Regime Responsiveness to Basic Needs: a Dimensional Approach

Regime Responsiveness to Basic Needs: A Dimensional Approach Alexander Korolev, Ph.D. Research Fellow National University of Singapore The final publication is available at link.springer.com: http://link.springer.com/article/10.1007%2Fs12116-015-9209-z For citations: Alexander Korolev, “Regime Responsiveness to Basic Needs: a Dimensional Approach,” Studies in Comparative International Development, Vol 51, Issue 4 (December 2016): 434-455, (DOI 10.1007/s12116-015-9209-z)

Abstract Despite a considerable amount of research over the last three decades, an unequivocal conclusion regarding democracy’s impact on social outcomes has not been reached. This paper attempts to enhance understanding of the impact of political regimes on social outcomes by applying a dimensional approach. Unlike previous studies, which have focused on the overall effect of democracy, this paper separates the dimension of elite competition from the dimension of popular participation and tests their relative effects on the satisfaction of basic needs. Cross-national statistical tests demonstrate that effective participation has a positive effect on need satisfaction, whereas excessive competition has a negative impact. Theoretical explanations of these different impacts are provided. It is argued that the best way to understand the relationship between democracy and social outcomes is to realize that democracy’s overall effect might conceal the existence of opposing effects of its component parts. This finding suggests more nuanced ways of reforming political systems that bypass the possible trade-off between democratization and social development. Introduction What types of regimes are most responsive to people’s basic needs? From the standpoint of the dichotomous classification of political regimes (democracy vs. no democracy), democracies are expected to be better promoters of human well-being. The global shift toward democracy under the “third wave” of democratization has solidified this conviction. Democracy has come to be seen not only as a good thing in itself but also as a promoter of general welfare. Numerous empirical studies have found that democracy has positive effects on actual welfare outcomes, such as health, education, and life expectancy (Besley & Kudamatsu, 2006; Boix, 2001; Franco, A´lvarez-Darde & Ruiz, 2004; Frey & Al-Roumi, 1999; Gerring, Thacker, Alfaro, 2011; Ghobarah, Huth & Russett, 2004; Lake & Baum, 2001; Lena & London, 1993; McGuire 2010, 2013; Moon & Dixon, 1985; Safaei, 2006; Wickrama & Mulford, 1996; Young, 1990; Navia & Zweifel, 2000), or that it may mitigate the adverse effects of other factors (Khaleghian, 2003; Shandra et al. 2004). Thus, investigations of the types of regimes in which basic needs have been met and the types of regimes in which the attainment of basic needs has lagged tend to find that basic needs are better met in democratic regimes. The logic of this argument is based on the idea that democratic institutions, mainly through competitive elections, create incentives for politicians to adopt the policies necessary for better social services. In contrast, non-democracies fail to motivate policy makers to take people’s needs seriously. 1

There are at least two factors that prevent social scientists from fully embracing this assumption. First, several large-sample cross-national studies find weak or no links between democracy and social outcomes or make counterintuitive conclusions. For example, Williamson (1987), using a sample of 80 developing countries, discovered that democracy has no effect on physical quality of life. Weede (1993) argued that democracy has almost no positive effect on the Human Development Index (HDI) and no effect on life expectancy. Gauri and Khaleghian (2002) revealed that, except among very poor countries, democracies have lower immunization coverage than autocracies. Keefer (2002) studied low-income countries and found that poor democracies perform similarly to poor non-democracies (sometimes better, sometimes worse) with respect to enrollment and school completion rates as well as six different indicators of health status. McGuire’s (2005, 2006) work on infant and child mortality reached rather uncertain conclusions regarding the effects of democracy. The arguably most challenging criticism of the theory of superior democratic performance was made by Ross (2006), who concluded that the positive effect of democracy on infant and child mortality is the result of systematic underrepresentation of high-performing nondemocratic regimes in widely used datasets. The second factor is the levels and trends of human development. In terms of infant mortality rate (IMR), for example, Belarus performs as well as Switzerland (4 and 4.1, respectively); the United Arab Emirates outperform the United States (6.1 and 6.5, respectively); and Cuba does as well as the United Kingdom (4.6 per 1000 live births in both countries). The HDI of Kazakhstan is higher than that of Brazil (0.745 vs. 0.718), and China far outperforms India (0.687 vs. 0.547). Global trends of human development have also demonstrated that all states, regardless of the nature of their political regime, have made good progress in social development over recent decades. Among the top HDI movers are China, Indonesia, and South Korea. Nepal, Oman, and Tunisia are countries where progress has also been remarkable. Ethiopia, Botswana, Benin, and Burkina Faso enter the top 25 (UNDP 2010). Basic needs have been met in a variety of political regimes and at different stages of economic development; South Korea, Singapore, post-reform China and Taiwan, the former Yugoslavia, Sri Lanka, Costa Rica, Cuba, and other countries all fare comparatively well and manage to satisfy basic needs within a relatively short time despite considerable differences in terms of the overall “democraticness” of their political regimes. The inconclusiveness of the academic debate and contradictory qualitative evidence poses a challenge to the theory that democratic regimes result in better social outcomes and leads us to assume that democracy’s impact is, in fact, much more complex and nuanced than current theorizing suggests. Specifically, different dimensions of democracy may pull in different directions in terms of their effects on basic needs satisfaction, and an overall effect of democracy may conceal different, and possibly opposite, effects of its constituent components. To date, existing studies have introduced different control variables and located democracy differently among other independent variables. They, however, have only used summary indices of democracy, such as Polity III or IV, Freedom House scores, or the Bollen index of democracy, which combine things as categorically different as freedom of speech and government regulatory capacity. Such a strategy would be helpful only if we had solid knowledge that all dimensions aggregated in an index interact and reinforce one another in their effect on government responsiveness to basic needs. Otherwise, using a single composite measurement would be counter-productive, if not misleading, for obtaining an accurate understanding of democracy’s impact on basic needs satisfaction because it does not allow a researcher to determine the relative impacts of different dimensions. In this context, the present study reconsiders the effects of democracy on social outcomes by applying a dimensional approach. Instead of asking whether democracy is good 2

or bad for social outcomes, this study asks which aspects of democracy are good or bad for social outcome. More precisely, this study tests and explains the separate effects of the two most basic procedural dimensions of democracy: competition and participation. If it is true that democracies perform better, is this because of competition or participation? What are the relative roles of the two? The Analysis below demonstrates that only one dimension – public participation – is beneficial for social outcomes. Excessive inter-elite competition, in turn, can hinder needs-oriented policies. This finding has substantial policy implications, especially for those transition countries that are at the early stages of democratization. The paper is organized as follows. Section 1 problematizes the relative impact of competition and participation on basic needs satisfaction. Section 2 elaborates the analytical strategy and conducts statistical tests. The analysis begins with cross-sectional regressions at one point in time as a first cut in the exploration of association between the variables of interest, and continues with the exploration of how the associations obtained cross-sectionally hold under time-series cross-section regression analysis. Section 3 presents the results of the analysis. Section 4 suggests theoretical explanations of the results. Section 5 concludes. 1. Competition, participation, and social outcomes According to the concept of polyarchy in Dahl’s theoretical framework, any political regime can be distinguished along the continuum of “contestation” and “inclusiveness” (Dahl 1971: 4). The former refers to the extent of political competition and opposition, whereas the latter denotes public participation in political life. According to Dahl (1971), these two dimensions work together to ensure the goal of democracy: “responsiveness” of leaders to non-leaders. However, causal links between democracy and basic needs satisfaction are complex, and different aspects of democracy may have different effects. An evaluation of the relative roles of competition and participation provides preliminary support for this assumption. 1.1. Political competition and social outcomes Competitive elections are widely regarded as the main mechanism linking democracy to better social outcomes. The pressure of competition for votes should produce a situation in which political elites are accountable to the citizenry and, therefore, are attentive to broad welfare issues. However, this assumption is supported by little evidence. It is surprisingly difficult to come to a conclusive answer regarding the effects of political competition on basic needs provision. Discussing elections as an institution in a broad sense, Ginsberg (1982: 102-106), demonstrates that authoritarian governments with a limited coercive capacity appear to exhibit, at least on the basis of one indicator, a level of responsiveness to their citizens’ needs that is greater than or equal to that of the electorally competitive nations. His research demonstrates, for example, that spending on healthcare and education by democracies is proportionally smaller than in authoritarian regimes with a limited capacity to forcibly suppress public dissent. He argues that this pattern implies that electoral competition can delimit mass influence in political life. Keefer (2007), in turn, notes that electoral competition and politicians’ concerns with vote maximization in young democracies lead to spending targeted at organized political constituents rather than broad welfare spending that enhances the livelihood conditions of population at large. McGuire (2010) argues that even in a mature democracy electoral incentives do not invariably promote the provision of mortality-reducing social services. He observes that one of the candidates for governor in the US state of Mississippi competed in the November 2003 elections on a platform of imposing more burdensome requirements on applications for Medicaid and the state’s Children’s Health Insurance Program. This candidate won, despite the fact that a large number of eligible voters in Mississippi are poor. As a result of the policies the newly-elected governor, the IMR for non-whites in Mississippi rose from 14.2 per 1000 in 2004 to 17.0 per 1000 in 2005. 3

McGuire notes this as a stark reminder that democracy may not lead to the provision of mortality-reducing social services and that politicians can sometimes win electoral competitions on anti-needs platforms (McGuire 2010: 308). We can also take a closer look at the quality of government literature. Rothstein (2011) provides the example of Jamaica and Singapore, which were similar with regard to social well-being before the independence but very different thereafter. By most measures, Jamaica has always been a more competitive regime than Singapore. In the 2007 Economist Intelligence Unit’s index of democracy, Jamaica scored 9.17 out of maximum of 10 for political competition, whereas Singapore had a despicable score of 4.33. However, compared to Singapore, Jamaica lags far behind in terms of prosperity, the alleviation of poverty, and standard measures of population health. The hope that electoral competition would produce good social outcomes was not fulfilled. In other words, the presence of high-level competition in Jamaican political regime failed to generate regime responsiveness necessary for considerable improvement of social outcomes. These observations call into question the role of competition in generating political impetus for basic needs satisfaction. When a positive correlation is observed between democracy and social outcomes, it is not clear whether it is because of competition or other dimensions of democracy the positive effect of which suppresses the actual negative effect of competition. 1.2. Political participation and social outcomes The relative effect of the second basic characteristic of democracy – public participation – on social outcomes is also not clear. The minimalist form of participation (voting) tends to have a positive impact on basic needs provision. McGuire (2010: 145) provides empirical evidence of electoral participation’s direct positive effect on the adoption of mortality-reducing policies. In Argentina, which demonstrated a slow infant mortality decline, electoral participation by the poor was limited. As a result, both the Peronist political party Partidod Justicialista (Party of Social Justice) and its main rival, Unión Cívica Radical (Radical Civic Union), were motivated to enact policies that responded to the needs of the middle classes and some segments of the working class, leaving the country’s least advantaged citizens unattended. In contrast, in Costa Rica in the early 1970s, political participation was higher among the poor than among the rich. Arguably due to this fact, the Costa Rican government introduced a range of basic health services, which guaranteed a conspicuous decline in the mortality rate in the country (McGuire 2010: 65). The impact of participation on social outcomes becomes unclear with regard to other ways via which citizens can participate in political life. On the one hand, there is evidence of a positive impact. Wong’s (2004: 8-10) work on welfare politics in Korea and Taiwan argued that democratization in Korea in the late 1980s and Taiwan in early 1990sled to a broadening of political participation by including a range of new actors that penetrated previously closed political space. In Taiwan, assertive legislators and social movement groups facilitated a greater presence of public concerns in social policy debate. In South Korea, similarly, civil society was able to penetrate elite circles and influence health policy making. Eventually, originally narrow social programs were expanded to universal coverage in 1989 in Korea and in 1995 in Taiwan. However, there is evidence that legal opportunities for mass participation can also inhibit state responsiveness to basic needs. McGuire (2010: 145) observes that in Argentina, participation groups influenced state elites to enact policies that benefitted urban formalsector employees while neglecting the rural poor and inhabitants of urban shantytowns. Unionized workers, university students, and other middle-class groups have more resources 4

than the destitute to make themselves heard. At the same time, by responding to the demands of these able groups, government was able to discharge their obligation to attend to the needs of “the people” while minimizing strikes, demonstrations, and criticism in the mass media. Similarly, in Brazil, the expansion of political participation from 1970s onward diverted public policies away from infant mortality reduction, because labor unions, business organizations, and professional associations did not always operate in ways that improved the provision of basic health services to the poor (McGuire 2010: 180). Thus, participation by the well off has added to the invisibility of the needs of the worse off. If political participation, similar to competition, can result in government policies that are less needs oriented, what can be concluded about the relative roles of these two dimensions of democracy? Dahl, who created a two-dimensional concept of polyarchy, stated that participation and competition are somewhat different theoretical dimensions of democratization, understood by Dahl as a movement toward responsive polity. He argued that “public contestation and inclusiveness vary somewhat independently” (Dahl 1971: 4) and that “a regime might change along one dimension and not the other” (Dahl 1971: 8). If the least competitive and least inclusive regime – “closed hegemony” – becomes more liberalized without becoming more inclusive (when the opportunities for public contestation increase, but only for a fraction of the population), it changes toward competitive oligarchy. At the same time, closed hegemony can evolve toward what Dahl calls “inclusive hegemony.” In this case, a regime becomes more inclusive without liberalizing (Dahl 1971: 5). Most of the existing real-world regimes are skewed towards greater competition or greater participation. Because no country in the real world is perfectly democratic, the practical questions are as follows: are a population’s basic needs better satisfied in competitive oligarchies or inclusive hegemonies? During the process of democratization, in which direction – toward more contestation or more participation – is earlier movement better for social outcomes? To answer these questions, the analytical model presented in the next section introduces competition and participation as two separate independent variables of interest. 2. Analytical strategy and measurements The analytical strategy involves two steps: cross-sectional analysis with various indicators of basic needs coming from different data sources, and time-series cross-sectional analysis with fixed effects, where infant mortality is taken as the indicator of the level of basic needs satisfaction. The cross-sectional analysis provides a useful pre-test of the relationship between a hypothesized cause and an outcome and makes use of some quantitative indexes that are good operational expressions of the concept of “basic needs” used in this study but for which reliable time-series data are not available. Time-series cross-section analysis, in turn, represents a more robust test of the theorized inter-variable associations. The cross-sectional model includes three layers of independent variables with potential influence on basic needs attainment. The first layer includes two independent variables of theoretical interest – competition and participation. Following the logic of the dimensional approach, these two variables are included in all models as separate variables. Layer two comprises socioeconomic and demographic variables that can influence social outcomes in a number of ways. The third layer includes state institutional capacity variables. The technique is ordinary least squares multiple regression analysis. The sample size includes both developing and developed nations and varies from 89 to 147 cases in 2010 or the nearest year depending on data availability.1 The data defines the sample size. For example, one indicator of basic needs attainment – the multidimensional poverty index (MPI) – covers only developing countries, which reduces the sample size in the models where MPI appears as the dependent variable. A similar case is the Bertelsmann Management Index, which is not 1

5

2.1. Dependent variable There are five operational expressions of the dependent variable “basic needs,” which are used interchangeably in this study: the Human Development Index (HDI), the Multidimensional Poverty Index (MPI), the infant mortality rate (the total number of deaths before age one per 1000 live births in a given year), healthy life expectancy (HALE) (the average number of years that a person can expect to live in “full health,” excluding years lived in less than full health due to disease and/or injury), and the maternal mortality ratio (number of maternal deaths during a given time period per 100000 live births during the same time period).2 Despite the fact that these indicators quantify technically different social outcomes, all of them perfectly express the same policy concept of basic needs that is important for understanding the causal mechanism of democratic responsiveness. Moreover, all indicators are highly correlated with one another and display similar characteristics in terms of dependency on GDP per capita, speed of change, and pattern of change. Shifting between them does not significantly affect the key findings produced by the model. The substantive interpretation of results, therefore, is the same. Using all these indicators alternatingly under the umbrella concept of “needs”, serves the purpose of general theorizing about “regime responsiveness to basic needs.” At the same time, the application of similar predictors to different indicators representing conceptually similar policy outcomes can be seen as one type of check of the robustness of the findings. The data for HDI comes from the Human Development Reports (UNDP ); the data on healthy life expectancy (HALE) are from Salomon et al. (“Healthy life expectancy for 187 countries 1990-2010: a systematic analysis for the Global Burden of Disease Study 2010”, The Lancet 2012); infant mortality rates and maternal mortality ratios are from UNICEF, WHO, World Bank, UN DESA, and UNDP. If data for 2010 are unavailable, data from the nearest time point are used. The MPI data is from Alkire and Santos (“Acute Multidimensional Poverty: A New Index for Developing Countries”, UNDP, Human Development Research Paper 2010). 2.2. Independent variables of analytic interest: competition and participation Operational expressions of political competition and political participation are taken from Vanhanen’s polyarchy index of democracy. This index was developed in parallel with Dahl’s theory of polyarchy and measures democracy with two indicators: competition and participation. Competition is measured by the percentage of the vote going to parties other than the most-voted party in parliamentary or presidential elections, or both, depending on whether a country has a strong parliamentary system or characterized by executive dominance (the same rules apply to calculating participation). It is calculated by subtracting the percentage of votes won by the largest (most-voted) party from 100. In parliamentary elections, the largest party is a party which receives the largest share of votes or seats in parliament (if data on the distribution of votes is not available), whereas in presidential elections the largest party refers to the party of the the presidential candidate who won the election.3 This approach allows measuring the intensity of electoral inter-party competition

available for developing countries. Variation with the data sources and sample sizes represents a check for robustness of the results. The common practice of including both developed and developing nations, in turn, allows maximizing variation in the variables of interest and captures both the successes and failures in basic needs attainment around the world. 2 See appendix 1 for links to all data sources used in this research. 3 See “Introduction” in Vanhanen’s dataset manuscript for more details on coding procedures. Available at: https://www.prio.org/Data/Governance/Vanhanens-index-of-democracy/Polyarchy-Dataset-Manuscript/.

6

that has been the main form of legal inter-elite competition and power sharing in many countries since the nineteenth century. Participation is measured by the actual voting turnout in each election. It is calculated as the percentage of the total, rather than adult (voting-age), population because there is more statistical data on the total population than on age structures of electorates. The degree of electoral participation indicates the extent to which “the people” participate in politics. If only a small minority of the population takes part in elections, electoral participation is restricted to the most active stratum of the population. Conversely, the higher level of participation indicates the greater involvement of the population in politics. There are expected shortcomings related to using of these crude measures of procedural democracy to capture phenomena as complex as competition and participation. The indicator of competition does not sufficiently take into account differences in electoral systems, which can affect the smaller parties’ shares. Neither does it take into account difference in party structures. The indicator of participation, in turn, does not take into account the variation in the age structure of the population. Given that the percentage of the adult population is higher in developed countries, the differences in the degree of electoral participation between rich and poor countries reflected in the indicators may be exaggerated.4 Nevertheless, this study believes that the lack of better and more convenient data on separate phenomena of competition and participation does not mean that the issue should not be addressed by the social scientists. The present analysis provides general insights into the workings of regime responsiveness to basic needs and indicates certain trends that may otherwise be camouflaged by strong existing beliefs regarding the effects of democracy on social outcomes. The goal is to call attention to the dimensional approach, hoping that others will follow up with their data in the future and either support or challenge the findings. Moreover, given that no data quantifying complex social phenomena is perfect, Vanhanen’s polyarchy dataset appears to be the most suitable for probing theoretical assumptions of this particular study as it has important advantages. First, in contrast to many existing measures that combine a number of categorically different indicators, the polyarchy dataset grasps simple but most crucial aspects of democracy and allows clear separation of the political regime from other factors. The minimalist operationalization of democracy avoids blurring the boundaries between the regime type and other political factors or aspects of the state (Munck & Verkuilen 2002: 54) and enables an extensive empirical analysis of causal relationships (Cheibub, Gandhi, and Vreeland, 2010). Parsimony and the least complicated methods make the establishment of causality more transparent. Second, the polyarchy dataset is more objective than its counterparts. Many expertcoded comprehensive democracy scales suffer from subjectivity and have problems with measurement errors, accuracy, and reliability. Indicators are often assigned by subjective judgments rather than measured. This, as some note, generates questionable ratings. According to Polity scores, for instance, the United States from 1845 to 1849 received the highest possible score of +10 despite the fact that slavery existed and women lacked the right to vote. Similarly, apartheid South Africa from 1910 to 1987 received a score of +7, which is higher than Chile from 1964 to 1998 (+6) and the same as Argentina from 1990 to 1998 (McGuire 2010: 32). Freedom House ratings have been criticized for bearing an intrinsic bias toward an American libertarian understanding of democracy (Bollen & Paxton, 2000; Munck & Verkuilen, 2002; Vreeland, 2003). Polyarchy’s scores for both competition and

4

Vanhanen himself estimates that in extreme cases this bias may be about 10-15 percentage points. See more detailed coding procedures available from https://www.prio.org/Data/Governance/Vanhanens-index-ofdemocracy/Polyarchy-Dataset-Manuscript/.

7

participation are derived from objective and transparent sources, such as electoral data, and are comparable cross-nationally. Third, and most importantly, all other indices make testing the independent effect of participation problematic because they neglect this dimension of democracy. Thus, in the Polity dataset, “what many would promote as a hallmark of democratic societies, namely the extent and character of popular participation in selection of leaders, is either totally absent or relatively unimportant in determining the degree of democracy” (Gleditsch & Ward 1997: 376). Freedom House , despite referring in its definition of political rights to “the right of all adults to vote,” excludes this aspect from its checklist of political rights (Ryan, 1994). Similarly, the Bollen index, while stressing the importance of universal suffrage, excludes it from its attributes. The result of these omissions is that quantitative studies may undervalue the effect of democracy on important social outcomes or misinterpret the aspect of democracy responsible for that effect (Moon 2006: 3). The polyarchy dataset alleviates these problems because it includes participation, although in a minimalist understanding, and measures the extent of actual participation, not just the right to participate. Fourth, in the polyarchy dataset participation and competition stand for two discrete dimensions of democracy and can be used separately –an aspect that is central to this research. 5 Polyarchy’s competition and participation dimensions are neither redundant nor highly correlated (Figure 1). Freedom House components of political rights and civil liberties, for instance, refer to identical phenomena and have a correlation higher than 0.95; this is similar to comparing lengths in inches and centimeters. The combined index of Polity IV, in turn, strongly depends on a single variable: XCONST (executive constraints). The correlations of democracy, autocracy, and polity with XCONST are 0.96, 0.86, and 0.95, respectively, which means that a high score on XCONST makes the polity index gravitate heavily upward. In the case of the polyarchy index, the character of the relationships between component variables makes disaggregation reasonable. [Figure 1 about there] 2.3. Control variables: state capacity and socio-economic factors An extra word should be said about the reason for including state capacity as a third-layer control variable. The focus of the present study is the impact of political regime, which is understood as a “form of government” rather than “degree of government” (Huntington 1968: 1). It refers to the type of rule rather than the specific capacities of the state. A distinction between the “state” and the “regime” is analytically justifiable because issues of state effectiveness and state power are logically prior to those concerning political regimes (Przeworski 1995: 13). Weberian components of the state, such as order and bureaucratic effectiveness, are independent of regimes. State effectiveness, or “infrastructural power” (Mann, 2008), is the foundation of any successful state, democratic or not, because it is crucial for maintaining order and making effective use of bureaucracies. Corruption and weak regulatory and administrative capacity will inevitably result in misallocations of resources and will make needs-oriented reallocations problematic in all types of regimes. It has been demonstrated, for instance, that healthcare spending indicators are associated with a population’s mortality levels only in countries with high-quality government institutions (Rajkumar & Swaroop, 2002). State effectiveness can also determine the functioning of democratic institutions, especially the ways people participate in political life and the types of techniques politicians use to compete for votes. In light of these considerations, any attempt Vanhanen (2002, 262) explains that “those who would like to challenge any of the assumptions of aggregation may classify governmental systems differently…. This dataset is not inextricably linked to my [Vanhanen’s] interpretations, but can provide data for many alternative formulations.” 5

8

to understand the independent effects of democracy should control for state institutional capacity. The cross-sectional model includes both objective and perception-based indicators of state capacity. The objective state capacity variables are murder rate, which represents the state’s ability to monopolize violence, and the size of the shadow economy, which represents the state’s ability to enforce economic regulations. Data for the first variable are from the United Nations Office on Drugs and Crime (UNODC) statistics on the international homicide rate (unlawful deaths purposefully inflicted on a person by another person per 100000 population). The second variable is derived from Schneider, Buehn, and Montenegro’s (2010) “Shadow Economies all over the World: New Estimates for 162 Countries” and includes monetary, labor market, and state of the official economy indicators to estimate the size and trend of the shadow economy. The inclusion of these indicators in the analysis is dictated by serious concerns with the objectivity of many expert-judgment-based state effectiveness indices. 6 At the same time, it is recognized that these two objective indicators may not be the best proxies for state capacity. This is why the model is also tested with two widely-accepted perception-based state capacity indexes – World Bank government effectiveness index and Bertelsmann Stiftung’s management index. Testing the model with both objective and perception-based indexes of state capacity increases the robustness of the model. The cross-sectional model also includes four socio-economic and demographic control variables. GDP per capita is controlled for because, according to the “wealthier is healthier” claim (Pritchett & Summers, 1996), it is the most important determinant of crossnational variation in mortality rates (the GDP data come from World Bank national accounts data and OECD National Accounts data files). Population density can also affect basic needs indicators: higher density can make the provision and utilization of basic services easier, thereby improving basic needs attainment, and information about unmet needs can spread faster in more populated areas voicing public demands for better social services (population density data are from the Food and Agriculture Organization and World Bank population estimates). Similarly, urbanization brings about improved housing conditions, better services and amenities, and better infrastructure, which make social service delivery easier. It also affects the organization of the social and cultural environment, improves social networking, and changes employment patterns (data on urbanization are from UN World Urbanization Prospects). Ethnic fractionalization is included in the battery of control variables because ethnically divided regions may exhibit poorer public policy performance along many dimensions (Easterly & Levine, 1997). Higher levels of ethnic division are also associated with less efficient public goods provision, even in highly industrialized nations (Alesina et al. 2003) (the ethnic fractionalization data comes from Alesina et al (2003) “Fractionalization)” 2.4. Time-series cross-section analysis When switching to time-series cross-sectional analysis, the number of variables is reduced due to the lack of data. The dependent variable is now the natural log of IMR, which is regressed on the political competition, participation, GDP per capita (natural log), and the level of urbanization, for which reliable time-series data for 1960-2012 is available. The resulting samples vary from 184 to 185 countries and 6980 to 7288 observations that is almost complete global coverage of sovereign nations and the world’s population. Similar to the above-described cross-sectional analysis, both developed and developing countries for which data is available are included to maximize variation on the variables of interest. The

6

For a more detailed analysis of these two objective indicators and the problems with other subjective indexes of state institutional capacity, see Popov, Vladimir: “Developing New Measurement of State Institutional Capacity,” PONARS Eurasia Policy Memo No. 158, May 2011.

9

model is tested with no lag, one time-period lag, and two-time period lag of independent variables to protect the analysis from X:Y endogeneity. The analysis includes country-fixed effects and time-fixed effects. Country-fixed effects technique helps to deal with the model specification problems typical for crossnational studies. Since it is impossible to include all potentially relevant control variables, the results may suffer from omitted variable bias. The country fixed effects wipe out the constant characteristics of states that may affect both dependent and independent variables and, therefore, help to control for unobserved state-level heterogeneity. Time-fixed effects are applied because repetitive observations of a parameter of an individual country are not independent and there are reasons to believe that part of time-series variation in Y is affected by overall time trends or other time-series patterns. This is particularly the case with IMR, which is relatively persistent over time and as such may have substantial explanatory power that may affect, or even erase, the contribution of other factors, which would mean that the estimation results are strongly affected. Therefore, repetition of observations of each country should be taken into account by the means of time-fixed effects that add time-series dummies to the model. 7 3. Results The results for cross-sectional analyses are reported in tables 1 and 2. The results for timeseries cross-sectional analysis are reported in table 3. In cross-sectional analysis, a similar set of control variables and independent variables of interest are applied to different indicators of basic needs: HDI, MPI, HALE, IMR, and MMR. The model is also tested with the abovementioned three different measures of state capacity. Such changes of indicators help to check the robustness of the association between competition, participation and basic needs. Table 1 presents cross-sectional regressions for 2010 with no time lag, whereas table 2 presents cross-sectional regressions with 2-year lag of independent variables. 8 The coefficients reported are standardized beta weights, which show the net contribution of each of the independent variables while holding all other factors constant. The potential multicollinearity issue is addressed by examining tolerance and variance inflation factors (VIF), which indicate the degree to which standard errors are inflated due to collinearity. Some suggest that a VIF value over five is a symptom of undesirable multicollinearity (Judge, Hill, Griggiths, Lütkepohl & Lee, 1988). Tables 1 and 2 report maximum VIF. In most cases it is much lower than the rule-of-thumb cutoff point. Only log of GDP per capita in some models has VIF slightly above five.9 Multicollinearity problems, therefore, are not severe, and there are no duplicate or redundant variables. Interestingly, the VIF for competition and participation are the lowest in all regressions (1.1–1.6), which means that the percentage of variance in either variable that is not accounted for by other independent variables amounts to more than 80 percent. This observation falls within the logic of the dimensional approach. Research findings involving socioeconomic and state capacity variables are, in most cases, not against intuition. GDP per capita is the most powerful predictor of basic needs 7

The Hausman test, which checks whether the unique errors are correlated with the regressors, has been applied, and the results show that country-fixed effects are necessary as the null hypothesis (that the random effects are preferred) was rejected with high level of significance. To see if time-fixed effects are necessary a test that checks if the dummies for all years are jointly equal to zero (the null hypothesis) was applied. This null was also rejected with high level of significance which shows that time-fixed effects is the preferred model (the results of both test are available from the author). 8 Theories do not tell us what the correct lag should be. The 2-year lag seems reasonable, but each model was also tested with lags ranging from 0–6 years. Varying the lag of variables had little effect on the magnitude or significance of the correlations between competition, participation, and basic needs. 9 It is particularly the case with HDI, because GDP is a part of it. To deal with this, all models were also retested with GDP excluded from the control variables. The VIF in this case drops below 2.

10

attainment, which is very much in accordance with the “wealthier is healthier” claim. Ethnic fractionalization has the expected strongly negative effect (p