Fiscal Decentralization and Regional Economic Growth - SSRN papers

Fiscal Decentralization and Regional Economic Growth: The Case of the Russian Federation

Master Thesis Presented to the Department of Economics at the Rheinische Friedrich-Wilhelms-Universität Bonn

In Partial Fulfilment of the Requirements for the Degree of Master of Science (M.Sc.)

Supervisor: Prof. Dr. Jürgen von Hagen

Submitted in August, 2014 by: Andrey Yushkov Matriculation Number: 2551507

Electronic copy available at: http://ssrn.com/abstract=2518319

Table of Contents 1. Introduction .................................................................................................................................1 2. Fiscal Decentralization and Economic Growth: Theory and Empirical Evidence .....................3 2.1. Fiscal Decentralization: A Conceptual Framework ..........................................................3 2.2. Fiscal Decentralization and Economic Growth: Analytical Framework ...........................9 2.3. Fiscal Decentralization and Economic Growth: Empirical Evidence ............................. 13 2.3.1. Cross-Country Studies ............................................................................................... 13 2.3.2. Regional Studies ........................................................................................................17 3. Fiscal Decentralization and Regional Economic Growth in Russia .........................................20 3.1. Russian Federalism: Historical Overview and Current State ............................................20 3.2. Fiscal Decentralization and Economic Growth in Russian Regions: An Empirical Analysis ....................................................................................................................................26 3.3. Results of the Empirical Analysis ......................................................................................31 3.4. Policy Implications ............................................................................................................36 4. Conclusion .................................................................................................................................38 References .....................................................................................................................................40 Appendix A. Davoodi-Zou Analytical Framework with Two Levels of Government .................43 Appendix B. Empirical Analysis ...................................................................................................45

Electronic copy available at: http://ssrn.com/abstract=2518319

1.

Introduction

Fiscal decentralization is one of the most popular concepts in the modern theory of public finance and a widely used policy measure. Its advantages and disadvantages are a topic of debates among theorists, empiricists and policy-makers around the world. On the one hand, fiscal decentralization can raise allocative efficiency, limit the size of the public sector by horizontal fiscal competition, and make local governments more democratic and accountable to their residents (Oates, 1999; Thiessen, 2003). On the other hand, decentralization may be dangerous since it can increase regional inequalities and create new opportunities for corrupt local officials (Prud’homme, 1995). Therefore, the success of decentralization depends critically on certain institutional, political and socio-economic conditions of a particular country. A possible link between fiscal decentralization and economic growth has not been explicitly considered in the classical theory of fiscal federalism. However, a thorough examination of such a relationship has become relevant since the beginning of the large-scale decentralization reforms in the last quarter of the twentieth century, implemented primarily by the former socialist states (China, former members of the Soviet Union and countries of Eastern Europe). In theory, several influential growth models, including Solow model, Barro’s endogenous growth model and Diamond OLG model, have been modified to encompass a potential influence of fiscal decentralization on steady-state values of economic growth (Brueckner, 2006; Davoodi, Zou, 1998; Thiessen, 2003). These augmented models have been widely used as analytical frameworks for empirical work. However, the results of cross-country and regional empirical studies are very controversial. The sign and significance of fiscal decentralization coefficient in growth regressions depend critically on the chosen measures of decentralization and the set of conditioning variables. Recent studies stress the importance of using both revenue and expenditure dimensions of fiscal decentralization simultaneously. Furthermore, major determinants of growth, summarized in Levine and Renelt (1992), should be necessarily controlled for to overcome potential omitted variable bias. The major goal of the present Master Thesis is to study the link between intra-regional decentralization and regional economic growth for the case of the Russian Federation. In order to achieve this goal, several intermediate steps are performed. First, theoretical foundations and empirical links of the relation between decentralization and growth are studied. Second, the analysis of the current state of the Russian federalism and regional development is conducted; several problems that can prevent efficient decentralization are 1

identified. Third, the empirical model is developed on the basis of Davoodi and Zou (1998) analytical framework. At the same time, data for the analysis is gathered and described. In this regression model economic growth depends on several measures of decentralization (intraregional revenue and expenditure decentralization, the municipal autonomy indicator and regional dependence on federal fiscal transfers), tax pressure and primary determinants of growth, identified in the literature. Fourth, different specifications of the model are estimated using a fixed-effects (FE) panel data analysis. Fifth, potential empirical problems are discussed; several robustness checks are proposed. Bayesian Model Averaging (BMA) approach, which is a widely used instrument in growth empirics, is implemented to show the relevance of the model of interest. Finally, major policy implications from the empirical analysis are suggested. The structure of the Master Thesis is as follows. Chapter 2 presents theoretical foundations of the link between fiscal decentralization and economic growth. In Section 2.1 a conceptual framework of fiscal decentralization is built up: its major advantages and disadvantages, as well as its prerequisites and measurement issues, are discussed. Section 2.2 presents possible channels, through which decentralization may affect growth, and considers three widely cited theoretical models. Also, main determinants of growth, which should be used in the empirical analysis, are summarized. In Section 2.3, an extensive review of crosscountry and regional empirical studies is provided. Chapter 3 is devoted to the analysis of the relation between fiscal decentralization and regional economic growth in the context of the Russian economy. Section 3.1 presents the development of the Russian federalism in 19902000s. In Section 3.2, the empirical analysis is conducted: panel dataset on Russian regions during 2005-2012 is described, and empirical methodology is presented. Section 3.3 summarizes major results of the study, additional specifications and robustness checks, while Section 3.4 presents possible policy implications. The conclusion is the final part of the Master Thesis. Appendix A presents the common analytical framework, based on the model of Davoodi and Zou (1998), for studying the link between decentralization and growth. Appendix B shows the regression tables and key results of the empirical analysis.

2

2. Fiscal Decentralization and Economic Growth: Theory and Empirical Evidence 2.1. Fiscal Decentralization: A Conceptual Framework Decentralization in general can be considered as the devolution of power from the center to the lower levels of decision-making. In a setting of the federal state, it means the transfer of authority from the central government to subnational (regional and local) governments. Fiscal decentralization is the devolution of revenue and expenditure responsibilities, i.e. shifting the power to raise taxes and to set spending priorities to the lower levels of government. Fiscal federalism is a more general concept, which, according to Oates (1999), deals with the vertical financial structure of the government (or public sector) and, in particular, with revenue and expenditure assignments among different levels of government and a system of intergovernmental fiscal relations between them. Thus, on one hand, fiscal decentralization represents an important policy measure within a fiscal federalism framework. On the other hand, it can be thought of as a necessary condition of fiscal federalism, because without (at least some) decentralization there is no vertical structure of the public sector – it is a purely centralized unitary government, which does not actually exist in reality. Two pioneering works in the field of fiscal federalism are monographs by Musgrave (1959) and Oates (1972), in which a normative approach to decentralization is developed. It has been the first serious attempt to build up a theory of the assignment of different functions between central and subnational levels of government. Musgrave (1959) defined three major goals of the public sector in general – efficiency, macroeconomic stability and income redistribution. While central government should be responsible for macroeconomic stabilization and income redistribution, regional and local governments, being better informed about preferences of their residents, can provide various public services within their jurisdictions more efficiently (Oates, 1999). According to the famous Decentralization Theorem (DT), the level of welfare will always be higher (or at least equal) if Pareto-efficient consumption levels of the public good are provided locally in each jurisdiction than if there is any uniform level of consumption maintained by the central government (Oates, 1972). However, the theorem requires an absence of externalities among regions as well as no economies of scale under centralized provision of the public good, which are quite restrictive assumptions in reality (Thiessen, 2000). Moreover, the traditional approach of Musgrave and Oates assumes that the government plays a role of a benevolent social planner or welfaremaximizer. An opposite assumption is presented by Brennan and Buchanan (1980): government officials are not benevolent, but rather behave as rent-seeking Leviathans, who are concentrated on their tax revenue maximization, instead of expenditure efficiency. In fact, 3

both of these theories are criticized due to their extreme assumptions on the role of the government, which are not observed in practice (Ambrosanio, Bordignon, 2006). An issue of expenditure and revenue assignments among different levels of government is probably the most essential problem in fiscal federalism literature. It is almost impossible to assign taxing and spending powers in such a way that three major goals of the public sector are achieved simultaneously. There is always a tradeoff between efficiency, equity and stability (Da Silva et. al, 2009). If primary objectives of the government are to stabilize the economy and reduce horizontal fiscal imbalances, it is necessary to centralize revenue responsibilities. However, if a centralized system is planning to increase efficiency of public services provision and minimize vertical fiscal imbalances, then spending and revenue responsibilities should be decentralized. Expenditure decentralization, thus, is driven primarily by efficiency concerns. Various types of public goods and services can be more efficiently provided by subnational governments, e.g. primary and secondary education, health care, housing and communal services, public transportation and communication, and certain types of social security. Being closer to citizens and possessing better knowledge about their preferences, regional and local governments can provide the most suitable mix of public services at lower costs. However, stability concerns and economies of scale force the federal government to have a monopoly over certain types of expenditures – equalization of resources among rich and poor regions, national defense, security and other national public services provision (e.g., higher education, hi-tech medical services). However, efficient decentralization requires not only a clever distribution of spending responsibilities among different levels of government, but also a close match between expenditure and revenue assignments (Gemmel, Kneller, Sanz, 2013; Oates, 1972). If this match is present, subnational governments will have incentives to preserve markets and foster their development (Jin, Qian, Weingast, 2005; Weingast, 1995). Otherwise, if expenditure responsibilities of regional and local governments are not self-financed by subnational taxes, but rather depend on intergovernmental grants, it may ruin positive incentives, since regions and municipalities have almost no benefits from developing local markets (as regional market development may lead to an increase in own tax potential and a proportional decrease in intergovernmental fiscal transfers). Hence, another concern of fiscal federalism literature is an efficient revenue assignment among different levels of government (Ambrosanio, Bordignon, 2006; Bird, 2011). Since income redistribution is a primary goal of the federal government, it should not decentralize taxes on income due to equity considerations. Thus, personal income tax is administered at 4

the federal level in most countries while revenues from this tax could be shared among lower levels of government. Corporate profit tax is also a poor candidate for being a decentralized one, since its assignment to lower levels of government can cause predatory competition among regions for tax revenues. Value added tax (VAT) and excises on natural resources are usually federal taxes because of equity and stability concerns. Property taxes on individuals and firms are often decentralized, being a major source of regional and local tax revenues. Bird (2011) summarizes in his study the main characteristics of subnational taxation:  Tax base should not be mobile to ensure possible flexibility in tax rates and to prevent capital outflow;  Tax revenues should be relatively stable and sufficient to finance a larger part of subnational expenditures;  A majority of taxpayers should be local residents, who observe benefits from paying taxes and perceive taxes as fair;  Collecting subnational taxes should be manageable and not too costly. In large federal states with many geographically, socio-economically and politically heterogeneous regions and a strong central government, the problem of horizontal and vertical fiscal imbalances leads to a greater role of the system of intergovernmental grants in supporting efficient decentralization. There exist two major groups of intergovernmental fiscal transfers: unconditional (equalization, block) grants and conditional (matching) grants. Unconditional grants are meant to equalize fiscal capacity among regions. Matching grants are used to co-finance different regional expenditure programs. As Oates (1972) points out, matching grants are needed, inter alia, to overcome interregional spillover effects while unconditional grants are meant to minimize horizontal imbalances. In reality, it is a difficult task to implement sound and efficient decentralization1 as there are certain institutional, economic and political barriers which may hinder from the desired goals of efficiency, equity and stability. Thus, it is necessary to identify and analyze potential advantages and disadvantages of decentralization. Arguments for and against decentralization The major theoretical argument for decentralization is an increase in consumer and producer efficiency (Martinez-Vacquez, McNab, 2003). Consumer (allocative) efficiency is higher for the decentralized provision of public goods, since subnational governments can increase welfare of residents by matching their individual preferences. A unified level of the 1

Sound decentralization, according to Dabla-Norris (2006), is conductive to macroeconomic stability; efficient decentralization enhances microeconomic efficiency (input/output mix of public service delivery).

5

public good provision, ensured by the central government, cannot satisfy special tastes of local taxpayers. Thus, it is associated with welfare losses compared to the decentralized provision of the same good. If citizens’ mobility is assumed to be high among regions and municipalities, it can lead to further improvements in resource allocation under decentralization. According to Tiebout (1956), mobile individuals can vote with their feet and choose the best jurisdiction in terms of their preferred tax-benefit package. Thus, regions and municipalities will be forced to compete for residents, and this horizontal competition may positively influence citizens’ satisfaction and overall allocative efficiency. Producer efficiency is also higher in decentralized systems since subnational governments, having knowledge advantages and experience in matching local preferences, can produce public goods and services at lower costs. Indeed, Oates (1999) and Thiessen (2003) argue that devolution of responsibilities to lower levels of government increases accountability of local officials and forces them to find innovative and cost-efficient ways to produce public goods and deliver them to citizens. Thiessen (2003) calls it “productivity enhancement hypothesis”. Increased horizontal and vertical fiscal competition, being a potential result of decentralization, can limit the size of the public sector and help restraint the predatory state – Leviathan (Brennan, Buchanan, 1980). If fiscal decentralization is accompanied by strong democratic institutions (transparent local elections, effective national parliament), it increases accountability of local officials, helps prevent corruption and improves the overall quality of government (Lockwood, 2005; Seabright, 1996). On the contrary, decentralization can be harmful in particular circumstances, for instance, in developing and transition economies. A critical overview of the dangers of decentralization can be found in (Prud’homme, 1995). Two major goals of the public sector, macroeconomic stability and income redistribution, are almost unachievable under full decentralization. Macroeconomic stabilization becomes problematic during periods of recession because the federal government does not have resources to stabilize the economy while powerful regional may have differing, often contradictory, fiscal policy priorities. Income redistribution also does not work under excessive decentralization. Resources are usually unevenly distributed among territories (at least in large federal states). Therefore, lack of centralized equalization policy can lead to bankruptcies of poor regions (Thiessen, 2003). Horizontal fiscal competition, which is usually regarded as one of the major advantages of decentralization, can be harmful as well and cause fiscal imbalances. For instance, if a rich region decreases corporate tax rate in order to attract new entrepreneurs and investors, its 6

neighboring territories may suffer from the outflow of capital and tax base destruction. Thus, horizontal competition also can increase inequality among regions. Subnational governments’ quality is usually a central problem in transition economies since they do not possess enough intellectual and informational resources to implement decisions efficiently. Moreover, if local officials are not elected, but are appointed by the central government, their accountability decreases. As a result, lower accountability and lack of administrative control from the central government create a room for corruption and opportunistic behavior (Rodriguez-Pose, Kroijer, 2009). Some authors claim that fiscal decentralization is neither beneficial nor harmful per se. A success of decentralization depends crucially on the proper design and realization of intergovernmental reforms (Da Silva et al., 2009). For instance, identical levels of expenditure decentralization in two similar countries can be accompanied by different levels of revenue decentralization, thus creating different incentives for subnational governments. Prerequisites for sound and efficient decentralization The effects of decentralization (positive and negative) differ among countries depending on their institutional, political and economic conditions. However, there are certain prerequisites for sound and efficient decentralization, which policy-makers should take into account. As discussed above, there is almost a consensus among researchers that macroeconomic stabilization and income redistribution should not be subject to decentralization because only central governments can implement them efficiently and equitably. It is pointed out in the literature that decentralization can be considered as a superior good, and its costs are justified only for states with a high level of economic development (Bahl, Linn, 1992; Martinez-Vazquez, McNab, 2003). Thus, attaining a particular level of development may be a requirement for successful decentralization. Also, it makes sense to implement fiscal decentralization only if a country consists of (at least) several economically or ethnically diverse regions (Thiessen, 2000). Otherwise, if a state is characterized by low per capita income or a small number of homogeneous regions, then centralized provision of local public goods can be more efficient. Sound system of intergovernmental grants (IGG) can also be viewed as a prerequisite for decentralization. In the absence of strict rules for fiscal equalization and bailouts of subnational governments, it is almost impossible to maintain financial discipline of the lower levels of government (soft budget constraint problem (Kornai, Maskin, Roland, 2003)). Thus, it can be dangerous to decentralize revenue and spending responsibilities to subnational 7

governments without sound IGG system since decentralization can cause increasing budget deficits and failure of the macroeconomic stabilization goal. Dabla-Norris (2006) determines three major prerequisites for sound and efficient decentralization – clarity, autonomy and well-functioning institutions. Clarity means transparent and stable budget framework – viable distribution of tax and spending assignments among different levels of government, clear IGG system and borrowing rules. Autonomy implies both a possibility for subnational governments to self-finance most of their public projects and an opportunity to spend their resources flexibly (within certain limits). Particular well-functioning institutions, which are crucial for decentralization prospects, include democratic representation (fair federal, regional and local elections), close and effective collaboration between central and subnational governments, maintenance or improvement of the quality of budget process, etc. Measurement of fiscal decentralization Before turning to the analysis of the relation between fiscal decentralization and economic growth, the issue of measurement has to be discussed. It is widely assumed that fiscal decentralization is a multidimensional phenomenon (Martinez-Vazquez, McNab, 2003). It includes revenue decentralization, expenditure decentralization (and discretion over spending purposes), different autonomy and self-dependence indicators. Therefore, a proper analysis of fiscal decentralization should possibly encompass its major dimensions. Some most widely used measures of fiscal decentralization include:  Subnational revenues (with or without IGG from other governments) as a share of total government revenues;  Subnational expenditures (with or without IGG to other governments) as a share of total government expenditures;  A share of subnational revenues, generated by own subnational taxes (or by own and shared taxes), over which regional and local governments have high discretion;  A share of IGG from the central government in subnational revenues (dependence on central government);  A share of subnational expenditures that is self-financed;  The difference between expenditure and revenue decentralization (to check the convergence hypothesis), etc. Choosing a particular measure of decentralization (and ignoring some other dimensions of this concept), as we can see in the empirical section of this chapter (2.3), can greatly 8

influence the results and conclusions made by researchers. Hence, an understanding of this multidimensionality of fiscal decentralization should be taken into account not only in theoretical discussions, but also in empirical research. 2.2. Fiscal Decentralization and Economic Growth: Analytical Framework The traditional theory of fiscal decentralization, concerned with efficiency, macroeconomic stability and income redistribution, has not considered the question of a direct relationship between decentralization and economic growth. Higher levels of decentralization can have a positive influence on efficiency (as decision-making authority is closer to citizens), but the goals of macroeconomic stability and equity become difficult to achieve, as regions are usually heterogeneous in terms of economic potential, costs of public services and overall standards of living. Thus, benefits of decentralization depend on whether gains from efficiency compensate losses from increasing instability and inequity. Recent studies examine the link between decentralization and growth more closely. The first extensive treatment of this issue is done by Martinez-Vazquez and McNab (2003). The authors argue that a bi-directional causality may exist between fiscal decentralization and economic growth. The causal link from growth to decentralization is considered by Bahl and Linn (1992). According to their study, decentralization may be considered as a superior good, i.e., it is more desirable and beneficial for countries with a particularly high level of economic development. In other words, economic growth in the past, which has led countries to being developed and prosperous, can have a positive influence on decentralization (both on its scope and success). Still, most of the theoretical and empirical studies examine the link from decentralization to economic growth. However, the initial level of economic development is usually controlled for in the models (Davoodi, Zou, 1998; Thiessen, 2003), as it is widely assumed that advanced economies tend to grow slower than their developing counterparts (conditional convergence hypothesis). The linkage between fiscal decentralization and economic growth can be either direct or indirect. A direct link is presented by Oates (1993). More probably, if it is not direct, then there may be several channels through which decentralization may affect growth. MartinezVazquez and McNab (2003) identify main economic indicators, which can, at least in theory, be affected by decentralization. These indicators have already been discussed in the first section and include efficiency, stability, income redistribution, horizontal fiscal competition, and political phenomena (corruption, capture and democracy). However, the relation between decentralization and each of these measures is unclear; it can be positive in some circumstances and negative in others. Thus, it is necessary to check empirically the direction 9

and robustness of two relationships: between decentralization and one of the indicators, and between this indicator and economic growth. If both of them are significant and causal, then an indirect impact of decentralization on growth is established. But the authors show that it may be complicated to establish such a link. For instance, there is almost a consensus that decentralization increases consumer efficiency. But is consumer efficiency directly related to growth? Not necessarily, because improved individual welfare is usually not equal to economic growth. In contrast, producer efficiency leads to an increase in quality or quantity of local public goods, thus promoting growth, but decentralization may not increase producer efficiency, as local officials are less qualified in reality than federal government employees. Thus, in establishing indirect links between decentralization and growth it is necessary to make certain assumptions. Feld, Baskaran and Schnellenbach (2007) specify four potential channels of the indirect relation between fiscal decentralization and economic growth: heterogeneity, market preservation, structural change and political innovation channels. Heterogeneity channel is related to the work of Brueckner (2006), which is discussed below. Market preservation channel implies that fiscal decentralization increases tax competition among regions, which, in turn, restricts Leviathan-type governments and creates better conditions for market development, thus contributing to economic growth. According to the authors, decentralization can also affect growth through a structural change channel. It means that during the period of a structural crisis (e.g., a particular industry faces permanent negative demand shocks), structural changes are easier to implement under federalism than in a unitary system, where risk-averse officials would oversupply incumbent industries with financial support, even if they are no longer efficient (Besley, Coate, 2003). Thus, decentralization may stimulate faster recovery and growth after structural shocks. Political innovation channel is the last one. The basic argument is that regions may be used as laboratories for economic experiments under federalism (Oates, 1999). If a policy innovation is successful in one region, it may be further disseminated among other regions. Rauscher (2006), within his theoretical framework, argues that decentralization may indeed be followed by an increase in experimentation and policy innovation under particular assumptions, e.g., sufficient heterogeneity among regions (Strumf, 1997). Decentralization in growth models Theoretical foundations of the relation between fiscal decentralization and economic grows rest on different growth models. The most common analytical framework that links expenditure decentralization to growth is a model developed by Davoodi and Zou (1998), 10

which is a modified version of Barro’s endogenous growth model (Barro, 1990). A CobbDouglas production function has two inputs – private capital and public spending by three levels of government – federal, state and local. Public expenditures are financed entirely through taxes on output. The authors show that maximizing a representative agent’s utility function with respect to a dynamic budget constraint gives a solution for the growth rate of output, which depends on the shares of different government levels in total public expenditures. Finally, it is possible to calculate from the model growth-maximizing shares of public spending. Davoodi and Zou (1998) conclude the following: whenever actual government shares are different from growth-maximizing ones, it is possible to increase output growth without changing the total level of public expenditures (for instance, if expenditures are excessively centralized, decentralization can be conductive to economic growth). In Appendix A, a simplified version of this general model, which is suitable for the present empirical analysis, is derived. Several versions of this model are used as a framework for the empirical analysis in the studies of Cantarero and Perez Gonzalez (2009), Xie, Zou and Davoodi (1999) and Zhang and Zou (1998). Another analytical framework, which is the basis for several empirical studies, is a model of Mankiw, Romer and Weil (1992), often denoted as MRW, based on the Solow growth model (Solow, 1956). The authors derive from the augmented Solow model an empirical specification, in which output per worker growth rate is a function of the initial value of output and determinants of the steady-state value of output, namely, human and physical capital accumulation rates and labor force growth rate. To study the impact of decentralization and other institutional factors on economic growth within this framework, Thiessen (2003) augments the MRW empirical specification with fiscal decentralization measures and conditioning factors. Lin and Liu (2000) also use a similar production-function based empirical specification. Brueckner (1999, 2006) adopts an overlapping-generations (OLG) endogenous growth model to compare growth rates under decentralized and centralized systems. Here a model from the latter article (Brueckner, 2006) is presented. A hypothetical world at time t consists of two overlapping generations – young and old (each agent lives for two periods, being young in the first and old in the second). Young individuals can invest part of their time in education since it raises their future income, and work the rest of the time. Also, a young generation can save a share of their income and invest it in physical capital. Old individuals devote all their time to work. A consumption bundle of each generation consists of two goods – private and public. Old generation, whose disposable income is higher (as their level of human capital is higher, and they do not spend their time on schooling), can consume more, 11

thus having higher demand for the public good. Further, Brueckner compares two systems – decentralized (federalism) and centralized (unitary). Under federalism, it is assumed that a perfect Tiebout-sorting mechanism (Tiebout, 1956) allows individuals to sort themselves in two demand-homogeneous jurisdictions with different levels of the public good provision (higher for old than for young). Under the unitary system, a common level of the public good is provided for all individuals. According to the proposition presented by the author, time spent on education and levels of physical capital are higher in the federalist equilibrium than in any unitary equilibrium (since multiple unitary equilibria may exist). Economic growth, determined by human capital growth rate, is hence faster under federalism. This model, even being too abstract to be implemented empirically, provides insights on how federalism (in the form of decentralized public good provision) may positively influence growth through heterogeneity channel, mentioned above. In the present Master Thesis a Davoodi-Zou model is used as an underlying analytical framework, since it shows in the most coherent way how expenditure decentralization may affect economic growth. However, an empirical model used in Chapter 3 is augmented by several dimensions of decentralization and a proper set of control variables. It is clear that fiscal decentralization, being an important institutional factor, is not a primary determinant of economic growth. Therefore, in order to avoid the problem of omitted variable bias in the empirical analysis, it is crucial to include a set of proper control variables in growth regressions. Levine and Renelt (1992) suggest several major determinants of growth, including the share of investment in GDP, population growth, initial levels of GDP per capita and human capital as well as a share of international trade in GDP (openness index). Omitting these variables from the analysis can bias estimates and lead to a wrong conclusion about the relation between fiscal decentralization and growth. Using almost the same conditioning variables in the regression analysis, Mankiw, Romer and Weil (1992) confirm in their study a conditional convergence hypothesis (lower initial levels of GDP per capita are associated with higher growth rates, assuming that other explanatory variables are held constant). Barro (2003) conducts an empirical analysis of the determinants of economic growth. He also confirms a conditional convergence hypothesis and finds that higher investment to GDP ratio, higher life expectancy (as a measure of health), a lower ratio of government consumption to GDP, and lower fertility are significantly associated with higher growth. Inflation and international openness ratio are only weakly related to growth. Democracy index has a nonlinear effect on economic growth, while the rule of law is positively related to growth. 12

2.3.

Fiscal Decentralization and Economic Growth: Empirical Evidence

Over the last decade, empirical analysis of the relation between fiscal decentralization and economic growth has become very popular among economists. Many studies are devoted to this question both in cross-country and regional perspective. Feld, Baskaran and Schnellenbach (2007) summarize major results of previous research using a meta-analysis methodology. However, their extensive treatment does not cover studies after 2007, which underlines the need for the updated empirical literature review. In this section, it is necessary to focus on empirical methods and techniques, results of the studies and robustness checks. For the purpose of the further analysis of decentralization in Russia, regional studies seem to be more relevant. However, cross-country studies give some critical methodological insights for the empirical analysis in Chapter 2. 2.3.1. Cross-Country Studies It is worth mentioning that cross-country studies are more popular than regional studies. The first reason, naturally, is the lack of reliable data on a regional level for many developing countries. At the same time, IMF, World Bank and OECD provide rich panel datasets for major developed as well as developing economies. Not surprisingly, most of the cross-country studies are devoted to empirical assessment of the relationship between decentralization and growth in OECD countries (based on IMF Government Finance Statistics). However, countries with incomparable socio-economic, historical and institutional backgrounds are often added to the sample in cross-country studies. Some authors find that it can be an important disadvantage of these studies as their results do not reveal the correct relationship between fiscal decentralization and economic growth (Akai, Sakata, 2002; Jin, Zou, 2005). One of the first studies on the link between fiscal decentralization and growth is the article by Davoodi and Zou (1998). In their empirical part, the authors use panel data analysis for 46 countries during the period of 1970-1989. The regression results show that fiscal decentralization (measured by the share of subnational expenditure in total government spending) is negatively related to economic growth in developing countries, but has almost no influence in case of developed economies. The authors provide several possible explanations for this negative relationship in developing countries: subnational governments may not allocate their expenditure in a growth-maximizing way (for instance, spending funds for different welfare programs instead of investing in infrastructure); revenue assignment may be wrong (subnational governments do not have enough taxing powers); local governments may 13

be less accountable to citizens (officials are appointed, but not elected, or mobility of people is restricted due to low income). Still, the authors do not verify all these explanations explicitly, thus it is an open question, why exactly is fiscal decentralization detrimental to growth in the study. Thiessen (2000, 2003) contributes to the empirical literature in several important directions. Firstly, using the sample of high-income OECD countries and selected developing countries, he finds, contrary to Davoodi and Zou (1998), that several measures of decentralization (both expenditure and revenue) are positively associated with growth. Secondly, the author reports that a nonlinear relation2 between fiscal decentralization and economic growth is present only for high-income states, while for states with lower income this relation seems to be linear. In other words, it is possible to find a growth-maximizing level of decentralization for rich countries. Thirdly, Thiessen considers the possibility of both direct and indirect influence of decentralization on growth (the latter – through changes in total factor productivity growth and investment share in GDP) and confirms that the indirect link between fiscal decentralization and growth is positive for high-income states. Thornton (2007) stresses the importance of finding a proper measure of fiscal decentralization. He proposes, following Stegarescu (2005), to use a share of subnational revenue over which lower levels of government have full autonomy as a measure of revenue decentralization. In his cross-sectional analysis Thornton finds that this indicator of decentralization has no significant effect on growth in 19 OECD countries. The disadvantage of this study is that only a limited number of conditioning variables are included in the regression (e.g., tax burden, inflation and openness indicator are not controlled for). Due to a limited number of observations in the sample and no time series dimension, it is not possible to include country-specific effects and different controls to the analysis. Baskaran and Feld (2012) enrich and improve the analysis of Thornton (2007) by using several measures of decentralization, including “own tax decentralization” measure, within a fixed-effects panel data framework for 23 OECD countries during 1975-2008. The authors argue that GFS-style revenue decentralization indicator, estimated by the IMF (available in Government Finance Statistics Database) and used in many previous studies, is imprecise and leads to biased estimation results since it measures a share of total subnational revenues in overall government revenues. This GFS-style measure is insignificant in the analysis, while the Stegarescu-style own tax decentralization measure, calculated by the authors, has a

2

Instead of taking decentralization variable squared to test for non-linearity (“hump-shaped relation” in his terminology), Thiessen constructs his own index – countries with the middle level of decentralization get higher scores, while countries with high or low levels of decentralization receive lower scores.

14

significantly negative effect on growth. Several robustness checks, e.g., including additional control variables, using a different proxy of economic growth – labor productivity growth, implementing instrumental variables (IV) technique to account for potential endogeneity of decentralization measure, show that the results of the study are robust – fiscal decentralization is negatively related to economic growth. Buser (2011), studying 20 high-income OECD countries during 1972-2005, reports that revenue autonomy of subnational governments has a significantly positive effect on economic growth, after controlling for institutional heterogeneity among countries in the sample. She includes an interaction term between fiscal decentralization and institutional quality (measured by Economic Freedom Index of Frazier Institute) into the regression analysis (pooled OLS and fixed effects model). This term is significantly larger than zero, meaning that positive effects of decentralization are higher when a country has sound institutional conditions for well-functioning markets. Iimi (2005) also controls for an institutional environment (namely, for political freedom, measured by Freedom House). Using a crosscountry dataset for the period of 1997-2001, he finds a positive relation between fiscal decentralization and economic growth. However, the interaction term of decentralization and political freedom has a negative sign. Analyzing decentralization experience in Central and Eastern Europe (16 countries during 1990-2004), Rodriguez-Pose and Kroijer (2009) find that expenditure decentralization has a negative effect on growth while revenue decentralization is positively related to the long-run growth prospects. Since expenditure decentralization is higher than its revenue counterpart in the countries under consideration, the authors confirm a convergence hypothesis (decentralization efficiency is higher when there is a close match between subnational spending and revenue) and conclude that countries of the former socialist bloc should give more revenue sources (e.g., suitable taxes) to lower levels of government and reduce their dependency on intergovernmental transfers. On the other hand, Rodriguez-Pose and Ezcurra (2011) report a negative relation between both expenditure and revenue decentralization and economic growth in the sample of 21 OECD countries during 19902005. As a robustness check, a distinction between current and capital expenditure (as well as between different categories of government spending) is made by the authors. They show that the relation of interest holds negative, except for the case when the level of decentralization is low (then, an increase in current expenditure may be growth-enhancing). In a recent article, Gemmel, Kneller and Sanz (2013) strengthen the findings of Rodriguez-Pose and Kroijer (2009), reporting similar results for OECD countries during the period of 1970-2005. Using a dynamic panel method (pooled mean group), they argue that 15

increasing revenue decentralization and simultaneously lowering expenditure decentralization can be growth-enhancing for the respective countries in the sample. Therefore, according to newest empirical studies, reducing the gap between subnational revenue and expenditure assignments can have a significantly positive effect on economic growth. Asatryan and Feld (2013) implement a perspective Bayesian Model Averaging (BMA) approach to study the robustness of the relationship between fiscal decentralization and economic growth for 23 OECD countries during 1975-2000. Several versions of the model averaging methodology have been used in empirical growth literature to resolve model uncertainty and find out major determinants of growth (Moral-Benito, 2012; Sala-i-Martin, 1997; Sala-i-Martin et al., 2004). Using the BMA approach3, the authors assume that there is uncertainty about the “true” model of economic growth, but many potential determinants of growth, including decentralization, are available. Then, a full set of regression models is considered, in which economic growth is regressed on a constant term and any subset of its potential determinants (so, if there are N potential regressors, then there are 2N possible models). Within BMA framework, it is possible to divide all regressors in two groups – focus and auxiliary. Focus regressors are always included, due to theoretical or practical considerations, for example, the authors include time and country fixed effects to three specifications as focus variables. For auxiliary regressors, unconditional BMA estimates are calculated as a weighted average of the estimates from each of the 2N possible linear models. Estimated coefficients, their t-statistics and posterior inclusion probabilities (pip) are reported as a result of the analysis. Asatryan and Feld (2013), reporting high pip for Levine-Renelt conditioning variables and some other determinants of growth, fail to find a statistically significant link between decentralization and growth. Finally, the authors question the validity of the previous research that has not dealt with the model uncertainty issue. It is necessary to point out that the results of the cross-country studies are mixed. Recent evidence suggests that it is important to take into account model uncertainty and check whether a relation between decentralization and growth is robust in a variety of models using non-standard methods, for instance, model averaging techniques (Asatryan, Feld, 2013). However, setting model uncertainty considerations aside, an important statement that a close match between revenue and expenditure assignments is needed for efficient and growthenhancing decentralization, firstly proposed by Oates (1972), seems to be confirmed (either intentionally or not) by the group of researchers.

3

For a theoretical discussion of BMA and WALS averaging techniques and their application to growth empirics, see Magnus, Powel and Pruefer (2010).

16

2.3.2. Regional Studies The most widely used cases in the regional studies of fiscal decentralization are China and the United States. These two countries, being major world economies in terms of GDP, possess large territories (which are divided into 33 provincial administrative divisions and 50 states, respectively) and population. China’s decentralization experience and its relation to growth are analyzed in the articles by Zhang and Zou (1998), Lin and Liu (2000) and Jin and Zou (2005). The relationship between fiscal decentralization and economic growth in the USA is considered by Xie, Zou and Davoodi (1999), Akai and Sakata (2002) and Akai, Hosoi and Nishimura (2009). Other countries examined in regional studies are Spain (Cantarero, Perez Gonzalez, 2009), Turkey (Tosun, Yilmaz, 2008), India (Zhang, Zou, 2001), Australia (Bodman, Heaton, Hodge, 2009), and Russia (Desai, Freinkman, Goldberg, 20054). Several other studies on Germany, Switzerland and Eastern European countries consider issues of fiscal decentralization in a more general context. Zhang and Zou (1998) analyze the link between expenditure decentralization and growth in 28 Chinese provinces during 1980-1992 and find a significantly negative relationship. Their major results remain the same after implementing several robustness checks, for instance, choosing different time periods, analyzing decentralization of specific government expenditure types, using random effects and averaging instead of the initial LSDV (least squares dummy variables) framework. The authors conclude that, conditional on the low level of provincial economic development, the Chinese central government seems to be “in a better position to undertake public investment with nation-wide externalities” (Zhang, Zou, 1998, p. 237). But the major methodological disadvantage of this study is that the authors examine only the expenditure dimension of decentralization, leaving aside other possible measures of it (revenue, autonomy, etc.). This problem is a typical drawback of the earlier studies of this issue – for instance, Davoodi and Zou (1998) also use a single dimension of decentralization. Working with a similar framework for India during 1970-1994, Zhang and Zou (2001) report the opposite relationship – different measures of decentralization (both revenue and expenditure) are positively associated with economic growth, however most of them are insignificant. Using a panel dataset of 28 Chinese provinces from 1970 to 1993 and a revenue-side measure of decentralization – a marginal retention rate of self-collected revenues by 4

The article by Desai, Freinkman and Goldberg (2005) is considered in the next Chapter.

17

provincial governments, Lin and Liu (2000) contradict the results obtained in (Zhang, Zou, 1998), as a positive link between decentralization and growth is found. However, their empirical approach is also questionable, since it is assumed that there was completely no decentralization before 1985. During 1985-1987, the marginal retention rate quickly changed from 0 to 100 percent for half of the provinces; after 1987, almost all provinces had a fiscal decentralization indicator equal to 100 percent. In other words, it seems that the authors do not analyze the relation between the level of revenue decentralization and growth per se, but rather test for the impact of the structural break in central government policy (namely, decentralization reforms of the 1980s) on GDP per capita growth. Xie, Zou and Davoodi (1999), using a framework of Davoodi and Zou (1998), analyze the relationship between the shares of different government levels (federal, state and local) in total public spending and economic growth using time series data for the USA during 19481994. The authors find no significant relationship and show that the existing shares are consistent with the aim of the growth maximization. Working with the sample of 50 states of the USA in 1997-2001 and using both crosssection (averaging technique) and panel data analysis, Akai and Sakata (2002) argue that revenue and expenditure decentralization are significantly positively related to economic growth. It is worth mentioning that the authors examine the relation between intra-state decentralization (shares of the local budget in consolidated state and local budget revenues and expenditures) and gross state product growth rates. This approach to decentralization is also used in the empirical section of the present Master Thesis. In a recent study, Akai, Hosoi and Nishimura (2009) confirm previous findings of Akai and Sakata (2002) for the case of the USA. The authors show empirically that fiscal decentralization is associated not only with higher regional economic growth, but also with lower volatility of growth. Thus, according to the authors, decentralization was conducive to stable growth for the case of 50 states of the USA in 1992-1997. Jin and Zou (2005) concentrate on the relation between convergence of revenue and expenditure decentralization and economic growth. They find that divergence of revenue and expenditure assignments has promoted economic growth in Chinese provinces under both fiscal contract system (1979-1993) and revenue assignment system (1994-1999). Thus, the traditional argument that a close match between subnational revenues and expenditures is conductive to economic growth is rejected in this study for the case of China. Tosun and Yilmaz (2008) study the Turkish decentralization experience and its relation to growth and development. The authors use questionable measures of decentralization – number of local governments per capita and per square mile. They find a negative relationship 18

between decentralization and growth, but it is not clear whether their measures, which represent geographical fragmentation of municipalities rather than their fiscal autonomy, are correlated with fiscal decentralization measures. Cantarero and Perez Gonzalez (2009), analyzing decentralization in Spain (17 regions in 1985-2004), show that a significantly positive relationship exists only between revenue decentralization and regional economic growth. Expenditure decentralization does not have a significant effect on growth. The authors use several robustness checks to confirm their findings. They perform instrumental variables (IV) technique (two-stage least squares procedure) to account for possible endogeneity between decentralization measures and economic growth. Instrumenting decentralization by its lagged values does not change its significance. Also, a dynamic panel data model (Arellano-Bond method) is estimated to control for the inclusion of the lagged dependent variable in the set of regressors. It confirms the major results of the study. Bodman, Heaton and Hodge (2009) implement a Bayesian Model Averaging methodology to examine the significance of 16 different decentralization measures in growth regressions for the case of Australia during 1972-2005 (time series data is used). Contrary to the results of Asatryan and Feld (2013), the authors stress that several revenue and expenditure decentralization indicators have high posterior inclusion probability and can be considered as important determinants of Australian economic growth, namely, the state share of total government expenditures, subnational (state and local) share in total government revenues, and the tax autonomy indicator. The most significant determinants of growth are population growth, human capital development, tax pressure, the openness ratio, and unemployment, which is perfectly in line with growth theory. Based on this extensive review of regional studies, it is possible to argue that the estimated relation between fiscal decentralization and growth depends crucially on the choice of decentralization measures and conditioning variables. Ideally, several dimensions of decentralization should be examined simultaneously. Early studies, however, fail to follow this principle and use either expenditure or revenue decentralization alone. Also some empirical papers suffer from omitted variable bias, as major determinants of economic growth are not controlled for. These lessons from regional studies will be taken into account in the next chapter, which is devoted to the empirical analysis of the relation between decentralization and regional economic growth for the case of Russia.

19

3. Fiscal Decentralization and Regional Economic Growth in Russia 3.1. Russian Federalism: Historical Overview and Current State Historical overview Since the fall of the Soviet Union in 1991, Russia has undergone a painful transition period – from a highly centralized socialist state through chaotic decentralization to a federal republic with several levels of government, de-jure independent in their own responsibilities. In the report Intergovernmental Relations in the Russian Federation by the World Bank (Da Silva et al., 2009), several important stages in the development of Russian federalism during 1990-2000s are identified:  Spontaneous decentralization (1991-1993);  Formalization of the rules (1994-1998);  Political recentralization and further fiscal decentralization plans (1999-2001);  Local government reform and new assignment of powers (2002-2004);  Fiscal recentralization (2005-2008). It is clear from this classification that intergovernmental relations among different levels of government have changed significantly during the last 20 years of transition. A period of spontaneous decentralization was characterized by a so-called “Parade of Sovereignties”, when regions, especially republics of Russia, received more political and fiscal autonomy and even wished to become fully independent from Moscow. Da Silva et al. (2009) point out that almost 20 percent of government expenditure assignments were transferred to subnational governments between 1992 and 1993, while the share of government sector in GDP was reduced from 52 to 42 percent in only two years (1992-1994), due to the start of privatization reforms. Before the second half of 1990s, when major institutions of fiscal federalism were created, intergovernmental grants were allocated without any strict rules, but rather according to the bargaining power of regions; taxes could be almost randomly established by the regions and municipalities; subnational borrowing rules were opaque. The adoption of the first part of the Tax Code and the Budget Code in 1998 played a role in the creation of a legal framework of intergovernmental relations – clear tax and spending assignments among regions was the major achievement of these documents. However, several problems, including unfinanced federal mandates (spending responsibilities of the federal government, executed by subnational governments but usually not compensated for), poor formula for equalization grants allocation, and a unified status of local self-governments (cities, municipal districts and settlements were considered equal under the law, but were not equal in reality), remained until the middle of 2000s. When V. Putin came to power in 2000, he announced a new local 20

government reform (which introduced a transparent two-tier system of local selfgovernments) and gradual political recentralization with the reassignment of responsibilities to strengthen the federal government and with the consolidation of regions in federal districts. However, political recentralization stopped to be gradual in 2004, when elections of governors were abolished (a formal ground was a terrorist threat and a necessity to ensure national security). A new “vertical of power” started to be implemented – regional leaders became appointed by the regional parliaments (in fact, candidates were proposed directly by the President), while the National Parliament (Duma) became dominated by the ruling party – the United Russia. Political centralization was also accompanied by fiscal recentralization. In terms of fiscal autonomy, regions have become less independent from the federal government: during 1990s, there was a plenty of regional taxes; moreover, republics remitted only a portion of locally-collected federal taxes, while today there are only three regional and two local taxes in Russia5. However, in 2012, it was decided to return direct governor elections, but with significant entry barriers shaped into party and municipal filters. Also, a project to introduce a new local tax (namely, a consolidated property tax) to increase fiscal capacity of local governments is now considered by the Russian government. So, the next stage of Russian federalism may be a new attempt to decentralize some taxing and political responsibilities. Some authors draw a parallel between decentralization experience in Russia and China (Blanchard, Shleifer, 2000; Shleifer, 2005; Zhuravskaya, 2000, 2007) since the initial conditions of both countries were similar: excessively centralized socialist states with large and inefficient public sectors. Analyzing Russian and Chinese federalism during 1990s, Shleifer (2005) argues that a key difference between these two systems lied in the nature of decentralization reforms. In China, gradual fiscal decentralization was accompanied by political centralization. Thus, the central government maintained control over local officials and forced them to implement growth-enhancing policies. As a result, provincial governments had incentives to preserve markets and stimulate local development, which, in turn, led to higher economic growth. On the contrary, Russia was spontaneously decentralized at the beginning of 1990s, both fiscally and politically. The federal government lost its control over subnational authorities. Growth and development were not primary objectives of the government, since the agenda at that time included the issues of territorial integrity, borders shaping, massive privatization of government property and relations with the former Soviet Union members. In this situation regional and local officials acted virtually as rent-seeking 5

According to the Russian Tax Code, regional taxes consist of corporate property tax, vehicle tax and gambling tax; local taxes include individual property tax and land tax.

21

Leviathans, establishing new taxes and increasing fiscal and administrative pressure on businesses. An opposite tendency has been present since early 2000s. The federal government under Putin’s presidency had intentions to implement the successful Chinese model of federalism, as pointed out by Zhuravskaya (2007). Still, political centralization reforms of 2000s have been accompanied by economic recentralization (regions and municipalities lost most of their revenue sources and hence flexibility in decision-making). It seems that local governments in modern Russia have never faced right incentives to foster local economic development. During 1990s, municipalities were subjects to unpredictable annual changes in regional grants and also faced lack of administrative control, extracting rents from local residents, whereas, during 2000s, they lost major taxing powers and became transferdependent. Using the dataset of largest Russian cities during 1992-1997, Zhuravskaya (2000) confirms the fact that local governments during 1990s lost their incentives to preserve markets because of the poor regional system of intergovernmental grants – whenever a local government showed excellent performance and collected more tax revenues, regional subsidies in its budget were partly reduced, thus destroying incentives of the municipality to develop its economic potential. Alexeev and Kurlyandskaya (2003) verify these results for the case of one particular region (Rostov Oblast) and come to similar conclusions. Dependence of regions on natural resource rents, which is a growing concern for the Russian economy, is also considered in several studies (Desai, Freinkman, Goldberg, 2005; Freinkman, Plekhanov, 2009). In the work of Freinkman and Plekhanov (2009), it is suggested that “rentier” regions, i.e. regions that depend heavily on natural resource production or intergovernmental grants, tend to be excessively centralized. So, intra-regional expenditure decentralization in their model depends negatively on transfer dependence and oil-gas production. A possible solution, presented by the authors, is to implement certain federal policies to increase intra-regional decentralization in these regions, as it can be below efficient levels. The authors argue that this measure can reduce rent-seeking activities of regional elites and improve overall performance. Desai, Freinkman and Goldberg (2005) analyze the impact of subnational fiscal autonomy on several measures of regional economic performance (the authors use data for 1996-1999). In general, they find that autonomy is positively related to major economic development indicators, including growth. However, rentier regions tend to benefit less than other regions from increased autonomy since their major sources of income are not own tax revenues, but rather windfall revenues from large oil-gas firms or intergovernmental grants.

22

Current state of Russian federalism Currently, the Russian Federation is subdivided into 83 regions of several types: republics, oblasts, krays, autonomous okrugs, one autonomous oblast and two federal cities (Moscow and St. Petersburg)6. Regions form eight federal districts (federalny okrug) – Central, Northwestern, Southern, Volga, North Caucasian, Ural, Siberian and Far Eastern. A comprehensive list of regions ranged by federal districts can be found in Table 1 (Appendix B). According to the Constitution of the Russian Federation, federal subjects have equal rights and responsibilities in their relations with each other and the federal government. However, de-facto regions are extremely diverse in terms of territory, population, natural resource allocation, economic development, ethnolinguistic fragmentation and other socio-economic characteristics. For instance, the Sakha Republic has the largest area in Russia, three million square kilometers (almost five times larger than France), but its population is less than one million. On the contrary, the Republic of Dagestan has an area of 50 thousand square kilometers with the total population of almost three million people. So, population density of Dagestan is 180 times higher than that of Sakha. Tyumen Oblast with two autonomous districts supplies more than 50 percent of Russian mineral resources, and its real GRP per capita is ten times higher than that of any North-Caucasian republic. Another example is an ethnic composition of regions: while Northern territories are populated predominantly by ethnic Russians (e.g., in Vologda and Kostroma Oblasts 97% of residents are Russians), Southern republics are much more diverse, usually having several strong ethnic groups and the Russian minority7. Naturally, these facts show that the inequality among Russian regions is high. Thus, their political power and bargaining position in relations with the federal government are not homogeneous in reality. Federal subjects are further subdivided into two major types of municipalities (local government entities) – cities (gorodskoi okrug) and municipal districts (raion). Municipal districts consist of smaller urban and rural settlements (gorodskoe / selskoye poselenie). Federal cities, Moscow and St. Petersburg, are divided into intra-urban municipal districts. In accordance with the Federal Law No. 131 “On General Principles of the Organization of Local Self-Government”, local governments are independent of federal and regional governments in their administrative and fiscal responsibilities. Still, in practice municipalities do not possess enough revenues and depend crucially on grants from the higher levels of government. 6

Since the status of the Republic of Crimea and Sevastopol is ambiguous and internationally unrecognized, these regions are not considered in the present thesis. 7 Dagestan has the largest number of ethnic groups: Avars (29%), Dargins (17%), Kumyks (15%) and Lezgians (13%) are the most populous of them. Russians are only 3.6%, according to the 2010 Census.

23

In order to understand the current state of fiscal federalism in Russia, it is necessary to examine the assignment of expenditure and revenue responsibilities among different levels of government and the system of intergovernmental grants. Expenditure assignment among federal, regional and local governments is regulated by the Constitution and Federal Laws No. 184 “On General Principles of the Organization of Government in Subjects (Regions) of the Russian Federation” and No. 131 (see above). National defense and security expenditures are a sole responsibility of the federal government (99.8% of national defense expenditures were federal in 2013). Expenditure on higher education, hi-tech health care, national cultural heritage, and military veterans’ social assistance are also assigned to the federal government. Other types of expenditures are distributed among different levels of government. Major regional spending responsibilities include specialized health care provision, vocational training provision, environmental protection and disaster prevention, provision of social services for elderly and handicapped people, intra-regional public transportation provision, maintenance of regional cultural heritage, museums and libraries, regional infrastructure maintenance (including roads) and some other minor responsibilities. Local government expenditure responsibilities are different for municipal districts (1st level) and settlements (2nd level). So, public transportation and infrastructure maintenance within a settlement is a responsibility of this settlement, while such inter-settlement services are provided by the districts. Municipal district expenditures also include the provision of primary and secondary education (as well as preschool education), municipal hospitals and maternity care centers maintenance, electricity provision, and waste management. Settlements are responsible, inter alia, for housing and communal utilities and maintenance of parks and gardens. Revenue assignment is regulated by the Tax Code and Budget Code of the Russian Federation and other laws. As pointed out above, currently there exist only five taxes, over which regional and local governments have full authority (their share in total tax revenues is less than five percent). At the same time, major taxes in terms of budget revenues are federal: value added tax (VAT), tax on mineral resource extraction, personal income tax, corporate profit tax, and excises. Their tax bases and rates are established solely by the federal government. Revenues from some of these taxes are shared between federal and lower levels of government. Regional governments get 100% of revenues from corporate profit tax, 85% of personal income tax revenues (or 50% for foreign citizens working under patent system of taxation), 40-100% of excises on different types of alcohol and gasoline, and 60-100% of revenues from taxes on extraction of different mineral resources (excluding oil and gas – these revenues go entirely to federal budget). Local governments, depending on their type 24

(cities, municipal districts, etc.), receive 5-15% of revenues from personal income tax, 50% of unified agricultural tax revenues, 0-100% of different license fees, and 0-100% of revenues from unified tax on imputed earnings. According to the Ministry of Finance, the share of consolidated regional revenues in the consolidated federal budget was only 34% in 2013. In turn, consolidated regional budget revenues in 2013 were primarily composed of personal income tax revenues (30.6%), corporate profit tax revenues (21.1%), individual and corporate property tax revenues (11%), and other non-tax revenues, including intergovernmental grants (27.2%). It is clear that regional and local governments are dependent on the decisions of the federal government to change a particular tax base or tax rate of shared tax. Inequality among regions and vertical fiscal imbalance (tax revenues are concentrated at the federal level) make the system of intergovernmental grants a very important policy instrument. According to the Budget Code, there are three major types of intergovernmental grants in the Russian Federation – equalization grants (dotacii), subsidies and subventions. Equalization grants are unconditional transfers from the federal government to regional governments, which are aimed at regions with below minimum fiscal capacity. Subsidies are conditional transfers, through which different federal and regional socio-economic programs are financed or co-financed. Subventions are also conditional transfers, aimed at financing responsibilities of the federal government, which are carried out by regional governments on its behalf. It is clear that maximal discretion of spending is related to equalization grants. However, certain penalties and increased budget control is introduced by the Budget Code for regions, which permanently depend on federal transfers. Using a set of prerequisites for sound and efficient decentralization, developed by Dabla-Norris (2006), it is possible to assess, whether the current state of Russian federalism satisfies these conditions. Clarity precondition is not fully achieved. The system of intergovernmental grants is much more transparent than during 1990s, but is still far from being perfect (equalization grants are distributed according to a clear formula, but subsidies for different federal programs are allocated in an ambiguous way); expenditure and revenue assignments are far from being matched, especially at the local level. Autonomy requirement is not satisfied, because regional and local governments, even rich ones, depend on intergovernmental transfers and cannot self-finance all of their expenditure programs. Lack of well-functioning institutions is one of the major problems of the Russian economy. For last eight years, there has been no democratic representation at the regional level. As a result, accountability of regional officials has decreased. Clearly, three main requirements for sound and efficient decentralization are far from being satisfied in Russia; thus, its implementation 25

becomes a complicated issue. In this situation, the dangers of decentralization may overweight its benefits. 3.2. Fiscal Decentralization and Economic Growth in Russian Regions: An Empirical Analysis Data Description A rich new panel dataset on Russian regions has been gathered for this study. It includes more than 20 variables for 78 regions8 during the period of 2005-2012. Major sources of the data are the Russian Statistical Agency (Rosstat), the Ministry of Finance and the Federal Treasury of the Russian Federation. The first source is the official Russian statistical database, which includes many socio-economic, demographic, geographical and political characteristics of regions. The second- and third sources possess an extensive set of budget documentation (e.g., consolidated federal and regional budget reports on annual and monthly basis). Time dimension of the dataset includes eight consecutive years from 2005 to 2012, thus the focus of this empirical study is on the latest stage of Russian federalism development, which has been accompanied by the global economic crisis. So, the primary goal of the empirical part of the present Master Thesis is to analyze the relation between intra-regional decentralization and regional growth during the period of economic instability and further fiscal recentralization. Gathering a larger dataset is problematic, because Russian budget standards have changed significantly during 2000-s (documentation for earlier years is often very difficult to analyze due to low budget discipline of particular regions and lack of centralized control). Moreover, mergers of the regions during 2003-20089 make the analysis even more complicated. Some databases consider merged regions separately, whereas others – as a single entity even before the year of acquisition. In the future, it is possible to update the present database and verify the robustness of the analysis by using more observations. Table 2 in Appendix B gives the definitions of the variables used in the study, while Table 3 presents the descriptive statistics for these variables on an annual basis. Clearly, there are some interesting patterns in the data. Firstly, average gross regional product (GRP) growth rate was unstable during 2005-2012. It achieved impressive 9.7% in 2007, but fell down to 4.4% in 2009 due to the global economic crisis, and then recovered to positive values in 20102012 (see also Graph 1). Thus, the present dataset includes the period of a severe economic 8

Moscow and St. Petersburg, being 2 federal cities, are excluded from the analysis, as their budget structure is incomparable with other regions; Tyumen Oblast includes 2 Autonomous Okrugs (Yamalo-Nenetsky and Khanty-Mansiysky), Arkhangelsk Oblast includes Nenetsky Autonomous Okrug. 9 According to the Constitution of 1993, the Russian Federation consisted of 89 regions; after administrative reforms of 2003-2008 only 83 regions survived (autonomous regions, often very small in terms of population, were acquired by bigger neighbors).

26

downturn, and it can be both an advantage and a drawback of the study. On the one hand, data give an opportunity to analyze the relation between decentralization and growth in a highvolatility period; on the other hand, it can be only a special case which does not reflect the “true” relationship in a long-term perspective. In the next section, this issue is examined in more detail. Average regional population growth, according to the descriptive statistics, is negative during the whole period of interest (ranging from -0.8% in 2005 to -0.1% in 2012). This unpleasant tendency has been present since the fall of the Soviet Union in 1991. It reflects several negative facts about Russian demographic policy: death rates have been higher than birth rates for the last 20 years (as a result, population has decreased by seven million people); population structure is unbalanced – there are more pensioners than young people (average share of pensioners in regions is about 30 percent); citizens are leaving regions in Central and Northwestern Federal Districts as well as in Siberia, and moving to Moscow, St. Petersburg or abroad, because of low living standards and poor opportunities for career development at home. One promising tendency is that average regional investment in physical assets is capturing a larger share of GRP. This share grew from 24% in 2005 to 29% in 2007, and since 2007 it has not fallen below 29%, even during the recession. Still, it is possible that a large share of capital investments is related to oil and gas exploration and to infrastructure development in this sector. Another trend that should be mentioned is the increased dependence of regional economies on natural resource production (especially, oil and gas): in 2005-2009 its average share in gross regional product fluctuated around 10.5-11.5%, but in 2011-2012, it increased to almost 14%10. A potential explanation is the following: oil and gas prices have grown steadily for the last years (with a short-term decline in 2008-09), distorting incentives of the Russian government to invest into other industries; after the crisis, oil and gas prices have recovered much faster than prices and demand in other major industries (steel, machinery, etc.), thus changing the relative share of industries in GRP and in GDP. This indicator is important to understand the current trend of the Russian economy, but a different measure is chosen for the regression analysis, namely, a regional share in total natural resource production.

10

The average share of natural resource production in GRP may seem too small for the Russian economy. Still, it is a simple average, i.e. an arithmetic mean ( ∑ ) among all 78 regions. If weighted average ∑ was calculated ( ∑ ), it would give a much higher result (~21%). For instance, in Tyumen Oblast, the largest producer of oil and gas, the share of natural resources in GRP fluctuates around 75%.

27

Tax pressure, measured by the share of regional tax revenues in GRP is relatively stable over time and is fluctuating around 11-12%. It reflects the fact that the rates of major taxes (personal income tax, corporate profit tax) have not changed during the period of interest. Openness ratio (export plus import as a share of GRP) dropped from 38% to 30% in 2006, and since then it has been close to 27%. A significant drop might be related to the centralization of natural resource trade, which reduced regional government foreign trade responsibilities. In Tables 2 and 3 in Appendix B, there is also a description of decentralization measures. Four indicators are used in order to reflect the fact that decentralization is multidimensional, which has not been taken into account in some previous studies. The first variable, DEC_1, is a regional revenue decentralization measure. Following Akai and Sakata (2002), it is calculated as a share of self-generated municipal revenues (net of intergovernmental grants from the other levels of government) in consolidated regional budget revenues. This measure could be even more precise – only conditional grants (e.g., transfers for the compensation of federal mandates or special-purpose transfers) should ideally be subtracted from total municipal revenues, as local authorities have a certain discretion over unconditional (equalization and gap-filling) grants. In fact, it is quite complicated to calculate the exact shares of conditional and unconditional grants for each region, because of the data structure, available from the Federal Treasury. It is assumed throughout the analysis that most of the grants are conditional, which seems a valid assumption in the case of Russia, since local governments are usually implicitly obliged to spend unconditional grants for different social and welfare programs. Revenue decentralization fluctuates around 28%, with a decreasing tendency after the peak of the economic crisis (from 31% in 2009 to 23% in 2012). The second variable, DEC_2, is a regional expenditure decentralization indicator. It is measured by the share of municipal expenditures in consolidated regional budget expenditures. Intergovernmental grants to the higher levels of government (“backward transfers”) are excluded from total municipal expenditures. Expenditure decentralization fluctuated around 50%, with the peak of 56% in 2007 and a declining trend after the crisis (53% in 2009, 45% in 2012). One important observation should be pointed out here: the gap between expenditure and revenue decentralization is large and significant (around 22-23%). For instance, Akai and Sakata (2002) report, that this difference for the USA in 1992-1996 was approximately 8%. Despite this gap is large in Russia, a decreasing trend is present – in 2008, it was 27% and in 2012 – 21%. Therefore, it is necessary to check in the empirical analysis, whether the 28

convergence of revenue and expenditure decentralization can be conducive to economic growth. The third decentralization variable, AU, is a municipal autonomy indicator. It is a share of self-generated revenues in total municipal revenues. It is a relatively stable measure, being close to 50-52% in all years of the analysis, except for 2012, when it falls to 45%. Despite its stability, the level of municipal autonomy is quite low (16-18% lower than in the USA). It reflects the fact that there are only two local taxes and a limited number of shared taxes, and local governments usually do not have full discretion over the majority of their financial resources. The fourth decentralization variable, IGG, is a measure of regional dependence on intergovernmental grants from the federal government. It is calculated as a share of federal intergovernmental fiscal transfers (equalization grants, subsidies and subventions) in consolidated regional revenues. As opposed to the first three measures, it is not an indicator of intra-regional decentralization, but rather a measure of regional (in)dependence. It is fluctuating around 27%, with the lowest dependence from the federal government in 2006 (24%) and the highest – in 2009 (31%).

Empirical Analysis The following empirical analysis is loosely based on the Davoodi-Zou analytical framework (see Appendix A for an exposition of the slightly modified original model from (Davoodi, Zou, 1998)). It is necessary to test whether shares of different levels of government11 and tax pressure are related to the per capita growth rate of output empirically. The general form of the regression is:

In this regression model

is a gross regional product per capita growth rate,

denotes unobserved regional fixed effects,

stands for the vector of time dummies.

one of the above-mentioned decentralization measures,

is

is a squared decentralization

variable, which is added in order to test for potential nonlinearity, while burden or tax pressure, measured by the share of tax revenues in GRP.

denotes tax is a set of basic and

additional control variables used in the analysis. Basic control variables, following the approach of Levine and Renelt (1992), include regional population growth, investment share 11

As the major goal of this thesis is to analyze a relation between intra-regional decentralization and regional economic growth, a government in a stylized Davoodi-Zou model is assumed to consist of two levels – regional and local (local governments consist of 2 further tiers - see Appendix A). Hence, a share of local governments in consolidated regional budget corresponds to the level of decentralization.

29

in GRP, a logarithm of initial (or lagged) values of GRP per capita and a proxy of human capital. Also trade openness ratio, inflation (regional consumer price index) and regional share of total natural resource production are used as additional conditioning factors. Unemployment is not considered as a control variable in this study due to two reasons: firstly, it is highly correlated with lagged GRP per capita; secondly, it may cause endogeneity problem due to its bi-directional causality with economic growth. Tables 4 and 5 in Appendix B show correlation matrices for decentralization measures and conditioning variables, respectively. Decentralization variables are highly correlated, except for rather low correlation between expenditure decentralization (DEC_2) and regional revenue autonomy AU. One interesting pattern should be mentioned. Transfer dependence of the region (IGG) is negatively correlated with intra-regional expenditure decentralization ( = –0.52). This can be a result of a flypaper effect – “money sticks where it hits” (Hines, Thaler, 1995), which induces regional governments to increase own spending when they get a grant, thus reducing the relative share of municipal expenditures in total regional budget. Major control variables, chosen for the empirical analysis, are almost uncorrelated (only two  exceed 0.3), which precludes a potential problem of multicollinearity. A sign of decentralization variables coefficient is the main interest of this study. It is expected from the previous empirical research (both cross-country and regional) that:  Tax pressure should have a negative impact on growth;  Whenever initial GRP per capita is negatively related to economic growth, it confirms a conditional convergence hypothesis (poorer regions tend to grow faster);  Investment share in GRP should be positively associated with growth;  Human capital should have a positive influence on growth; Other conditioning variables (population growth, openness, inflation), according to previous studies, have an ambiguous effect on economic growth. The panel data regression model above is estimated in the fixed-effects (FE) framework. The results of the robust Hausman specification test show that random-effects (RE) estimation is inconsistent for the model of interest (a null hypothesis is strongly rejected at 1 percent level)12. A drawback of FE model is that it is impossible to analyze an impact of time-invariant independent variables (e.g. initial GRP per capita or dummies for federal districts) on the variable of interest. However, a nice feature of FE is that unobserved heterogeneity is controlled for by regional fixed effects ( ). As Baskaran and Feld (2012) put 12

A standard Hausman test assumes that RE is efficient, which is a quite strong assumption. Still, it is possible to relax it by using a robust Hausman test (Wooldridge, 2010).

30

it, by using fixed effects it is possible to “implicitly control for various largely time-constant institutional features … and specific characteristics” of a particular country or region. Cantarero and Perez Gonzalez (2009) point out that RE estimator behaves adequately in a similar model (with data on Spanish regions) as long as the regression specification does not include the initial GDP per capita (or lagged values of GDP p.c.) as an independent variable. Since lagged GRP per capita is a must-have control variable in our study (to test a conditional convergence hypothesis), FE is preferred to RE. Since fixed-effects analysis is used, it is impossible to include time-invariant log of initial GRP per capita variable. Thus, following Jin, Qian and Weingast (2005), a log of GRP per capita, lagged by two periods, is included in the regression model. In order to deal with potential endogeneity problem, the inflation variable is also lagged by one period. Time dummies are added in each regression specification to control for exogenous macroeconomic shocks, which have been present during the global economic crisis in 20082010 (see Graph 1). Modified Wald test shows the presence of groupwise heteroscedasticity in the model. Thus, cluster-robust standard errors are reported in the regression tables. 3.3.

Results of the Empirical Analysis

Main Results Tables 6-8 in Appendix B present the results of the regression analysis within a fixed effects panel data framework. Column 1 in each table reports the results for the basic specification – real GRP per capita growth is regressed on fiscal decentralization (revenue in Table 6, expenditure in Table 7, autonomy and IGG dependence in Table 8), population growth, investment and tax shares in GRP, lagged values of real GRP per capita (logarithm) and human capital, measured by the ratio of higher educational institutions’ alumni of the current year to total regional population. Columns 2 and 4 in Tables 6 and 7 (for DEC_1 and DEC_2) include a square of decentralization to test for a nonlinear relationship13. Columns 3 and 4 in Tables 6 and 7 contain additional control variables – inflation, natural resource production indicator and openness ratio. Squared decentralization variables are not reported in Table 8 for autonomy and transfer dependency indicators due to their insignificance. Revenue decentralization (DEC_1) has a significantly negative coefficient (at 10 percent level) only in the first specification (see Table 6). Throughout columns 2, 3 and 4 its significance vanishes, therefore, this negative relationship is not robust to the inclusion of 13

When decentralization and its square are included in the regression simultaneously, a multicollinearity problem may arise. It is possible to overcome this difficulty by implementing a zero-mean transformation ( – . The transformed measures are used in the analysis.

31

additional control variables. Squared revenue decentralization is also not significant, rejecting the hypothesis of non-linearity. Expenditure decentralization (DEC_2), on the contrary, is significantly negatively associated with economic growth (at 5 percent level) in all four specifications (see Table 7). The square of expenditure decentralization in columns 2 and 4 is also negative but not significant; thus, a nonlinear relationship is not found between major decentralization variables and growth. Autonomy indicator is not significant (columns 1 and 2 in Table 8). One possible explanation of this weak result is an issue of measurement. If own tax autonomy were calculated (without shared taxes), the coefficient could have been more significant. A share of intergovernmental grants in total regional revenues (IGG) in columns 3 and 4 of Table 8 is positively related to economic growth and significant at 10 percent level. So, transfers to regional budgets in the period of economic crisis can help to overcome serious recession. Graphs 2 and 3 show the scatterplots of average expenditure decentralization and transfer dependence versus GRP per capita growth. These results give evidence that government expenditures within Russian regions are excessively decentralized, which is destructive for regional economic growth. Self-generated revenues form only 50% of overall local revenues (see AU indicator in Table 3, Appendix B); another half consists of intergovernmental grants from regional and central governments. Therefore, local governments do not have enough own resources to finance their spending programs and hence rely heavily on transfers. The majority of their spending programs is designed and financed by different levels of government in the form of conditional grants. It means that local officials have no incentives to spend less than they are given; in other words, local expenditures usually go beyond efficient levels. Not surprisingly, expenditure decentralization is negatively related to economic growth. At the same time, larger shares of intergovernmental grants in regional revenues correspond to higher growth. One possible explanation is that a flypaper effect reduces excessive expenditure decentralization within a region. For instance, a regional government receives an additional unconditional equalization grant during a crisis. Then, by the flypaper effect, it is spent directly by the regional administration for a particular anti-crisis program instead of sending this grant further to less efficient local governments. Thus, unconditional grants can have a twofold effect: firstly, a share of local expenditures in total regional budget is reduced, which is conducive to growth; secondly, financial resources are spent by more qualified regional officials. Almost all of the control variables have expected signs, which support the validity of the model. Lagged real GRP per capita is negatively related to growth and significant at 1 percent level, confirming the conditional convergence hypothesis. Indeed, poor regions, even 32

in the period of the crisis, tend to grow faster than their richer counterparts. Tax pressure, measured by the share of tax revenues in GRP, is significant and also negatively associated with growth. It supports the argument that excessive taxation, especially during economic downturns, can have negative consequences for both small and medium enterprises and large industries, creating obstacles for economic recovery and provoking a rise of shadow (illegal) economic activities. On the other hand, investment to GRP ratio, being a proxy for physical capital formation, is positively related to regional economic growth. The same is true for human capital development. Regions with a higher share of university graduates in the total population tend to have faster economic growth, even in the presence of the crisis. This result may be explained by the fact that strong regions usually invest in higher education in order to train qualified employees for the regional economy. Higher inflation rates correspond to lower growth, supporting the argument that macroeconomic instability may be detrimental to economic growth. Population growth rates do not have a significant effect on economic growth. This fact can be attributed to a poor demographic situation in Russia (with a slightly positive trend) – population decline is a distinctive feature of the majority of the Russian territories (exceptions are Moscow and St. Petersburg, not considered here), although growth rates in these regions may be quite different. Additional specifications According to previous studies on Russian federalism (Desai, Freinkman, Goldberg, 2005; Freinkman, Plekhanov, 2009), the category of rentier regions, which have high natural resource endowments or receive a significant part of their revenues in the form of intergovernmental grants, is quite different from other regions in terms of their development patterns. In the present study, rentier regions are characterized by above average natural resource share in GRP (>0.11) or above average share of intergovernmental grants in total budget revenues (>0.27). Non-rentier regions have both below average natural resource endowments ( 0.05, exept for unemployment rate. Thus, the results for the original determinants of growth and unemployment are presented.

Statement of Authorship

I hereby confirm that that the work presented has been performed and interpreted solely by myself except for where I explicitly identified the contrary. I assure that this work has not been presented in any other form for the fulfillment of any other degree or qualification. Ideas taken from other works in letter and in spirit are identified in every single case.

Date: Signature: