University of Groningen
FACULTY OF ECONOMICS AND BUSINESS
A study on the role of social trust and the welfare state in
equalization of incomes
RESEARCH MASTER THESIS
Andreas Katsikidis
S2093618
Abstract
The main objective of the present study has been an assessment of a potentially equalizing effect of social trust on income distribution. Initially, and after an extensive review of the relevant literature, I evaluate any theoretical linkages of social trust to public welfare policies that advance vertical redistribution and social mobility. At a later stage, I examine empirically such relationship in a wide longitudinal sample of 88 countries, and in combination with the public welfare state system in OECD countries. Results suggest a strong and independent equalizing effect of social trust to income inequality, but in addition such an impact appears to be moderated by the presence of an extensive public welfare state.
SUPERVISORS
Prof. Dr. Sjoerd Beugelsdijk Dr. Robbert K.J. Maseland
1
I)
Introduction
Sharp income disparities, once considered a feature of a distant pre-War past, have been increasingly realized in the course of the last three decades. According to The Economist1, by the late 2000’s the fraction of US national income yielded to the richest 1% of the American population has more than doubled compared to that in the 1980s, and matched approximately to one quarter. Far from being solely a US development, the respective percentages for the UK stood at 15% and in relatively more egalitarian France and Sweden to almost 10%. Such disparities are recorded as even steeper as we move upper in the wealth pyramid; 421 US individuals, each with net wealth above 1 billion dollars are in possession of 10.5% of the US GDP, and for the richest 0.01% of the total population this corresponds to 5%.
Rising income inequalities stand of utter significance as they bring into the fore of social discourse issues of normative, instrumental and functional nature (Salverda et al. 2009). Public sentiment regarding social justice and fulfillment of the “social contract” may turn sour under a landscape of surging inequalities and limited social mobility. In addition to concerns of this sort, the potential social and economic consequences stemming from an increasingly unequal distribution have led to a renowned interest on the topic of income inequality in various disciplines of the social science (Neckerman and Torche, 2007). Correspondingly, various studies have suggested that income inequality may result to increased mortality rates (Wilkinson, 1992; Kennedy et al. 1996), corruption (You and Khagram, 2005), rise in violent crime (Kelly, 2000), decreased educational attainment (Mayer, 2001) and lower levels of happiness (Alesina et al., 2004). Moreover, acute income disparities may be proven quite detrimental to intergenerational mobility, sustaining in this way a poverty trap, usually via the attainment and/or political process2.
From an efficiency perspective, various theoretical arguments for a harmful influence of inequality to economic growth may be provided, but nevertheless, the evidence remain largely inconclusive and depend principally on the underlying quality of the data and statistical method utilized (Voitchovsky, 2009). Yet, economic growth has been usually placed forward as the main culprit when one starts exploring the causes of income inequality. Such thesis is traced originally to the seminal work of Simon Kuznets (1955), where the argument advanced is that in the course of development, income inequality will rise before commencing to follow a reverse trajectory, i.e., the inverted-U hypothesis.
Whether the existence of any systematic link between inequality and growth can be established in the relevant literature or not, the empirical investigation on the determinants of inequality should focus on other factors which may be of significance in affecting inequality directly or via the growth channel. Such factors may be identified as the level of economic liberalization (i.e., size of government, rule of law and regulation of credit and business environment in general, financial development, trade openness and primary exports) and the structure and quality of the institutional environment (Papanek and Kyn, 1986; Dollar and Kraay, 2002).
In the present study, the focus will be directed to social trust as one unexplored potential determinant of income distribution. Social or generalized trust can be defined as the confidence that others in society will
1 Special Report on World Economy, For richer, for poorer, 13/10/ 2012, http://www.economist.com/node/21564414 2
2
not behave opportunistically, if they have the option to do so (Delhey and Newton, 2005). Ceteris paribus, trusting societies have been indicated to attain higher levels of growth (Knack and Keefer, 1997; Zak and Knack, 2001; Beugelsdijk et al., 2004), have superior governance performance (Putnam, 1993; Boix and Posner, 1998; Knack, 2002; Uslaner, 2008) and even better public health (Kawachi et al., 1997).
Trusting communities possibly provide better opportunities for the poor via safety net arrangements such as risk sharing and informal credit transactions (Alesina and La Ferrara, 2002). Micro-level analysis on informal networks seems to support such argument (Fafchamps and Lund, 2003). At macro level, the role of such safety net is usually delivered by the welfare state institutions. Modern welfare states serve mainly two functions, that of a ‘piggy bank’ and ‘Robin Hood’ (Barr, 2004). The former refers to insurance against shocks and redistributes each individual’s income across their lifespan (i.e., horizontal redistribution); the latter occurs via income transfers from the rich to the poor through progressive taxation (i.e., vertical redistribution) and therefore is relevant for alleviating poverty and income inequality (Esping-Andersen and Myles, 2009).
High levels of social trust may be regarded as a prerequisite for the implementation of welfare policies with strong vertical redistributive character, as these may discard free riding considerations, adequately reduce monitoring and regulative costs stemming from such considerations and prevent tax revolts and evasion (Rothstein, 2001, Nannestad, 2008; Bergh and Bjornskov, 2011; Jensen and Svendsen, 2011; Bjornskov and Svendsen, 2012). Thus, to the extent that social trust facilitates such policies, it can be regarded as a potential determinant of income inequality. Such thesis has been rarely assessed theoretically or tested empirically in the vast literature of income inequality. Whereas the reverse link between social trust and income inequality has been suggested in various studies (Knack and Keefer, 1997; Zak and Knack, 2001; Uslaner, 2002; Delhey and Newton, 2005; Rothstein and Uslaner, 2005; Berggren and Jordahl, 2006; Bjornskov, 2006; Leigh, 2006), concerns regarding its strict exogeneity have been also raised (Bjornskov, 2006, Leigh, 2006). The main contribution of the study is thus to evaluate both theoretically and empirically the role of social trust as predictor of income inequality.
In order to appraise the standing of such link, a wide panel data sample, including almost all the available data on the World Values Survey (WVS) trust measure, covering all the five survey waves, and concerning 88 countries has been utilized. Furthermore, an assessment of such impact in combination with the presence of an extensive public welfare support network, as this captured by the levels of public social expenditure, in the subsample of OECD countries has been attempted. In both instances, the empirical investigation has yielded results that suggest an equalizing effect of social trust to income distribution. However, that effect appears to be moderated by the existing levels of public social expenditure and wanes off under high levels of the latter. Still, an instrumental variables analysis suggested that any potential equalizing impact of trust seems to be robust of any endogeneity biases, proposing in that sense that the latter may stand on its own right. Findings of this sort indicate that successful redistributive welfare policies may be established more difficulty due to low trust levels.
3
II)
Income Inequality
Income inequality has been on the surge during the last three decades within a number of countries both in the developed and developing world. The conventional measure of income inequality is expressed in the GINI coefficient, which ranges from 0 (perfect equality, i.e. everyone has the same income) to 1 (perfect inequality, i.e., all income is in the possession of only one person/household). Figure 1 presents the percentage change of net and gross household GINI in various countries from the mid-80s to late 2000s, just on the onset of the global financial crisis. Both measures show a steep increase in ex-communist countries probably due to their transition to a liberal market economy but nonetheless, the majority of developed countries depicted have also observed non-trivial increases especially, but not exclusively, in gross income inequality. Only in a handful of countries, mostly Latin American ones, inequality has followed a reverse trend.
Before starting looking to income inequality more thoroughly, one has first to define the kind and the reference group that this inequality concerns (OECD, 2011). The starting point is thus the labor market and the respective dispersion of earnings among individual workers. However, members of a household usually pool their resources together and consequently before proceeding to earnings inequality at household level we need to take into consideration changes in the structure of the latter. Furthermore, by incorporating the distribution of income derived in other markets (e.g., capital rents) we obtain a measure of household market income inequality (i.e., gross). Finally, household disposable income inequality (i.e., net) takes into account taxation and welfare support in the form of cash transfers and further amendment for benefits in-kind and public services yields measures of household adjusted disposable income inequality.
Distribution of labor market earnings during the last three decades has been characterized by steep increases in the wages and salaries of top earners relative to those in the middle of distribution (Atkinson, 2007). Respectively, earnings at the bottom of distribution have declined relative to the middle but to lesser extent than the surges documented at the top (OECD, 2011). Explanations for these patterns are usually attributed to skill-biased technological change, global integration of labor markets due to the opening to world trade or due to an interaction of both (Goldberg and Pavcnik, 2004; Atkinson, 2007; Freeman, 2009).
Changes in household structure as these are reflected by increases in the number of single-headed households and “assortative mating” and which may appear relevant for earnings inequalities at household level, seem to have played a much modest role than labor market effects; similar is the impact of capital rents which nevertheless have become an important source of income for households at the top of market income distribution only (OECD, 2011).
4 Figure 1 Data: Solt (2009) Own Calculations -30.00 -10.00 10.00 30.00 50.00 70.00 90.00 Turkey Peru Ireland Egypt Venezuela El Salvador Korea Brazil Chile Singapore France Switzerland Norway Greece Mexico Denmark Philippines Colombia Italy Netherlands Indonesia Japan Argentina Uruguay Sweden Canada Belgium United States Spain Germany Australia Poland Potugal Austria Azerbaijan Finland Luxembourg Bangladesh United Kingdom Hungary Estonia Bulgaria Latvia Lithuania Moldova Russia
% Change - Gross GINI
5
a ratio below 100% is attained, indicating in that manner a diminishing redistributive effort. On the other hand, continental European countries observed increased redistribution in comparison to that of the mid-80s. In regard to developing countries depicted, the numbers should interpreted with caution as in most of these actual redistribution is quite small and thus the ratio may be quite sensitive to even small absolute changes.
Antecedents of Income Inequality
There is no standard theory of income inequality (Salverda et al. 2009) but nonetheless, the literature on its antecedents is quite extensive and mainly focuses on the effects of growth/development, economic liberalization, political environment and institutional structure, human capital and demographics.
Growth
Since the exposition of the inverted-U hypothesis by Kuznets (1955), the examination of the relationship between income inequality and growth holds central place in the literature of income distribution. However, after decades of research, the literature remains inconclusive regarding its validity. Some empirical studies point to the existence of such pattern (Papanek and Kyn, 1986; Tsai, 1995; Barro, 2000) though they provide a word of caution regarding its strict stability over time (Papanek and Kyn, 1986) or its ability to explain most of the variation in income inequality (Barro, 2000). In addition, studies focusing on the effect of economic dualism indicate a positive association between the two suggesting in this way that the transition from an agricultural to industrial economy may yield a skewed income distribution (Nielsen and Alderson, 1995; Bourguignon and Morrison, 1998; Gustafsson and Johansson, 1999). Other evidence however, do not support the presence of such empirical regularity (Deininger and Squire, 1996b) suggesting in this way that growth and equality are not necessarily mutually exclusive (Dollar and Kraay, 2002; Lundberg and Squire, 2003).
Economic Liberalization
The literature on the relationship between the level of economic liberalization and inequality is somewhat ambiguous regarding its main findings as it is dependent on certain aspects of economic liberalization under investigation. Studies that focus on general levels of economic liberalization (i.e., composite indices), tend to regard these beneficial to equality (Berggren, 1999; Scully, 2002) or to growth and thus equality (Dollar and Kraay, 2002). Carter (2007) however, uncovers empirically a U-shaped relationship between inequality and economic liberalization, arguing that increases above a critical point may be proven detrimental to equality. Gustafsson and Johansson (1999) on the other hand, claim that higher levels of government spending and trade unionization are conducive to equality.
7
Concerning the relationship between trade openness and inequality the literature seems to indicate a negative association between the two (Lundberg and Squire, 2003; Reuveny and Li, 2003) or the positive association between trade protection and inequality (Bourguignon and Morrison, 1990) or at least the absence of any link (Edwards, 1997). Spilimbergo et al. (1999) examine the impact of trade openness on inequality, contingent on the factor endowments residing within a country. Specifically, they found that trade openness is detrimental to equality in those countries relatively endowed with land and physical capital whereas it is beneficial in countries relatively endowed with human capital.
The role of primary exports (mineral resources etc.) in driving inequality in the developing world is explored by Bourguignon and Morrison (1990) where countries in which the comparative advantage lies within natural resources or agriculture seem to have a more skewed distribution of income. However, such effect is mitigated by a relatively more equal distribution of land. Fum and Hodler (2010) explore the same link between natural resources and inequality and show that the direction of this relationship is dependent on the level of ethnic fractionalization in a country, with polarization pushing towards greater inequality and vice versa.
Institutional Environment, Human Capital and Demographics
Moving away from strictly economic factors, the literature on inequality has focused more on institutional, educational and demographical characteristics. The focus on institutional environment mainly concerns the role and structure of political democracy, the level of political liberalization and the effect of corruption. On the role of political democracy on inequality, the relevant literature usually identifies no such link (Bollen and Jackman, 1985; Nielsen and Alderson, 1995; Dollar and Kraay, 2002) with the study of Reuveny and Li (2003) being the exception. Birchfield and Crepaz (1998) investigate the role of political institutions in reaching political decisions via either competition or consensus and their potential impact on inequality, showing that a competitive political process favors a more unequal distribution compared to one that fosters collective political bargaining.
Whereas the link between political democracy and inequality is usually absent, that is not the case for political liberalization as this is encapsulated by the level of civil liberties. Higher levels of civil liberties are negatively associated with inequality (Li et al., 1998; Lundberg and Squire, 2003) with the study of Lundberg and Squire (2003) warning of a potential detrimental effect of civil liberties on growth.
Finally, low levels of institutional quality, as these are expressed by high levels of corruption, seem to exert a negative effect on the distribution of income (Gupta et al., 2002). Nevertheless, Li et al. (2000) provide evidence that this relationship is more complex and specifically is characterized by an inverted-U pattern, where intermediate levels of corruption are more detrimental for equality, as countries with the highest levels of corruption observe lower levels of income inequality.
8
Demographics also seem to play a role, with the natural rate of population increase (Nielsen and Alderson, 1995) or the percentage of the population under 15 (Gustafsson and Johansson, 1999) to be unfavorable factors for income equality3.
A comprehensive overview of the discussed literature can be found on page 9.
Social Trust and Income Inequality
The main thesis elaborated below is that social trust stands as a prerequisite for the formation of safety net institutions facilitating redistribution and providing better opportunities for social mobility. These arrangements may reflect social pooling of risks and resources to great extent due to the high tax and public spending rates that may imply. The establishment of such institutions may constitute a “collective action problem” that is impossible to be solved under “inadequate” levels of social trust (Rothstein, 2001, Rothstein and Uslaner, 2005). In order to establish theoretically the link between social trust and inequality, I focus initially on the potential role of social trust in providing a solution to such “problem” and subsequently on modern welfare states as manifestation of such safety net arrangement and on the latter’s role in assisting equalization of incomes and social mobility. Accordingly, I develop 2 sorts of hypotheses: firstly, whether social trust facilitates income equalization and secondly whether such an effect is moderated by the very presence of an extensive welfare state.
High trusting societies potentially provide better opportunities for social mobility via the establishment of safety net institutions reflecting greater social pooling of risks and resources (Alesina and La Ferrara, 2002). Rothstein and Uslaner (2005: 42) go further and argue that: “Trust reflects a sense of social solidarity that they [people] believe that the various groups in society have a shared fate, and that there is a responsibility to provide possibilities for those with fewer resources”. Yet, realization of such pooling implies high levels of taxation and transfers among the members of a society. This in turn exhibits the presence of a “collective action problem” (Rothstein, 2001) encompassing free riding considerations and high regulative costs which may subsequently induce tax revolts and evasion (Bjornskov and Svendsen, 2012).
Nevertheless, social trust emerges as a potential solution to such “collective action problem” (Rothstein, 2001; Rothstein and Uslaner, 2005; Jensen and Svendsen, 2011; Bjornskov and Svendsen, 2012). First, social trust implies that free riding considerations will be largely discarded as people will trust that others will not abuse the common resources and so will do themselves4. Consequently, solving the free-riding problem under the presence of a trusting environment would imply low regulative and monitoring costs. Finally, “trust increases the willingness to accept one’s taxes being spent on nonmonitored transfers and benefits extended to strangers” (Bjornskov and Svendsen, 2012, pp. 4) thus mitigating pressures for tax revolts and evasion. Under those theoretical considerations, social trust appears as a crucial determinant in
3
Except for the role of the independent variables under focus in the aforementioned studies, a wide list of control variables is employed in the statistical investigation on the determinants of inequality. These usually include (in an order of frequency and when they do not serve the role of the main independent variable): income, education, demographics, urbanization, unemployment, inflation and investment. Income and education can be undoubtedly labeled as “universal” control variables, whenever they do not reserve the role of the main independent variable. The abovementioned list of control variables is by no means exhaustive and can be expanded to a considerable degree.
4 It is assumed here that trust and trustworthiness “coincide”. Whereas theoretically these two notions should not be
9
Literature Overview Effect on Income
Inequality
Growth and
Development Other Economic Variables Institutional Variables
Human Capital and Demographics Kuznets Curve Dualism Level of Economic Liberalization Financial
Development Inflation FDI
Trade Openness Primary Exports Political Democracy Civil
Liberties Corruption Education Demographics Alderson and
Nielsen (1999) Positive
Barro (2000) Present
Beck et al. (2004) Negative
Breggren (1999) Negative Bollen and Jackman (1985) Absent Bourguignon and Morrison (1998) Positive Bourguignon and
Morrison (1990) Negative Positive Negative
Carter (2007) U-shaped pattern Clarke et al. (2006) Negative Deinenger and Squire (1996b) Absent Dollar and Kraay
(2002) Negative Negative Absent
Edwards (1997) Absent
Fum and Holder (2010) Contingent on ethnic fractionalization Gupta et al. (2002) Positive Gustafsson and Johansson (1999) Positive Positive (pop. >15) Li et al. (2000) Inverted-U pattern
Li et al. (1998) Negative Negative Negative
Lundberg and
Squire (2003) Absent Negative Negative Negative
Nielsen and
Alderson (1995) Positive Absent Negative
Positive (rate of pop.) Papanek and Kyn
(1986) Present Reuveny and Li
(2003) Positive Negative Negative
Romer and Romer (1998) Negative Scully (2002) Negative Spilimbergo et al. (1999) Contingent on factor endowments Negative
10
the distribution of income, as in societies with low trust levels it would be impossible for social pooling to be materialized and better opportunities for the poor to be provided. Hence:
H.1: Ceteris paribus, high (low) levels of social trust would imply a more (less) equal distribution of
incomes.
A manifestation of such safety net arrangement in modern economies is mirrored to the welfare state institutions. Those are by definition redistributive via their taxation and spending pillars with the bulk of redistribution (i.e., two thirds) occurring via the spending functions of the welfare state (Esping-Andersen and Myles, 2009). Except for the determination of net/disposable incomes, welfare states affect in non-trivial way the labor markets via social policies and provision of public services that influence employment participation rates (Korpi, 2000) and the earnings capabilities of the respective participants (Esping-Andersen and Myles, 2009). Consequently, the very presence of the welfare state exerts strong countervailing influences on the distribution of market incomes and a plain comparison between gross and net inequality will produce biased conclusions concerning the actual redistributive effect of the welfare state (Bergh, 2005). Nevertheless, what is crucial for a successful redistribution is the implementation of policies that aim towards intensive provision of human capital, inclusive employment promotion and successful tax/transfer policies (OECD, 2011).
One of the most important influences of the welfare state on the distribution of market incomes derives from the provision of public education (Bergh, 2005). Compulsory education may stimulate growth and facilitate intra-generational equalization of incomes (Eckstein and Zilcha, 1994). Empirical evidence suggest that countries with high spending on public education present a more equitable distribution of income (Sylwester, 2002) with expenditures on primary and secondary education being more conducive to income equality (Bergh, 2005). In that sense, a welfare state arrangement providing universal coverage of primary and secondary education to its citizens is of utmost importance in tackling with labor market earnings inequalities. In addition, of great significance to income equality appear also welfare policies that promote maximum labor market participation (Kenworthy, 2007). For instance, the influx of women into the labor market during the post-war era was facilitated to large extent by the provision of public family services (e.g. child-care) and increased state employment, which both rendered among many women the “dilemma” between family and career as irrelevant (Mandel and Smyonov, 2006).
Nonetheless, equalization of incomes via the delivery of public services and welfare benefits does not depend only of the plain size of the welfare state (i.e., size-redistribution thesis) but also of its institutional design; namely, the degree of equalization depends on where transfers, in the form of welfare benefits but also public services go and on the progressivity of the tax system (Esping-Andersen and Myles, 2009). Institutionally, the former regards whether transfers will be universally available to all the citizens of a country irrespective of need or will be means-tested (i.e., selective) and targeted exclusively to those in need; the latter concerns whether such transfers will be in flat-rate or earnings-related (Korpi and Palme, 1998, Rothstein, 2001).
11
conducive to equality than a flat-rate one (Castles and Mitchell, 1992). However, both theses outlined above seem not to stand to scrutiny with welfare states characterized by universality and earnings-relatedness (i.e., progressivity) being more redistributive, identifying in this way what has been labeled by the welfare state regimes literature as the paradox of redistribution: “The more we target benefits at the poor only and the more we are concerned with creating equality via equal public transfers to all, the less likely we are to reduce poverty and inequality” (Korpi and Palme, 1998, p.681).
Nevertheless, a universal and progressive welfare state implies great social pooling of risks and resources via high taxation and public social expenditure. This further raises the issue of unsustainability of such institutional arrangement due to economic inefficiency, budget deficits and tax revolt considerations (Korpi and Palme, 1998, Rothstein 2001). However, while free-riding incentives hypothetically abound, no such free riding takes place in an environment supposedly providing disincentives for active labor market participation (Bjornskov and Svendsen, 2012). Free riding would also imply huge regulative and monitoring costs making a universal and progressive welfare state eventually fiscally unsustainable and at the same time exacerbate any propensity to tax evasion due to large transfers among strangers (Bergh and Bjornskov, 2011).
However, as argued previously the role of social trust appears as pivotal in dismissing such disintegrating considerations and thus providing solution to this “collective action problem”. Yet, we may expect that the presence of extensive safety net institutions, as those manifested in modern welfare states, to have a moderating impact on the direct effect of social trust to income inequality. That is sensibly to be expected as the very existence of the latter will render the impact of social trust on the equalization of incomes marginally weaker. Conversely, the absence of such common pool will constitute the equalizing effect of social trust relatively stronger. Thus:
H.2: The presence of an extensive welfare state would relatively mitigate the impact of social trust on
income distribution.
Schematically, the two hypotheses developed above are presented in the following figure:
12
III) Data and Descriptive Statistics
Raw data constitute an unbalanced panel with observations covering 88 countries5 for a period of 3 decades (1981-2010). Longitudinal series were preferred to cross-sectional only so to achieve greater variability in time and not depend on observations of one specific year. The sample of countries and timespan has been mainly determined by the availability of data on social trust measure, which is the main independent variable under concern6. The first issue to be acknowledged here is that the dataset is highly unbalanced due to sporadic data on trust, available for specific years only during the 5 waves of the WVS, and of data on human capital/schooling which exist only on a 5-year time interval. In order to render the panel more balanced, I transformed the existing annual observations of each data series to 5-year averages and collapsed the time dimension of the panel to 6 periods7. Below, I discuss the choice of measurements regarding the dependent and independent variables8.
Dependent Variable
One of the crucial issues in investigating empirically, in a cross-country setting, the causes and consequences of income inequality has been the absence of long, comparable and reliable data series on GINI coefficients (Deinenger and Squire, 1996a). Comparability has been hampered to large extent by the different measuring approaches and reference units employed by the various national statistical agencies. Datasets developed by Deinenger and Squire (1996a) and subsequently by the United Nations University (i.e., WIID) have tackled to some extent the problem of coverage by providing long GINI series, but nevertheless, issues of comparability remain basically present (Solt, 2009). On the other hand, GINI series from the Luxembourg Income Study Database (LIS), which are considered of high quality in terms of cross-country comparability, remain narrow in their coverage both in time and cross-section dimension. To minimize this trade-off problem between coverage and comparability, I make use of the GINI series drawn from the Standardized World Income Inequality Database (SWIID) developed by Solt (2009). SWIID is a standardization of the WIID via the employment of a custom missing-data algorithm utilizing data on GINI coefficients from the LIS database. Such data have been preferred as they provide the largest possible coverage at the lowest relatively cost in terms of cross-sectional comparability. For the purposes of this study, such series comprise of net (i.e., after taxes and transfers) and gross GINI household coefficients.
Yet, it would be sensible to utilize exclusively measures of inequality after taxation and transfers. The theoretical discussion has pinpointed to the potential role of trust in primarily facilitating equalization of incomes via extensive redistribution policies. Though, in most of the developing world gross and net GINI scores are usually of approximate magnitude - possibly due to absence of extensive welfare policies - that is not the case in the developed one where redistribution might be broad. Graph G1 in Appendix A illustrates such pattern for the whole set of countries, with observations regarding developing countries
5 See table T1 in Appendix B for the list of countries.
6 Iraq, Saudi Arabia and Zimbabwe, for which data on social trust are available, were excluded from the sample due
to unavailability of data on income inequality or income per capita.
7 1981-1985, 1986-1990, 1991-1995, 1996-2000, 2001-2005, 2006-2007 respectively. Human capital data have
remained intact of any transformation as they already regard the aforementioned periods.
8
13
lying approximately upon a 45 degrees line. Clearly, that is not the instance for richer countries where the scatterplot appears below such line9.
Independent Variables
Social Trust
The main independent variable under focus is that of social trust. Normally, cross-sectional studies on social trust utilize the WVS measure (Nannestad, 2005), which responds to question A165 of the survey asking: “Generally speaking, would you say that most people can be trusted or that you need to be very careful in dealing with people?” with two possible answers: “Most people can be trusted” and “Can´t be too careful”. Trust scores show the percentage of interviewees in the respective countries suggesting that most people can be trusted. There are some concerns on whether such measure actually reflects the theoretical definition of trust, but evidence seems to suggest that it is indeed a relatively good proxy (Bjornskov, 2006). Here, I make use of as extensive trust data as possible, covering all the survey waves and 88 countries in total.
Public Social Expenditure
In order to assess the equalizing impact of trust in combination with that of the welfare state I make use of data on public social expenditure as proxy of the latter’s extensiveness. Gross government consumption constitutes an inappropriate measure as it also includes many other elements of public spending which may be irrelevant to equalization of incomes (e.g. military consumption). Thus, data on public social spending retrieved from Adema et al. (2011) were preferred for the empirical analysis, as they appear more appropriate for such purposes (Alesina et al., 2001). Such series are part of the OECD social expenditure database (SOCX) and consequently an additional set of regressions were run for testing the second hypothesis restricted to all 34 OECD countries.
Total gross social public expenditure constitutes the sum of cash transfers in the form of pensions and income support to working-age population (e.g., unemployment benefits), social services expenditure (health and other services) and active labor market participation (ALMP) programs. The series included in the regressions is that of total gross public social expenditure minus social spending on pensions as the latter should be considered more relevant for horizontal redistribution (Espring-Andersen and Myles, 2009). Due to unavailability of longitudinal data series on net public social spending only gross data were utilized. Net series would have been preferable as they take into account taxation policies which are pertinent for redistribution purposes but nevertheless, as the larger part of taxable cash transfers is already excluded (i.e., pensions), potential biases introduced by the use of gross data should not be of significant concern10.
9 Sets of developing and developed countries were based on the current World Bank categorization. see also graphs
G2 & G3
10 The only cash transfers included in the measure of social expenditure used are those for supporting working-age
14 Rest of Independent Variables
The review of the literature on the determinants of inequality suggested a number of important control variables identified as income/level of development, trade openness, financial development, size of government and rule-of-law, civil liberties, human capital and demographics.
Income is operationalized by the log of real GDP per capita in 2005 international dollars, at purchasing power parity. In turn, the impact of government size and intervention is captured by the percentage of government consumption in GDP11 and rule-of-law indices (Barro, 2000; Dollar and Kraay, 2002). In this study, as measure of rule-of-law I use data from Fraser Institute’s economic freedom index and specifically its second dimension (EF2) which is relevant for such notion (Bergh and Nilsson, 2010), with higher values indicating greater rule-of-law. Financial development is operationalized respectively by the percentage of money and quasi-money to GDP (Li et al, 1998, 2000; Lunderg and Squire, 2003) and trade openness as the sum of exports and imports to GDP (Barro, 2000; Dollar and Kraay, 2002; Reuveny and Li, 2003; Fum and Holder, 2010). Inflation is captured by the annual change of GDP deflator (Dollar and Kraay, 2002) and political liberalization by Freedom House’s civil liberties index12 (Li et al., 1998; Lundberg and Squire, 2003). Lastly, demographics are reflected on the percentage of population belonging to the 0-14 age group (Gustaffson and Johansson, 1999) and human capital on the average years of secondary and tertiary schooling (Barro, 2000) drawn from Barro and Lee (2012).
In addition, throughout the empirical analysis a number of regional dummy variables are included, so to check for potential idiosyncrasies not captured by the rest of independent variables. In that manner, the sample of countries is coded to 6 major regional categories including Eastern Europe and Central Asia (i.e., ex-communist countries), Latin America, Sub-Saharan Africa, East Asia (i.e. Developing Asia), Middle East and North Africa, leaving as reference group developed countries, mostly Western ones. Additionally, in various specifications a Nordic dummy variable is included in order to control for Scandinavian outliers, which are characterized by great scores of social trust and low levels of income inequality13.
Instruments
Generally, the literature on the determinants of social trust has indicated a potentially detrimental effect of income inequality to trust levels (Nannestad, 2008). Consequently, the issue of endogeneity between the dependent and main independent variable becomes apparent and an instrumental variables analysis would be required at a later stage.
Two instruments are utilized for the purposes of such analysis. The first is that of ancestral trust which expresses the trust levels of 2nd and 3rd generation US immigrants. Specifically, I make use of individual trust data from the cumulative US General Social Survey (GSS) database, in order to derive the instrument of ancestral trust by controlling for the family origin of the respondents. The percentage of interviewees with family origin from a certain country, which indicated that most people can be trusted, was assigned as the ancestral trust level of that particular country. Table T3 in Appendix B shows in detail the data series on ancestral trust. The rationale behind the use of such instrument is that if social trust is considered
11 Such measure is substituted when the sample is restricted to OECD countries by public social expenditure. 12 Here inverted so that higher values reflect greater level of political liberalization.
13
15
of path-dependent nature and if such values are transmitted in inter-generational manner (Tabellini, 2008), then ancestral trust levels will be satisfactory predictors of the current trust levels in the countries of origin, and at the same time be free of any current socio-economic influences within the latter, fulfilling in this way the exclusion restriction.
As second instrument I include a constitutional monarchy dummy variable14. Following Bjornskov (2006) and Bergh and Bjornskov (2011), such instrument may capture historical trust levels in the sense that in countries where such levels were high, a transition to (semi-)democratic state would have occurred with the maintenance of monarchy, even in ceremonial fashion. On the contrary, in countries where historical trust levels were low, such transition would have taken place abruptly with the concurrent abolishment of the monarchial regime. As such, the contemporaneous presence (absence) of monarchy would be expected to reflect high (low) historical levels of trust and at the same time be independent of any income distribution developments.
Descriptive Statistics
Pairwise correlations along with the means and standard deviations among the complete panel dataset series can be found on table 1 and for the OECD subsample in table 2. Detailed summary statistics of raw data are presented in table T4, Appendix B; respectively, for the transformed series and for the entirety of the sample in table T5 and OECD subsample in table T6. As it is evident, distributions of transformed data series have not changed in any substantial manner from those in table T4. In addition, as net GINI and trust scores are considered relatively stable in time, a transformation of this kind should be expected to alter only trivially the corresponding data series. Such stability is also existent for averaged data as it is obvious from Tables T7 and T8, which present the inter-period pairwise correlations for net GINI and trust scores respectively.
A rough indication in regard to the negative relationship between net GINI and trust values can be identified in the analogous pairwise correlations, for both the entirety of the sample and the OECD subsample, with a magnitude of -.5. Furthermore, such relationship is well evident in Graph G4 which tabulates the particular scatterplot for all 88 countries, with the only exception a cluster of observations in the southwestern part of the graph and below the fit-line. Nonetheless, most of this cluster can be characterized as ex-communist, coming in terms with what have also been identified in the relevant literature: ex-communist countries present low levels of inequality, due their former regime, and at the same time low trust levels, perhaps for the very same reason (Rothstein and Uslaner, 2005).
What appears as rather interesting is the close association of higher social expenditure levels, in the sense identified previously, with lower levels of inequality. The corresponding correlation is almost -.7 and graph G5 pictures such relationship. In addition, this association becomes tighter when replacing inequality with the measure of redistribution mentioned earlier. Then correlation increases, in absolute sense, to .8 and graph G6 visually suggests a strongly equalizing effect of social expenditure15.
14 Source: The CIA World Factbook
15 Regressing social expenditure only on redistribution, apart from yielding a coefficient of great magnitude (2.2) and
16
Table 1 – 88 countries Summary
Statistics Net GINI
Gross GINI GDP capita GDP capita^2 Pop. under 15 Second. School. Tertiary School. Social Trust Trade
Open. Fin. Dev.
17
Table 2 – OECD sample Summary Statistics Net
18
IV) Results
OLS
88 countries
Pooled OLS regressions for the whole sample of 88 countries are presented in table 3. Four main specifications were employed: one quite basic with the inclusion of the logarithm of GDP and its square, one further with human capital and demographics variables, one with the addition of social trust and finally with the insertion of the rest of control variables indicated by the review of empirical studies on inequality. Each specification was reproduced by controlling for regional effects, so 8 specifications in total16.
Consistent with the first hypothesis that high trusting environments facilitate equalization of incomes, social trust coefficients appear with the right negative sign and highly significant in all circumstances. Addition of a number of control variables does not seem to affect in any meaningful way the statistical impact or significance of social trust, neither does the inclusion of regional controls17. Furthermore, approximately in any case, controls appear of trivial size and insignificant. Finally, one can stress the high fit of the models as this is reflected in R-squared, especially when regional controls are included. Even in the quite basic specification where only income is included, addition of regional controls more than doubles the magnitude of R-squared. In the last specification almost 85% of the variability in the dependent variable can be explained by the model.
OECD countries
Restricting the analysis to OECD countries, I am interested not only in evaluating the impact of trust as previously, but additionally in assessing also any equalizing effect of social expenditure, both individually and in combination with that of social trust. Therefore, next to a basic model, controlling merely for the level of development, demographics and human capital, I employ 3 more specifications with the inclusion of social trust, social expenditure, both variables, and further with their interaction. All models were reproduced with the addition of the same set of control variables as in the wider sample. Results are presented in table 418.
Focusing on the models including the interaction term of social trust and expenditure (models 5 and 9), we observe that such term turns out positive and significant. Those findings are in line with hypothesis 2 and suggest that any equalizing effect of trust would appear stronger in countries with lower levels of public social expenditure. Correspondingly, public social expenditure’s impact would be greater at low trusting environments. Graphs 1 and 2 visualize respectively these relationships; specifically, graph 1 shows that under quite low levels of social expenditure, the marginal effect of trust on net GINI appears the greatest in absolute terms. Conversely, in countries where social expenditure is already high, the impact of trust on income distribution appears as minimal or even insignificant. Similarly, in graph 2 the marginal effect of
16 Specifications 3-8 are reproduced with the exclusion of squared income in table T9, Appendix B. The results are
largely the same with those in table 3.
17 The same holds even when controlling for the Scandinavian outliers, see specifications 7-8 in table T9. 18
19
social expenditure is noted as greatest in low-trusting environments whereas the opposite seems to hold under high-trusting ones19.
Control Variables
In concern to control variables, regression analyses suggest the presence of an inverted-U pattern of income inequality in the course of development. Such finding appears to be consistent with the Kuznets thesis and other results acknowledged in the literature which also purpose the manifestation of such regularity (Papanek and Kyn, 1986; Barro, 2000)20. From the measurements of human capital employed, tertiary education appears robust and conducive to income inequality in the majority of the estimated models, thus being in line with similar findings which link public spending on higher education to increasing net income disparities (Bergh, 2005). Estimates on secondary schooling however, enter in most of the cases as insignificant, something which does not seem to be the case for the demographical measure utilized and which suggests that countries wherein the share of younger population is relatively larger would present a more skewed income distribution (Gustafsson and Johansson, 1999).
Regarding the various measures that possibly reflect the levels of economic liberalization, the results indicate in most circumstances a statistically insignificant role of the latter in relation to the measure of income distribution. Specifically, government consumption seems as statistically irrelevant in any case as is also the rule-of-law index. In turn, trade liberalization as this is mirrored in the sum of imports and exports to GDP yields in most of the models insignificant coefficients suggesting in that manner the absence of any strong link between the latter and income inequality (Edwards, 1997). Yet, whenever the estimate turns out significant it suggests a mitigating effect to income disparities. Potentially, any sort of such impact, positive or negative, may be characterized by contingencies and specifically on the kind of factor endowments residing within a country (Bourguignon and Morrison, 1990; Spilimbergo et al., 1999). Financial development as this is expressed in the share of money and quasi money to GDP and inflation in GDP deflator, appear also statistically trivial in contrast to empirical findings that indicate respectively any sort of statistical association (Li et al., 1998; Romer and Romer, 1998; Dollar and Kraay, 2002; Beck et al., 2004).
2SLS
The second stage pooled 2SLS regressions are presented in table 5. In connection with the OLS specifications, two main models were regressed one with and one without the rest of control variables. Regional controls were also included in half of the specifications; the last two ones control in addition for the Nordic outliers.
The estimates in regard to social trust appear quite satisfactory, highly significant and with the expected sign in all regression models. A number of follow-up statistics were computed to assess the robustness of the results. Specifically, a Durbin-Wu-Hausman endogeneity test, suggests the presence of endogeneity in all but two specifications, with p-values well below 5 percent and thus rejection of the null
19 Graphs G7 and G8, Appendix A, illustrate the corresponding marginal effects derived from specification 9. Those
differ in trivial manner with the previously discussed graphs 1 and 2.
20 Certainly, any comparison of the findings of the present study regarding the rest of independent variables with the
20
Table 3
Sample 88 countries, see appendix for the list of countries and.regional controls. ***, **, *, significant at 1, 5, 10 percent level respectively. Robust SEs in parentheses.
Net GINI – OLS – 88 countries 1 2 3 4 5 6 7 8
Log GDP capita 0.509*** 0.485*** 0.749*** 0.952*** 0.697*** 0.935*** 0.731*** 1.032*** (0.111) (0.152) (0.107) (0.142) (0.185) (0.222) (0.251) (0.202) Log GDP capita^2 -0.078*** -0.068*** -0.096*** -0.126*** -0.087*** -0.124*** -0.086** -0.129*** (0.015) (0.020) (0.014) (0.019) (0.024) (0.028) (0.035) (0.026) Population under 15 0.640*** 0.086 0.751*** -0.024 0.713*** -0.061 (0.063) (0.067) (0.104) (0.097) (0.113) (0.097) Secondary Schooling -0.001 0.011*** 0.006 0.013** 0.000 0.003 (0.004) (0.004) (0.006) (0.005) (0.006) (0.004) Tertiary Schooling 0.021 0.025** 0.037* 0.049*** 0.043** 0.067*** (0.014) (0.013) (0.021) (0.017) (0.018) (0.013) Social Trust -0.165*** -0.119*** -0.151*** -0.108*** (0.032) (0.029) (0.036) (0.028) Trade Openness 0.023 0.010 (0.020) (0.012) Financial Development 0.006 0.001 (0.009) (0.004) Government Spending -0.096 0.123 (0.135) (0.082) Rule of Law -0.008 -0.003 (0.006) (0.004) Inflation 0.000 0.000** (0.000) (0.000) Civil Liberties -0.003 -0.012*** (0.007) (0.004)
Regional Controls No Yes No Yes No Yes No Yes
Observations 474 474 437 437 179 179 154 154
21
Table 4
Net GINI – OLS – OECD sample 1 2 3 4 5 6 7 8 9
Log GDP capita -0.129*** -0.002 -0.038 0.040 0.064 -0.049 -0.073 0.026 0.026 (0.028) (0.048) (0.027) (0.050) (0.050) (0.082) (0.049) (0.077) (0.072) Population under 15 0.549*** 0.771*** 0.424*** 0.582*** 0.508*** 0.807*** 0.427*** 0.613*** 0.570*** (0.094) (0.121) (0.072) (0.118) (0.128) (0.152) (0.106) (0.181) (0.179) Secondary Schooling 0.010** 0.007 0.005 0.002 0.002 0.007 0.003 0.001 0.001 (0.005) (0.006) (0.004) (0.006) (0.006) (0.006) (0.004) (0.006) (0.006) Tertiary Schooling 0.036*** 0.049*** 0.026*** 0.036** 0.039** 0.032* 0.027** 0.031* 0.031 (0.012) (0.015) (0.009) (0.015) (0.015) (0.019) (0.011) (0.018) (0.019) Social Trust -0.197*** -0.140*** -0.359*** -0.199*** -0.174*** -0.403*** (0.030) (0.035) (0.093) (0.042) (0.044) (0.123) Social Expenditure -0.666*** -0.471*** -1.112*** -0.562*** -0.387*** -1.101*** (0.070) (0.122) (0.303) (0.100) (0.142) (0.401) Trust*Social Expenditure 1.551*** 1.592** (0.578) (0.759) Trade Openness -0.051*** -0.022** -0.022 -0.022 (0.018) (0.009) (0.016) (0.016) Financial Development 0.006 0.012** -0.002 -0.001 (0.009) (0.006) (0.009) (0.010) Rule of Law -0.160 -0.005 -0.001 0.003 (0.802) (0.005) (0.008) (0.008) Inflation -0.029 0.000 0.000 0.000 (0.046) (0.000) (0.001) (0.001) Civil Liberties 1.120 0.003 0.012 0.015 (0.853) (0.010) (0.012) (0.011) Observations 195 103 183 98 98 90 156 85 85 R-squared 0.4166 0.563 0.5872 0.6292 0.658 0.6345 0.6351 0.6634 0.6888 F-statistic 23.84 26.67 48.76 28.45 25.76 19.25 42 26.23 19.09
22
Graph 1 – Marginal Effect of Trust on Net GINI
Specification 5 – OECD countries
23
Graph 2 – Marginal Effect of Social Expenditure on Net GINI
Specification 5 – OECD countries
24 Table 5 Net GINI – 2SLS 1 2 3 4 5 6 Log GDP capita -0.721* 0.245 -0.281 0.481*** -0.776* -0.280 (0.398) (0.423) (0.409) (0.277) (0.463) (0.404) Log GDP capita^2 0.093* -0.039 0.047 -0.063* 0.099 0.047 (0.053) (0.054) (0.054) (0.035) (0.061) (0.053) Population under 15 0.474** -0.296* 0.560*** -0.333* 0.475** 0.555*** (0.218) (0.166) (0.133) (0.187) (0.227) (0.132) Secondary Schooling 0.003 -0.001 0.007 0.001 0.005 0.007 (0.006) (0.005) (0.006) (0.005) (0.007) (0.007) Tertiary Schooling 0.066*** 0.090*** 0.047** 0.079*** 0.062** 0.050** (0.026) (0.020) (0.020) (0.018) (0.027) (0.021) Social Trust -0.330*** -0.225*** -0.207*** -0.177*** -0.377*** -0.187** (0.073) (0.052) (0.074) (0.050) (0.128) (0.081) Trade Openness 0.036* 0.002 -0.038** (0.020) (0.022) (0.020) Financial Development 0.084 -0.021* -0.008 (0.017) (0.011) (0.017) Government Spending -0.146 -0.203* -0.129 (0.151) (0.111) (0.150) Rule of Law -0.011* -0.006 -0.011* (0.007) (0.006) (0.006) Inflation 0.000 0.000 0.000 (0.000) (0.000) (0.000) Civil Liberties -0.015* -0.006 -0.014 (0.009) (0.007) (0.009) Regional Controls No Yes No Yes Nordic Nordic
Observations 92 92 77 77 92 77
R-squared 0.339 0.6919 0.6153 0.7929 0.2909 0.6289 Durbin-Wu-Hausman (p-value) 0.0012 0.1005 0.0204 0.6346 0.002 0.0115 Hansen J Statistic (p-value) 0.2729 0.3356 0.5222 0.2222 0.218 0.6019 First Stage F 27.4065 25.2669 21.0857 22.3598 11.6912 14.7686
Sample 88 countries, see appendix for the list of countries and.regional controls. ***, **, *, significant at 1, 5, 10 percent level respectively. Robust SEs in parentheses. Instruments: percentage of
25
i.e., exogeneity). A Hansen-J-statistic for testing of over-identified restrictions was also computed and with all p-values significantly higher than 10 percent, it indicates a non-rejection of the null of orthogonality between the error term and the instrument(s). Thus, we may be confident concerning the validity of at least one instrument. Lastly, such confidence is further enhanced by the first stage F-statistics, all significantly higher than 10, pointing in this manner that instruments are sufficiently strong. In sum, the analysis strongly indicates the existence of an independent negative impact of social trust to disposable income inequality, irrespective of the findings in the literature that indeed establish an influence in the opposite direction21.
V)
Concluding Remarks
The focus of the present study has been a theoretical and empirical assessment on the role of social trust in equalization of incomes. Abiding to such objective two interrelated hypotheses were formulated after reviewing the existing literature on the determinants of income inequality and evaluating the potential theoretical linkages between social trust and the distribution of incomes. Namely, the first hypothesis proposed that high trusting environments will contribute to a more egalitarian distribution through the formation of safety net institutions that facilitate redistribution and provide better opportunities for social mobility. Social trust potentially delivers a solution to the “collective action problem” - which implies social pooling of risks and resources and may inhibit the establishment of such institutions - by discarding any disintegrating concerns accompanied by such arrangement. The second hypothesis in turn, argues that any potential equalizing effect of social trust would be moderated by the very existence of such safety net institutions, as these are manifested in modern welfare states. Specifically, the presence of an extensive welfare state should be expected to relatively mitigate any equalizing effect of social trust.
In order to evaluate empirically the standing of these two hypotheses I utilized a wide panel data sample covering 88 countries and including all the available data on social trust available throughout all the five waves of the WVS. However, due to absence of data for social expenditure for the totality of the sample, the evaluation of the second hypothesis has been restricted to OECD countries only. In any case, regressed models included a wide range of control variables drawn from the respective literature on the determinants of income inequality and regional dummies checking for potential idiosyncrasies not captured by the included independent variables. In addition, 2SLS regressions models were incorporated in the analysis so to assess to what extent any impact of social trust is of independent nature or due to causality in the opposite direction.
Turning now to the key findings, I may suggest the following: First, the empirical patterns identified from the OLS analysis for both the wider and OECD sample as well as in concern to the 2SLS results indicate a strong equalizing effect of the measure of social trust to net GINI, being supportive in this sense to the first hypothesis: a high (low) trusting environment would imply a more (less) equal income distribution, all else equal. Those findings were robust to the inclusion of a number of control variables and also of a wide set of regional dummies. In addition, association of this kind does not appear to be driven due to
21 Complementary regressions with the inclusion of only one instrument, that of monarchial dummy, were run and
26
causality running in the opposite direction as the instrumental variables analysis has also yielded strongly negative and statistically significant coefficients. In sum, empirical findings identify social trust as an important predictor of income inequality, a pattern not established in the respective existing literature. However, such impact may be moderated by the presence of a welfare state as the second hypothesis claims. Social trust has been argued that possibly affects income inequality via its role as a prerequisite for the establishment of welfare state policies with vertical redistributive character. Sensibly, the equalizing role of social trust should appear as relatively stronger under the absence of such policies. Conversely, under a high trusting environment any further emphasis on welfare spending should be expected to be marginally trivial. Such hypothesized combined effect seems to be supported by the statistical investigation in the present study, with the interaction effect of the respective measures of social trust and public social expenditure to enter highly significantly and positive in all regression models. Furthermore, the marginal impact of social trust and social expenditure appear waning as the levels of social expenditure and trust increase respectively (see graphs 1, 2; G7 and G8 in Appendix A).
From a policy perspective, what would appear as relevant for strong vertical redistribution would be an improvement of either trust levels or alternatively via higher levels of public social spending. However, the picture may be rather complicated. First, social trust may not be subject to manipulation as it may be contingent on structural characteristics of a society which are hard to change in the course of time. Those may constitute existing levels of income inequality, implying in this way the existence of a poverty trap that reproduces itself through time (Rothstein and Uslaner, 2005), cultural traits such as the dominance or not of a hierarchical religion, communist past and ethnic homogeneity of the populace (Knack and Keefer, 1997; Zak and Knack, 2001; Uslaner, 2002; Delhey and Newton, 2005; Rothstein and Uslaner, 2005; Berggren and Jordahl, 2006; Bjornskov, 2006; Leigh, 2006).
27
References
Adema, W., Fron, P. and Ladaique, M. (2011), Is the European welfare state really more expensive? Indicators on social spending, 1980-2012; and a manual to the OECD Social Expenditure Database (SOCX), OECD Social, Employment and Migration Working Papers, No. 124, OECD Publishing Alderson A.S. and Nielsen F. (1999), Income Inequality, Development, and Dependence: A
Reconsideration, American Sociological Review, 64, No. 4, 606-631
Alesina, A., Giuliano, P., (2009). Preferences for redistribution, Working Paper No. 14825, National Bureau of Economic Research
Alesina, A., Glaeser, E., (2004), Fighting poverty in the US and Europe: A world of difference, Oxford University Press, Oxford, UK
Alesina, A., La Ferrara, E., (2005), Preferences for Redistribution in the Land of Opportunities, Journal of Public Economics, 89, 5, 897-931
Alesina A., La Ferrara E., (2002), Who trusts others?, Journal of Public Economics, 85, 207–34
Alesina, A., Di Tella, R. and MacCulloch, R., (2004), Inequality and happiness: are Europeans and Americans different?, Journal of Public Economics, 88, 9, 2009-2042
Alesina, A., Glaeser E., Sacerdote B., (2001), Why Doesn’t the United States Have a European-Style Welfare State?, Brookings Paper on Economics Activity, 187-278
Atkinson, A. B., (2007), The distribution of earnings in OECD countries, International labour review, 146, 1‐2, 41-60
Barr N. (2004), The Economics of the Welfare State. Oxford University Press, Oxford, UK
Barro R. J. (2000), Inequality and Growth in a Panel of Countries, Journal of Economic Growth, 5, 5-32 Barro, R. J., Lee, J. W. (2012). A new data set of educational attainment in the world, 1950–2010., Journal
of Development Economics, in press
Beck, T., Demirguc-Kunt A., Levine R., (2004), Finance, Inequality, and Poverty: Cross-Country Evidence, NBER Working Paper 10979
Bergh, A., (2005), On the counterfactual problem of welfare state research: how can we measure redistribution?, European Sociological Review, 21, 4, 345-357
Bergh, A., and Bjørnskov, C. (2011), Historical trust levels predict the current size of the welfare state. Kyklos, 64, 1, 1-19
Berggren, N. (1999), Economic freedom and equality, friends or foes? Public Choice 100, 203–223 Berggren N. and H. Jordahl, (2006), Free to trust? Economic freedom and social capital, Kyklos, 59, 141–
69
28
Birchfield V. and Crepaz M. (1998), The Impact of Constitutional Structures and Collective and Competitive Veto Points on Income Inequality in Industrialized Democracies, European Journal of Political Research, 34, 2, 175-200
Bjornskov C., (2006), Determinants of generalized trust: a cross-country comparison, Public Choice, 130, 1–21
Bjørnskov, C., and Svendsen, G. T. (2012), Does social trust determine the size of the welfare state? Evidence using historical identification, Public Choice, 1-18
Boix C., Posner D.N. (1998), Social capital: explaining its origins and effects on government performance. British Journal of Political Science, 4, 686–93
Bollen, K. A. and Jackman R. W., (1985), Political Democracy and the Size Distribution of Income, American Sociological Review, 50, 438-57
Bourguignon F. and Morrison C., (1998), Inequality and development: the role of dualism, Journal of Development Economics, 57, 233–257
Bourguignon F. and Morrison C., (1990), Income distribution, development and foreign trade: A cross-sectional analysis, European Economic Review, Volume 34, Issue 6, Pages 1113–1132
Carter, J.R. (2007), An empirical note on economic freedom and income inequality. Public Choice 130, 163–177
Castles, F. G. and Mitchell, D., (1992), Identifying welfare state regimes: the links between politics, instruments and outcomes, Governance, 5, 1, 1-26
Citrin, J., Green, D. P., Muste, C., Wong, C., (1997), Public opinion toward immigration reform: The role of economic motivations, Journal of Politics, 59, 858-881
Clarke G.R.G., Xu L. C., Zou H. (2006), Finance and Income Inequality: What Do the Data Tell Us? Southern Economic Journal, 72, No. 3, 578-596
Deininger, K., and Squire, L., (1996a), A new data set measuring income inequality, The World Bank Economic Review, 10, 3, 565-591
Deininger, K., and Squire, L. (1996b), New Ways of Looking at Old Issues: Inequality and Growth. Journal of Development Economics 57, 259-287
Delhey J., Newton K. (2003), Who trusts? The origins of social trust in seven societies, European Sociological Review, 5, 93–137
Dollar D. and Kraay A., (2002), Growth is good for the poor, Journal of Economic Growth, 7, 195-225 Eckstein, Z. and Zilcha, I., (1994), The effects of compulsory schooling on growth, income distribution
and welfare. Journal of Public Economics, 54, 3, 339-359
Edwards S., (1997), Trade Policy, Growth, and Income Distribution, American Economic Review, 87, 2, 205-210
29
Esping-Andersen G. and Myles, J. (2009), Economic inequality and the welfare state, The Oxford Handbook of Economic Inequality, Oxford University Press, New York
Fafchamps M., Lund S. (2003), Risk-sharing networks in rural Philippines, Journal of Development Economics, 71, 2, 261–287
Freeman, R., (2009), Globalization and Inequality, The Oxford Handbook of Economic Inequality, Oxford University Press, New York
Fum RM and R Hodler, (2010), Natural resources and income inequality: The role of ethnic divisions, Economics Letters, Volume 107, Issue 3, Pages 360–363
Gilens, M. (1995), Racial attitudes and opposition to welfare, Journal of Politics, 57, 994-1014
Glaeser, E.L. (2005), The Political Economy of Hatred, The Quarterly Journal of Economics, 120, 1, 45-86
Goldberg, P. K., and Pavcnik, N., (2004), Trade, Inequality, and Poverty: What Do We Know?. In Brookings Trade Forum (Vol. 2004, No. 1, pp. 223-269). Brookings Institution Press.
Gupta S., Davoodi H., Alonso-Terme R., (2002), Does corruption affect income inequality? Economics of Governance, 3, 23-45
Gustafsson B. and Johansson M., (1999), In Search of Smoking Guns: What Makes Income Inequality Vary over Time in Different Countries? American Sociological Review, 64, 4, 585-605
Jensen, C. and Svendsen, G. T. (2011), Giving money to strangers: European welfare states and social trust. International Journal of Social Welfare, 20, 1, 3-9
Kelly, M., (2000), Inequality and crime, Review of Economics and Statistics, 82, 4, 530-539
Kennedy, B.P., Kawachi, I. and Prothrow-Stith, D., (1996), Income distribution and mortality: cross sectional ecological study of the Robin Hood index in the United States, British Medical Journal, 312 (7037), 1004-1007
Kesler, C., Bloemraad, I., (2010), Does immigration erode social capital? The conditional effects of immigration-generated diversity on trust, membership, and participation across 19 Countries, 1981– 2000., Canadian Journal of Political Science, 43, 2, 319-347
Kenworthy, L., (2007), Egalitarian capitalism: Jobs, income, and growth in affluent countries, Russell Sage Foundation Publications
Knack S. (2002), Social capital and the quality of government: evidence from the United States, American Journal of Political Science, 46, 772–85
Knack S., Keefer P. (1997), Does social capital have an economic payoff? A cross-country investigation, The Quarterly Journal of Economics, 112, 1251–88
Kawachi I., Kennedy B.P., Lochner K., Prothrow-Stith D. (1997), Social capital, income inequality, and mortality, American Journal of Public Health, 87, 9, 1491-1498
