The Eэect of Inequality.

The E¤ect of Inequality, Democracy, and Economic Development on Institutions: A Dynamic Panel Study

Nadezhda V. Baryshnikova and Maria Monica Wihardja y

University of Adelaide and University of Indonesia

We study how inequality, democracy, and economic development a¤ects government stability, investment pro…le, the level of corruption, and bureaucratic quality using a dynamic panel data across 58 countries from 1998 to 2006, controlling for population size, natural resource endowment, and con‡icts/tension. Savoia, Easaw and McKay (2010) argue that inequality is detrimental to institutional variables because inequality directly advances rent-seeking behaviors of political and economic elites that perpetuates the unequal distribution and political power. Meanwhile, inequality works indirectly through democracy since voting rights are granted to poorer citizens who demand redistribution. Using the Arellano-Bond (1991) 2-step GMM estimator with a heteroskedastic-consistent error correction by Windmeijer (2005) to control for endogeneity biases, we …nd that inequality has a negative e¤ect on investment pro…le but a less negative e¤ect in a more democratic country. The direct e¤ect of polity, a measure of democracy, on investment pro…le is mixed. There is a strong evidence that the level of development has a positive e¤ect on government stability and investment pro…le, which supports the theory of vicious cycle (Myrdal, 1944). Moreover, natural resource endowment has a negative e¤ect on investment pro…le, which suggests the phenomenon of a resource curse. These e¤ects are either non-robust or insigni…cant on the level of corruption and bureaucratic quality. Policy interventions on enhancing investment climate and ensuring political stability on the one hand, and increasing the level of development and reducing inequality, on the other hand, are therefore crucial in breaking the vicious cycle.

Keywords: endogenous institutions, inequality, democracy, dynamic panel.

Address for correspondance: Nadezhda Baryshinkova, School of Economics, University of Adelaide, Adelaide, SA 5005; Email: nadezhda.baryshnikova@adelaide.edu.au; Phone: +61 8 8303 4821; Fax: +61 8 8223 1460

yor Maria Monica Wihardja, Department of Economics, University of Indonesia, Indonesia. Email: mariamonicawihardja@gmail.com. Phone; +62-81-21220313. Fax: +62-21-5365-4607.

2

1. Introduction

Using a dynamic panel data, we look at cross-country evidence of how inequality, democracy, and GDP per capita may a¤ect qualities of institutions - government stability, investment pro…le, corruption, and bureaucratic quality - using an empirical model suggested by Savoia, Easaw, and McKay (2010), controlling for natural resource endowments, population, and military in politics/ethnic tension/religious tension/internal con‡icts/external con‡icts. Savoia, et al. (2010) asserts that inequality a¤ects quality of economic institutions (de…ned as a set of rules constraining incentives) directly through the persistent rent-seeking policies and indirectly through democracy. In areas with high inequalities, the rent-seeking behaviors of political and economic elites perpetuate the unequal distribution of wealth and political power. Democracy indirectly a¤ects economic institutions through voting rights granted to poorer citizens demanding redistribution. In other words, democracy promotes greater participation of the voters that serves as a whistle blowers to poor-quality institution, while inequality works in the opposite direction namely greater inequality translate into weaker economic and political power of the poor and therefore their voices for better institutions are not heard of. The possibility of resource curse in which natural resource abundance results in more rent-seeking behavior and hence lower institutional qualities is also studied in our paper. Moreover, this paper will study how the level of development, proxied by GDP per capita, a¤ects institutional quality. A number of studies show that a good institution delivers economic prosperity, including IMF (2005) and Kau¤man and Kray (2010). But, what actually shapes institutions remains a question. Institution is often misunderstood to be an exogenous factor that a¤ects economic performance and other factors with little thoughts given to the endogeneity issues on the formation of institutions. The theory of endogenous institution contends that institution is endogenously determined and not exogenously given, by showing factors that a¤ect the quality of institution. This theory may help us understand why some countries are "trapped" as a low middle-income countries with poor institutions.

The de…nition of institution is broad and multiple. Plott (1979) equate the sum of preference, institutions, and physical possibilities to outcomes. Schotter (1981) de…nes institutions as the alternative equilibrium standards of behavior or conventions of behavior that evolve from a given game described by its rule. Schotter contends that institutions are about the properties of the equilibrium of games and not properties of the game’s description. They are about what the agents do with the rules of the game, not what the rules are. Reformulating Plott’s equation, Ostrom (1986) equates the sum of rules, physical laws, behavioral laws, and model of a decision maker to outcomes. The most wellknown de…nition of institution probably comes from North (1990) that de…ne institution as rules of the game consisting of formal rules, informal constraints, and enforcement. Similarly, Savoia et al (2010) de…nes institutions as a set of rules to constrain incentives, and the absence of these rules will generate undesired outcomes. More recently, an economic historian, Greif (2006), sees institution as a system that consists of rules, beliefs, norms, implied behaviors, players/organization, in equilibrium.

3

We will de…ne institution as that of Schotter (1981) and Greif (2006): it is a system in equilibrium. This allows us to study the strategic behaviors of individuals given the incentive and informational structures of the societies. However, Ostrom’ equation agrees in substance with Schotter’s and Greif’s de…nition of institution. Ostrom’ second equation (1986) emphasizes two things: what we want to study is how rules change the incentive structure leading to a change in outcomes, and these rules change the incentive structure given that physical and behavioral laws are invariant. In practice, we will use some governance indicators as proxies to institutional qualities.

There are two interesting properties of institution namely endogeneity, which is reiterated by Myrdal’s reinforcing cycle of Negro oppression and White prejudice (1944) and North’s address on endogenously determined incentive structures in slow-growth economies (1990), and indeterminacy. In what he calls "the theory of the vicious cycle," Myrdal argues that white prejudice and discrimination keep the Negro oppressed and more oppressions in turn exacerbate white prejudice. Moreover, he argues that the original change of either white prejudice or Negro oppression could set a di¤erent future trajectory that spirals either upward or downward. North (1990) wrote, "What makes for e¢ cient markets? If poor countries are poor because they are the victims of an institutional structure that prevents growth, is that institutional structure imposed from without or is it endogenously determined or is it some combination of both?" (p.134-135).

Endogeneity means that not only does institution a¤ect welfare but also welfare a¤ects institutions through, among others, behavioral and physical law. This latter hypothesis is also what we study in this paper namely how the level of development proxied by GDP per capita a¤ects institutional qualities. Being hungry and is unable to eat because a poor individual could not a¤ord buying food would a¤ect this individual’s preferences between rice and the right to vote. Educated and uneducated citizens will have di¤erent capacities in participating in a democratic system. There are a lot more examples to these. A more rigorous analytical framework of how welfare could a¤ect preferences, choices, and implied behaviors of individuals is called a set-dependent choice framework. Suppose a household has a choice of saving some income for future higher education of the children, spending some income on better education of the children now, or saving some income for food in times of crises. Given that the head of the household cares about the future of the children but also cares about having enough money for food in times of crises, he might choose to spend some income on better education for the children now. Now, suppose the same household has a choice of spending some income on better education of the children now or saving some income for food in times of crises. In this latter case, the household does not perceive any opportunity for their children to attend higher education and hence, saving some income for future higher education is not in their choice set. Given only the two choices, He might choose not to send their children to school at all. It might seem inconsistent with what he chooses in the prior case, but a set-dependent choice framework agrees that people’s choice and preference depend on the choice set. This story tells us that how the “almost poor”and the “very poor”people may react very di¤erently to a government poverty program because their choice sets are not completely identical.

In regards to inequality and democracy, we may as well argue that institution may evolve with inequality and democracy, just as it does with welfare. Political and economic institutions that are captured by a few political and economic elites while leaving the rest

4

of the society poor may perpetuate unequal distribution of wealth, which in turn, giving more de facto power to the elites to capture those institutions in the next period, and so on, regardless of the de jure political regime (Acemoglu, 2008; Acemoglu and Robinson, 2008). This evolution is socially unstable in the long term and may result in a revolution, such as what happens in Tunisia, Egypt, Libya, Bahrain, Yemen, and Syria in beginning of 2011, where people revolted against decades of dictatorship regimes, ignited by some worsening socio-economic conditions.

Indeterminacy in the context of institutions means that there is no unique outcome that comes out from a certain incentive structure. Indeterminacy arises because of multiplicity of equilibria - the same set of rules, payo¤s, and players may result in multiple outcomes of implied behaviors by the players. This also means that two di¤erent institutions can arrive at the same outcome without necessarily having the same institutional trajectory. This is why studying institutions must involve both deductive and inductive methods done iteratively. A lot of times, it will involve also historical analysis. Indeterminacy of institutions make traditional social science di¢ cult because neither deductive nor inductive analysis alone is su¢ cient to explain the complexity of institutions (Greif, 2006). Moreover, institutions are context-speci…c and there is no "one-size-…ts-all" remedy to improve an institution.

Future works in this direction may focus on individual country-level panel studies across states, districts, municipalities, and provinces.

2. Literature Review

The study of endogenous institution goes back to Myrdal (1944) and more recently North (1990). More recent literatures link together more explicitly political democracy, distribution of wealths, and economic welfare, which intertwine resulting in persistency of institutions. Acemoglu (2008) discusses the reasons of why a dysfunctional institution continues to persist, and relates to the theory of endogenous institutions. He argues that distributions of resources and initial political institutions a¤ect the de facto and de jure political power respectively that a¤ects the economic institutions and political institutions of the next period. Economic institutions and political institutions further a¤ect the economic growth and the distributions of resources in the period after next. This framework illustrates the evolution between distribution of resources and institutions. Acemoglu and Robinson (2008) show an equilibrium that is called captured democracy. The de facto investment on political power by the elites is high enough that the economic institution is captured by the elites despite democratic political institution. Savoia, et al. (2010) argue that inequality a¤ects quality of economic institutions (de…ned as a set of rules constraining incentives) directly through the persistent rent-seeking policies and indirectly through democracy. In areas with high inequalities, the rent-seeking behaviors of political and economic elites perpetuate the unequal distribution of wealth and political power. Democracy indirectly a¤ects economic institutions through voting rights that are granted to poorer citizens demanding redistribution.

Resource-curse phenomenon has been studied by a series of literature. Leite and Weidmann (2002) study the extent to which natural resource abundance, trade openness, rule of law, and political instability a¤ect corruption, and hence economic growth. By using a

5

Two-stage Least Squares, they show that natural resource abundance increases opportunities for rent-seeking behavior, while trade openness, rule of law, and political stability reduces corruption. In the growth regression, they …nd that natural resource abundance tends to reduce long-run growth rate and long-term growth is negatively a¤ected by the level of corruption. Busse and Groning (2011) analyze the impact of natural resource abundance on governance indicators using a dynamic GMM panel estimator introduced by Blundell and Bond (1998). They show that natural resource abundance leads to an increase in corruption, but not for other governance indicators. In this paper, we will study using the same econometric model and control variables as that of Busse and Groning (2011), adding inequality, democracy, and tension/con‡ict as extra control variables. Acemoglu, Johnson, and Robinson (2002) shows a reversal of fortune that weigh against a view linking economic development to geographic factors. Instead, they show that reversal took place through institutional reversal resulting from European colonialism. They argue that Europeans were more likely to introduce good institutions in previously poor regions, encouraging investment and allowing industrialization to take place.

There is little empirical study that has been done on the e¤ect of both inequality and democracy on institutions. Keefer and Knack (2000) estimate how land and income inequality a¤ect property rights security, controlling for regime type. However, they conduct the estimation using OLS that fails to deliver unbiased and consistent estimates because of the possible endogeneity of institutions. Other papers focus on one of the variables, either democracy or inequality alone (Savoia et al., 2010). Our paper will contribute to the literature by analyzing the e¤ects of democracy and inequality on various institutional indicators controlling for endogeneity biases.

Societal participation, social pressure, and accesses to information to voters are studied in a number of literatures. Von Luebke (2009) discusses the importance of local leadership and societal pressures in shaping local governance. He argues that variations in local governance are better explained by local leadership than it is by societal pressure. The importance of participation level in curbing corruption is ingeniously captured by Serra (2008), who conducts an experimental study on bribery games that show that combining bottom-up and top-down monitoring on corruption the is most e¤ective way, even if the institution is weak. Trust-based, informal institutions may also help to attenuate institutional constraints (Della -Giusta, 2008). Providing access of information to voters, thus avoiding a narrow ‡ow of information going to only speci…c interest groups, is necessary (Economic Review, RIETI, 2009).

Ivanyna and Shah (2010) argue that four widely used indicators including the World Bank’s Worldwide Governance Indicators and Overseas Development Institute’ Word Governance Assessments are lacking citizen-based evaluations. This paper provides a framework for governance quality measurement across countries and overtime based on citizens’evaluations.

In terms of policy implication, Myrdal (1944) wrote, "a rational policy will never work by changing only one factor" (p.77). This is particularly true in the case of an endogenous system. In order for the system to work well, existing institutions and socioeconomic conditions must reinforce each other. Moreover, Greif (2006) writes that,"...rather than focusing only on helping countries specify rules, it will have to seek to change organizations, beliefs, and intertransactional linkages" (p.403). Central government must be

7

3.Log of GDP per capita is used as a welfare measure Source: World Development Indicator Database and CIA World Factbook.

4.Endowment is GDP share of natural resource exports. Natural resources include mineral fuels (SITC section 3), ores and metal (27), metalliferous ores (28), and nonferrous metals (68). We compute Endowment using the following equation:

Endowment = RES MES

where RES is resource exports as a share of total merchandise exports, MES is merchandise exports as a share of GDP. These, in turn, are computed

as

RES = (F E + OE)=100

MES = ME=GDP

where F E is Fuel Exports (as % of merchandise exports), OE is Ores and metal exports (as % of merchandise exports), ME is Merchandise export (in current US$), GDP is GDP (in current US$), taken from the World Development Indicator Database. Source: World Development Indicator Database.

5.Mineral Depletion is the product of unit resource and the physical quantities of minerals and energy extracted. It refers to bauxite, copper, iron, lead, nickel, phosphate, tin, zinc, gold, and silvers as well as crude oil, natural gas, and coal. It is expressed as a percentage share of Gross National Income (GNI). Source: World Development Indicator Database.

6.Log of Population is used to control for population size. Source: World Development Indicator Database.

7.Internal Con‡ict assesses political violence in the country and its actual or potential impact on governance. It consists of three subcomponents namely civil war/coup threat, terrorism/ political violence, civil disorder. Source: International Country Risk Guide.

8.External Con‡ict assesses both the risk to the incumbent government from foreign actions, ranging from non-violent external pressure (diplomatic pressures, withholding of aid, trade restrictions, territorial disputes, sanctions, etc.) to violent external pressure (cross-border con‡icts to all-out war). It consists of three subcomponents: war, cross-border con‡ict, foreign pressures. Source: International Country Risk Guide.

9.Military in Politics assesses military involvement in politics. Source: International Country Risk Guide.

8

10.Religious Tension assesses various religious tension from the domination of society and/or governance by a single religious group that seeks to replace civil law by religious law and to exclude other religion from the political and/or social process; the desire of a single religious group to dominate governance; the suppression of religious freedom; the desire of a religious group to express its own identity, separate from the country as a whole. Source: International Country Risk Guide.

11.Ethnic tension assesses the degree of tension within a country attributable to racial, nationality, or language divisions. Source: International Country Risk Guide.

The summary statistics of the dependent and independent variables as well as control variables are given in Table 1.

4. Model

We test a version of the model proposed by Savoia et al. (2010) that asserts that inequality a¤ects quality of economic institutions directly through the persistent rentseeking policies and indirectly through democracy.

yit = i + t + 1yit 1 + 1giniit + 2polityit + 3giniit polityit + 4lgdpit + xit + wit +"it

where:

yit is one of the governance indicators (government stability, investment pro…le, corruption, and bureaucratic quality)

i is a country …xed-e¤ect

t is a time-speci…c e¤ect, which include changes in the governance variables over time xit is a set of control variables (population, endowment or mineral depletion, inter-

nal/external con‡icts, military in politics, religious tension, ethnic tension)

wit is a set of method and survey dummies for Gini (see the Data section above) "it are the errors

We take all variables to be endogenous except for population size (and endowment/mineral depletion, for a robustness check), and take the …rst di¤erence and use past levels lagged two periods of more as instruments for the endogenous variables (e.g. y(t-1)-y(t-2) is instrumented with y(t-3), y(t-4), etc., and x(t)-x(t-1) is instrumented with x(t-2), x(t-3), etc.). We use the Arellano-Bond (1991) 2-step GMM estimator, with a heteroskedasticconsistent error correction by Windmeijer (2005). Sargan test of overidentifying restrictions is used to check the appropriateness of the instruments. The Arellano-Bond test is used to test second-order serial correlation. To avoid over…tting the endogenous variables, we reduce the number of instruments to be below the number of countries. For some of the variables, we add the second lag of the dependent variable to avoid second-order serial correlation.

9

5. Results

Table 2-21 show the regression coe¢ cients and their signi…cance. The coe¢ cients in these tables are robust and pass the Sargan and Arellano-Bond tests (with the Sargan and Arellano-Bond p-values being greater than 0.05). The main results suggest that the level of development (proxied by GDP per capita) has a positive e¤ect on government stability and investment pro…le and this is consistent throughout di¤erent control variables namely internal and external con‡icts/ military in politics/ religious tension/ ethnic tension (Table 2 to Table 11). These are robust throughout regressions with endowment or mineral depletion. This con…rms our hypothesis on endogenous institutions. However, this result somehow contradicts the Kau¤man and Kraay’s (2010) empirical study that suggests that the level of income negatively a¤ects governance. While this paper suggests the existence of "vicious" and "virtuous" cycles, Kau¤man and Kraay’s result suggests the absence of those cycles.

Inequality has a negative e¤ect on investment pro…le, but it has a less negative e¤ect in more democratic countries in both regressions with endowment (in this case, with military in politics, Table 7) and mineral depletion (in this case, with ethnic tension, religious tension, and internal con‡icts, Table 8-10). This con…rms the hypothesis proposed by Savoia et al. (2010) that inequality may exacerbate the rent-seeking behaviors of political and economic elites and perpetuate unequal distribution of wealth and political power. Rent-seeking behaviors are certainly harmful to investment climate, including pro…t repatriation. However, this negative e¤ect is smaller if a democratic political regime can guarantee people’s representation, who may serve as whistle blowers to a bad government and demand for a more equal distribution.

Endowment has a negative e¤ect on the investment pro…le throughout di¤erent control variables namely internal and external con‡icts/ military in politics/ religious tension/ ethnic tension, which suggests a "resource curse" phenomenon (Table 7-11). This result is consistent with Busse and Groning (2011) and Leite and Weidmann (2002), although our …nding shows no signi…cant e¤ect on corruption but on investment pro…le. As argued in both papers, natural resource abundance increases the opportunities of rent-seeking behaviors, including obtaining concessions to mining areas or permits to exploit forest areas whose issuances are controlled by a few political elites in the government. The existence of such costly rent-seeking behaviors is damaging to contract viability, pro…t repatriation, and other elements of investment pro…le.

Moreover, fewer internal con‡icts improve government stability (Table 5). More military involvements in politics means better investment pro…le (Table 7). This may be surprising but one may perceive military involvement in politics as a way to secure and protect certain business interests.

The e¤ects of polity and gini on corruption are both positive when religious tension is included (Table 14). The positive e¤ect of gini on corruption (more unequal society means less corruption) is surprising, but this result is not very robust since the signi…cance goes away in the robustness check regression with mineral depletion. Moreover, less religious tension means less corruption (Table 14).

In the robustness check regression with mineral depletion, there are some evidence that polity has an e¤ect on investment pro…le, although this e¤ect is mixed being positive

10

in some regressions and negative in others (Table 8-10). This may not be surprising since there are many anomalies with respect to the e¤ect of autocracy - for example, Singapore has thrived under an authoritarian political regime, and so has China, while autocracy in some of the Middle East and other Northern African countries may be blamed for poor socio-economic conditions. Moreover, as mentioned earlier, in a captured democracy, a democratic political institution may coexist with an economic institution that is captured by elites. Hence, the direct e¤ect of democracy on institutional indicators is ambiguous.

Other coe¢ cients are either insigni…cant, non-robust, or both. From these results, we …nd that only investment pro…les are most responsive to changes in inequality, polity, GDP per capita and endowment although government stability is better with a higher level of development. We contend that other institutional variables are more dependent on other factors excluded in our regression speci…cations. For example, corruption and bureaucratic quality may depend more on personal integrity rather than economic incentives such as higher payments. From the results, we can argue that reinforcing factors behind the vicious cycles between low welfare and low institutional qualities may not be corruption nor bad bureaucracy but lower investment pro…le and government instability.

One may ask why some of the lagged dependent variables are not signi…cant. Only the lagged two periods variable on government stability is signi…cant while only lagged one period variable on corruption is signi…cant. Meanwhile, both lagged one and two periods variables on investment pro…les and bureaucratic quality are signi…cant. We argue that some institutional indicators are more "persistent" than others while others are more prone to abrupt shocks. It makes sense that bureaucracy and investment pro…les are institutional qualities that cannot change very rapidly, while government stability and corruption in the government are more random in a sense that one big political con‡ict or one big corruption case in one year can make government stability and corruption indicator to rise rapidly, etc.. In other words, lagged dependent variables only a¤ect "structural" parts of institutions, such as bureaucracy and investment pro…le, but they do not a¤ect parts of institutions that are more abrupt to changes, such as government stability and corruption.

We try the same data on di¤erent years namely 1984-1997 using the same empirical model, and surprisingly, we …nd no signi…cant coe¢ cient. We argue that there are a lot of structural changes on these years on some of our sample countries, including the fall of the Soviet Union. Further researches on this is encouraged.

6. Conclusion

We study how inequality, democracy, GDP per capita a¤ect institutional variables controlling for endowment, population, and military involvement in politics/ethnic tension/religious tension/internal con‡ict/external con‡ict. We test the hypothesis proposed by Savoia et al.(2010) that inequality a¤ects quality of economic institutions de…ned as a set of rules constraining incentives directly through the persistent rent-seeking policies and indirectly through democracy. Based on the theory of endogenous institutions that contests that not only do institutions a¤ect economic performance but also the level of development a¤ects institutional variables, we test whether GDP per capita as a proxy