Effect of Inequality on Government Spending (percentage of GDP)

Latin American Countries All Available Countries General Central General Central Government Government Government Government Expenditure, Expenditure, Expenditure, Expenditure, % of GDP % of GDP % of GDP % of GDP

EBSCOhost - printed on 9/7/2021 10:30 AM via UNIVERSITY OF SOUTH ALABAMA. All use subject to https://www.ebsco.com/terms-of-use

66 Melissa Ziegler Rogers

was thoroughly documented for the Latin American cases and held true for the larger sample of all available countries with regional inequality data on the right side of Tables 4.3 and 4.4 (Rogers 2016). Whether the dependent variable was general or central government spending, a rise in interregional inequality was associated with a reduction in government output.

Interpersonal inequality, measured with the Gini coefficient, appeared to have a strong positive effect on government spending in the Latin American sample. In the global sample, the effect was highly inconsistent and often negative.

The control variables performed largely as expected. The personal vote was associated with increased spending in the Latin American sample, but had no clear effect in the full sample. Importantly, the personal vote was not a proxy for regional conflict, but for territorialized political institutions.12 Territorial orientation of politics may very well incentivize overspending on pork by districted politicians who have little regard for the national budget. However, nationalizing political institutions may also encourage overspend- ing as politicians collude to extract rents (Persson, Roland, and Tabellini 1997). There are conflicting expectations for how the personal vote should impact government spending overall, but district-targeting incentives are likely to drive down social spending in the national budget (Milesi-Ferretti, Perotti, and Rostagno 2002).

The effect of trade on spending was negative, although capital open- ness had an ambiguous, and sometimes positive, effect on spending. There was no evidence of the Wagner effect—that higher GDP per capita drives increased spending—in the Latin American or general sample. Contrary to broad expectations, in the Latin American sample, ethnic fractionalization was associated with higher government spending relative to GDP. The association was negative in the full sample, but not significant. The effect of ethnic fractionalization may have been driven in the Latin American sample by the relatively high spending and high fractionalization in Brazil, Bolivia, and Colombia. GDP growth was associated with reduced spending, likely because the denominator grew and spending did not keep pace.

The effect of interregional inequality on spending in Latin America was the opposite of that on interpersonal inequality. While differences in income across individuals drove higher government spending, as regions became more unequal, the generosity of national expenditure declined. This is an important difference in the nature of inequality and its relation to govern- ment that has not been previously theorized or tested in Latin American countries. The important question becomes, Why, in Latin American coun- tries, does distributive conflict manifest itself in two distinct ways? Why do the preferences of rich regions win out in Latin America while the desires of

EBSCOhost - printed on 9/7/2021 10:30 AM via UNIVERSITY OF SOUTH ALABAMA. All use subject to https://www.ebsco.com/terms-of-use

Inequality and Democratic Representation 67

the poor individuals appear to shape changes in government spending? I discuss two of many possible answers here. First, it is possible that rich regions are also unequal regions (such Buenos Aires), so rich regions may align with poor regions for social spending that will benefit their poor indi- viduals but dampen spending that will redistribute to other, poorer, regions. This is unlikely in Brazil and Mexico in which the rich regions are also

Regional inequality −12.457*** −5.883* −27.505*** −18.187*** (3.971) (3.193) (6.436) (4.161)

Gini coefficient 0.509*** 0.623*** 0.647*** 0.610*** (0.090) (0.166) (0.106) (0.210)

Regional inequality x 1.163* 1.379 personal vote rank (0.694) (1.229)

Gini coefficient x −0.036 −0.004 personal vote rank (0.028) (0.045)

Personal vote rank −0.124 2.086 0.166 0.85 (0.241) (1.451) (0.437) (2.340)

GDP growth −0.099** −0.095** −0.097** −0.099** (0.042) (0.038) (0.044) (0.045)

Population over age 65 1.499*** 1.093* 3.517*** 3.052*** (0.519) (0.580) (0.540) (0.423)

GDP per capita (logged) −0.568 −0.348 −0.466 0.006 (2.041) (2.459) (2.592) (3.028)

Population (logged) 0.47 0.346 0.587 0.97 (0.584) (0.628) (0.678) (0.695)

Capital openness 0.488 0.641** 0.035 −0.256 (0.328) (0.327) (0.312) (0.264)

Trade openness −0.006 −0.014 0.053*** 0.059*** (0.016) (0.017) (0.017) (0.017)

Ethnic fractionalization 12.666** 10.632* 24.463*** 23.389*** (5.123) (5.555) (6.209) (6.755)

Federalism 1.505*** 1.666** 0.968 (0.552) (0.661) (0.600)

Constant −22.371 −28.398 −39.195 −41.979* (20.120) (21.023) (25.675) (25.398)

R2 0.545 0.525 0.845 0.836 Observations 127 127 83 85 Countries 9 9 8 8 Country fixed effects No No No No χ2, F [Prob > F] 90.782 (0.00) 58.4 (0.00) 254.528 (0.00) 197.75 (0.00)

Note: GDP is gross domestic product. *p<0.1, **p<0.05, ***p<0.01, two-tailed test.

Table 4.4 Interactive Effect of Inequality and Electoral Institutions on Government Spending (percentage of GDP)

Latin American Countries

General Central General Central Government Government Government Government Expenditure Expenditure Expenditure Expenditure

EBSCOhost - printed on 9/7/2021 10:30 AM via UNIVERSITY OF SOUTH ALABAMA. All use subject to https://www.ebsco.com/terms-of-use

68 Melissa Ziegler Rogers

more equal. It is plausible in Argentina, but unknown in the other countries in the sample (Beramendi, Rogers, and Diaz-Cayeros, forthcoming).

Another possibility that I could test more directly is how the electoral rules shape political winners and losers. Does the political system give incentives for politicians to cater to geographic regions or to social classes? If politicians are rewarded for representing the poor as a group, for exam- ple, we should expect interpersonal inequality to drive higher government spending. If, however, all politics is local, then politicians should advocate geographically based resources as interregional inequality increases. I examine this below with the electoral system variables from Latin America.

Conditional Effects: Inequality Under Different Electoral Systems

The simple theory I offered above is that politicians should care about dif- ferent constituencies depending on electoral rules. If electoral rules incen- tivize politicians to deliver goods to national social groups, regional con- cerns should be less important. If electoral rules motivate locally oriented thinking, politicians will debate geographically oriented spending with more fervency. What is unclear from this theory is what the general effect of inequality should be, given the electoral system. Does regional inequal- ity drive more or less spending in regionally oriented political systems? Does interpersonal inequality drive a wedge between voters and politi- cians only in party systems oriented to social class? The interactive results of inequality based on electoral system help to sort these dynamics.

Figure 4.2 Marginal Effect of Regional Inequality by Personal Vote Rank, Latin America

$ $

$

!

! !

-25

-20

-15

-10

-5

0

5

10

15

20

0 1 2 3 4 5 6 7 8 9 10 11 12

E st

im at

ed E

ff ec

t o f M

ea n

R eg

io na

l I ne

qu al

it y

Personal Vote Rank

EBSCOhost - printed on 9/7/2021 10:30 AM via UNIVERSITY OF SOUTH ALABAMA. All use subject to https://www.ebsco.com/terms-of-use

Inequality and Democratic Representation 69

The conditional effects of inequality, based on the electoral system, are featured in Table 4.4. Recall that the personal vote was measured as a rank- ing of electoral systems by Joel W. Johnson and Jessica S. Wallack (2006). The higher the rank, the more personal vote oriented was that political sys- tem. The lower the rank, the stronger was the party vote. Accordingly, when interpreting the statistical effect of interregional and interpersonal inequali- ty under different electoral systems, I looked for opposite signs to show similar effects and similar signs to show distinct effects. I expected interper- sonal inequality to matter most in party vote systems (low personal vote rank) and interregional inequality to matter most in personal vote systems (high personal vote rank). Mattering within an electoral system, of course, may mean higher or lower spending or no effect at all. This is what I explored in the marginal effects. The results of interaction effects were not easily read through statistical output. Accordingly, I graphed the marginal effects of inequality, based on the electoral system, in Figures 4.2 and 4.3.

The results in Table 4.3 suggest several important things for the rela- tionship between inequality and electoral systems in Latin America. First, the strong negative effect of regional inequality that I observed in all models is driven primarily by nationally oriented electoral systems. Figure 4.2 shows that in nationally oriented political systems (personal vote value is low), the effect of interregional inequality on government spending was significantly lower. Regionally oriented political systems (personal vote value is high) did not respond to regional inequality with lower spending—their spending was unchanged. Moreover, the overall

Figure 4.3 Marginal Effect of Interpersonal Inequality by Personal Vote Rank, Latin America

$ $

$

!

-0.6

-0.4

-0.2

0

0.2

0.4

0.6

0.8

1

1.2

0 1 2 3 4 5 6 7 8 9 10 11 12

E st

im at

ed E

ff ec

t o f M

ea n

G in

i

Personal Vote Rank

EBSCOhost - printed on 9/7/2021 10:30 AM via UNIVERSITY OF SOUTH ALABAMA. All use subject to https://www.ebsco.com/terms-of-use

70 Melissa Ziegler Rogers

effect of regional inequality to reduce spending was robust in all models. Second, the strong positive effect of interpersonal inequality on govern- ment spending in Latin America was also driven by those nations with nationally oriented electoral institutions, shown in Figure 4.3. In geo- graphically oriented systems, increased interpersonal inequality had no statistical effect on government spending.

What do these results suggest about the relationship between electoral systems and inequality on government spending in Latin America? First, it is clear that interpersonal and interregional inequality have divergent effects on government spending in Latin America and these effects differ depend- ing on a nation’s electoral system. The results in the Latin American sample suggest that nationally oriented electoral rules (in all except Brazil, Colombia, and recently Bolivia) depressed spending as regional inequality grew. These same nationally oriented systems push spending up to deal with increasing interpersonal inequality.

Essentially, there was a null result for geographically oriented political systems to changes in interregional inequality. This finding is important in itself because it suggests a particular dynamic in political bargaining: name- ly, of resolving political conflict with pork barrel spending. Where politi- cians in nationally oriented political systems feel little motivation to redis- tribute to poor regions, in geographically oriented political systems it appears that rich regions are not able to cut off the poor regions. The most straightforward explanation for this outcome may be in party bargaining. While most national political parties are diverse in regional representation, they are often more coherently arranged on the issue of interpersonal inequality and redistribution. More simply, parties form precisely on the basis of individual income redistribution with broad regional representation. Accordingly, matters of interpersonal inequality are most often dealt with across parties while interregional inequality is a distinctly intraparty as well as interparty matter. These bargaining dynamics cannot be examined in this chapter, but are considered in detail in Rogers (2016).

Conclusion

What do these findings mean for the quality of democracy in Latin America today? Overall, I argue that regional economic differences impede national responses to inequality, and this may undermine the qual- ity of democracy. Regions with distinct preferences have difficulty agree- ing on policy reform and this may result in policy stagnation, even on pressing concerns over income inequality. At the same time, there is some reason for hope that governments will make serious efforts to lower

EBSCOhost - printed on 9/7/2021 10:30 AM via UNIVERSITY OF SOUTH ALABAMA. All use subject to https://www.ebsco.com/terms-of-use

Inequality and Democratic Representation 71

income inequality. The quality of democracy in the region, and the parties representing lower income constituencies appear to be stronger than in the past. Nonetheless, significant structural barriers stand in the way of major redistributive reform in Latin America.

The results of my study suggest that regional inequality is an impedi- ment to redistributive government spending. Indeed, in complementary results, regional inequality is shown to be a strong predictor of lower fiscal redistribution, measured as the percentage decrease in the Gini coefficient that results from government taxes and transfers (Beramendi, Rogers, and Diaz-Cayeros, forthcoming). The theoretical mechanism that links regional inequality to lower redistribution is differences in policy preferences. Economically dynamic regions want different policies out of national gov- ernments than do stagnating regions, and they have trouble coming together on a common policy. Not surprisingly, therefore, Latin American govern- ments’ responses have been limited in comparative perspective.

Yet Latin America’s democracies have recently shown signs that addressing inequities is on their policy agenda. In the past decade, Latin American democracies have reduced income inequality and poverty to a significant degree through the efforts of their governments. Levy and Schady (2013) document how government programs to subsidize income to the poorest individuals in Latin America have resulted in substantial reductions in poverty (from 28 percent of the population below the pover- ty line in 1996 to 13 percent in 2011). Nora Lustig, Luis F. Lopez-Calva, and Eduardo Ortiz-Juarez (2013) show that higher and more progressive government transfers have contributed to reductions in income inequality in Argentina, Brazil, and Mexico. Some of the economic conditions that made these reforms palatable, such as a commodities boom that has put more money in government coffers, may be ephemeral. Nonetheless, the passage and implementation of these reforms reflects a potentially impor- tant political change in the region. Democratic leaders appear more atten- tive to the demands of the majority poor in the region than they have been in the recent past.

The sources of those changes, and the reasons for hope, come from improvement in the quality and functioning of many of Latin America’s democracies. After decades of concern that parties in Latin America were populist machines devoid of programmatic content, we now see increasing evidence that parties offer distinguishable platforms and voters are able to discern those differences (Baker and Greene 2011; Zechmeister 2006). This is particularly important for advancements in redistributive policies. For the interests of the relatively poor to be articulated in the political system, they need parties able and willing to press for their demands. The rise of leftist parties and improved income distribution in recent decades suggest that

EBSCOhost - printed on 9/7/2021 10:30 AM via UNIVERSITY OF SOUTH ALABAMA. All use subject to https://www.ebsco.com/terms-of-use

72 Melissa Ziegler Rogers

democratic politics has encouraged the representation of a broader segment of the population (Huber and Stephens 2012).

This positive outlook needs to be tempered by severe lingering inequali- ties in the systems of representation that have clear implications for the likeli- hood of redistributive policy change. In particular, legislative malapportion- ment and enclaves of subnational autocracy limit the voice of the poorest in many nations of Latin America (Gervasoni 2010a; Gibson 2005). Moreover, both are inextricably linked to the structure of regional inequality.

Latin America, especially Argentina and Brazil, is a region with substan- tial malapportionment (Samuels and Snyder 2001). Not coincidentally, those countries with uneven development are also those in which political represen- tation is often disconnected from the one person−one vote democratic ideal (Stepan 2004). Less populated and small regions have disproportionate politi- cal voice in these systems. This appears to be important for the distribution of interregional transfers and redistributive reform. For example, Martin Ardanaz and Carlos Scartascini (2013) show that nations with high malappor- tionment have low personal income tax collection. Low tax revenue and low progressivity in the tax structure limit redistributive options for governments. These authors argue that malapportionment allows antiredistributive elites to buy the support of politicians from less populated (and thus cheaper to buy) regions (see also Gibson 1996; Gibson and Calvo 2000).

Related to this, if regions are powerful actors in national politics, it is important who is representing the interests of those regions. Despite national- level democracy, some subnational regions continue to be governed by autocrats (Gervasoni 2010b). The quality of subnational democracy is much more variable than that of national democracy in Latin America (Gibson 2005). If politicians at the national level have incentive to treat all politics as local, and their local political environment is highly autocratic, then subna- tional autocracy may influence national democracy. One specific way this may occur is through decisions on interregional transfers.

While government support for the neediest individuals has often been lacking, many Latin American nations transfer considerable income across regions. In fact, intergovernmental transfers in Argentina, Brazil, and Mexico, as a percentage of GDP, are comparable to those found in highly redistribu- tive nations such as Germany (Beramendi, Rogers, and Diaz-Cayeros, forth- coming). Economically productive regions subsidize the spending of less pro- ductive regions through centralized taxation. However, these transfers have not led to convergence in regional income (Russo and Delgado 2000). Malapportionment and subnational autocracy help to explain the limited redistributive effects of intergovernmental transfers. Transfers are not pro- gressive, but reflect inequities in representation. Regional income is not a good predictor of regional transfers, but (low) population is a very strong pre-

EBSCOhost - printed on 9/7/2021 10:30 AM via UNIVERSITY OF SOUTH ALABAMA. All use subject to https://www.ebsco.com/terms-of-use

Inequality and Democratic Representation 73

dictor. Regional autocrats capture a high share of these transfers, which act like rentier resources (Gervasoni 2010a). Redistribution across regions thus reflects a specific political logic in many Latin American nations that is likely inimical to major initiatives to change existing levels of income inequality.

Notes

1. These behavioral assumptions have been questioned. See Ansell and Samuels (2010) and Dion and Birchfield (2010).

2. This is not to say that the preferences of individuals for redistribution are uncontroversial or straightforward. A lively debate in the US and comparative litera- ture is working to sort out the psychological and political economy motivations for individual preferences for government spending.

3. A summary of variables and sources is shown in Appendix 1. The sample of the central government spending is smaller because of missing data in the dependent variable. No central spending data are available for Ecuador.

4. All fifty countries cannot appear in the empirical results because of missing data in the independent variables.

5. The results are also robust to alternative measures of regional inequality— the regional Gini coefficient and the unweighted coefficient of variance.

6. An arithmetical transfer from rich to poor regions reduces inequality. 7. With care taken to address the challenges of the data, discussed more below. 8. For a thorough description of the calculation of the Gini coefficient and the

data collection process and methods, see Solt (2009). 9. Vote type = 1 if a vote for an individual is observationally equivalent to a

vote for a party such as in single-member districts. 10. Carey and Shugart (1995) show that district magnitude interacts with ballot

structure in proportional representation systems. As district magnitude rises in closed list systems, the personal vote declines. As district magnitude rises in open list systems, parties have more competitors from which to distinguish themselves, so the personal vote rises.

11. In additional models not shown here, I control for possible nonstationarity in the dependent and independent with moving averages in the inequality variables, and full estimations with five-year fixed and moving averages. A full empiric treat- ment of these data, including an instrumental variables approach, is available in Rogers (2016).

12. In the Latin American sample, in fact, the personal vote is negatively corre- lated with regional inequality. This may suggest that pork encourages regional con- vergence, or that regionally disparate societies purposefully choose nationalizing voting rules, or that they are unrelated.

EBSCOhost - printed on 9/7/2021 10:30 AM via UNIVERSITY OF SOUTH ALABAMA. All use subject to https://www.ebsco.com/terms-of-use

EBSCOhost - printed on 9/7/2021 10:30 AM via UNIVERSITY OF SOUTH ALABAMA. All use subject to https://www.ebsco.com/terms-of-use