Graduate School of Social Sciences
Ethnic Diversity and Public Goods in Bolivia
Testing the diversity debit thesis at the national, sub-national and individual level
Thesis
Msc. Political Science: International relations Name: F.J.R. (Floris) van Dort Student number: 10153292
E-mail: FJRvanDort@gmail.com Supervisor: Mr. A.F. (Alexandru) Voicu Second reader: Mr. Dr. S. (Seiki) Tanaka Date of submission: June 23, 2017
iii Abstract
The negative relationship between ethnic diversity and public goods provision (i.e. the diversity debit
thesis) has recently been subject to empirical and theoretical critique. This study tests the original
theory and the contesting explanations in Bolivia, an ethnically diverse country with poor public goods provision, which has not yet been analyzed in-depth. At the national level, the most-similar cases method is used for Bolivia and Paraguay, and finds that ethnic diversity can only serve as an explanatory factor when the historical context of the countries is considered. At the sub-national level, ordinary least squares (OLS) regression models are used on a newly created dataset, which contains geographic and demographic data of all 339 Bolivian municipalities (2012). This analysis finds a weak relationship between ethnic diversity and public goods provision. Ethnic diversity thus affects public goods provision at the sub-national level in Bolivia only to a very small extent. At the individual level, this study contributes to the existing literature by adding a new approach centered around the concept of institutional racism. Using OLS and logistic regression models on survey data of the Americas Barometer (2012), this study finds strong evidence of unequal provision of public goods for indigenous groups, which make up 24 percent of the population. This indicates the existence of institutional racism in Bolivia. In sum, the findings of this study call for the revision of all three levels of analysis of the diversity debit thesis.
Keywords: ethnic diversity, public goods provision, diversity debit, Bolivia, Paraguay, institutional racism
v Table of Contents
1. Introduction ... 1
1.1 Ethnic diversity and public goods ... 1
1.2 Academic relevance, research questions and methods ... 1
1.3 Bolivia: ethnically diverse and poor public goods provision ... 2
1.4 Structure of the thesis ... 3
2. Theoretical framework ... 4
2.1 Introduction ... 4
2.2 The concepts of ethnicity, ethnic diversity, and public goods ... 4
2.3 Diversity debit and its critique at the national level ... 8
2.4 Sub-national diversity: debit versus merit ... 11
2.5 Diversity and public goods at the individual level ... 14
3. Diversity debit at the national level ... 18
3.1 Introduction and research method ... 18
3.2 Different levels of public goods provision in Bolivia and Paraguay ... 19
3.3 Alternative explanations ... 20
3.4 Ethnic diversity: outcomes that require context ... 26
3.5 Discussion of results and sub-question ... 27
4. Ethnic diversity and public goods at the sub-national level ... 29
4.1 Sub-national level methods, data and model... 29
4.2 Discussion of results ... 32
4.3 Diagnostics and sub-question... 36
5. Ethnicity and public goods at the individual level ... 38
5.1 Individual level methods, data and models ... 38
5.2 Discussion of results ... 41
5.3 Diagnostics and sub-question... 44
6. Conclusion ... 46
6.1 Introduction ... 46
6.2 Ethnic diversity’s effects on public goods provision in Bolivia ... 46
6.3 Theoretical contributions ... 48
6.4 General implications and future research ... 49
7. References ... 50
1
1. Introduction
1.1 Ethnic diversity and public goods
Scholars have aimed to explain why some countries are stable and prosperous, providing their citizens high quality infrastructure, education, and healthcare, while others seem stuck in poverty and have a hard time supplying the most basic public goods as security, paved roads and sewers. Different branches in political science and economics all have their own explanations for what causes these great divides among states, such as their colonial history (e.g. Mamdani, 2001), culture (e.g.
Huntington, 1997), geographical conditions (e.g. Sachs, Mellinger and Gallup, 2001), civil wars (e.g. Collier, 2003), or corruption (e.g. Kaufmann, Kraay and Mastruzzi, 2009).
In this thesis I focus on ‘ethnic diversity’ as explanation for the level of public goods in countries. Prominent scholars in this field claim that there is a negative relationship between ethnic diversity and public goods provision (e.g. Easterly and Levine, 1997; Alesina, Devleeschauwer, Easterly, Kurlat and Wacziarg, 2003). Mechanisms suggested by these authors are that ethnic groups have divergent preferences, which prevent achieving consensus on what public goods to provide, and are unwilling to pay for public goods that are also enjoyed by members of other ethnic groups. This
diversity debit is considered as “one of the most powerful hypotheses in political economy
scholarship” (Banerjee, Iyer and Somanathan, 2005: 639).
Recently, the negative relationship between ethnic heterogeneity and public goods provision has been subject to criticism. Singh and Vom Hau (2016) argue that the thesis is overstated, and criticize its lack of historical and political context, which they believe to be essential for current public goods provision. Soifer (2016) claims that it is not ethnic diversity, but regionalism that explains levels of public goods provision. Economic importance of different areas in the country are in this sense regarded as omitted variable in the diversity debit thesis. Gisselquist, Leiderer and Nino-Zarazúa (2016) criticize the thesis because - although it is able to explain differences at the national level - its assumed logic does not hold when applied to sub-national levels. Gao (2016) argues that there is indeed a correlation between ethnic diversity and public goods, but that this is positive. She claims that, at least in in semi-democratic countries, political competition among ethnic groups results in better public goods. Officials who cannot take their power for granted must provide better policies and public goods than those who rely on automatic support from their majority ethnic group.
Both the supporters of the diversity debit thesis and its critics have called for case studies to learn more about the relationship between ethnic diversity and public goods. This thesis analyzes the effects of ethnic diversity on public goods provision in Bolivia, a country that has not yet been subject to an in-depth analysis. Bolivia is an ethnically diverse country with one of the lowest levels of public goods provision in South America. By studying the relationship between ethnic diversity and public goods provision in Bolivia, this thesis contributes to the academic debate about the diversity debit thesis, and provides more insight into the specific reasons of Bolivia’s poor provision of public goods.
1
1.2 Academic relevance, research questions and methods
In this thesis, I show that the current discussion about the diversity debit thesis can be structured at three levels of analysis: national, sub-national and individual. At the national level, the original diversity debit literature is questioned by scholars who claim that the theory and analyses are too simple: either not providing important context or missing important variables. At the sub-national level, the discussion is less theoretical and has a rather empirical character. Here the original diversity debit scholarship is contradicted by studies that find neutral or even positive effects of ethnic diversity on public goods provision. At the individual level, existing studies are centered on mechanisms of the diversity debit thesis. I contribute to the academic debate by adding a new approach to this level of analysis, focused on the effects of ethnicity for individuals. The existing literature at national, sub-national and individual level is only able to explain overall public goods provision in the unit of analysis: how public goods provision is negatively, positively, or not at all affected by ethnic diversity. These studies do not explain differences in public goods provision for white, black or indigenous individuals that live in the same city or country. Based on the concept of institutional racism, I provide a theoretical basis and a statistical approach for investigating whether individuals are disadvantaged or privileged in the provision of public goods as result of their own ethnicity, rather than the degree of ethnic diversity in their city or country.
The existing debate about the diversity debit thesis, and my own theoretical contribution, lead to the following research question: “To what extent does ethnic diversity cause poor public goods provision at the national, sub-national and individual level in Bolivia?” Each of these levels is discussed in its own chapter, with corresponding sub-question and matching research methods.
For the analysis of the diversity debit at the national level I compare Bolivia and neighboring Paraguay by using the most-similar cases method. These are two countries with similarities on variables that according to critics of the diversity debit are important, but with differences on ethnic diversity and public goods outcomes. By testing alternative theories on both Bolivia and Paraguay, I am able to detect whether it is just ethnic diversity that affects public goods provision, or if other causes are also present. The sub-question that will be answered is: “Does ethnic diversity
autonomously affect public goods provision at the national level in Bolivia?”.
The analysis of the diversity debit at the sub-national level will be executed in a statistical research design. I execute ordinary least squares (OLS) regression analyses on a newly created dataset with demographic and public goods data from all 339 municipalities in Bolivia, based on the National Census of 2012. I created this dataset, which is provided as supplement to this thesis, based on 339 separate PDF-files of the Bolivian census. The sub-question that will be answered is: “To what extent
does ethnic diversity affect public goods provision at the sub-national level in Bolivia?”.
The analysis of the effects of ethnicity on public goods provision at the individual level will also be executed in a statistical research design. I execute OLS and logistic regression analyses on
2 existing survey data from the Americas Barometer of 2012. The sub-question that will be answered is:
“To what extent does ethnicity affect public goods provision at the individual level in Bolivia?”.
1.3 Bolivia: ethnically diverse and poor public goods provision
Bolivia is one of the most ethnic diverse and poor countries in South America. According to a representative survey of the population in 2012, 77 percent of Bolivians self-identify as being of “mixed” white-indigenous origin, 17 percent as indigenous, and 6 percent as white (Americas Barometer, 2012). According to Canessa (2007: 199), racism and weakening community ties in Bolivia have led to a decline in self-identification as “indigenous” and cause people to identify as “mixed” instead. On the other hand, Canessa also finds signs of people that re-identify as indigenous, mostly as strategy to gain recognition and resources from the state or NGOs, especially since the election of “the world’s first indigenous president” Evo Morales in 2004 (Canessa, 2014: 156). Among urban intellectuals there is a trend to self-identify as indigenous, as political act, while other (perhaps more ‘traditional’) people regard that qualification as “primitive”, and therefore do not wish to be included in that category (2007: 204)1.
Approaching the population according to its native language, which is much less receptive for changes in self-identification, also shows an ethnically diverse Bolivia. In 2012, 76 percent of the respondents stated that Castellano (Spanish) was the language spoken at home when they grew up, for 16 percent this was the indigenous language Quechua and 7 percent learned their first words in the indigenous language Aymara (Americas Barometer, 2012). In addition, 36 other indigenous languages are officially recognized by the Bolivian state, but none of these are spoken by more than a few thousand people (Ministry of the Interior, Bolivia, 2017).
Bolivia has, especially when compared to other countries in South America, a low level of public goods provision. Measures that indicate public infrastructure, public health and public education show that Bolivia is one of the worst performing countries of the continent. Bolivia is the country that has the least number of people connected to the sewer system and has the highest infant mortality rate in South America (World Bank, UNICEF, & WHO, 2015). In addition to this, the poor living conditions in the country are illustrated by the fact that Bolivia has the highest extreme poverty rate on the continent (World Bank, 2015). These grim indicators make my thesis, in addition to its scientific relevance, also of much social relevance: being able to understand better what causes the poor public goods outcomes in Bolivia can help to improve policy. If there are clear signs of a diversity debit in Bolivia, solutions to minimize its negative effects can be designed, while indicators of institutional racism might provide disadvantaged groups a tool to address it, and policymakers insight where it occurs and who it affects.
1
Canessa (2007: 196) describes a meeting with a shaman, who only speaks Aymara and often gets into contact with ancestral spirits of the Incas, but denies to be indigenous. Canessa also argues that urban intellectuals or political activists, who do not speak Aymara and have a white skin (descendant of Europeans), are most-likely to self-identify as indigenous (2007: 209).
3
1.4 Structure of the thesis
In chapter 2, the theoretical framework, I start by discussing the three most important concepts of this thesis: ethnicity, ethnic diversity, and public goods. Thereafter I present the literature -both in favor and critical- of the diversity debit thesis, according to their corresponding national, sub-national or individual level of analysis. Since I use mixed research methods and different datasets, I present the methods for each level of analysis at the beginning of the analytical chapters. The diversity debit thesis at the national level is discussed in chapter 3. The diversity debit at the sub-national level is discussed in chapter 4. The effects of ethnicity on public goods provision at the individual level is discussed in chapter 5. In the final conclusion, in chapter 6, I answer my research question and discuss main findings, weaknesses and broader theoretical implications of this thesis.
Figure A: Levels of analysis, research questions and methods (presented as funnel)
National level (chapter 3)
Sub-question: Does ethnic diversity autonomously affect public goods provision at the national level in Bolivia?
Method: Most similar-method (Bolivia and Paraguay) Sub-national level (chapter 4)
Sub-question: To what extent does ethnic diversity affect public goods provision at the sub-national level in Bolivia?
Method: ordinary least squares (OLS) regression analysis Individual level (chapter 5)
Sub-question: To what extent does ethnicity affects public goods provision at the individual level?
4
2. Theoretical framework
2.1 Introduction
This chapter sets out the theoretical debate and main concepts of the diversity debit thesis. Many scholars within this debate jump right at measurements, variables and datasets without deliberating on the meaning of underlying concepts (e.g. Easterly and Levine, 1997; Gisselquist et al., 2016; Sing and Vom Hau, 2016). I start by taking one step back to discuss the concepts that are central in the
literature: ethnic diversity and public goods. After this brief conceptual discussion I present the literature of the diversity debit thesis. Agreeing with critics, such as Gisselquist et al. (2016: 309), who argue that theory and results of the thesis are often misused at different levels (findings at sub-national levels that are used to explain cross-country differences), I structure this theoretical framework along three levels of analysis: the national, sub-national and individual level.
The sub-sections about the national and sub-national level start by explaining the main literature and concepts in favor of the diversity debit thesis, and are followed by its critique. At the national level, the diversity debit arguments are critiqued by scholars who have alternative theories about public goods provision. At sub-national level, the discussion is not so much about alternative theories, but rather empirical: scholars find contradicting results. Since the aim of this thesis is to add to the academic debate by testing all existing theories that are presented in this chapter, my explicit objective is to remain neutral in the presentation of the existing literature. In addition, the critique that opponents of the diversity debit thesis have on the original literature is quite sharp, and comes best to its right in the clear structure of this chapter. After I tested all presented theories at the national and sub-national level in my research of Bolivia (chapters 3 and 4), I will be able to pick sides and make an original contribution to the existing debate.
In the section about the individual level I use a slightly different approach. I start to present the current literature at the individual level, which is centered on individual mechanisms of the diversity debit. After introducing this, I posit an alternative approach for the individual level, which focusses on the effects at the individual level, by making use of literature about institutional racism.
2.2 The concepts of ethnicity, ethnic diversity, and public goods
2.2.1 Ethnicity and ethnic diversity
Ethnicity and ethnic diversity are concepts that are used in a variety of ways. In this section I briefly discuss the appearance of ethnicity in the social sciences and hence offer a useful conceptualization for this thesis. Thereafter, I provide a brief discussion on the concept and mathematical calculation of ethnic diversity.
In political science, the umbrella-concept for ethnicity coined by Donald Horowitz (1985) is widely accepted and applied. He argues that ethnicity is about groups with “different skin-color, language, or religion”, and thereby “covers tribes, races, and nationalities” (Horowitz, 1985: 53). This
5 all-encompassing concept leads to divergent use of, and application of, ethnicity in academic (political science) studies. According to Marquardt and Herrera (2015: 689), this ambiguity on the concept leads to “unacknowledged theoretical implications”, because researchers’ understandings of ethnicity are often not represented in their operationalization and use of data. To limit the wide scope of the umbrella-concept, some scholars have aimed to demark the concept. Wimmer (2008) provides a typology of “forms” of ethnicity, which he bases on race, language, and religion. Each of these forms is applicable to other parts of the world. For example: race is useful in the United States (U.S.), language in Eastern Europe, and religion in the Middle East (Wimmer, 2008: 1028). I argue that, although this typology of ethnicity is constructive and useful for my conceptualization, Wimmer’s eagerness in applying these typologies to different parts of the world is too blunt. To find nuance, I turn to anthropological works that equally aimed to deal with the concept’s elusiveness
In anthropology, ethnicity has been widely understood in accordance with Barth (1969). For him, ethnicity is not so much about specific characteristics of people, such as skin-color or language, but rather how these characteristics are differentiated. Ethnicity is therefore mainly about the process of defining the boundaries between groups (Barth, 1969: 9-11). In a more recent (widely-cited) anthropological work, Eriksen (2001: 262) shares this view by emphasizing that ethnicity concerns “the relationship between groups whose members consider each other distinctive”. A similar stance is taken by Jenkins (2008), who argues that ethnicity is a highly political concept: “it is situationally defined, produced and reproduced in the course of interactions that occur at or across the ethnic
boundary in question” (Jenkins, 2008: 54, italics added). I regard these (anthropological) nuances as
constructive and use them for my conceptualization by emphasizing that context is necessary for deciding which typology is useful to capture ethnicity.
Taking into consideration the concept’s elusiveness, and sharing the concerns that leaving out a conceptualization has great theoretical implications (Chandra, 2006: 398), and conceptualizing it unclearly leads to problems with application and validity (Marquardt and Herrera, 2015: 689), I provide a demarked conceptualization of ethnicity. I define ethnicity as referring to shared
characteristics of people, based on language, religion, or racial background, that allow either members or non-members to perceive them as a distinguished group. The specific context is of great
importance for deciding which characteristics are used to distinguish groups. Ethnicity should therefore be captured by different indicators in different contexts. In this thesis, I therefore conduct a thorough examination of the historical context of the countries that are studied (Bolivia and
Paraguay), to decide which indicator(s) are useful to capture ethnicity.
The importance of the contextual element can be illustrated with examples of how each of the characteristics is deemed valid or invalid to capture ethnicity. In the U.S., differences between
Catholics and Protestants are not useful to capture ethnicity, because (hardly) no-one uses it to
distinguish ethnic groups according this branch of Christianity. In Northern-Ireland, with its history of violent religious conflicts, Catholic and Protestant characteristics capture ethnicity to a high degree,
6 because people distinguish each other accordingly. In many African countries, such as Uganda, where 43 tribal languages are spoken (Katushemererwe and Nerbonne, 2013: 113), language is an important ethnic characteristic. Contrasting, in Burundi and Rwanda, there can be little disagreement that Hutus and Tutsis should be considered separate ethnic groups, although they speak the same language (McDonnell, 2016; 1519). In the U.S.-context, the racial characteristic (black versus white) is a more valid lens to view ethnicity than language or religion.
For clarity, I follow the typology of Wimmer (2008: 1026), and make a division of three sub-concepts that each represent one of the indicators of ethnicity: the linguistic indicator, the racial indicator and the religious indicator. In the diversity debit literature, it is these indicators that are (often as synonyms for ethnicity) used to distinguish groups. Depending on the context, one (or a combination) of these indicators can be used to capture the concept of ethnicity. I present examples of these indicators based on the Bolivian case, and provide references to literature that use thesee indicators.
The linguistic indicator refers to groups with their own native language (in the case of Bolivia: Spanish, Quechua, Aymara, etc.). Linguistic indicators are used in the diversity debit literature by (among others) Easterly and Levine (1997), Alesina et al. (2003) and Gisselquist et al. (2016). The racial indicator refers to groups with a distinct race (in the case of Bolivia: white, mixed, indigenous etc.). The racial indicator is used in the diversity debit literature by (among others) Jackson (2013) and Gisselquist et al. (2016). The religious indicator refers to groups with a distinct (branch of) religion (in the case of Bolivia: Catholicism, Protestantism, nature/indigenous religions etc.). Religious diversity is used in the diversity debit literature by (among others) Alesina et al. (2003) and Gerring, Thacker, Lu and Huang (2015). Again, thorough examination of the Bolivian context determines which of these indicators are useful to capture the concept of ethnicity (chapter 4).
Rather than ethnicity, it is ethnic diversity that is the main independent variable of the diversity debit thesis. All authors - the adherents and critics - use fractionalization to measure ethnic diversity. Whether these studies use the linguistic, racial, or religious indicator, and whether they look at neighborhoods, cities, or states, the calculation is always configured according the Herfindahl index. Ethnic fractionalization is calculated as:
𝐸𝐸𝐸𝐸ℎ𝑛𝑛𝑛𝑛𝑛𝑛 𝑓𝑓𝑓𝑓𝑓𝑓𝑛𝑛𝐸𝐸𝑛𝑛𝑓𝑓𝑛𝑛𝑓𝑓𝑓𝑓𝑛𝑛𝑓𝑓𝑓𝑓𝐸𝐸𝑛𝑛𝑓𝑓𝑛𝑛𝑢𝑢= 1 − � 𝑞𝑞𝑔𝑔𝑢𝑢2 𝑁𝑁 𝑔𝑔=1
Where 𝑞𝑞𝑔𝑔𝑢𝑢 is the share of ethnic group 𝑔𝑔 in unit of analysis 𝑢𝑢. This index results in scores from 0 to 1, which indicates the probability that two randomly chosen individuals within a unit of analysis (e.g. state or municipality) are members of different ethnic groups (e.g. Easterly and Levine, 1997: 1206; Alesina et al., 1999: 1255; Jackson, 2013: 443; Gerring et al., 2015: 168; Gisselquist et al., 2016: 1312). In my analyses, the indicator used for fractionalization is equal to the best suited indicator of
7 ethnicity, and is therefore picked based on the in-depth analysis of the context (for Bolivia and
Paraguay, see chapter 3).
2.2.2 Public goods
Public goods is a concept that is also often overlooked in the diversity debit literature. Without discussing the underlying meaning, most of the articles jump straight at their variables and exact measures (e.g. Easterly and Levine, 1997; Gisselquist et al., 2016; Sing and Vom Hau, 2016). In this section I take one step back and discuss the concept more thoroughly.
Public goods are often divided into two kinds: ‘pure’ and ‘impure’ public goods. Pure public goods are non-rival and non-excludable, which means that people cannot be excluded from
consumption of the good and the consumption of one does not affect the consumption of others (Stiglitz, 1999: 308). Examples of these goods are clean air, the military who protects borders against foreign invaders, and dikes that prevent a country from flooding. Impure public goods are much wider defined, and fall in between pure public and pure private goods (Saylor, 2014: 17). The benefits of these goods are unevenly distributed and are not always available to the entire population. Examples of these goods are electricity networks, public transportation systems and public schools. Stiglitz (1999: 310) calls the public goods that are geographically limited “local public goods”. Adding another requirement for these impure, or local public goods, is that these must generate positive effects for society and (economic) development. Such public goods associated with economic growth and productivity are considered “core public goods” (Aschauer, 1989: 193), or “productive public goods” (Duranton and Déo, 1999: 266). Examples of these goods are airports, electricity networks, gas pipes, water systems, and sewers. In one of the most influential articles of the diversity debit thesis, Alesina et al. (2003) also use the term “productive public goods” as overarching concept in their analysis of public schools, roads, sewers and trash-pickup.
In this thesis, I use the concept public goods in its impure and productive meaning. Since the existing literature (and my own study) is about provision of public goods, the first requirement is that the good is provided by the state. Secondly, it must be considered productive, which means that it has a positive influence on (economic) development. This criteria is somehow vague, and to a high degree all-encompassing; therefore I limit my definition of public goods to three particular categories: public infrastructure, public education and public health2.
Examples of infrastructural public goods are paved roads and drinking water systems. Measuring their provision is straightforward and widely done in the diversity debit literature: kilometers of paved roads (e.g. Easterly and Levine, 1997: 1212), and share of population with
2
One could theoretically consider all goods to be productive since they could somehow have a positive impact on people, making them work harder or more efficiently. Duranton and Déo discuss this issue by giving the example of public parks, which could be considered productive public goods because they provide leisure, which eventually could have a positive impact on productivity (1999: 267). An in-depth discussion about this is beyond the scope of my thesis, in which I focus on clearly productive public goods which fit the categories infrastructure, education and health. These categories are not jointly exhaustive, but cover a wide range of productive public goods that are provided by the state. Variables that represent these categories of public goods are the used in the diversity debit literature, and the ones that I select for my own analyses.
8 connection to the drinking water system (e.g. Gerring et al. 2003: 167; Jackson, 2013: 447). The provision of public education and public health is slightly harder to measure. One can count schools or hospitals, but this does not tell which share of the population is provided health services, or which share of the population is being educated, and for how long. Therefore, these goods are often
measured based on their outcomes: for public health, the number of child deaths is often used as indicator of its (non) provision (e.g. Alesina et al., 2003: 173; Gerring et al. 2003: 173; Gisselquist et al., 2016: 312), while for public education literacy rates (e.g. Alesina et al., 2003: 173), or average years of education of the population (e.g. Jackson, 2013: 448; Gerring et al., 2003: 167; Gisselquist et al., 2016: 312) are often used.
2.3 Diversity debit and its critique at the national level
2.3.1 National level diversity debit thesis
The central claim of the diversity debit thesis is that ethnic heterogeneity leads to lower provision of public goods. One of the first and most cited scholars who discussed this causal mechanism were Easterly and Levine (1997). They argue that ethnic diversity is a major determinant for Africa’s poor performance, and that it explains cross-country differences in public goods, policies and economic growth. Ethnic diversity is conceptualized as the probability that two randomly selected individuals in a country belong to different linguistic groups (1997: 1206). According to their theory, societies that are “ethnolinguistically” diverse have more difficulty in reaching consensus, which results in lower public goods provision and worse economic policies (1997: 1241). Indicators of public goods levels are the number of telephones in a country, percentage of roads paved, efficiency of the electricity network and average years of schooling of the population. Easterly and Levine (1997: 1241) find a strong negative correlation at the state-level between ethnic diversity and public goods. High heterogeneity results in poor education, insufficient infrastructure and high infant mortality rates. Though the focus of the paper is on Africa, the authors argue that the results are not particular to the continent and they apply their thesis to explain differences in economic performance and public goods provision of states worldwide.
Alesina et al. (2003) argue that the concept of ethnic diversity should be about more than just the variety of languages spoken in a country. They broaden the concept of Easterly and Levine (1997) by adding racial and religious characteristics. Alesina et al. (2003) measure the chance that two randomly drawn individuals belong to different linguistic, religious or racial groups. They call this diversity-score “fractionalization”, which reaches its theoretical maximum (at value of 1) when each person belongs to a different group. Racial and linguistic fractionalization are negatively associated with public goods provision and government quality, while religious fractionalization is positively related (Alesina et al., 2003: 158). According to Alesina et al. these results are a confirmation of the diversity debit thesis. The positive correlation of religious diversity and public goods is found because
9 religious diversity tends to be higher in more tolerant and free societies, in Europe and the U.S., which also have high levels of public goods (2003: 158). This touches the sensitivity of the concept of ethnicity: it makes much difference how it is defined and which measure is used per country (language, ethnicity or religion). An example the authors give are blacks and whites in the U.S., or various groups in Latin America, all speaking the language of their former colonizer, but having different racial characteristics. In most of Africa on the other hand, the concept of ethnic diversity is better captured when measures of language are used, instead of racial characteristics. Although Alesina, et al. (2003) update the measures of ethnic diversity, they confirm the findings of Easterly and Levine (1997): ethnic diversity (or, at least linguistic and racial diversity) negatively affects public goods measures as infrastructure quality and public education.
Jackson (2013) finds a negative correlation between ethnic diversity and household access to water and electricity networks and whether children are attending school or not. He studies diversity based on self-identified race in eighteen countries in Sub-Saharan Africa. Jackson admits that
measuring ‘access’ is not equal to “provision” of public goods, but defends this choice by stating that “the decision to access the public good should reflect the benefit received from the good, relative to potential alternatives” (Jackson, 2013: 448).
2.3.2 Critique at the national diversity debit thesis
Singh and Vom Hau (2016) argue that it is essential to take a pause and re-examine the diversity debit thesis. According to them, ethnic diversity is important, but is the thesis overstated because it does not consider the role of history and politics: “In fact, a doggedly ahistorical perspective prevails in this scholarship” (2016: 1304). Contemporary diversity and public goods provision are better explained by the historical processes of nation-state formation, the authors argue. Past state intentions and
capabilities to provide public goods shaped both current patterns of ethnic diversity and state
provision of public goods. Therefore, the analytical lens should be extended back in time to the period after independence to investigate context (2016: 1305). Singh and Vom Hau (2016: 1308) present a theory where current levels of public goods are directly connected with three historical modes of nation-building: assimilation, accommodation and exclusion. States with an accommodationist strategy aimed to establish a multicultural political community by recognizing ethnic distinctions: universal public goods had to be available for all citizens, and distinct goods had to be provided based on preferences of ethnic groups (2016: 1308). Assimilationist states aimed to limit ethnic diversity and discouraged or prohibited the institutionalization of ethnic differences. Public goods as education or health care were provided on universal basis, often with implicit aim of establishing congruence between groups (2016: 1308). Exclusionist states limited national membership to the dominant ethnic group and targeted public goods only to them. In some contexts, basic and low-quality public goods
10 were also provided to non-dominant groups (2016: 1309)3. According to Singh and Vom Hau it is these historical nation-building strategies, in combination with capacity to execute them, that explains current public goods provision (see table 1). The main critique at the diversity debit thesis is therefore that it is not about how ethnically diverse countries are or were, but how the state engaged with these different groups that explains contemporary levels of public goods provision (2016: 1328).
Table 1: Public goods provision as result of nation-building strategy and state capacity (based on Singh and Vom Hau, 2016: 1307-1314)
Strategy: Accommodation Strategy: Assimilation Strategy: Exclusion
Strong state capacity
Supportive policies increase ethnic diversity. Incentivizes pressure by ethnic groups for distinct public goods, and collective pressure for universal public goods.
Example: Singapore. High overall public goods provision.
Decline of ethnic diversity due to efficient strategy to unify population in single national identity.
Example: France. High overall public goods provision.
Divisive state policies increase ethnic diversity and grievances. State is able to (violently) oppress non-dominant groups. Though most (and first-class) public goods are targeted to dominant ethnic group, excluded groups might have access to some goods.
Example: South Africa.
Medium/high overall public goods provision.
Weak state capacity
Supportive policies increase ethnic diversity. Incentivizes pressure by ethnic groups for distinct public goods. Due to limited capacity to deliver distinct and universal public goods, perceptions of unfairness arise. Collective pressures increase public goods provision over time.
Example: India.
Medium overall public goods provision.
Deficiency of state-power results in favoritism towards dominant ethnic group and increases levels of ethnic diversity.
Example: Mexico. Medium overall levels of public goods provision.
Divisive state policies increase ethnic diversity and stimulate recurring violent conflict. State not able to effectively oppress non-dominant groups, or provide public goods to any group.
Example: Togo.
Low overall public goods provision.
Historical processes of national-building are also regarded as main explanation for current levels of public goods provision by McDonnell (2016). Looking at Africa, she argues that initial strategies adopted by first post-independence leaders are a decisive factor for public goods provision (2016: 1514). In states with a “modest plurality” (one ethnic group that is large enough to dominate but without ethnic proximately sized ethnic groups) diversity encouraged provision of public goods (2016: 1514). Building on Singh and Vom Hau (2016) she blends the assimilationist and
accommodationist nation-building strategies and comes up with the concept of “conciliatory states”. This strategy recognizes ethnic differences, but avoids favoritism and only provides universal public goods (2016: 1515). According to McDonnell, states that followed this strategy are now the most stable and prosperous ones in Africa. Her main problem with the diversity debit literature is the deterministic negative effect of ethnic diversity. McDonnell argues that it is not diversity per se, but
3
An example of this is South Africa during the Apartheid-era (1948-1991), where first-rate health care was available for whites, while non-dominant groups had some access to basic health care (Singh and Vom Hau, 2016: 1314).
11 rather how the state engaged with its heterogeneous population, that determines provision of public goods (2016: 1518). Additionally, she claims that the operationalization and quantification of ethnic diversity is so context-specific, that many problems occur when grouping (supposedly similar) groups in umbrella categories (2016: 1519)4.
According to Soifer (2016) it is regionalism that determines variation in public goods
provision. This concept is defined as the existence of multiple large and economically important urban centers within a country (2016: 1343). Regional demographic and economic independence and
importance is regarded as omitted variable in the diversity debit thesis (2016: 1324). Regions that are self-sufficient, and separately integrated into the world market, have a smaller dependence on the central state. This makes that there is fewer to gain from public goods that transform the country towards one unitary national unit; thus, state leaders will not invest in these goods and rather support regional economies to flourish (Soifer, 2016: 1346). Support for this argument is provided by
evidence from countries in Spanish South America which, according to Soifer, are especially suitable for studying causes of public goods provision levels due to their similarities: “they have some core properties in common, including common colonial background, broadly similar political economies, and common absence of the sorts of wars and territorial conflicts that promoted state-building and public goods provision elsewhere” (Soifer, 2016: 1357). Therefore, these can be ruled out as sources of variation. Soifer argues that public goods provision is higher in states in Spanish South America that had a single dominant urban center, and is lower in states with multiple large cities (2016: 1342). He studies data of regionalism of the earliest post-independence date available, to rule out that public goods provision affected regionalism5.
2.4 Sub-national diversity: debit versus merit
2.4.1 Sub-national diversity debit
Alesina et al. (1999) argue that diversity debit also exists at the sub-national level and present an analysis of local-level public goods provision in the U.S. The authors assume the same causal
mechanism as the one applied at the national level: heterogeneity of preferences among ethnic groups leads to lower public goods provision (Alesina et al., 1999: 1243). Like the state-level articles, the authors again do not clearly conceptualize what public goods are. As long as they are provided by the state, considered “productive” and in line with preferences of citizens, public goods can anything (1999: 1248)6. When it comes to measuring them, the article is much clearer: ranging from levels of education of the population to frequency of trash collection. Ethnic diversity is measured as
fractionalization, using the same calculation as Easterly and Levine (1997), though they use ethnic
4 For example, the grouping of Hutus and Tutsis in Rwanda and Burundi by linguistic ethnic diversity measures (McDonnell,
2016: 1519).
5 For example, by drawing population to the capital city to make use of available the public goods (Soifer, 2016: 1358). 6
12 self-identification instead of language (as used by Easterly and Levine). Alesina et al. defend this break in the least extensive way possible: “These classifications are somewhat arbitrary, but they reflect which ethnic groupings are politically salient” (1999: 1255). It is not a hard guess why they use self-identification, however; it is evident that language (or religion, as did scholars who published later) is not a useful indicator to investigate differences between two major groups in the U.S.: Afro-Americans and whites7. Once again, this proves the salience and context-specify of choosing fitting indicators of ethnicity. The core theoretical assumption of Alesina et al. (1999: 1246-1251) is that a (sub-national) society consists of heterogeneous groups, which are formed by members with
homogenous preferences and only care about wellbeing of members of their own group. When voters have different preferences, they want to pull fewer resources together for public projects, and
subsequently want less public goods when a significant fraction of tax revenues collected on their own ethnic group is used to provide public goods shared with (or only provided for) other ethnic groups (1999: 1244). To illustrate this theoretical model, the authors tell an anecdote about public education funds in Oakland, California. Inner-city Afro-American parents want “black English” to be
recognized as separate language, so their children with severe language-deficiency get access to special programs for non-native speakers. Hispanic parents argue that the recognition of “black English” is an unjust effort to enable Afro-Americans to grab bilingual funds. Asian parents, in their turn, complain that Hispanic children get more bilingual resources than their children. Many whites object to the diversion of resources to any non-standard English instruction. Alesina et al., argue that this struggle leaves all ethnic groups dissatisfied, and finally results in overall less spending on public education than if the city would have had a more homogenous population (1999: 1251). The article presents results of an analysis in U.S. cities, metropolitan areas and urban counties. To be sure that it is ethnic diversity that causes public goods provision, the authors control for various socioeconomic and demographic indicators that are also connected to development and public goods (1999: 1259): The income of citizens, to make sure that it is not simply richer cities that have more public goods, or that it is a matter of ‘income inequality debit’ rather than an ethnic diversity debit. The size of the city, to be sure scale factors are not causing the levels of public goods. Education of citizens, because educated citizens might demand better goods, especially when it comes to public education. Age structure, because of large groups of old people could also affect public goods demands, this time most likely leading to less education (1999: 1257). The main finding of the study, in line with their theoretical model, is that sub-national communities with higher ethnic diversity, this time based on self-identified race, have lower levels of public goods provision, even after controlling for their socio-economic and demographic situation.
7
Alesina et al. (1999: 1255) do encounter problems with the ‘Hispanic’ category. Many (according to the authors) Hispanics self-identify as ‘other’ because they “do not feel accurately represented in the multiple racial choice provided by the census”.
13 2.4.2 Critique at the sub-national diversity debit
Trounstine (2015) revisits the model of Alesina et al. (1999) and argues that it is not just diversity, but segregation along racial lines that leads to lower public goods provision. Segregation, the degree to which groups are disproportionally distributed across geographic space (2015: 710), is measured based on waterways that flow through cities and physically divide parts of cities. Trounstine’s conclusions are a clear response to the situation in the U.S., and only applicable there: “Racially segregated American cities are also politically polarized cities, making collective investment more challenging and public goods expenditures lower” (Trounstine, 2015: 709). Analyzing election- and finance data from U.S. cities, she finds that segregated cities, rather than just ethnic diverse cities, raise fewer dollars from their residents and spend less on roads, parks, sewers, welfare and housing (2015: 720).
According to Gisselquist, et al. (2016), even if the methods and results of Alesina et al. (1999) are correct, these do not say anything about other parts of the world: “To the best of our knowledge,
no similar findings have been replicated outside of the U.S. and particularly not in Sub-Saharan Africa
where their hypothesis is so often applied” (Gisselquist et al., 2016: 310, italics added). Because of the unique governance system and ethnic relations in the U.S., the findings are not externally valid, they argue. Gisselquist et al. (2016) do find a negative correlation between ethnic diversity and public goods at the national level, but a positive one at the local level. Therefore, their main critique on the diversity debit thesis is that its logic is falsely used to explain both variation in public goods at the national and sub-national level. “Studying variation in provision of education and health services at the municipal level (as a function of municipal-level factors alone) would be potentially misleading in a country where education on health policies are highly centralized at the national level, or where decisions about education are made primarily at the local level and those about health policy at the state level” (2016: 309). This highlights a straight-forward, but important context-specific factor of the debate: it is of much importance that ethnic diversity is measured at the same level as where the public goods are provided. In their case study of the sub-national level in Zambia, Gisselquist et al. (2016: 319) find that ethnic diversity (both racial self-identification and linguistic fractionalization) is positively related to welfare outcomes. They regard this as evidence of either diversity driving improved outcomes, or improved outcomes driving higher diversity. According to the authors, future research should therefore shift focus why ethnic diversity undermines public goods, but instead when, and under what conditions it does so (2016: 321).
Gerring, et al. (2015) also find evidence for the existence of a diversity debit at the national level, and contradicting outcomes at the sub-national level. They conducted their research with survey data from 36 developing countries. Ethnic diversity is measured as (self-identified) racial and
religious fractionalization. Unlike most other scholars, Gerring et al. do explain their choices for operationalizing ethnicity according to self-identification. “Self-identification presumably reflects the subjective nature of ethnic identity as it exists in a particular time and place” (2015: 167). The authors
14 state that they do not make assumptions about what causes ethnicity to be defined in certain ways, and admit that it is possible that different measures could lead to different findings, though they doubt that those would not overturn overall conclusions of the study (2015: 167). “These categories reflect current understandings in a country. We make no assumptions about the permanence of these
categories; indeed, we suspect substantial variability across countries and through time” (2015: 167). The authors also discuss the definition of the dependent variable in the diversity debit thesis, which leads to a slightly different formulation than in other articles. Gerring et al. look at human
development, arguing that if ethnic diversity has consequences for public goods, then those should also be reflected in these measures. Their measures include fertility rates of women, child mortality rates, levels of education and wealth. “These are easier to measure, as well as being of greater ultimate importance than many of the other outcomes of the original diversity debit literature” (2015: 166). None of the tests in 36 developing countries show a negative and statistically significant correlation at the sub-national levels between ethnic or religious fractionalization and human development, making the diversity debit thesis either false or negligible at the sub-national level (Gerring et al., 2015: 174).
The study of Gao (2016) of Jordanian municipalities finds the strongest evidence for a positive relationship between ethnic diversity and public goods at the sub-national level. Arguing that in semi-democratic countries, with a weak political system, ethnic diversity can be an important source of electoral competition that leads to better public goods (2015: 1373). Electoral completion is higher if political entities have a diverse constituency. If leaders have to compete for votes of people outside of their ethnic group (because only members of their own group are too few to win elections), this leads to election of more competent public officials who provide better public goods (Gao, 2016: 1377). According to this study, ethnic diversity therefore has a positive effect on public goods provision at the sub-national level in countries with a weak or semi-democratic system.
2.5 Diversity and public goods at the individual level
2.5.1 Individual mechanisms of diversity debit
Existing individual level research on the diversity debit thesis is mainly concerned with its micro mechanisms. Central question is how ethnic diversity affects public goods provision.
Habyarimana, Humphreys, Posner and Weinstein (2007) study mechanisms through which diversity debit works. They employ experimental games with 300 citizens from an ethnic diverse neighborhood in Uganda’s capital Kampala to compare policy preferences and cooperation among ethnic groups, while controlling for age, education, income and gender (2007: 720). According to the study, there is little variation in public policy preferences (such as security and garbage collection), but is there better cooperation between members of the same ethnic group (2007: 724). Rather than shared preferences, it is in-group reciprocity norms that lead to better outcomes: when individuals believe that their behavior is observed by others, they put more effort in their tasks (2007: 724). The
15 existence of norms and institutions therefore, “police” the defection of non-contributors, which leads to more efficient outcomes (Habyarimana et al., 2007: 724). This finding contradicts expectations of Alesina, et al. (1999) who put heterogeneity of preferences central in their model. Gerring, et al. (2015: 174) criticize the study of mechanism in Kampala, by arguing that it is easy to design experiments that measure distrust, but much harder to measure possible positive aspects of ethnic diversity such as innovation. Lieberman and McClendon (2012) also find contradicting evidence in their survey study of individuals in 18 Sub-Saharan African countries. According to their research, ethnicity is strongly associated with policy preferences (2012: 589): which they do find to vary based on ethnic group membership. According to Lieberman and McClendon, heterogeneous societies are therefore likely to provide less public goods. Just as at the national and sub-national level, the discussion at individual level is ongoing and far from solved.
2.5.2 New approach: effects of ethnicity at individual level
The existing diversity debit literature at national, sub-national and individual level all follow the same line of reasoning. Both defenders and critics of the thesis aim to explain whether, why or how ethnic diversity has a negative, positive, or no effect on public goods provision in the unit of analysis that they are studying. These studies do not explain differences in public goods provision for these distinct ethnic groups within the unit of analysis. At the national level, for example, Easterly and Levine (1997) find that countries with more diversity have less public goods, but do not explain how these goods are –equally or unequally- distributed among these ethnic groups. Singh and Vom Hau (2016) do discuss the historical distinctions between dominant and non-dominant ethnic groups, but they solely discuss how this effects current overall national public goods provision. Singh and Vom Hau argue that, depending on historical nation-building strategies pursued by the state, distinct ethnic groups have been provided different public goods. However, Singh and Vom Hau do not engage in discussing what this means for these ethnic groups’ current public goods provision. At the sub-national level, Gao (2016) finds that Jordanian municipalities with more diversity have more public goods, but she neither looks at the distribution of these goods towards ethnic groups within the municipality. At the individual level, Habyarimana et al. (2007) find that group-mechanisms lead to higher efficiency, which serves as mechanism to explain better performance and higher public goods provision of more ethnically homogenous municipalities or states. Again, this study does not consider differences in public goods provision for distinct ethnic groups, or individuals with a certain ethnicity, within these units of analysis, but only explains variation in overall public goods provision of the unit of analysis.
Existing diversity debit literature is therefore unable to explain differences in public goods provision for white, black or indigenous people that all live in the same city or country. Based on the concept of institutional racism, I provide a new approach for the individual level diversity debit literature. The aim of this approach is to find the effects of ethnicity in the provision of public goods
16 for individuals, rather than individual mechanisms that explain overall public goods provision at an aggregate level.
The concept of institutional racism first appeared in the book Black Power (1968) of Afro-American civil rights activists Stokely Carmichael and Charles V. Hamilton. They argue that racism is deeply embedded in U.S. society and is part of everyday life. Institutional racism relies on anti-black
attitudes of inferiority, even if individual whites do not actively discriminate against individual blacks (Phillips, 2011: 173). According to Williams (1985), institutional racism is the “cumulative
disadvantage experiences across interrelated welfare experiences (housing, education, employment and so on), produced through institutions’ routine operations, regardless of intentionality of individual actors” (quoted in Phillips, 2011: 177). Although institutional racism is mainly used to explain
inequalities in the U.S., the concept has also been applied in the United Kingdom. After the racist murder of a young black man by a policeman in 1993, the British government ordered a report on institutional racism in the U.K. In this report institutional racism was defined as “the collective failure of an organization to provide an appropriate and professional service to people because of their color, culture, or ethnic origin” (MacPherson, 1999: 28). According to Gillborn, a prominent British scholar of the concept, is institutional racism the “historical discrimination against minority ethnic groups that is institutionalized so that even when contemporary barriers to equality are removed, persistent inequality remains” (2008: 64).
Henry, Houston and Mooney (2004: 517) researched institutional racism in public health care of Australia and argue that the system is “institutionally racist” for Aboriginals. Due to worse public health provision, these people have a life-expectancy that is much lower than that of white
Australians. They refer to institutional racism as “the ways in which racist beliefs or values have been built into the operations of social institutions in such a way so to discriminate against, control and oppress various minority groups” (2004: 517). Came (2014) argues that institutional racism is
embedded in the public health system of New Zealand as well. Uneven provision and access of public health services contribute to health inequities among Maori people (2014: 214).
Aligning the basic elements of the definitions used by scholars who study institutional racism, I argue that the concept is also applicable in non-Western contexts. Central elements are (a history of) oppression and dominance by one ethnic group over others, and a persistent disadvantage for non-dominant ethnic groups, even when inequalities are formally removed. Although the theory is mostly applied in the West, these are surely not unique features of the American, British or Australian reality. An integration of institutional racism in diversity debit thesis implies lower levels of public goods provision for members of non-dominant ethnic groups, compared to members of a (historically) dominant ethnic group, even after controlling for ‘regular’ diversity debit effects8 and the other usual
8
I.e. the effects of the diversity debit (at national or sub-national level) on overall provision of public goods in the city or country that the individual lives in.
17 socio-economic and geographic controls. Unequal provision of public goods for non-dominant ethnic groups is investigated by measuring the levels of public goods provision (dependent variable) for individuals from dominant and non-dominant ethnic groups (independent variable). Although this means that the independent variable is different than at the national and sub-national level, were I analyze ethnic diversity, this is highly connected because dominant and non-dominant ethnic groups can only exist if there is ethnic diversity.
Although mechanisms of institutional racism should be analyzed by using qualitative research methods, statistical research methods can detect unequal provision for non-dominant groups. Unequal provision implies “a persistent disadvantage for non-dominant ethnic groups” and is therefore regarded as indicator of institutional racism.
18
3. Diversity debit at the national level
3.1 Introduction and research method
In section 2.3, I showed that critique on the presumed negative relationship between ethnic diversity and public goods provision at the national level is (mainly) theoretical. In this chapter, I analyze whether arguments of both adherents and opponents of the diversity debit thesis are able to explain the outcomes in Bolivia. The sub-question that will be answered is: “Does ethnic diversity
autonomously affect public goods provision at the national level in Bolivia?”. In order to answer this
question, I use the most-similar cases method, which is described by political scientist John Gerring in his book ‘Case Study Research’ (2007). I compare Bolivia with neighboring Paraguay, a country that is very similar in important ways, but has much higher levels of public goods provision, the dependent variable in this study.
In its purest form, the most-similar method uses a pair of cases that are similar on everything, except the variable of interest (Gerring, 2007: 131). Ideally, these studies start with the observation of two (or more) cases, that at first glance seem identical, but end up differing in outcome. This is clearly the case for Bolivia and Paraguay, that for example share colonial history, geographic conditions, are both at peace, and have similar levels of perceived corruption. According to Gerring (2007: 131), a thorough analysis of the cases can reveal one -or several- factors that differ across the cases, and hence explain the different outcomes.
Gerring considers the similarities between the cases “control variables” (2007: 132). In my research these are the alternative explanations, which attribute low provision of public goods to causes that are similar in Bolivia and Paraguay. Gerring (2007: 133) argues that it is not necessary to explain these ‘variables’ extensively, since they simply cannot explain the different outcomes: “It is not usually necessary to measure control variables (at least not with a high degree of precision) in order to control for them. (..) One can simply assert that, whatever they are, they are more or less constant across the two cases” (Gerring, 2007: 133). I follow this method, providing concise argumentation of the alternative theories’ essence, and subsequently why they fail to explain the low public goods provision in Bolivia compared to Paraguay. The final alternative theory that I discuss, the historical processes of nation-building (Singh and Vom Hau, 2016), is able to explain the different outcomes and is therefore discussed more extensively.
In section 3.2, I show that Bolivia and Paraguay differ on public goods provision. Section 3.3 provides a discussion of the alternative theories. Section 3.4 provides an analysis of the diversity debit thesis in both countries. In the concluding section 3.5, I discuss why the theory about historical processes of nation-building and the diversity debit thesis have to be integrated in order to be able to interpret ethnic diversity-scores and derive meaningful conclusions, this discussion answers my sub-question.
19
3.2 Different levels of public goods provision in Bolivia and Paraguay
This section shows, based on the most recent World Bank data, that compared to Paraguay, Bolivia performs worse on public goods provision. I include six measures for the three categories of productive public goods that I presented in section 2.2 (infrastructure, public health and public education). The justification for the selection of these indicators is threefold. First, each of them clearly measures one of the three specific productive public goods categories. Second, all indicators are used in previous studies of adherents and critics of the diversity debit thesis. Third, the World Bank only offers a selective number of variables for Bolivia and Paraguay, which limits the selection possibilities. Especially the lack of multiple indicators for public health and public education is a weakness. Still I regard the variables that are presented in table 2 as a profound image of the level of public goods provision in both countries.
For infrastructural public goods, table 2 shows that almost all Paraguayans have access to an improved water source, while this is only the case for nine out of ten Bolivians. Access to improved sanitation facilities is secured for nine out of ten Paraguayans, but only for half of the Bolivians.
For public health, the table shows that, on average, the chance that Bolivians die before the age of five is double compared to Paraguayan children. Bolivian mothers are twice as likely to suffer pregnancy related death. The life expectancy for the people of Bolivia is four years shorter than their neighbors in Paraguay.
For public education, the table shows that in Bolivia and Paraguay, a similar share of the population is able to read and write a short story about activities in their daily life. This difference in adult literacy rate is 0.4% in the advantage of Paraguay.
In sum, the six indicators that are presented in table 2 give a clear image of lower overall public goods provision in Bolivia compared to Paraguay.
Additionally, I also present three income indicators, which I do not consider as measures of public goods and therefore do not deserve much attention. The reason to include the variables is that Easterly and Levine (1997) argue that, in addition to public goods provision, economic growth and income levels are also negatively affected by ethnic diversity. The data obtained from Bolivia and Paraguay confirm this theory. What might be confusing however, is that both the adherents and critics of the diversity debit thesis at the sub-national and individual level, consider income and poverty-levels as control variables for public goods provision. I follow the existing literature (at all three poverty-levels of analysis) by regarding the income indicators at the national level as dependent variable, and at the sub-national level and individual level as control-variables. The presentation of GDP per capita, share of population below the poverty line, and share of population that lives in extreme poverty in table 2 serves to give complete openness, and give the reader a better understanding of Bolivia and Paraguay.
20
Table 2: Public goods and income indicators Bolivia and Paraguay
Bolivia Paraguay Source
Population with access to improved water source
90% 98% World Bank, WHO & UNICEF (2015a)
Population with access to improved sanitation facilities
50% 89% World Bank, WHO & UNICEF (2015b)
Child mortality rate (before age 5).
38/1,000 21/1,000 World Bank, WHO & UNICEF (2015c)
Maternal mortality rate (during or as result of pregnancy)
206/100,000 132/100,000 World Bank & WHO (2015)
Life expectancy 69 years 73 years World Bank & United Nations Population Division (2015) Adult literacy 95.1% 95.5% World Bank & UNESCO (2015) GDP per capita (in current
USD)
3,076 USD 4,081 USD World Bank (2015a)
Population below national poverty line
38.6% 22.2% World Bank (2015b)
Population in extreme poverty (less than 1.90 USD per day)
6.8% 2.8% World Bank (2014)
* The full description of the indicators is provided in Appendix 1.
3.3 Alternative explanations
This section tests alternative theories that aim to explain differences between economic development and public goods provision in different countries, regarding the cases of Bolivia and Paraguay. As suggested by Gerring (2007: 133), I do not extensively discuss the theories that point to ‘variables’ that are equal in both countries. This section shows that all the alternative explanations that are discussed in the introduction and theoretical framework of this thesis fail to explain the lower level of public goods provision in Bolivia compared to Paraguay, except for Singh and Vom Hau’s ‘historical processes of nation-building’ (2016). This last alternative explanation is therefore discussed more extensively.
3.3.1 Colonial history
According to Mamdani (2001), the colonial history of countries is connected to a high degree with their current economic and political performance. Bolivia and Paraguay were both Spanish colonies from the early sixteenth century till the early nineteenth century. In both, the Spaniards set up extractive institutions, centered at depriving them of their resources (Acemoglu and Robinson, 2013: 11, 19 and 115). Absolute power was in the hands of a small elite, whose main aim was to obtain as
21 many recourses as possible, for their own and the Spanish crown’s benefit. Since the Spanish colonial empire had a similar policy in Bolivia and Paraguay, current differences in levels of public goods in the countries cannot be explained by this shared colonial history9. Therefore, theories that connect the colonial history of countries to their current level of development and public goods provision are unable to explain the differences in Bolivia and Paraguay.
3.3.2 (Civil) war
According to Collier (2003), war, and more specifically: civil war, is an important hindrance on economic development, state capacity and thus public goods provision. Even though the causal relationship between development and (civil) war works in both ways, it is indisputable that violence and destruction wreak havoc on public goods provision. The Upssala Conflict Data Program (2016) shows that in both Bolivia and Paraguay, interstate wars as well as civil wars are absent. The available data for both countries was collected from the year 1989 until 2016. It reports zero casualties in Paraguay, and fifty in Bolivia which are qualified as non-state violence. The absence of (civil) war dismisses it as a viable independent variable to explain levels of public goods provision.
3.3.3 Corruption
Corruption is negatively associated with economic development, state capacity and thus public goods provision (Kaufmann et al., 2009). Although corruption is, almost by design, hard to measure, indexes have been created to somehow capture the phenomenon. The Corruption Perception Index is an often-used indicator, which measures society’s perceptions of corruption and combines this with expert assessments (Transparency International, 2016). Corruption-scores are determined for 176 countries, of which a classification is published. Lower-ranked countries are perceived as more corrupt, higher-ranked countries (closer to #1) are perceived as less corrupt (Transparency
International, 2016). As table 3 (on the next page) shows, over the last five years citizens of Bolivia perceived (slightly) less corruption compared to citizens of Paraguay. The 113th rank of Bolivia and 123th rank of Paraguay indicate that the countries perform (compared to the rest of the world and South America) quite similarly, which indicates that corruption is not fit to serve as an independent variable explaining different public goods levels in both countries. When one does put emphasis to the slightly better performance of Bolivia, it even contradicts the theory that corruption negatively affects development and public goods provision.
9 The Spanish colonies in South America were characterized by extractive political and economic institutions, which main
purpose was – as the term suggests- to extract resources. Inclusive institutions on the other hand, which developed for example in British colonies in North America distributed power more broadly in society, making it harder to extract resources for the benefit of a small elite (Acemoglu and Robinson, 2013: 81-82).
