Does political decentralization affect income inequality? The role of governance quality

ABSTRACT This paper examines whether an increase in political decentralization, which is a global trend, affects income inequality. It argues that the effect of decentralization on equality lies with the quality of governance, that is, how well the traditions and institutions of an authority in a country are exercised. Using an instrumental variables (IV)/two-stage least squares (2SLS) method, this paper shows that an increase in decentralization reduces inequality. However, this reduction is stronger for countries with a relatively low quality of governance, possibly because private investments in these countries reduce income inequality. A policy implication is that institutional-specific decentralization strategies can reduce inequalities.


INTRODUCTION
International organizations (e.g., the World Bank, International Monetary Fund (IMF), Organisation for Economic Co-operation and Development (OECD), United Nations and European Commission) have long recognized the developmental role of local/regional authorities, and have generally supported political decentralization 1 as a public policy instrument to achieve lower income inequality (Tselios & Rodríguez-Pose, 2020 ) and poverty (Tselios & Rodríguez-Pose, 2022). Although decentralization in decision-making is a very controversial topic (Levaggi & Menoncin, 2017;Oates, 2008;Weingast, 2009), there is a growing global trend for central governments to transfer political authority to subnational tiers of government (Rodríguez-Pose & Gill, 2003;Tselios & Rodríguez-Pose, 2020) as well as a growing and widely accepted recognition of the important role of governments in reducing inequalities. Over the last two decades, in both developed and less-developed countries, for example, powers to tax and spend have been transferred from the central government to subnational governments (Cavusoglu & Dincer, 2015). Central authority has the responsibility for formulating, planning and implementing economic policy, but controlling all aspects of the decision-making process is an extremely demanding task (Grove, 1978). While the traditional theory in federalism argues that the federal government should take the primary, if not the sole, responsibility for the income redistribution, aiming to reduce income inequality (Gordon & Cullen, 2012), the reduction of inequality seems to be, in practice, an unavoidable concern of subnational governments (Bahl et al., 2002).
Although political decentralization has globally accelerated, the level of decentralization varies enormously from country to country. Some countries have taken limited actions towards political decentralization, while others have transferred a substantial amount of political power to their regions, which enjoy a high degree of regional autonomy and/or determine central decision-making (Tselios & Rodríguez-Pose, 2020). For example, Germany, Spain, Belgium, the United States, and Bosnia and Herzegovina have high levels of political decentralization, while Luxembourg, Malta, Iceland, the Bahamas and Estonia are centralized countries (Hooghe et al., 2016). There is also a global awareness of rising income inequalities within countries. Inequality is one of the key challenges facing supranational organizations as well as national and local governments and a persistent cause for concern for policymakers. It is, along with political decentralization, at the top of the political agenda. A Sustainable Development Goal (SDG) of the United Nations Development Programme (UNDP) is 'reduced inequalities'. The degree of household income inequality across countries is uneven. Some countries, such as Haiti, Colombia, Honduras, Peru and Bolivia, have high levels of household income inequality, while other countries, such as Slovenia, Denmark, Sweden, Iceland and Norway, have low levels (Solt, 2016(Solt, , 2020. But does an increase in political decentralization, which is a global trend (Rodríguez-Pose & Gill, 2003;Tselios & Rodríguez-Pose, 2020), affect income inequality? Some scholars (e.g., Carreras, 2016;Khan et al., 2017;Kristal & Cohen, 2007;Morelli & Seaman, 2007;Sacchi & Salotti, 2014) have examined the direct effect of decentralization on income inequality, while others (e.g., Bojanic & Collins, 2021;Canare et al., 2020;Cavusoglu & Dincer, 2015;Makreshanska-Mladenovska & Petrevski, 2019;Neyapti, 2006;Sepúlveda & Martínez-Vázquez, 2011;Stossberg & Blochliger, 2017;Tselios et al., 2012) argue that the effect of decentralization on inequality is moderated by economic, political, institutional or geographical factors. The decentralization-inequality association has been examined using a variety of different variables and proxies, different methods and different countries and time periods, but it is clear that the existing empirical studies have yielded mixed results.
Using a cross-country analysis, this paper examines whether an increase in political decentralization affects income inequality. I argue that the quality of governance (i.e., how well the traditions and institutions of an authority in a country are exercised) is a key factor in the decentralization-inequality relationship, and private investments act as a transmission channel. Therefore, I test the conditional hypothesis that the effect of a change in the political decentralization of a country on household income inequality within the country is contingent on the quality of the governance of the country, and the hypothesis that an increase in the quality of governance reinforces private investments, which, in turn, affect income inequality.
This paper makes contributions to understanding the political and institutional determinants (i.e., the joint effects of political decentralization and governance quality) of income inequality from a global perspective, that is, for 91 countries from 1996 to 2018. The benefits of undertaking analysis at this scale are that I can identify the global conditional effect of decentralization on inequality. To the best of my knowledge, this paper makes four contributions to the literature. First, institutions matter for income inequality (Beramendi, 2012). The central government, through the transfer of political power to subnational levels and the higher quality of governance, has the capacity to shape the distribution of household income. The effects of political decentralization are examined in conjunction with those of governance quality, because both institutional effects go hand in hand (Tselios & Rodríguez-Pose, 2022). Second, both the self-and shared ability of a subnational tier of government (Hooghe et al., 2016) can affect income inequality within a country, but this ability is shaped by the quality of governance. The distinction between the self-and shared responsibility for the provision of public goods within a country is vital because the regional government of the country can either independently manage the provision of these goods or affect the authority of the other regional governments and/or the authority of the central government for the allocation of these goods (Carreras, 2016). Hence, regions can enjoy a considerable degree of political autonomy (self-ability or self-rule) and/or determine central decision-making (shared ability or shared rule) to reduce income inequality. The particular type of political decentralization depends on the country's and its regions' specific political, economic, and cultural and historical institutions and traditions (Oates, 2005). Third, it is very important to control for endogeneity, because income inequality may be a driver for higher levels of political decentralization. Fourth, compared with the current empirical studies, this is the first empirical study to find that political decentralization reduces income inequality, but this reduction becomes weaker as governance quality increases. An increase in decentralization seems to have relatively more desirable consequences on income distribution for the low-quality of governance countries, because private investments lower income inequality in these countries. Fifth, this paper and that by Tselios & Rodríguez-Pose (2022) examine whether the quality of governance moderates the potential relationship between decentralization, measured by the self-and shared ability of a subnational tier of government (Hooghe et al., 2016), and socio-economic outcomes. However, there are many differences between these papers. While this paper examines the effects on national income inequality for 95 countries, Tselios and Rodríguez-Pose (2022) analyse the effects on national and regional poverty and social exclusion in Europe. Both papers use the Kaufmann et al. (2010) indicators to proxy the quality of governance within a country, but Tselios and Rodríguez-Pose (2022) also use Charron et al.'s (2019) indicators to proxy the regional governance quality. This paper examines the role of private investments in the inequality-decentralization relationship, while Tselios and Rodríguez-Pose (2022) analyse potential differences in the poverty-decentralization relationship between urban and rural areas. As for the outcomes, both papers find that decentralization influences socio-economic outcomes. Nevertheless, this paper finds that an increase in decentralization reduces income inequality, but this reduction is stronger for countries with a relatively low quality of governance, since private investments in these countries seem to reduce inequality, whereas Tselios and Rodríguez-Pose (2022) find that increases in decentralization reduce poverty and address social exclusion, but this mainly happens in countries with a high degree of governance quality and, fundamentally, in urban areas.
The remainder of the paper is organized as follows. Section 2 provides the theoretical background of the conditional hypothesis that the effect of political decentralization on income inequality depends on the quality of governance. Section 3 displays the data and the variables and proxies used to test this conditional hypothesis, presenting some descriptive statistics. It also introduces the empirical model and provides its justification. Section 4 is devoted to the results that arise from regression analysis 830 Vassilis Tselios and provides some robustness tests. Section 5 summarizes the main points of the analysis, discusses some potential limitations and suggests some implications for policy.

Political decentralization
Central authorities have the responsibility for policy formulation and implementation, but the problem facing these authorities in the strategic planning process is the selection of the most appropriate instruments to achieve the politically designated targets (Grove, 1978). Since the central authorities find it difficult to control all aspects of the decision-making process, they assign authority for subsets of the instruments to subnational tiers of government (Grove, 1978). This process is known as decentralization and is clearly defined and analysed in the classic model of Tinbergen (1952). According to Sepúlveda and Martínez-Vázquez (2011, p. 321), 'decentralisation is not a policy goal per se, but a national strategy pursued in order to reorganise the operation of the public sector and make it more efficient and accountable to the needs and preferences of the population'. Bardhan (2002, p. 202) argues that: the logic behind decentralization is not just about weakening the central authority, nor is it about preferring local elites to central authority, but it is fundamentally about making governance at the local level more responsive to the felt needs of the large majority of the population.
However, in the absence of a governance and institutional framework, the analysis is incomplete (Oates, 2005). Conceptualizing and measuring political (de)centralization has been a long-standing challenge (Dardanelli, 2019). The regional authority index (RAI), proposed by Hooghe et al. (2016) and Shair-Rosenfield et al. (2021), yields an multidimensional measure of (de)centralization, considering the degree to which regions not only enjoy a considerable degree of political autonomy (self-rule) but also determine central decision-making (shared rule).
Political decentralization has taken on increased importance all over the world over many years. The global trend towards higher levels of political decentralization can influence the targets of public socio-economic policies and, therefore, socio-economic outcomes, such as income inequality. Hence, a key question is: Does an increase in political decentralization reduce income inequality? 2.2. Political decentralization and income inequality: the key role of public officials Whether political decentralization affects income inequality depends on public decision-makers (public officials), that is, on the quality of governance of a country, such as on the level of voice and accountability, political stability, government effectiveness, regulatory quality, rule of law, and control of corruption (Kaufmann et al., 2010). The quality of governance can provide the backdrop for the distribution of political power from the central state to the subnational tiers of government, shaping the distribution of income inequality.
If public officials are benevolent welfare maximizers, political decentralization may affect (either increase or decrease) income inequality, usually through allocative efficiency, preference heterogeneity, regional competition, voice, accountability and participation.
First, the transfer of authority to subnational tiers of government can lead to improvements in the equity of an economy in the long-run (i.e., both between regions and between individuals), as resources are used more efficiently than under a more centralized system (Oates, 1972;Tiebout, 1956). This is due to the information advantage that the local governments have over the central government, allowing them to apply more specific public policies (Kyriacou et al., 2015), and due to the mobility of people who select a local area based on their choices for local public goods and services (Cavusoglu & Dincer, 2015;Tiebout, 1956). Canare et al. (2020) argue that since the regional governments of a country have better information from the ground than the central government, they can design and implement economic policies and they can provide public goods and services that are tailored to local needs, improving social wellbeing across jurisdictions. Neyapti (2006) states that decentralization can lead to more efficient and/or equitable revenue collection, but under good governance. Cavusoglu and Dincer (2015) claim that income inequality is more efficiently managed at a subnational level. Carreras (2016) argues that if regional governments interact with the central government (i.e., more shared rule) and decide together on redistributive issues towards more equity, this is expected to be associated with lower income inequality.
Nevertheless, higher allocative efficiency may increase income inequality because increasing decentralization reduces the resources (in cash and in services) available for redistribution at the central level. According to the classical theory of fiscal federalism (Musgrave, 1959), income redistribution is more efficiently performed by the central government than the regional governments. More regional authority (i.e., more self-rule) hinders the capability of the central government to redistribute in a more standardized way through taxes and transfers and thus the discrepancies of redistribution may lead to higher income inequality (Carreras, 2016). The central government may not be able to extract enough resources, such as tax revenues from income, sales and property, to pay for income redistribution, which implies that the size of government, measured by the share of income redistributed, is small (Meltzer & Richard, 1981). Beramendi (2012) argues that the quality and quantity of redistribution in a political union rests on centralization (i.e., the central government collects taxes and controls interpersonal and interregional redistribution), on decentralization (i.e., the regional governments control taxation and interpersonal income redistribution), and hybridization (i.e., decentralized interpersonal redistribution combines with centralized interregional redistribution).
Second, decentralization is to be preferred when tastes are heterogeneous and there are no spillovers across jurisdictions (Bardhan, 2002;Oates, 1972). Political decentralization can match public spending to the heterogeneous preferences and needs of the citizens living in different regions in terms of their socio-economic, political and cultural characteristics, and to the systematically different standard of living of citizens and the way they live, more appropriately than central governments (Follesdal, 2001;Rodríguez-Pose & Ezcurra, 2010). Decentralization can be justified by differences in local preferences (i.e., demand for local public goods) and information asymmetry (Oates, 2005). Preference heterogeneity, which can be reduced by higher decentralization, accounts for a considerable part of income inequality (Decancq et al., 2017). Hence, decentralization can reduce inequality through the lower heterogeneity of citizens. Pauly (1973) argues that decentralization increases income equality if there is limited mobility of income, capital and citizens between regions. However, heterogeneity also matters for redistribution (Borissov et al., 2020), which is a policy tool of central governments for higher equality. For example, Hodler (2009) finds that redistribution leads to higher equality by transferring income and wealth from high-to low-ability individuals, but lower equality by transferring income from consumption-loving to leisure-loving individuals.
Third, decentralization can mobilize underused resources, creating spatial competition, delivering better socio-economic policies and, thus, reducing inequalities (Rodríguez-Pose & Ezcurra, 2010). However, spatial competition may 'restrain public spending by inducing a competition for lower taxes among subnational units' (Kleider, 2018, p. 357) and thus reduce social provision. This implies that decentralization may lead to a race to the bottom in the provision of public goods, worsening income inequality (Pierson, 1995). Stossberg and Blochliger (2017) point out that fiscal decentralization may raise income inequality by exacerbating regional competition and compromising national government's ability to redistribute. Moreover, regions are endowed with different resources and characteristics (e.g., natural resources) that affect tax revenue and regional competition, and thus income inequality. Fiscal equalization aims to transfer funds to reduce differences in the tax revenues and fiscal capabilities between regions (Brenton, 2020), but it does not necessarily imply lower interpersonal inequalities. According to Kleider (2018), self-rule is a potential driver of competition among regional governments, while shared rule provides incentives for policy coordination.
Fourth, political decentralization can bring government closer to the people, thus promoting a stronger voice, more accountability and more participation. Decentralizing political power and resources to subnational governments increases the accountability and responsibility of public officials to citizens and, at the same time, decreases social solidarity (Khan et al., 2017). Higher responsibilities of subnational governments push local policymakers closer to their citizens, favouring policies more sensitive to the existence of interpersonal inequalities (Sacchi & Salotti, 2014). People can maximize control and self-management over the decisions which have to do with their lifestyle and affect their lives, allowing a cooperative community of equals (Brown, 1975). Decentralization can generate 'greater transparency and bring to the fore differences in the provision of goods and services across jurisdictions' (Rodríguez-Pose & Ezcurra, 2010, p. 624). Moreover, participation empowers the poor, reducing inequalities, because by giving ordinary people a political voice and participation in local decision-making, democracy is thought to strengthen incentives for government performance and generate fundamental benefits for citizens (Casey, 2018).
If public officials have self-interested goals, political decentralization usually aggravates existing income inequalities through elitism and corruption. In the absence of strong local accountability, local vested interests are powerful, increasing social, economic and ethnic fragmentation and inequality (Neyapti, 2006). Local bureaucrats may possess detailed knowledge of any opportunities for corruption that might arise in their jurisdictions (Dincer et al., 2010). This means that the beneficial effects of political decentralization can be undermined by local elites with access to social, economic and political power (Khan et al., 2017). In countries with a low-quality of institutions (e.g., high political corruption), local governments are more prone to capture by local sociopolitical elites, who have strong clientelist networks and receive a disproportionate share of spending on public goods and services (Bardhan, 2002;Rodríguez-Pose et al., 2016). What is more, officials of local governments tend to follow their own political agenda rather than the agenda of their principal (Cavusoglu & Dincer, 2015;Prud'homme, 1995). Any discussion of the delivery of public goods and services 'has to grapple with issues of the capture of governments at different tiers by elite groups more seriously than is the custom in the traditional decentralization literature' (Bardhan, 2002, p. 188). Therefore, if 'public officials have self-interested goals, which they pursue under incentives set by political institutions, leading to behaviours that [often] diverge significantly from social optima' (Khan et al., 2017, p. 328) and are in favour of rich people, decentralization can increase income inequality.
Overall, the quality of governance of a country moderates the impact of a change in decentralization of the country on its income inequality level. However, the governance quality is positively associated with private investments (e.g., the inward foreign direct investment (FDI) and the financial resources provided to the private sector), which, in turn, may affect income inequality. First, the efficiency of institutions in an economy helps to boost private investments, entrepreneurial initiatives and the factor productivity of firms (Cebula et al., 2016;Khanna & Sharma, 2018;Le et al., 2021). Cebula et al. (2016), using data for OECD nations, find that higher quality public regulation and higher economic freedom improve the business environment in which corporations and other private business entities operate, and promote an entrepreneurial spirit and performance. Second, private 832 Vassilis Tselios investments affect income inequality (Giuliani, 2019;Helpman et al., 2004;Sun et al., 2021). On the one hand, multinational enterprises may contribute to high inequality through the abnormal accumulation of wealth by chief executive officers, top managers and shareholders, and on the other hand, they may benefit the less skilled, reducing inequality (Giuliani, 2019;Helpman et al., 2004). Sun et al. (2021) show that inequality in China increased with the reduction of state-owned enterprises, because the wages of these enterprises are determined by a national standardized grading system, which is based on the equality criterion. Using a meta-analysis, Huang et al. (2020) find that FDI is associated with higher inequality for the low-income countries, but with lower inequality for the high-income countries. Figure 1 displays the theoretical framework of this paper, described above.

Data sources and variables
Political decentralization (DEC) is proxied using the RAI for 95 countries over the period 1950-2018, as defined by Hooghe et al. (2016) and Shair-Rosenfield et al. (2021). Regional authority is a regional legitimate power, which is recognized as binding and is derived from accepted principles of regional governance (Dahl, 1968). A regional government may exercise authority in its own jurisdiction and/or exercise authority over a larger jurisdiction (even the whole country) of which it is part (Hooghe et al., 2016;Hooghe & Marks, 2013;Shair-Rosenfield et al., 2021). Hence, the RAI is the sum of both the self-rule (SELF) and the shared rule (SHARED) authority (Elazar, 1987). Self-rule is the authority by a regional government over those within the regional territory, while shared rule is the authority exercised by a regional government or its representatives in the country as a whole (Hooghe et al., 2016). In other words, self-rule measures the degree of independence of the regional governments from the influence of the central authority, while shared rule measures the capacity of the regional government to determine central decision-making (Ezcurra & Rodríguez-Pose, 2013). Self-rule rests on the underlying notion of autonomy applying to individual subcentral units, whereas shared rule rests on that of the influence applying to subcentral units collectively (Dardanelli, 2019). 2 As far as shared rule is concerned, subnational governments are not given direct responsibilities, but they can influence the results of the policies carried out at the central level. With respect to the depth of authority, RAI measures the degree to which a regional government has an independent legislative, fiscal and executive organization (Hooghe et al., 2016;Hooghe & Marks, 2013). The unit of analysis is a country in a given year of evaluation. Country scores aggregate scores for each regional tier and individual regional governments in a country (Hooghe et al., 2016). Hence, the political decentralization of a country refers Does political decentralization affect income inequality? The role of governance quality to the average legitimate power of all regional governments of the country to undertake the political functions of governance (Hooghe et al., 2016;Hooghe & Marks, 2013;Shair-Rosenfield et al., 2021). All political decentralization proxies are transformed into natural logarithmic (ln) form to make them more normally distributed (Tselios & Tompkins, 2017).
Income inequality (INEQ) is measured using the Gini index of disposable income inequality (net income inequality), which is assumed to be the best suited index, for 196 countries from 1960 to 2019, provided by the Standardized World Income Inequality Database (SWIID) (Solt, 2020). The disposable income inequality index is inequality in equivalized (square root scale) household disposable (post-tax, post-transfer) income, using Luxembourg Income Study data as the standard (Solt, 2020). The Gini index ranges from 0 (perfect equality) to 1 (perfect inequality).
The quality of governance (GOV) is proxied using the World Governance Indicators (WGI) for 214 countries for 1996for , 1998for , 2000for -19 constructed by Kaufmann et al. (2010. The governance indicators summarize the views about the capacity of a government to formulate and implement effective and efficient policies. The views are those of a large number of enterprise, citizen and expert survey respondents, gathered from several survey institutes, think tanks, non-governmental organizations, international organizations and private sector firms (Kaufmann et al., 2010). Six dimensions of governance are used: (1) voice and accountability, (2) political stability and absence of violence, (3) government effectiveness, (4) regulatory quality, (5) rule of law and (6) control of corruption (Kaufmann et al., 2010). All these indicators are general subjective indicators of the quality of governance because they are derived from questions asked to enterprise, citizen and expert survey respondents about the broad concepts of voice and accountability, political stability and violence, government effectiveness, regulatory quality, rule of law and corruption. These indicators range from −2.5 (weak governance performance) to 2.5 (strong governance performance) (Kaufmann et al., 2010;Rodríguez-Pose & Tselios, 2019). GOV is proxied using the sum of the six indicators.
Combining these databases, I generate an unbalanced database of 93 countries 3 for 1996, 1998, 2000 and 2002-18. Data on the control variables were obtained from World Bank data (WB), 4 the Penn World Table data (PWT) 5 (Feenstra et al., 2015) and the historical index of ethnic fractionalization index dataset (Drazanova, 2020), which is available until 2013. The control variables deal with some important sources of heterogeneity and make the estimated parameter of political decentralization more precise. The control variables were chosen after considering the existing literature on the determinants of income inequality as well as the year and country-level data availability. The control variables are: (1) gross domestic product (GDP) per capita and its squared term (WB) in order to test the Kuznets' inverted-'U' hypothesis that as an economy develops, market forces first increase and then lower income inequality (Kuznets, 1955); (2) the agriculture, industry and services value-added as a percentage of GDP (WB), which measure the sectoral composition of a country (Rodríguez-Pose & Tselios, 2009); (3) the annual GDP per capita growth (WB) in order to see whether there is a trade-off between economic growth and income inequality (Rodríguez-Pose & Tselios, 2010); (4) total factor productivity (Tfp) (PWT), which is a measure of productive efficiency and technological progress (Hanson & Rose, 1997); (5) the urban population as a percentage of total population (WB), the annual urban population growth (WB) and the population density in ln form (WB), which measures urbanization economies (Castells-Quintana, 2018); (6) trade as a percentage of GDP (WB), because international trade may entail wage and job polarization (Furusawa et al., 2020); (7) expense as a percentage of GDP (WB), which includes cash payments for the operating activities of the government in providing goods and services and, thus, it could be considered as a proxy for the size of government (Dotti, 2020); (8) the unemployment rate (WB), as unemployment has an aggravating impact on income inequality (Cysne & Turchick, 2012); (9) the social contributions as a percentage of revenue (WB), because payments to social insurance schemes (e.g., unemployment insurance benefits and supplements, accident and sickness benefits, family allowances) affect income distribution (Hairault & Langot, 2008); and (10) the ethnic fractionalization index, which measures diversity as a steadily increasing function of the number of groups in a country (Drazanova, 2020). In order to reduce potential heterogeneity bias, due to the missing observations of control variables, I examine the impact of political decentralization on income inequality, initially without and then with the control variables. Finally, the private investment variables, which are obtained from the WB, are (1) the net inflows of FDI as a percentage of GDP; and (2) the monetary sector credit (e.g., loans, purchases of non-equity securities and trade credits) to the private sector as a percentage of GDP. Table 1 reports the number of observations, mean, standard deviation (SD), minimum and maximum values for the variables of the merged database. A first inspection of the unbalanced dataset indicates that the within variation (time series) of the main variables (i.e., income inequality, political decentralization and governance quality) is very low. Most of the variation is cross-country.

Econometric specification
In order to examine the conditional hypothesis that the effect of political decentralization on income inequality within a country depends on the value of the governance quality of the country, after controlling for some characteristics of the country, I will use the following multiplicative interaction model with panel data: Vassilis Tselios at time t; DEC it is political decentralization (ln) for country i at time t, which is the main independent variable; GOV it is governance quality for country i at time t, which is the conditioning variable (moderator); DEC it GOV it refers to a variable as the simple observation-by-observation product of DEC it and GOV it ; CONTR it is a vector of control variables; u t is a vector of time dummies, which controls for all time-specific national invariant variables (such as global technological effects); and u it is the error term. The coefficient on decentralization in this multiplicative interaction model does not indicate the average effects of decentralization as it does in additive models. The marginal effect of decentralization on income inequality is: Hence, the interaction model asserts that the effect of a change in political decentralization on income inequality depends on the value of the governance quality. 6 An instrumental variable (IV) is needed to deal with any endogeneity, because political decentralization might be correlated with the error term. That might be present because: (1) income inequality and political decentralization are simultaneously determined; (2) there are omitted variables that are correlated with both inequality and decentralization; and (3) decentralization is measured with errors. For instance, endogeneity could potentially arise at least because national governments may increase political decentralization in order to reduce household income inequalities. In other words, high-income inequalities are likely to be the cause of higher political decentralization. For example, Hatfield and Miquel (2012, p. 608) argue that 'higher inequality can lead in the constitutional vote to higher decentralisation and, as a consequence, to lower redistribution'. The IV needs to be exogenous, it must be highly correlated with political decentralization, once the other independent variables are controlled for, but not correlated with the error term, and it must be made on the basis of a theory. It should be mentioned here that the time-series variation of the political Does political decentralization affect income inequality? The role of governance quality 835 decentralization variable is very low: changes in decentralization are long-run processes and do not happen frequently and, when they do, they do so in steps (Rodríguez-Pose & Tselios, 2019). Hence, the time-series variation of the IV should also be very low. I use the following IV it for country i at time t: where Size i is the size of country i; and Globalisation t is a globalization index at time t.
. Size i : the size of countries, which is an exogenous variable, can be an effective instrument for decentralization. Political decentralization is much higher in large countries (e.g., the United States and the UK) than in small ones (e.g., Luxembourg), because the heterogeneity of preferences, needs, culture and attitudes of the population is higher in larger countries (Alesina, 2003). Larger countries are more likely to transfer political authority to subnational tiers of government for better matching of public spending to the heterogeneous preferences of their citizens living in different regions than in smaller countries, while government in smaller countries is more likely to be closer to the people, promoting a stronger voice, more accountability and participation than that in larger countries. Moreover, small countries do not need high decentralization because decentralization is to be preferred not only when tastes and needs are heterogeneous but also when there are no spillovers across jurisdictions (Oates, 1972), and spillovers are usually stronger in smaller countries than in larger ones (Bardhan, 2002). The size of a country, which is measured by the natural logarithm (ln) of the area (km 2 ) of the country, is highly correlated with its level of political decentralization (see Figure A1 in Appendix A in the supplemental data online). . Globalisation t : international calls for political decentralization within countries coexist with globalization. As was noted in the introduction, there is a global trend for higher levels of political decentralization, which is reinforced by international organizations. According to the World Bank (1999), decentralization alongside globalization is the main force shaping the world in the first decade of the 21st century (see also Sacchi & Salotti, 2014). In Europe, for instance, calls for decentralization within countries coexist with a drift toward the Europeanization of some relevant functions, such as fiscal policies (Levaggi & Menoncin, 2017). Khan et al. (2017) argue that decentralization is an international movement that happens in almost all the world's countries and regions. And most countries are moving to higher levels of decentralization because decentralizing a country may increase its democratic strength by making it more 'supple' (Faguet et al., 2015). The centralized state in almost all countries has lost a great deal of legitimacy (Bardhan, 2002), and both globalization and decentralization are widely believed to promise a range of benefits for the economies and the societies in general. For instance, global technological change along with globalization help to provide services relatively efficiently in small areas, which implies that lower levels of government have a greater capacity, such as resources, to handle certain tasks and to deliver a broader range of services than before (Bardhan, 2002). Finally, globalization has reached the political sphere, with many countries taking significant steps toward introducing democratic principles and freedom (Work, 2002). Hence, I do not use a globalization index for each country (i.e., Globalisation it ), but a globalization index for all countries (i.e., Globalisation t ), known as world globalization, which is a time-specific country-invariant variable that measures globalization at the international level. However, a distinct feature of globalization is its supra-territoriality (Gygli et al., 2018). To proxy this variable, I use the KOF Globalisation index, which is the most heavily adopted and cited globalization index in the literature (Gygli et al., 2018). This index identifies three dimensions of globalization: economic (trade and financial), social (interpersonal, information and cultural) and political. The KOF Globalisation index increased from 1996 to 2018 (see Figure A2 in Appendix A in the supplemental data online), which shows that the process of globalization has accelerated.
For the estimation of the empirical specification, I use IVs and two-stage least squares (2SLS) for panel data models. More specifically, I use the G2SLS randomeffects estimator with cluster-robust standard errors. In the first-stage model, the political decentralization variable is the dependent variable, while in the second-stage model income inequality is the dependent variable. Potentially effective IV should have a high correlation with the decentralization variable, but a relatively low correlation with the error term of the second stage model. Hence, the IV's influence on income inequality is transmitted uniquely through their effect on decentralization.
Following the empirical study by Barro (2000), the empirical specification does not include country fixed effects, because they would eliminate the cross-country information in the data. A fixed-effects model has the advantage that it reduces the risk of obtaining biased estimation of the regression results, but this is not without cost, because the reduction in bias removes the crosscountry variation from the data, which is likely to affect the efficiency of the parameter estimates (Higgins & Williamson, 1999;Rodríguez-Pose et al., 2012;Tselios & Tompkins, 2019). The empirical specification incorporates random effects, and thus the error terms can be correlated over time within countries (Barro, 2000). Using cluster-robust standard errors, I keep the assumption of zero correlation across countries as with fixed effects, but allow the within-countries correlation to be anything at all. This implies that the results are robust to arbitrary heteroscedasticity and within-country correlation. Moreover, some scholars (Durlauf & Quah, 1999; Griliches & Table 2. Impact of the regional authority index (RAI) on income inequality.
(1)   Mairesse, 1984;Mairesse, 1990) argue that the generalized least squares (GLS) along with the ordinary least squares (OLS) coefficients reflect long-run effects, because they allow for both within-country and cross-country variation from the data. Therefore, the results reflect crosscountry variations as well as within-country variations over time (Barro, 2000). Table 2 presents the G2SLS random-effects regression results of the conditional impact of the RAI on disposable income inequality, with cluster-robust standard errors. 7 The impact of political decentralization on income inequality is examined both without and with the control variables of GDP per capita and its squared term, sectoral composition, GDP per capita growth, Tfp, urban population, growth of urban population, population density (ln), trade, expense, unemployment, social revenue and ethnic fractionalization, in order to test the robustness of the results with the inclusion of these control variables. It should be noted here that not only is political decentralization (RAI(ln)) an endogenous variable, but also the interaction variable (RAI(ln) × GOV) because it includes an endogenous term. Hence, Table 2 reports (1) the first-stage model when RAI(ln) is the dependent variable;

REGRESSION RESULTS
(2) the first-stage model when RAI(ln) × GOV is the dependent variable; and (3) the second-stage model when INEQ is the dependent variable.
The first-stage regressions show that the IV has a statistically significant effect on RAI, and the interaction term of the IV with the governance quality (IV × GOV) has a statistically significant effect on the interaction term of RAI with the governance quality (RAI(ln) × GOV). Thus, the interaction term of the size of the country with the globalization level can serve as an effective instrument for political decentralization. The second-stage regressions show that the coefficient on RAI(ln) is negative and the coefficient on RAI(ln) × GOV is positive, which implies that the higher the governance quality, the lower the negative effect of political decentralization on income inequality. This finding is robust to the proxy for political decentralization, that is, the degree of independence of the regional government from the influence of central authorities, and the capacity of the regional government to determine central decision making (see Table A2 in Appendix A in the supplemental data online). In other Table 3. Impact of governance quality on foreign direct investment (FDI) inflows and the private sector.
(1)  Does political decentralization affect income inequality? The role of governance quality 839 words, political decentralization (RAI, SELF and SHARED) has a negative effect on income inequality, which becomes weaker as governance quality increases. Higher decentralization seems to reduce inequality, which is likely to imply that regional governments have an information advantage over the central government, allowing them to design policies and public goods that are tailored for the local needs of low-income people (Canare et al., 2020;Kyriacou et al., 2015), and that the transfer of authority to subnational tiers of government can bring government closer to the low-income and ordinary people, promoting their voice, and favouring policies for lower interpersonal inequalities (Sacchi & Salotti, 2014). However, political decentralization has more desirable consequences on the income distribution for countries with a relatively low-quality of governance, possibly because private investments, which are positively associated with the quality of governance, lower income inequality in these countries. Indeed, Table 3 shows that the net inflows of FDI within a country and the monetary sector credit to the private sector of the country are positively associated with the quality of governance of the country. High quality institutions seem to boost private investments and entrepreneurial activities (Cebula et al., 2016;Khanna & Sharma, 2018;Le et al., 2021). Table 4 also shows that an increase in the net inflows of FDI or in the monetary sector credit to the private sector reduces income inequality in countries with a relatively low quality of governance. Private investments create jobs which are likely to benefit the less skilled in these countries, reducing income inequality (Helpman et al., 2004). Finally, as for the controls, there is evidence that countries with high social revenue and low ethnic fractionalization are more egalitarian than others (Table 2).
Overall, the results fail to reject the conditional hypothesis that the effect of a central government transferring political authority to subnational tiers of government on income inequality within the country depends on how well the traditions and institutions of an authority in the country are exercised. This finding is robust to the proxy for political decentralization (RAI, SELF and SHARED) as well as to the inclusion of different control variables. I also evaluate the stability of this finding in the measurement of income inequality. I replicate the analysis using the market-income inequality as a dependent variable. 8 The Gini index of market-income inequality 'measures the distribution of income before taxes (e.g., tax on wages, salaries, dividends, interest and rents) and transfers (e.g., publicly provided in-kind transfers, such as public spending on education and health care)' (Tselios & Tompkins, 2019, p. 165). This difference is important for the robustness of the results because countries differ with respect to the level of both taxes and transfers (Tselios & Tompkins, 2019). Although market-income inequality takes taxes and transfers into account, it cannot be considered 'pre-government' because many non-redistributive government policies (e.g., public education) shape the income distribution (Iversen & Stephens, 2008;Solt, 2016;Tselios & Tompkins, 2019). The results show that political decentralization has a negative effect not only on disposable income inequality, but also on market income inequality, which becomes weaker as governance quality increases. Hence, the main finding of this paper is robust to the measurement of income inequality.

CONCLUSIONS
This paper has sought to examine whether the effect of a change in the political decentralization of a country (i.e., the degree of independence of the regional government from the influence of the central authority of the country and/or the capacity of the regional governments of the country to determine central decision making) on household income inequality within the country is contingent on the quality of governance of the country. Since there is a growing global trend for central governments to transfer legitimate power to subnational tiers of government, this paper has explored whether an increase in this transfer reduces income inequality. The results suggest that the decentralized institutions in a country have played a significant role in reducing income inequalities within the country, but this reduction seems to be stronger for countries with a low quality of governance, possibly because private investments (i.e., the net inflows of FDI within a country and the monetary sector credit to the private sector of the country) in these countries reduce income inequality.
There are some limitations that this study was not able to overcome as these limitations result from the data availability and the quality of the data. Hence, some questions remain open to analysis in greater depth. More specifically, although six different indicators of the quality of governance were used, all these indicators summarize the views on the quality of governance of a large number of enterprise, citizen and expert survey respondents, which implies that they are subjective indicators. These indicators might be subject to measurement error as they depend on the subjective view of those who responded to the survey. Despite the fact that these indicators have been widely used by many scholars, it would be valuable to refine the estimates of the analysis by experimenting with alternative proxies for governance quality.
Although restricted by the aggregated nature of the study, some broad policy implications can be discussed. Political decentralization strategies can reduce not only poverty and social exclusion (Tselios & Rodríguez-Pose, 2022), but also income inequalities. Higher responsibilities of subnational governments push local policy makers closer to their citizens. Governance at the local level seems to be more responsive to the socio-economic needs of the large majority of the population, but the social benefits of decentralization are likely to be weaker as governance quality increases. Overall, decentralization-promoting strategies for equity tend to be institution specific, because there is strong evidence that the effect of the higher legitimate power of regional governments to undertake the political functions of governance on ACKNOWLEDGEMENT I thank James Robinson (University of Chicago), the seminar participants at the University of California -Irvine, and the two anonymous referees for their useful comments and suggestions on an earlier version of this paper. I also thank Fulbright for my six months' visit at the Harris School of Public Policy of the University of Chicago.

DISCLOSURE STATEMENT
No potential conflict of interest was reported by the author.

FUNDING
This work was supported by the Fulbright Foundation in Greece.

NOTES
1. 'Political decentralisation refers to the power of subnational governments to undertake the political functions of governance ' (Tselios & Rodríguez-Pose, 2020, p. 5). It involves devolving political power from the central state to elected local/regional administrations, which inherit either direct or indirect responsibility and duties for the design, provision and allocation of specific public services (Khan et al., 2017). Hence, the elected local/regional administrations have self or shared responsibility for the public service provision. 2. For a review of RAI, see Dardanelli (2019 it is only when GOV = 0, and this hypothesis may or may not be supported at other levels of GOV (Braumoeller, 2004). Hence, 'one cannot determine whether a model should include an interaction term simply by looking at the significance of the coefficient on the interaction term' (Brambor et al., 2006, p. 74). 7. The regression results of the unconditional hypothesis do not show evidence that decentralization directly affects inequality (see Table A1 in Appendix A in the supplemental data online). 8. These results can be provided upon request.