Report No: ACS22573 . Russian Federation Spatial Differences in Living Standards Convergence without Equity: A Closer Look at Spatial Disparities in Russia . June 2017 . POVERTY AND EQUITY GLOBAL PRACTICE EUROPE AND CENTRAL ASIA . 1 . Standard Disclaimer: . This volume is a product of the staff of the International Bank for Reconstruction and Development/ The World Bank. The findings, interpretations, and conclusions expressed in this paper do not necessarily reflect the views of the Executive Directors of The World Bank or the governments they represent. The World Bank does not guarantee the accuracy of the data included in this work. The boundaries, colors, denominations, and other information shown on any map in this work do not imply any judgment on the part of The World Bank concerning the legal status of any territory or the endorsement or acceptance of such boundaries. . Copyright Statement: . The material in this publication is copyrighted. Copying and/or transmitting portions or all of this work without permission may be a violation of applicable law. The International Bank for Reconstruction and Development/ The World Bank encourages dissemination of its work and will normally grant permission to reproduce portions of the work promptly. For permission to photocopy or reprint any part of this work, please send a request with complete information to the Copyright Clearance Center, Inc., 222 Rosewood Drive, Danvers, MA 01923, USA, telephone 978-750-8400, fax 978-750-4470, http://www.copyright.com/. All other queries on rights and licenses, including subsidiary rights, should be addressed to the Office of the Publisher, The World Bank, 1818 H Street NW, Washington, DC 20433, USA, fax 202-522-2422, e-mail pubrights@worldbank.org. 2 Convergence without Equity: A Closer Look at Spatial Disparities in Russia Executive Summary Addressing regional disparities is key to unlocking Russia’s potential to achieve stronger gains in growth and equity outcomes as well as to improve its institutional environment. While spatial disparities have been an important policy concern in Russia for a long time, inequalities across its vast territory remain stark. This report explores the current state of regional disparities at the macro and micro-level, updating existing literature to reflect recent trends and providing new insights into household-level drivers of welfare. The report stresses that addressing spatial disparities does not necessary imply “balancing” growth across a geographic territory – but rather focusing on creating opportunities for all people, regardless of where they live. The following three key messages emerge from the analysis. Message 1: Over the last decade spatial inequalities have declined in terms of incomes, although despite “convergence” in monetary outcomes across regions, the country remains characterized by significant disparities in access to services. Russian regions have become more similar in terms of real income per capita and poverty incidence over the last decade – in other words they have experienced “convergence”. However, these developments are tempered by two major qualifications which arise from a deeper look at the underlying drivers of monetary disparities as well as observations of disparities in non-monetary dimensions of welfare. First, households in different regions remain very different in terms of their demographic and socioeconomic factors which are important drivers of regional disparities. In Russia, around half of the difference in incomes between the cities of Moscow and St. Petersburg and other parts of the country are attributable to differences in endowments (vis-à-vis differences in returns to endowments). The results are similar for comparisons between urban and rural areas within each type of regions. Second, non-monetary indicators of living standards show that one’s region of residence influences the opportunities one has to enjoy access to good quality public services. Message 2: Spatial inequality today is mostly driven by inequality within regions, and particularly by inequality in richer regions. While the welfare gaps across regions appear to be decreasing, the welfare gaps within regions remain large and explain an increasingly larger share of overall inequality. Inequality is driven by the richest and most populous regions in Russia, including Moscow City, Moscow Oblast and St. Petersburg (which are the top three contributors to overall inequality and are also the three largest regions in terms of population) and some of the resource-rich regions. One of the reasons why richer regions are also more unequal is that a large share of the poor is concentrated in richer regions. While the percentage of people living in poverty is very high in poorer regions, because of the low population in these regions, the number of people living in poverty account for a small share of all poor people in Russia. Richer regions are home to a much larger number of poor people despite their low poverty incidence because of their high populations. For example, despite having some of the lowest poverty rates in the country (around 7.5%), the cities of Moscow and St. Petersburg together account for almost 10% of the poor people in Russia. Another reason richer regions tend to have higher levels of inequality is because households in richer 3 regions rely a lot more on labor income. Conversely, one of the reasons poorer regions tend to have lower levels of inequality is partly because of government transfers. Message 3: Regional governments may seek to prioritize different policies as determined by their conditions in terms of endowments, access to markets and fiscal constraints. Policies should focus on providing equal opportunities so that individuals are equipped with assets (such as human capital) that allow for gainful employment in the labor market. In particular, regional governments may seek to prioritize different policies as determined by their conditions in terms of endowments, access to markets and fiscal constraints. For example, rich and poor regions face very different fiscal constraints--with the latter in particular having limited fiscal space which would manifest especially during economic downturns. Rich regions, where a lot of poverty and inequality is concentrated, have more resources to spend on innovative approaches that could help address the challenges of more inclusive growth. 4 Table of Contents 1. Introduction: A territorial quest for development ........................................................................... 8 2. Why do we care?: Spatial disparities matter for efficiency, equity, and governance ................. 11 Spatial disparities matter for efficiency .................................................................................................. 12 Spatial disparities matter for equity ........................................................................................................ 14 Spatial disparities matter for governance ................................................................................................ 16 Beyond markets: Russia’s regional policies ........................................................................................... 18 3. Regional inequality in Russia: Trends and composition ............................................................... 23 4. Are the gaps closing? Regional convergence in income and poverty ........................................... 32 Data and methodology ............................................................................................................................ 33 Conditional beta-convergence: Results ................................................................................................... 35 Convergence in real income per capita: the role of structural variables ............................................. 36 Convergence in poverty rate: the role of structural variables ............................................................. 37 5. What factors explain spatial disparities at the household-level? .................................................. 37 Data, methodology, and regional groupings ........................................................................................... 38 Understanding welfare gaps across and within regions: Endowments and returns to endowments ....... 40 A closer look at metropolitan areas......................................................................................................... 45 6. Unbalanced growth and inclusive development in Russia: Lessons from the spatial analysis .. 48 References ................................................................................................................................................ 51 Annex ........................................................................................................................................................ 55 A. Inequality Trends ............................................................................................................................ 55 B. Convergence Results (Real income) ............................................................................................... 56 C. Convergence Results (Poverty) ....................................................................................................... 57 D. Oaxaca (Means of Covariates) ........................................................................................................ 58 E. Oaxaca Results: Household budget survey (2005-2014) ................................................................ 59 5 List of Figures Figure 1. Inequality in the top 9 largest countries (Gini index) ...................................................................................9 Figure 2. Poverty incidence by region ......................................................................................................................... 9 Figure 3: Economic Dispersion and Agglomeration in Russia .................................................................................. 13 Figure 4: Herfindahl index: Growth and specialization by region ............................................................................. 14 Figure 5: State density by region: Infant mortality, education, and utilities (2014) ... Error! Bookmark not defined. Figure 6: Days to obtain a construction permit, by region ......................................................................................... 17 Figure 7: Share of total inequality driven by inequality between or within regions .................................................. 23 Figure 8: Regional change in inequality (Gini), 2005-2015 ...................................................................................... 25 Figure 9: Per capita income by region ....................................................................................................................... 26 Figure 10: Number of income sources by quintile ..................................................................................................... 27 Figure 11: Share of income source in total household income, by quintile ............................................................... 27 Figure 12: Distribution of household characteristics by quintile in richest and poorest regions in Russia ................ 31 Figure 13: Occupational distribution by quintile in richest and poorest regions in Russia ........................................ 31 Figure 14: Degree of variation in wage, income, GDP, and poverty incidence across regions (1998-2015) ............ 33 Figure 15: Welfare and endowment gaps by region .................................................................................................. 41 Figure 16: Inequality across regions .......................................................................................................................... 42 Figure 17: Inequality within regions: Oaxaca decomposition of urban-rural areas within region ............................. 43 Figure 18: Average wages by sector, 2015 ................................................................................................................ 44 Figure 19: Mean income by region and city size (rubles) .......................................................................................... 46 Figure 20: Distribution of cities and towns by region (%) ......................................................................................... 47 Figure 21. Oaxaca decomposition: Metro vs Other large urban, Large vs small urban areas .................................... 48 List of Tables Table 1. Disparities in regional GDP per capita ................................................................................... 10 Table 2. Paradigm shift of regional development policy........................................................................... 21 Table 3. Poverty in selected rich and poor regions............................................................................... 24 Table 4. Gini decomposition by income source ........................................................................................ 28 Table 5: Conditional beta convergence coefficients: Real income per capita and poverty rate ................ 35 Table 6: Regional groupings ..................................................................................................................... 40 Table 7: Sectors of activity for household heads in highly-skilled occupations (manager or specialist) by education level........................................................................................................................................... 44 Table 8: Contribution of Covariates to welfare gap within and across regions......................................... 45 Table 9: Demographic and Socioeconomic Characteristics of Households in Metro vs Other Large Urban Areas.......................................................................................................................................................... 46 List of Boxes Box 1: Russia’s regions: Basic facts ............................................................................................................................ 8 Box 2: The state of the debate on regional disparities in Russia: The “Four Russias” and a tale of convergence ..... 10 Box 3: Poorer regions are most dependent on transfers ............................................................................................. 19 Box 4: Center-region relationship since the transition ............................................................................................... 20 Box 5: Micro-level analyses on welfare disparities in Russia .................................................................................... 23 Box 6: Estimating regional conditional beta-convergence using a dynamic panel data model with spatial effects .. 33 Box 7: Using the Oaxaca Blinder decomposition to understand sources of regional inequality in Russia ................ 39 6 Acknowledgments This report was prepared by a team led by Caterina Ruggeri Laderchi (Senior Economist, GPV03) and Yeon Soo Kim (Economist, GPV03), and including Mikhail Matytsin (Research Analyst, GPV03), Kimberly Bolch (Consultant, GPV03) and Elena Vakulenko (Assistant Professor, National Research University Higher School of Economics). The work was carried out under the overall guidance of Andras Horvai (Country Director, ECCRU) and Luis-Felipe Lopez-Calva (Practice Manager, GPV03). The team is thankful for the comments received by peer reviewers Nancy Lozano Gracia (Senior Economist, GSU10) and Aleksandra Posarac (Lead Economist, GSP03). 7 Convergence without Equity: A Closer Look at Spatial Disparities in Russia 1. Introduction: A territorial quest for development 1. An understanding of Russia’s quest for development begins with an understanding of its expansive geography and the natural difficulties of governing such vast territory. Russia is the world’s largest country by landmass and its geographic endowments encompass a diverse range of terrain and climate. These have shaped Russia’s development policies in the past, starting three hundred years ago when Russia’s modernization from a predominantly agrarian society began, through Soviet times and the present. During the Soviet era, in a quest to explore Siberia’s vast natural resources and to develop military capabilities, labor and capital was moved towards the East to support an even distribution of population and economic activities, against market forces. The resulting economic structure was physically more dispersed throughout the territory, yet inefficient and distorted. While almost thirty years have passed since Russia’s transition from a centrally planned command economy to a market-based one, the physical legacy of seventy years of Soviet rule remains large. Efforts to reverse these historical remnants have been often undermined by the inherited economic, social, physical, and relational networks that hindered progress toward more efficient and equitable regional development (Ericson 2013). To date, there are immense disparities in living standards between the prosperous West and the remote Far East (see Box 1 for an overview of Russia’s regional structure). Box 1: Russia’s regions: Basic facts Russia has a complex structure of subnational government: • The country is divided into over 80 federal subjects (termed as oblasts and federal cities. Territorial subdivisions also include krais (administrative territories), republics, autonomous okrugs (territorial divisions) and autonomous oblasts. The administrative units are grouped into eight federal districts, each headed by a Presidential Plenipotentiary appointed by, and representing, the President of the Russian Federation, who monitors the performance of the regions in each federal district. • There are large cities (formerly known as cities of oblast subordination) and rural raions (districts), the latter contain various forms of small towns and village governments, collectively known as second-tier municipalities. There are more than two thousand first-tier municipalities comprising more than five hundred cities and more than 1,800 raions. There are more than 20,000 second-tier municipalities, comprising more than 1,600 townships and more than 18,000 rural communities. • All municipalities (including rural settlements with small populations) are legally obliged to establish local governments, employ municipal office staff, formulate and execute budgets, and conduct an independent borrowing policy. The law assigns expenditure responsibilities to each tier of municipal government. The budget code specifies their revenue sources. Source: “Pathways to Inclusive Growth: The Russian Federation Systematic Country Diagnostic.” Box 4, chapter 4. (World Bank 2016a). 2. Russia achieved significant gains in poverty reduction since the late 1990s, but overall inequality remains high and regional disparities are stark (Figure 1). Poverty rates range from less than 10 percent in resource-rich Tatarstan and the large metropolitan areas of Moscow and St. Petersburg to 8 1 0 5 10 15 20 25 30 35 40 45 Tatarstan republic 7.6 St. Petersburg city Honduras. Belgorod oblast Moskow oblast Lipetsk oblast Voronezh oblast Moscow city Nizhny Novgorod oblast Sverdlovsk oblast Sakhalin oblast Chukotka autonomous okrug Leningrad oblast Kursk oblast Tula oblast Tambov oblast Yaroslavl oblast Dagestan republic Argentina Kazakhstan China Australia United States Canada Brazil India Russia Kaluga oblast Krasnodar krai 0 Udmurtia Republic Perm krai Bashkortostan republic Ryazan oblast Tver oblast Bryansk oblast Murmansk oblast Figure 2. Poverty incidence by region Source: World Development Indicators. Khabarovsk krai Source: Rosstat, Regions of Russia 2015 Orenburg oblast 0.1 Stavropol krai Note: Data from most recent year available. Magadan oblast Adygeya republic Omsk oblast Oryol oblast Chelyabinsk oblast Vologda oblast Poverty is measured using Rosstat’s official poverty line. North Osetiya republic Astrakhan oblast 0.2 Rostov oblast Samara oblast Kostroma oblast Novgorod oblast Tyumen oblast Kaliningrad oblast Volgograd oblast Amur oblast 0.3 Komi republic 0.26 Figure 1. Inequality in the top 9 largest countries (Gini index) Vladimir oblast Ulyanovsk oblast Primorskii krai Kirov oblast Penza oblast Ivanovo oblast 0.34 Arkhangelsk oblast 0.35 0.35 Chechnya republic 0.4 Kemerovo oblast Karelia republic Saratov oblast Smolensk oblast 0.41 0.42 Buryat republic 0.42 0.43 Altai krai Khakasia republic Chuvash republic 0.5 Tomsk oblast Novosibirsk oblast Sakha (Yakutia) republic Krasnoyarsk krai Kamchatka oblast 0.51 Pskov oblast Kurgan oblast Mordovia republic Zabaikal krai 0.6 Irkutsk oblast Kabardino-Balkar republic Karachaevo-Cherkess republic Mari-El republic Altai republic Evrei autonomous oblast Kalmyk republic Ingush republic Tuva republic 38.8 experience living standards similar to those found in Singapore, while those living in Ingush republic means that those households living in Sakhalin oblast (which has the highest regional GDP per capita) (which has the lowest regional GDP per capita) experience living standards closer to those found in 9 almost 40 percent in the poorest regions in the North Caucasus, Siberia and the Far East (Figure 2).1 Regional disparities in GDP per capita are even starker, differing by a factor of over 17 (Table 1). This Table 1. Disparities in regional GDP per capita Russian regions 2015 GDP per capita (2011 PPP) Countries with similar GDP per capita Top five regions Sakhalin oblast 87,426 Singapore Tyumen oblast 69,412 United Arab Emirates Chukotka autonomous okrug 54,187 Hong Kong SAR, China Moscow city 47,105 Netherlands Magadan oblast 36,132 New Zealand Bottom five regions Tuva republic 6,414 Bolivia Kabardino-Balkar republic 6,214 Cabo Verde Karachaevo-Cherkess republic 6,138 Congo, Rep. Chechnya republic 4,957 Myanmar Ingush republic 4,952 Honduras Source: Authors based on UNDP (2011). Data from WDI, Rosstat, WB calculations. 3. While there is an extensive Russian scholarship analyzing regional inequalities in Russia, relatively few quantitative studies have looked at their evolution over the past decade (see Box 2 for some highlights from recent academic debates). Earlier studies on regional inequalities in Russia have mainly looked at trends in regional convergence, but evidence on more recent trends is largely lacking (see Gluschenko 2010 for an extensive review of the literature on inter-regional income inequality in Russia). At the micro-level, few studies have used survey data to understand the drivers of spatial disparities at the household level. Box 2: The state of the debate on regional disparities in Russia: The “Four Russias” and a tale of convergence There are extensive debates in the Russian scholarship on classifying regional grouping to analyze disparities in welfare. Recently, Natalia Zubarevich, a prominent Russian academic, articulated the idea of “Four Russias,” in which the country is divided into four distinct categories: the post-industrial and urban towns of “First Russia”, the industrial blue collar towns of “Second Russia”, the rural and semi-urban towns of “Third Russia”, and finally the underdeveloped “fourth Russia” of the Northern Caucasus republic and Southern Siberia. While Zubarevich’s distinction is based on population size instead of geographic divisions, it yields a grouping that is distinguished by its level of economic development, demographic trends and socioeconomic characteristics of the residents, leading Zubarevich to argue for differentiated and decentralized policies that could lead to unequal growth across territories (Zubarevich, 2012). From a dynamic perspective, the Four Russias typology provides some useful insights for analyzing the present crisis, as it shows how the areas that are suffering tend to also be the richest areas which were better integrated with the global economy. In this view, the cosmopolitan and rich “First Russia” is now on a path to converge with the “Second Russia” of struggling indu stries, in a reversion of recent economic development. Zubarevich’s emphasis on disparities along a center -periphery axis which cut across variously defined administrative borders for Russia’s regions echoes an important idea that has emerged in recent debates—namely that the spatial disparities that matter in Russia today are the disparities within regions, be it across groups variously represented in different contexts or specific areas within the regions. This idea is at the basis of an ongoing study of the faces of exclusion across Russia’s regions, and is one that potentially lends itself to qua ntitative empirical testing (World Bank, forthcoming). 10 4. The potential heterogeneous impact of recent economic developments and policy changes in the country call for an updated diagnostic on spatial disparities. After a period of significant progress in poverty reduction between 1998 and 2008 driven mostly by rapid economic growth, Russia faced two economic crises that have had heterogeneous impacts across regions. While Russia’ economy managed to recover quite effectively from the 2008/09 crisis, the effects of the recent 2014 drop in oil prices and the sanction regime have pushed the economy into a recession. The effects of this recession have been felt differently in different parts of the country given its uneven economic structure. Some regions saw declining living standards, including some of the leading regions, which has led to reduced disparities through some “convergence to the bottom.” 5. Moreover, the federal government’s response to the financial crisis of 2008-09 and the oil price crisis/sanction regime that started in 2014 have been different. During the former, pensions and transfers became key engines of income growth for those at the bottom of the income distribution. However, during the more recent crisis, social assistance spending remained mostly constant (Gorina 2016, World Bank 2016). In the face of a shrinking economy and fewer resources to distribute, regional fiscal transfers to the regions have not increased in real terms and are at risk of being cut –with possibly large and unequal impacts across regions. For example, due to differences in population, many cuts could unintentionally be disequalizing –such as cuts to transfers to pensioners or low-income households that may end up disproportionally hurting regions with older or poorer populations. 6. This paper seeks to contribute to the current debate on regional disparities by exploring their underlying drivers. The availability of a new detailed survey on household income sources representative at the regional level allows for a unique opportunity to analyze in greater detail the challenge of regional disparities in Russia. By analyzing regional convergence in income and poverty from a macro-perspective and by analyzing regional inequalities due to differences in returns and endowments from a micro- perspective, this papers seeks to inform the policy debate on the types of interventions that are more likely to bring about improved living standards in lagging regions. The larger policy debate on regional development is often framed over whether to target resources toward investments in “poor places” or “poor people” (Ravallion and Wodon 1999). Investments in poor places, for example, would focus on developing non-portable assets such as infrastructure, while investments in poor people would focus on developing more portable assets like education. Assessing the implications of the recent economic developments on regional disparities, and identifying the key levers to address them can help ensure that resources are spent where they can contribute most effectively to improve livelihoods and restore growth. 7. The rest of this paper is structured as follows. Section 2 provides an overview of why spatial disparities matter for issues of efficiency, equity, and governance. Section 3 provides a diagnostic on regional inequality to investigate trends and between/within-region breakdown in the past ten years. Section 4 continues the diagnostic at the macro-level through a convergence analysis of real income per capita and poverty incidence. Section 5 uses survey data to explore household-level drivers of differences in living standards across regions. Section 6 concludes with policy implications. 2. Why do we care?: Spatial disparities matter for efficiency, equity, and governance 8. Addressing regional disparities is key to unlocking Russia’s potential to achieve stronger gains in growth and equity outcomes as well as to improve its institutional environment. In order to 11 achieve sustainable growth and equity outcomes, countries need to generate more diversified economic activity and a more homogeneous provision of public goods and services throughout their territory. The World Development Report (WDR) 2009 on Reshaping Economic Geography suggests that as economies develop, they transform along three dimensions: density (where the growth of cities leads to agglomeration and higher human concentration), distance (which is shortened with development, as workers and businesses migrate closer to more densely populated areas), and division (along lines such as political borders which become more porous as countries enter international markets to benefit from trade and specialization) (World Bank, 2009). To the extent that economic gains are translated—or not—into spatial convergence in living standards, they can lead to equal opportunities for individuals regardless of where they live. Indeed, the main message of the WDR 2009 is that while economic growth is by its very nature unbalanced—as inequalities in income and production across regions are an intrinsic aspect of the development process—development can still be inclusive (ibid). 9. From a policy perspective, the WDR 2009 identifies three priorities for overcoming economic geography challenges. To support the economic forces needed to transform economies in terms of density, distance or division (the “3 Ds” discussed above), three policy approaches (“3 Is”) can be considered: interventions (spatially-targeted approaches), infrastructure (spatially-connective approaches), and institutions (spatially-blind approaches). Spatially-targeted and spatially-connective approaches are more related to place-based policy investments, where poor or lagging regions are the focus of the engagement. These may be particularly relevant for supporting economic forces like agglomeration which can lead to higher efficiency and growth. Spatially-blind approaches on the other hand are more related to people-based policy investments, where poor people are the focus. In this way, investments in improving individual access to services like education can foster more equitable outcomes. Spatial disparities matter for efficiency 10. Spatial disparities matter for efficiency and growth. As Venables (2003) argues, spatial disparities in economic development are rooted in both ‘first nature’ and ‘second nature’ geographical reasons. First nature geographical reasons include natural endowments and physical geographic features— such as proximity to the coast or the presence of a mountain range, whereas second nature geographical reasons emphasize the interaction between economic agents and increasing returns to scale which come with greater density. Agglomeration effects are a key force shaping the geographic concentration of development, as illustrated by the New Economic Geography literature. Krugman (1991) shows how in the presence of low transportation costs, and where there is a large share of manufacturing (vis-à-vis agriculture), and economies of scale, a process of circular causation takes place. In this process, large local markets become attractive for firms to produce—due to the larger demand for their products—and for people to live close to—due to the availability of goods and services produced—influencing how economic activity is distributed geographically. 11. Russia’s economic landscape, however, is very geographically dispersed (World Bank 2011). This is an issue which has been shaped by centuries of territorial policy decisions, driven by Russia’s changing economic, social, and political environment. In Soviet times, massive subsidies to lagging and remote regions and the location of industrial facilities towards the East tried to equalize living standards across regions. These policies led to a more “balanced” distribution of economic activity across Russia— producing a certain degree of inefficient dispersion (Figure 3, panel a). This has led to a current level of 12 agglomeration that is very low relative to other economies with the same level of development (Figure 3, panel b). Figure 3: Economic Dispersion and Agglomeration in Russia a. Russia’s economic activity is very dispersed b. Agglomeration in Moscow vs Europe and Japan Source: World Bank (2011), Maps S2 (left panel) and S6 (right panel) Note: Height is proportional to economic output measured as GDP per unit area 12. Greater dispersion of economic activities impedes the benefits from agglomeration economies where spatial concentration of production and population promotes economic diversification and innovation. Diversification in a resource-rich country implies a greater concentration of non-extractive economic activities and skilled people in large cities. These are advantages that cannot be conferred through government action (World Bank 2011). In Russia, the benefits of diversification are further supported by a simple relationship between diversification and economic performance at the subnational level. Figure 4 displays the Herfindahl index of employment as measure of diversification against regional real GDP growth per capita. Concentration is minimized when all sectors employ equal shares of workers, which tends to occur when economic activities are less spatially concentrated and less diversified. In Russia’s regions, as expected, higher levels of diversification are observed in Moscow city, St. Petersburg and Nizhny Novgorod which also exhibit better economic performance; on the other hand, resource-rich regions are located towards the bottom end in terms of both diversification and growth. 13 Figure 4: Herfindahl index: Growth and specialization by region 115 Chukotka 110 GDP p.c. growth 105 Sakha (Yakutia) Magadan Nizhny Novgorod Kemerovo Arkhangelsk Tatarstan Sakhalin Udmurtia St. Petersburg city 100 Tomsk Moscow city Orenburg Tyumen Komi 95 90 .08 .1 .12 .14 .16 Herfindahl of sectoral employment share Source: Authors calculations using data from Rosstat (2014). Data for employment by sector and real GDP growth rate by region. Note: Herfindahl is calculated as the sum of the squared employment share of each sector and ranges between 0 and 1. Moscow city, St. Petersburg city, Moscow oblast, Nizhny Novgorod and all resource-rich regions are marked in red. Spatial disparities matter for equity 13. Spatial disparities also matter for equity reasons, as inequality between the regions of a country is a significant component of overall inequality between individuals. The notion of ‘equity as fairness’ implies that the geographic location of a person should be independent of other outcomes such as income, educational attainment, or ownership of assets, as well as of the opportunities of that individual to pursue a life of her choosing (World Bank, 2017). No matter where people live, they should have equal opportunities to accumulate assets and access quality services. Yet, inequalities related to place of birth persist and one’s geographic location is often correlated with other “circumstances.” Whether a person is born in an urban or a rural area in the Europe and Central Asia (ECA) region, for example, explains close to 20 percent of the inequality in opportunities to access tertiary education, jobs and in income (EBRD, 2016). In addition, as Kanbur and Venables (2005) note, territorial inequality “has added significance when spatial and regional divisions align with political and ethnic tensions to undermine social and political stability.” 14. Inequality in access to services across a country’s territory matter particularly for equity outcomes in non-monetary dimensions of well-being. To better analyze these challenges, Ceriani and Lopez-Calva (2016) propose a methodology for quantifying regional disparities in state effectiveness for key service domains. Applying this methodology in Russia reveals that some regions are systematically affected by a low state presence. Infant mortality incidence, education outcomes, and access to utilities vary significantly across regions (Error! Reference source not found., panels a-c). Taking into account the weighted average for these indicators (the ‘composite density of the state’), the top five regions with the highest state presence are Moscow city, Murmansk oblast, Tatarstan republic, St. Petersburg city, Chuvash 14 republic (Error! Reference source not found., panel d). The bottom five regions with the lowest state presence are Chukotka autonomous okrug, Tuva republic, Evrei autonomous oblast, Altai republic, and Sakha (Yakutia) republic. The outcome for Chukotka is surprising given the household income in Chukotka in 2014 is the third highest, only behind Moscow and St. Petersburg. This is mainly a result of having the highest infant mortality rate and its students being some of the worst performers in the university entrance exam in the country (third only after Amur and Magadan oblasts). On the other hand, composite state density seems to be correlated with monetary indicators of poverty, at least at the bottom, as Tuva, Evrei, and Altai republic are among the top 5 regions with the highest poverty rates. The other two regions with the highest poverty rates (Ingush republic and Kalmyk republic) also have relatively low composite state density. In this context, it is useful to know that there have been interventions in the past that helped decrease the inequality of outcomes: for example, the expansion of primary health care in 2006 led to a slight decrease in male mortality in those regions where it was previously the highest. Figure 5. State density by region: Infant mortality, education, and utilities (2014) a. Infant mortality b. Education 15 c. Utilities d. Composite density Source: Authors’ calculations based on Ceriani and Lopez-Calva (2016). Data from Rosstat (infant mortality), Federal Testing Center (education), and HBS (utilities). Note: “Infant mortality” is measured as the number of deaths under one year of age per 1000 live births. “Education” is measured by the mean test score from Russia’s university entrance exam (Unified State Examination). "Access to utilities" is measured as the average of three sub-indicators: whether the household experienced any electricity outages in the past year, whether the household is connected to the district sewage system, and whether the household receives piped water through the central supply system. Each indicator is normalized between 0 and 1. Darker blue shading indicates higher state effectiveness (indicator closer to 1). Composite density represents the equally weighted average of infant mortality, education, and utilities. Spatial disparities matter for governance 15. Finally, spatial disparities matter for institutional reasons—as they are both the result and a driver of institutional dynamics. Institutions, broadly understood as the rules and organizations which emerge from the agreements among state and non-state actors, differ across regions. The differences in de jure policies and laws as well as de facto social norms can have large impacts on the effectiveness of local interventions. This can have significant impacts on equity outcomes, as the above analysis on state 16 effectiveness reveals, but also on efficiency outcomes. For example, vast disparities in regulatory indicators suggest significantly differential institutional treatment of firms across regions: while it takes less than month on average to obtain a construction permit in Moscow City, it takes over a year in the Tver region (Figure 6). Understanding how the institutional environment shapes service provision and economic activity at the regional level is critical for developing a more effective territorial policy agenda. Figure 6: Days to obtain a construction permit, by region Tver Region 369.5 Krasnodar Territory Khabarovsk Territory Kaliningrad Region Perm Territory Voronezh Region Leningrad Region Samara Region Kursk Region Republic Of Mordovia Republic Of Tatarstan Moscow Region Omsk Region Ulyanovsk Region Volgograd Region Saint Petersburg Sverdlovsk Region Rostov Region Republic Of Sakha (Yakutia) Kaluga Region Lipetsk Region Novosibirsk Region Stavropol Territory Belgorod Region Moscow City 29.3 Russian Federation Europe & Central Asia All Countries 0 50 100 150 200 250 300 350 400 Source: World Bank Enterprise Surveys (2012). Note: Data available for 25 regions. 16. Spatial disparities are also a driver of institutional dynamics as differential outcomes impact the ability of local actors to influence future policy decisions. When circumstances are not independent from outcomes, not only do they play a part in determining the socioeconomic achievement of people— they also affect the bargaining power of actors. Being born in a region with lower access to services, economic opportunities, and access and voice to influence the policymaking process, limits the ability of individuals not only to shape their present, but also that of their offspring, contributing to lock in households in inter-generational traps. Inequalities may be reflecting the ability of certain groups of actors to influence policymaking and the allocation of resources in society, systematically favoring their interests (World Bank, 2017). This unequal bargaining power (through lobbying) feeds back into a more inefficient allocation of resources and the further entrenchment of existing inequalities over time (Esteban and Ray, 2006). 17. At the level of policy coordination, this has important implications for the role played by federal vs regional authorities and is consistent with an increasing focus on institutional development as key variable to explain spatial disparities in economic growth in Russia. Using regional data related to entrepreneurship, perceptions of corruption, and trust in institutions, Kaliuzhnova (2011) finds evidence of an institutional trap at the “federal level of rules which allow governors solving current problems to some extent and do not allow them implementing of the proclaimed policy objectives of improving the 17 regional competitiveness” (p 67). The author argues that enabling greater economic independence at the regional level is critical for fostering competitiveness and regional development. Beyond markets: Russia’s regional policies 18. Fiscal transfers are at the heart of the relationship between the center and the regions in Russia. The federal budget transfers large amounts of resources to the regions to support federal programs, and own mandates of regional budgets. There are three main categories of transfers: grants, subsidies and subventions. Grants are not earmarked and can be spent at the discretion of the recipient. Subsidies are federal matching grants. These support a wide range of federal programs, some —but not all—of which involve capital investments. The typical matching arrangement is 70 percent federal/30 percent regional, although this varies. The third major category of transfers –subventions—consists of compensation for functions that subnational governments perform on behalf of the federal government. These include unemployment subsidies, rent subsidies granted to certain categories of federal beneficiaries (such as war veterans or victims of radiation catastrophes), benefits paid to blood donors, and the costs of running civil registration offices (World Bank. 2016b). Out of all transfers, most are in the form of grants (36 percent on average) and more specifically grants for equalization (25.6 percent on average). Subsidies account for 31 percent of all transfers on average, subventions account for 21.5 percent, and the rest is made up of other transfers. 19. The regional budgets are financed from a combination of shared taxes, exclusive local taxes, non-tax own revenues, and intergovernmental transfers (World Bank 2016b). The two biggest sources of regional revenues are Personal Income Tax (PIT) and Corporate Income Tax (CIT). PIT is exclusive subnational tax and accounts for about 30 percent of subnational budgets, from which 85 percent goes to regional budgets and the rest to lower level subnational budgets. CIT is a shared tax between regional and federal budgets. 90 percent of CIT goes to regional budgets and this constitutes about 23 percent of their revenues. Another 8 percent of subnational revenues are coming from various property taxes, including corporate asset tax and land tax (World Bank 2016b). More than half of subnational budget spending are linked to social expenditures (education, health and social protection). Education is the largest category of expenditures and accounts for 26 percent of total expenditures (World Bank 2016b). 20. Russia’s prevailing policy stance has been described as more “equalizing” than in other developing countries (Zubarevich 2009, p. 3). Still, the country is destined to be the “the country of perpetual catch-up development” (Ibid. p 6) if federal policies do not contribute to greater agglomeration effects, the development of human capital and the modernization of institutions. These factors are particularly crucial for middling regions, those that are currently most resource constrained as they are neither dynamic leaders, nor the target of special interventions for the poorest regions. Even lagging regions whose income is greatly dependent on federal transfers (Box 3), however, suffer from the lack of a developmental emphasis in the transfers they receive. 18 Box 3: Poorer regions are most dependent on transfers Many regions heavily depend on transfers from the federal budget. According to the latest data (2011-2015) there are 11 regions that had average share of transfers from federal budget in their fiscal revenues over the 5-year period higher than 50 percent. And there are another 18 regions that have a share of transfers between one third and one half of their total revenue. The median share of transfers exceeds one quarter. The highest shares of transfers are in the republics of the North Caucuses (Ingush – 85 percent, Chechnya – 84 percent, Dagestan – 71 percent) and South of Siberia (Tyva – 76 percent and Altai – 74 percent). There is also a strong relation that the richer the region, the lower the share of transfers in its budget (see figure B2.1). The lowest share of transfers is in the resource-rich regions of Khanty-Mansiysk, Yamalo-Nenetsk and Sakhalin and metropolitan areas of Moscow and St. Petersburg. Figure B2.1: Share of transfers from federal budget in the regional revenues and GDP per capita 90 80 Share of transfers in revenue 70 60 50 40 30 20 10 0 0 100 200 300 400 500 600 700 800 900 1,000 GDP per capita (thousand rubles, annually) Source: Treasury data, Rosstat and authors’ calculations. 21. More fundamentally, the plurality of objectives followed over time, and the growing dependency of regions from the central government appear to have hindered the ability of regions to build on their competitive advantages. Since the beginning of the 1990s the governance of the relations between the central government and the regions has evolved (Box 4) and so have the policy objectives pursued. The regional policy concept adopted in 2005 for example focused on existing leading regions as engines of growth for the federal districts in which they lay. At the same time infrastructural and industrial projects were planned for the Far East, though their economic feasibility was in doubt. More recently, the policy direction has been focused on “agglomerations”, which has led to a proliferation of plans for “agglomerations” throughout the regions. Investments driven by geopolitical consideration or high visibility (such as those around Sochi for the Olympic Games) seem to have become bigger priorities now. Against the background of shifting goals, a rebalancing of fiscal resources towards the center is contributing to make regions increasingly dependent from the federal budget and eroding their incentives to compete to attract private investment. 19 Box 4: Center-region relationship since the transition Regional development in the first years of transition was characterized by a focus on greater autonomy, including the creation of free economic zones throughout the country as areas where central regulations did not apply. The rest of the decade saw a greater emphasis on regulating the relations and transfers between a stronger center and the regions, including the creation of the formula based Regional Fund for Financial Support of the Regions in 1994. The 2000s saw further strengthening of the center vis-à-vis the regions, including the introduction of 7 (later the number was increased to 8) “macro -regions” (federal districts) and municipalities as key structures of governance. Achieving greater uniformity in the development strategies and planning of different regions was a key objective. Between 2004 and 2014 the existence of a Ministry for Regional Development provided a locus for thinking systematically about regional policy, as opposed to giving it a secondary role with respect to organizational or political goals. The Regional Development Concept proposed by the Ministry and adopted in 2005 was however never implemented. Since the Ministry was abolished in 2014, its responsibilities have been transferred to other ministries with a territorial portfolio, such as the Ministry for the Development of the Far East, the Ministry of Economic Development and the Ministry of Construction and Housing. Source: Leonid Limonov.2 22. While Russia’s regional policies have been centered around equalization by means of inter - budgetary transfers, policies in OECD countries and the EU have evolved over time to focus on boosting regional competitiveness and growth. In many OECD countries, the policy framework moved away from a top-down, subsidy-focused approach to one designed to build regional potential and broader territorial development. The objective comprises both an equity and efficiency (growth) angle, with the latter receiving increasingly more attention. The approach is increasingly context- specific taking into account local conditions and assets, with the central government’s role being redefined into one that provides an overarching framework and coordinates across sectors and regions. In the EU, the Cohesion Policy with the objective to reduce disparities across regions and member states aims to make them more competitive, fostering growth and creating jobs. Funds under this Policy are used to improve the broader business environment, research and innovation, key infrastructure, education and training, social inclusion and professional adaptability of workforce (Box 5). Box 5. Regional development policies in OECD countries Paradigm shift of regional development policy Regional policy in most OECD member countries started in the 1950s and 1960s with the objective of achieving greater equity and balanced development during a time of rapid industrialization and increasing regional disparities. Until the 1980s, regional policy predominantly focused on regional investment aid and infrastructure support, with policy interventions heavily targeting lagging regions. In the EU, Cohesion policy also mainly focused on infrastructure as the assumption was that convergence was not always ensured through market mechanisms. Despite various efforts, regional disparities were not significantly reduced. Against a background of increasing globalization, decentralization, and budget strain2 since the 1980s, large allocations for regional programs have become unsustainable in a period of successive economic recessions, generalized higher levels of 2 http://www.regionalstudies.org/uploads/Regional_Studies_in_Russia.doc 20 unemployment and increasing pressure on public expenditures. This prompted regional policies to evolve from top-down subsidy-based interventions designed to reduce regional disparities into much broader policies designed to improve “regional competitiveness”. National governme nts are increasingly favoring regional growth over redistribution, in pursuit of national or regional competitiveness and balanced national development. Territorial development instruments have become broader and have adapted to the requirements of individual regions, against a growing trend of decentralization to the regional levels. Regional strategic programs have grown in prominence, reflecting a general policy shift towards support for endogenous development and the business environment, building on regional potential and capabilities, and aiming to foster innovation-oriented initiatives. In a move away from previous approached dominated by central government, multi-level governance approaches involving national, regional and local governments as well as third-party stakeholders (e.g. private actors and non-profit organizations – NPOs) have increased in importance (Table 2). Objectives, policy framework, tools and governance for regional policy Most countries implement regional policies with both equity (regional balance) and efficiency (growth and competitiveness) objectives, with the latter receiving increased attention. The equity dimension is typically discussed from the perspective of allowing all citizens equal opportunities and human rights- the latter concern satisfactory living conditions especially in relation to access to public services such as basic education and transport infrastructure. In most OECD countries today, regional policies are focused on regions’ growth potential based on regional assets and not just limited to the challenges by declining regions. Reflecting this change in the general framework, regional policies tend to now have an all-region focus, moving beyond the dichotomy of leading versus lagging regions, as all regions need to develop unique strategies for regional development. For example, though the regional development agencies of Canada have undertaken similar activities at a broader level, programming varies from region to region in order to be responsive to local conditions and address specific gaps. In terms of policy tools, efforts to improve the business environment have been key: in order to support the development of firms, there has been a move away from aid (income transfer to residents or subsidies to firms in poor regions) towards support to increase availability of skills, access to information and access to network infrastructure. Finally, the expansion of regional policy requires coordination across sectors and borders vis-à-vis previous models that targeted specific sectors in specific territories. The role of the central government is being redefined rather than diminished to provide an overarching framework and oversee coordination mechanisms within which regional policy is formulated: e.g., facilitate consensus-building and coherence between regions and sectors, gather and analyze data and information and coordinate discussions and concerning needs and opportunities, develop legal, fiscal and administrative frameworks, help weaker entities establish capacity-building strategies, evaluate and monitor policy results. Table 2. Paradigm shift of regional development policy 21 Old paradigm New paradigm Problem recognition Regional disparities in income, Lack of regional competitiveness, infrastructure stock and employment underused regional potential Objectives Equity through balanced regional Competitiveness and equity development General policy framework Compensating temporally for Tapping underutilized regional location disadvantages of lagging potential through regional regions, responding to shocks (e.g., programming (Proactive for industrial decline) (Reactive to potential) problems) - Theme coverage Sectoral approach with a limtied set Integrated and comprehensive of sectors development projects with wider policy area coverage -Spatial orientation Targeted at lagging regions All-region focus -Unit for policy intervention Adminstrative areas Functional areas -Time dimension Short term Long term -Approach One-size-fits-all approach Context-specific approach (place- based approach) -Focus Exogenous investments and transfers Endogenous local assets and knowledge Instruments Subsidies and state aid (often to Mixed investment for soft and hard individual firms) capital (business environment, labor market, infrastructure) Actors Central government Different levels of government, various stakeholders (public, private, NGOs) Source: OECD, 2010. Regional Development Policies in OECD Countries. Regional policy of the European Union EU cohesion policy started around 30 years ago and was necessitated by disparities across member countries and their regions. Managed by the Directorate General for Regional Policy, it is the EU’s main investment policy which explicitly targets all regions and cities to make them more competitive, fostering growth and creating jobs. Funding is delivered through 1) the European Regional Development Fund which invests in growth-enhancing sectors to improve competitiveness and create jobs, 2) the European Social Fund which invests in people with a focus on education and employment and 3) the Cohesion Fund which invests in green growth and sustainable development and improves connectivity in member states with a GDP below 90% of the EU-27 average. The funds are used to co-finance public investments: between 2014 and 2016, the Funds are expected to account for approximately 14% of total public investment on average, and even reach up to 70% in some member countries. In the 2014-20 policy cycle, a total of 454 billion Euros from the EU budget combined with 183 billion Euros in national co-financing will support over 500 programs to support investments in research and innovation, key infrastructure, education and training, social inclusion and professional adaptability of the workforce. Source: OECD (2010) “Regional Development Policies in OECD Countries“, European Commission’s Regional Policy website (http://ec.europa.eu/regional_policy/en/, accessed June 19, 2017) 22 3. Regional inequality in Russia: Trends and composition 23. This section provides an initial diagnostic of regional inequality in the past decade. Micro- level analyses using survey data to explore regional difference in welfare in Russia are relatively scarce in the literature (see Box 6 for a brief overview of previous studies). In this section, 2005-2014 data on consumption expenditure from Household Budget Survey (HBS) and 2014 data on income from the Survey of Income and Social Program Participation is used to provide an assessment of more recent trends and composition of regional welfare disparities. Box 65: Micro-level analyses on welfare disparities in Russia A few studies have used survey data to explore drivers in regional welfare disparities in Russia. Most of these studies rely on data from the 1990s or early 2000s. For example: Commander et al (1999) attribute rising inequality during the transition period to major changes in households’ income structure, as the dispersion of wages increased and social assistance declined, and find that the between component in regional inequality widened over time. Specifically, the increase in inequality is attributed to increased dispersion in wages, driven by liberalization of wage settings and intensified by growth of the private sector, and accumulation of wealth following the privatization of public assets that started towards the end of the Soviet era. Spending on social assistance and subsidies also declined. Using HBS data, Yemtsov (2003) finds that the relative share of inequality between regions (compared to within regions) was large and increased during 1994-2000. An absence of interregional convergence is demonstrated as well, while inequality within regions appeared to be converging towards an internationally high level. Kolenikov and Shorrocks (2005) also investigate some proximate explanations for differences in poverty across regions on the basis of a Shapley decomposition analysis —exploring the possible role of per capita income, inequality and prices. Contrary to expectation, regional poverty variations turn out to be due more to differences in inequality across regions than to differences in real income per capita. However, when real income per capita is split into nominal income and price components, differences in nominal incomes emerge as more important than either inequality or price effects for the majority of regions. 24. Inequality at the regional level varies substantially across regions and is the main driver behind overall inequality (Figure 7). Regional Gini estimates range from about 0.28 in Vladimir, Karachaevo-Cherkess and Dagestan to greater than 0.4 in Irkutsk, Moscow and Tyumen. Moreover, A Theil decomposition of household per capita consumption reveals that over 90% of current levels of inequality are due to intra-regional inequality. While this is consistent with findings that within-region inequality is typically greater than between-region inequality (Shorrocks and Wan, 2004; Kanbur, 2006), this is a notable increase from the previous decade in Russia. During the period 1994-2000, only about two thirds of the increase in inequality was driven by intra-regional distributions (Yemtsov, 2003). In the last ten years, with the exception of 2015, inequality seems to have decreased very little and the within-region dispersion appears to have grown even more. Figure 7: Share of total inequality driven by inequality between or within regions 23 100% 0.45 90% 0.4 80% 0.35 70% 0.3 60% 0.25 50% 0.2 40% 0.15 30% 20% 0.1 10% 0.05 0% 0 2005 2006 2007 2008 2009 2010 2011 2012 2013 2014 2015 %Within %Between Theil (GE(1)) Gini Source: Staff estimation using HBS 2005-2015. Note: “%Within” and “%Between” indicators refer to the Theil measure of inequality 25. Inequality is driven by the richest and most populous regions. The top 15 regions (out of the 79 regions for which data is available) that contribute the most to inequality explain 21% of total inequality. These regions are home to 46.2% of the population and tend to be either metropolitan areas or resource- rich regions. The top three contributors are Moscow City, Moscow Oblast and St. Petersburg. Of the remaining 12 regions, 7 regions had a mining share of over 20% in 2015. 26. Poor places in Russia are not where most of the poor people live. One of the reasons why richer regions are also more unequal is that a large share of the poor is concentrated in richer regions. While the percentage of people living in poverty is very high (34.7%) in poorer regions such as Tuva republic and Kalmyk republic, because of the low population in these regions, the number of people living in poverty only account for 0.6% of all poor people in Russia. In absolute terms, richer regions are home to a much larger number of poor people despite their low poverty incidence because of their high populations. For example, despite having some of the lowest poverty rates in the country (around 7.5%), the cities of Moscow and St. Petersburg together account for almost 10% of the poor people in Russia, (Table 3). This is not uncommon as similar patterns are observed in other large countries such as Brazil and China. Table 3. Poverty in selected rich and poor regions Region Poverty rate 2014 Share of poor people living GRP 2014 in this region Moscow city 7.5 6.6% 12,808,573 Moskow oblast 7.6 3.3% 2,705,579 St. Petersburg city 8.3 2.6% 2,652,050 Krasnodar krai 10.1 3.3% 1,792,048 24 Tyumen oblast 12.1 2.6% 5,178,490 Tuva republic 34.7 0.6% 46,707 Kalmyk republic 34.7 0.6% 46,044 Ingush republic 24.9 0.7% 52,168 Evrei autonomous oblast 21.4 0.2% 41,742 Altai republic 20.7 0.3% 39,135 Source: Rosstat and authors’ calculation. 27. While the majority of regions in Russia saw a substantial decrease in levels of inequality over the period 2005-2015, many of the richest and populous regions saw only very small decreases – and in some cases, actually experienced increases in inequality (Figure 8). This is reflected in an aggregate national trend of inequality levels that has remained largely stable, with a slight decrease recently in 2015.3 As regions across Russia have been differently impacted by the global crisis as well as the recent fall in oil prices, some of the broad reduction in inequality may be explained by decreases in the incomes of those at the top, rather than increases in the incomes of those at the bottom. Thus, it remains to be seen whether this decrease in inequality is structural or just a temporary narrowing of inequality due to the crisis. Figure 8: Regional change in inequality (Gini), 2005-2015 Source: Staff estimation using HBS data. Note: Size of circle proportional to the size of the population. 3 While inequality measured using household survey data is shows a slightly decreasing trend, there is evidence that wealth inequality has increased (consistent with subjective perceptions that that inequality in general is increasing). According to estimates from Credit Suisse (2016), the top decile in Russia owns 89% of the nation’s wealth –this is higher than the corresponding figures for the United Sates and for China. 25 28. Newly available survey data on household income is utilized in order to provide a fuller characterization of intraregional inequality and how the regional welfare distributions compare across regions. This section compares the richest and poorest regions in Russia to draw a clear contrast in the levels and composition of income among households in those regions and their statistical profiles. To select comparator regions, the analysis considers the “poorest regions” the three regions with lowest mean per capita income, and considers the “richest regions” the three regions with the highest mean per capita income.4 Respectively, the groupings include the regions of Ingush, Altai republic and Adygeya and Moscow city, St. Petersburg and Chukotka (highlighted in red and green in Figure 9). Figure 9: Per capita income by region 585643.3 600000 500000 400000 300000 193194.1 200000 100000 0 Leningrad oblast Kamchatka krai Krasnodar krai Vologda oblast Altai krai Bryansk oblast Kurgan oblast Smolensk oblast Vladimir oblast Penza oblast Tambov oblast Belgorod oblast Novosibirsk oblast Komi republic Ingush republic Tuva republic Mordovia republic Khakasia republic Stavropol krai Novgorod oblast Arkhangelsk oblast Tyumen oblast Murmansk oblast Tver oblast Samara oblast Lipetsk oblast Amur oblast Kirov oblast Ivanovo oblast Kostroma oblast Rostov oblast Oryol oblast Bashkortostan republic Buryat republic Voronezh oblast Khabarovsk krai Kalmyk republic St. Petersburg city Evrei autonomous oblast Mari-El republic Ulyanovsk oblast Orenburg oblast Tula oblast Chelyabinsk oblast Pskov oblast Saratov oblast Nizhny Novgorod oblast Udmurtia Republic Kaluga oblast Moscow city Kaliningrad oblast Primorskii krai Perm krai Sakha (Yakutia) republic Kemerovo oblast Krasnoyarsk krai Sakhalin oblast Adygeya republic Irkutsk oblast Chuvash republic Ryazan oblast North Osetiya republic Dagestan republic Sverdlovsk oblast Volgograd oblast Yaroslavl oblast Tomsk oblast Omsk oblast Zabaikalskiy krai Astrakhan oblast Magadan oblast Kursk oblast Moskow oblast Chukotka autonomous okrug Altai republic Kabardino-Balkar republic Karelia republic Tatarstan republic Karachaevo-Cherkess republic Source: Staff estimation using Income Survey 2014 29. Households’ reliance on income sources varies a lot by quintile but also across regions (Figure 10, Figure 11). Household income consists of broadly labor income (L), income from financial and non- financial assets (A) and transfers (T). The biggest contrast lies in the share of households that rely primarily on labor income vs transfers. While in the richest regions, at least 20-30% of households live off labor income only regardless of income quintile, that share is significantly smaller across all quintiles in the poorest regions. This is also reflected in the low share of labor income out of total income across all quintiles but particularly for the bottom quintile in the poorest regions. On the other hand, more than a quarter of households in the top quintile in the richest regions draw income from a diversified mix (labor, transfers and assets) and with the exception of the bottom quintile, the income is drawn predominantly from labor with all other sources accounting for only about 10 %. Unsurprisingly, within both the poorest and the richest regions the bottom quintile has the largest share of households that rely on transfers as their only 4The dataset used in the rest of this section and for the Oaxaca decomposition is the Survey of Income and Social Program Participation of 2014. 26 source of income. However, in the poorest regions this share is closer to 60 percent whereas it is only about 25 percent in the richest regions. Figure 10: Number of income sources by quintile Figure 11: Share of income source in total household income, by quintile 100% 100% 90% 90% 80% 80% 70% 70% 60% 60% 50% 50% 40% 40% 30% 30% 20% 20% 10% 10% 0% 0% Q1 Q2 Q3 Q4 Q5 Q1 Q2 Q3 Q4 Q5 Q1 Q2 Q3 Q4 Q5 Q1 Q2 Q3 Q4 Q5 Top 3 regions Bottom 3 regions Top 3 Bottom 3 L A T L+A L+T A+T L+A+T % labor inc % asset inc % transfers Source: Staff estimation using Income Survey 2014. Source: Staff estimation using Income Survey 2014. Note: Figure shows share of households by income source. L (labor income only), A (asset income only), T (transfer income only), L+A (labor and asset income), L+T (labor and transfer income), A+T (asset and transfer income), L+A+T (income from labor, assets and transfers). 30. A greater reliance on labor income in the richest regions helps explain why income inequality tends to be greater in these regions. Labor income usually drives income inequality because it exhibits high levels of inequality (as opposed to transfers) and accounts for a large share of household income (as opposed to asset income). A Gini decomposition by income source using data from the 2014 Income Survey reveals that more than 90 percent of overall income inequality in Russia is in fact determined by labor income, both through its high share in overall income (69%) and its high level of inequality (Table 4). Asset income inequality tends to be very high but because of its very small share in overall income, its impact on total inequality is almost negligible. This can also be seen from Lorenz curves which show the cumulative percentage of total income, pension and labor income against the cumulative percentage in the population, ranked by total income (Figure 12). Social transfers are deemed relatively progressive when the concentration curve lies between the 45-degree line and the market income Lorenz curve, which is the case of pensions in Russia. Pensions are much more equally distributed compared to total income with the Lorenz curve located closer to the 45-degree line and are therefore inequality-reducing. This is also consistent with results from a World Bank report which showed that the whole redistribution system in Russia is dominated by pensions that accounts for two thirds of total reduction of Gini (Matytsin, Popova, Sinnott, 2017). 27 Table 4. Gini decomposition by income source Source Sk Gk Rk Share Labor income 0.678 0.575 0.891 0.908 Non-labor income 0.322 0.530 0.206 0.092 Asset income 0.004 0.973 0.509 0.005 Transfers 0.225 0.534 0.208 0.065 Pensions 0.154 0.595 0.013 0.003 Other transfers 0.071 0.726 0.462 0.062 Total income 0.382 Source: Staff estimation using Income Survey 2014. Note: Sk: Share in total income, Gk: Source Gini, Rk: Gini correlation of income source with total income distribution, Share: Percentage contribution to overall inequality Figure 12. Lorenz curve of incomes 1 0.9 0.8 0.7 0.6 0.5 0.4 0.3 0.2 0.1 0 0 0.2 0.4 0.6 0.8 1 45-degree line Total income Labor income Pension Source: Staff estimation using Income Survey 2014. Note: X-axis shows the cumulative proportion of the population; y-axis shows the cumulative proportion of total income, labor income and pensions. 31. The distribution of pensions is particularly progressive and equalizing in Moscow and St. Petersburg, two of the richest regions in Russia. Figure 13 contrasts the Lorenz curves of total income, labor income and pensions in the two cities against those in all other regions combined. In the case of Moscow, where the absolute progressivity of pensions is especially strong, regional social pension supplements play a large role as they are provided on the basis of special rules and are financed 28 from the city’s own revenues.5 Moscow established its own rules with the key distinction being that pensioners who are registered in Moscow at the place of stay or residence for less than 10 years are granted supplements on the basis of general Russian rules. For nonworking pensioners who are registered in Moscow at the place of their residence for 10 years or more, the city government established a supplement to the city’s higher social standard of 14,500 rubles, which exceeds by 27 percent the special pensioner subsistence minimum established in Moscow for 2016. Besides, the assessment of supplements for this category of pensioners takes into account only the amount of pension benefits, excluding all other components of total income as defined by the federal law. As a result, the number of social supplement beneficiaries in Moscow exceeds 2 million people (Gorina, 2016). Figure 13. Lorenz curves in Moscow, St. Petersburg and other regions a. Moscow 1 0.9 0.8 0.7 0.6 0.5 0.4 0.3 0.2 0.1 0 0 0.2 0.4 0.6 0.8 1 45-degree line Total income Labor income Pension b. St. Petersburg 5 Absolute progressivity is measured by the Lorenz curve’s location relative to the 45 -degree line. If it is above it, social transfers are deemed to be progressive in absolute terms. 29 1 0.9 0.8 0.7 0.6 0.5 0.4 0.3 0.2 0.1 0 0 0.2 0.4 0.6 0.8 1 45-degree line Total income Labor income Pension c. All other regions 1 0.9 0.8 0.7 0.6 0.5 0.4 0.3 0.2 0.1 0 0 0.2 0.4 0.6 0.8 1 45-degree line Total income Labor income Pension Source: Staff estimation using Income Survey 2014. Note: X-axis shows the cumulative proportion of the population; y-axis shows the cumulative proportion of total income, labor income and pensions. 32. Household characteristics also tend to differ quite substantially across regions and quintiles. Home ownership tends to be very high in Russia overall, and is actually higher in the poorest regions (92%) compared to the richest regions (80%). The high home ownership partly reflects a legacy from the transition when homes were transferred to their occupants at little to no cost (World Bank, 2014). This also explains why there is relatively little variation across quintiles in regards to home ownership. More than 90% of the bottom quintile in the poorest regions and more than 70% in the richest regions are home owners. On the 30 other hand, land ownership is substantially higher in the poorest regions (55% vs 26% in the richest regions). However, the nature of land ownership is likely different across regions, as land ownership in the poorest regions likely represents agricultural land holdings used for subsistence farming whereas in the richest regions it likely represents ownership of non-agricultural assets such as financial income. Almost half of the households in the poorest regions are headed by a person that is not working, and for the bottom quintile this figure is at a staggering 85%. In the richest regions, the comparable estimate is at about a third. Finally, a very high share of households (32%) in the poorest regions work in the public sector, with about 57% in the bottom quintile holding a public sector job. This is in stark contrast to the allocation in rich regions where only about 10% of household heads are working in the public sector, with little variation across income quantiles. Figure 14: Distribution of household characteristics by quintile in richest and poorest regions in Russia 1.00 0.80 0.60 0.40 0.20 0.00 Q1 Q2 Q3 Q4 Q5 Q1 Q2 Q3 Q4 Q5 Top 3 regions Bottom 3 regions House ownership Land ownership Educ: less than secondary Educ: Primary professional Educ: Secondary professional Educ: University and higher Works in public sector Source: Staff estimation using Income Survey 2014 Figure 15: Occupational distribution by quintile in richest and poorest regions in Russia 1.00 0.80 0.60 0.40 0.20 0.00 Q1 Q2 Q3 Q4 Q5 Q1 Q2 Q3 Q4 Q5 Top 3 regions Bottom 3 regions Not working Managers Professionals (high qualificatin) Professionals (medium qualification) Clerical support workers Service workers Skilled agricultural workers Plant, machine operators, assemblers Source: Staff estimation using Income Survey 2014. Note: The shares across all occupations sum up to 100 percent for each quintile. 33. In sum, high levels of intraregional inequality are the main drivers of overall inequality and determined by significant differences in households’ productive characteristics (education) and labor market outcomes, including labor force participation. Poor households in the most lagging regions are significantly less likely to work, attain less education and hold less-rewarding jobs – relative to households in richer quintiles but also compared to households in the bottom quintile of the richest regions. This disadvantage in terms of endowments inevitably translates into differences in observed welfare. Section 5 31 formally investigates the relative contribution of differences in endowments and returns to endowments to differences in welfare. 4. Are the gaps closing? Regional convergence in income and poverty 34. Unlike the case of income inequality (which relies on the Gini coefficient), there is no standard approach to estimate spatial disparities. Instead, the literature contains different approaches and indicators. Some studies look at convergence of nominal and real incomes, wages or GDP. Other research compares poverty levels across regions, including in the form of poverty maps. Others look at the distribution of access to services, or at the agglomeration of economic activity. In addition to the indicator used, spatial disparities—and the policies to address them—are related to the geographical scale utilized. For example, in the case of Russia, it seems that the literature has largely ignored the “macro -regions” (federal districts) that were instituted in the 2000s, while lack of small scale estimation indicators (such as poverty maps) makes the 80+ regions discussed above the most disaggregated level at which quantitative analysis is possible. 35. Several studies document the rise of overall and regional inequality and the lack of convergence throughout the 1990s in Russia (Gluschenko 2010; Guriev and Vakulenko, 2012). Guriev and Vakulenko (2012) find evidence of weak convergence throughout the 2000s when using regional GDP per capita as outcome and strong convergence when using incomes and wages. While there is some evidence of subsequent convergence in living standards across regions in the early 2000s (Federov, 2002; Kholodilin et al, 2009), the evidence is less clear on what happened over the past 10 years. 36. In this note, individual regions in Russia are used as the scale of analysis to understand recent trends in convergence for real income per capita and for poverty incidence. From a relative perspective, a simple estimation of the degree of variation across regions suggests that over the last fifteen years regions have experienced sigma convergence (convergence in the level of dispersion) in income and poverty across regions in terms of real income per capita but not in terms of poverty incidence ( Figure 16). While in 1998 regions were more heterogeneous in terms of real income per capita than poverty, this position had reversed by 2008 – as the regional variation of poverty incidence remained relatively stable and income continued to converge. However, in order to understand interregional convergence from an absolute perspective, this analysis estimates conditional beta-convergence to understand whether poorer regions have been able to ‘catch up’ with richer regions in real income per capita or poverty incidence due to higher growth rates. 32 Figure 16: Degree of variation in wage, income, GDP, and poverty incidence across regions (1998-2015) 0.4 0.35 0.3 Gini Coefficient 0.25 0.2 0.15 0.1 0.05 0 1998 2000 2002 2004 2006 2008 2010 2012 2014 Wage Real income Nominal income GDP Poverty Source: Staff estimation using Rosstat data. Note: In this figure, the Gini coefficient is used to measure the degree of variation for each indicator across regions. Data and methodology 37. This analysis uses an estimation dynamic panel data model with spatial effects to see if Russia’s regions experienced conditional beta-convergence for real income per capita and poverty incidence over the past decade (see Box for a more technical description of the estimation model). The analysis uses official data from the Russian Statistics Service (Rosstat)6 for 78 Russian regions for which we have data during the period 2004-2015. 38. The analysis takes into account the role of key socio-economic variables and regional spillover effects that influence convergence. The conditional convergence approach controls for the effect of variables such as demographic indicators (population growth rate, share of youth, share of pensioners), net migration rate, unemployment rate, sectoral structure of the economy (share of mining, manufacturing, and construction in GDP), share of transfers in budget income, and real investments per capita. The two last variables allow us to evaluate the role of government in income convergence and the contribution of capital mobility. Regional spillovers are measured in this analysis by the physical distance between the respective regional capitals by railway (see Box 7 for more details). Box 7: Estimating regional conditional beta-convergence using a dynamic panel data model with spatial effects In order to estimate absolute regional convergence of real income per capita and poverty rate, this analysis uses a conditional -convergence model similar to Barro and Sala-I-Martin (1991). However, it extends their approach 6www.gks.ru, Russian Regions. 33 in two key ways by exploiting the basic model’s data structure. The basic estimation model is extended in this analysis specifically to incorporate (i) dynamic panel data and (ii) spatial effects. The basic estimation model (1) is:  y  K ln  i ,t    i   t   ln  yi ,t 1  + k X k ,i ,t   i ,t y   i ,t 1  k 1 In the basic estimation model (1), yi ,t is the dependent variable for region i in year t ,  i is a regional fixed effect, t is a time effect, X k ,i ,t is the set of explanatory variables, i is the region’s index, k is the index of an independent variable, and  and  k are estimated coefficients.  represents the convergence. If >0, then there is conditional beta convergence. This means that poorer regions have higher growth rates than richer regions — which is why they are able to ‘catch up’. The dynamic panel data estimation model (2) is: K ln  yi ,t    i   t  1    ln  yi ,t 1  +  k X k ,i ,t   i ,t k 1 In order to incorporate a lag of the dependent variable as well additional independent variables, we can extend model (1) in such way that it becomes a dynamic panel data model (2). For this estimation model we use Blundell and Bond (1998) system generalized method of moments (GMM). In this model two equations, in levels and in first differences, are jointly estimated with additional moment conditions. The equation in levels is instrumented with lagged differences, and the equation in differences instrumented with a lagged variable in levels. To test the joint statistical validity of instrumental variables, we use the Sargan test for over identifying restrictions (Sargan, 1975). Since by requirement, we must assume the absence of second and third order autocorrelation in errors ε_(t,i), we test for this using the conventional Arellano-Bond test (Arellano, Bond, 1991). The dynamic panel data with spatial effect estimation model(3) is: ln  yi ,t   i   t  1    ln  yi ,t 1     i , j ln  y j ,t + k X k ,i ,t   i ,t J K j 1 k 1 In order to take into account spatial autocorrelation, we can extend model (2) by adding a spatial lag to create estimation model (3). In this estimation model, we analyze a spillover effect including the weighted average of the values of our dependent variable for all regions, without the region for which dependent variable is in the left side of the model. The weight for this variable ij is the inverse distance between the region i and all other regions, as measured by the physical distance between their capitals by railway. To test spatial correlation significance for our dependent variable we use Moran’s I statistics. Kukenova and Monteiro (2008) show that it is possible to use the system GMM estimates for the analysis of models involving spatial components. As a robustness check, we also use the maximum likelihood (ML) approach to estimate this model. Regional spillover effects are captured in the following way. In spatial econometric models, the change in the independent variable in a particular region causes a change in the dependent variable not only in this region, but also in neighboring regions, which also affect this region. The marginal effect is described by a matrix of coefficients in this case. Consistent with the literature, this analysis presents the results by summarizing the matrix elements and calculating the so-called direct, indirect and total effects (LeSage and Pace 2009). The direct effect is defined as the average (across all regions) change in the dependent variable in the region with a change in independent variable in the same region. In other words, this is the average value of the diagonal elements of the matrix. The indirect effect is the average change in the dependent variable in the region with a change in independent variable in all other regions (i.е. the mean value of the sum of off-diagonal elements of the matrix of marginal effects). The indirect effect is also called as a spillover effect. The total effect is the sum of direct and indirect effects (i.e. the average change in a real income per capita in the region with a change in independent variable in all regions). 34 Conditional beta-convergence: Results 39. Table 4 presents the results of the conditional-beta-convergence model for real income per capita and poverty incidence (using estimation model 3 described in Box ). Respectively, columns 1 and 5 present the results of the model using the maximum likelihood (ML) estimation method for income and poverty, and Columns 2-4 and 6-8 present the direct, indirect and total effect for the ML estimator. See annex B and C for the full presentation of results, including standard errors and estimation results using the ordinary least squares (OLS) with fixed-effects (FE) and generalized method of moments (GMM) methods. 40. The estimation results find that there is convergence for both real income per capita and poverty incidence.7 For both income per capita and poverty the results are consistent with convergence. This suggests that – conditional on their other characteristics – poor regions are growing faster than rich regions. The results also find that the spatial lag is significant and positive for both variables. Therefore, real income per capita and poverty rates in neighboring regions are positively correlated. This means that there are geographical clusters of poor and rich regions. As a result, the signs for direct, indirect and total effects are the same, so the factors that increase incomes in home region also drive up incomes in the neighboring ones. Table 5: Conditional beta convergence coefficients: Real income per capita and poverty rate Dependent variable: Dependent variable: Real income per capita Poverty rate (1) (2) (3) (4) (5) (6) (7) (8) Control Variable ML Direct Indirect Total ML Direct Indirect Total 0.641** Time lag 0.665*** * 0.323** Spatial lag (rho) 0.224*** * Net internal migration rate (t-1) 0.013* 0.014 0.004 0.018 -0.169 -0.150 -0.074 -0.224 Unemployment rate (t-1) -0.004** -0.004* -0.001* -0.005* -0.003 0.000 -0.001 -0.001 Share of transfers in region -0.002*** -0.002*** -0.000*** -0.002*** 0.022* 0.022* 0.010* 0.032* budget income Real investments per capita (log) 0.039*** 0.042*** 0.012*** 0.054*** -0.181 -0.099 -0.050 -0.149 Share of mining in GDP -0.002** -0.002** -0.001** -0.002** 0.001 0.006 0.003 0.010 0.272** Share of manufacture in GDP -0.007*** -0.007*** -0.002** -0.009*** 0.290*** 0.137** 0.427*** * Share of construction in GDP -0.003*** -0.004*** -0.001*** -0.005*** -0.002 -0.003 -0.001 -0.003 Share of youth (0-15) 0.004 0.004 0.001 0.005 0.072 0.086 0.035 0.131 0.413** Share of pensioners (65+) -0.010*** -0.010*** -0.003** -0.013*** 0.394*** 0.193** 0.586*** * Population growth -0.002 -0.002* -0.000 -0.002* 0.056 0.060* 0.028 0.088 Notes: *** p<0.01, ** p<0.05, * p<0.1 7 The specific type of convergence is “conditional beta-convergence” (Barro and Sala-I-Martin, 1991). 35 Convergence in real income per capita: the role of structural variables 41. A higher unemployment rate in the past period is associated with lower income today (there is persistence). This suggests that problems in the labor market lead to high labor supply and as a result low wages. However, wages are only one source of household income. Therefore, regional labor market conditions such as a demographic crisis or a lack of labor supply increase income. The direct effect of the unemployment rate is three times larger than the spillover effect. As far as migration is concerned, the literature has highlighted that it can either foster convergence (by reducing wage differentials as in the human capital model of migration) or rather amplify disparities (as in the New Economic Geography models – see also below), yet in this analysis it turns not significant.8 The spillover (indirect) effects for the net migration rate are also insignificant, possibly as we can only control for registered migration in the analysis. 42. Some features of the regional economic structure (share of mining, manufacturing, and construction in GDP) also have significant and negative signs, while real investment per capita is associated with higher real income. Resource and industrial regions have lower real income per capita if controlled for other factors. However, as will be shown later in Section 5, this does not necessarily mean that real incomes are lower in resource-rich regions. The only conclusion form this analysis is that having a high share of mining in regional GDP separately from other factors does not guarantee a high level of income in the region. The real investment per capita is associated with higher real income in regions, as it is an important source of capital income. 43. Demographic factors also matter for regional income levels. Faster growing populations are correlated with lower income per capita, possibly because of labor supply effect on wages as well as higher dependency ratios. Similarly, older populations (here proxied by the share of pensioners in the region) have lower income levels. This is not a surprising result, as pensions are lower than wages and pensioners on average have lower incomes per capita than working age adults or mixed families with children. 44. While the increase in the working age population in especially poorer regions has supported income convergence since the 1990s, this trend is set to change soon. Emerging demographic trends could worsen regional inequalities as Russia’s dependency ratio is projected to increase rapidly with recent declines in fertility and continuing improvements in life expectancy. Absent major changes in labor supply or savings behavior or policies to support old-age population, this does not bode well for regional inequalities going forward given that poorer regions have less favorable demographics. Conditional convergence analysis suggests that without behavioral or policy changes, aging will likely negatively affect growth through reduced savings and productivity and therefore convergence (World Bank, 2015). 45. Regions more dependent on transfers have lower income levels, pointing to the important role that transfers have on convergence. Indeed, redistribution through the federal budget has an important role in income convergence, because transfers go to poor regions. The result is confirmed by more detailed 8 A more in depth investigation in Vakulenko (2016) shows that migration has only a very small effect on absolute convergence (and no impact on relative convergence). 36 estimations which break down transfers by major type - subsidies, subventions, and grants. – showing that each type is allocated to poorer regions. Convergence in poverty rate: the role of structural variables 46. The share of transfers in budget income, the share of manufacturing in GDP, the share of pensioners, and population growth matter for poverty convergence. Additional analysis on transfers reveals that –transfers have a positive and significant coefficient for poverty incidence, suggesting that government interventions support poorer regions and help convergence. This result is stable for both total transfers and transfers by components. Major types of transfers have the same effect as total transfers. 47. The apparent positive association between poverty convergence and the share of pensioners in the poverty model is quite puzzling. While pensioners usually have lower incomes than other adults – as was discussed previously—they are also less often poor. This happens because the design of pension system guarantees them pensions at least as high as the poverty line. Thus, unless pensioners have dependents, they are usually non-poor. This is also supported by the negative correlation between the share of pensioners in the region and the poverty rate. In order to find an explanation for this puzzle, the model was tested for a structural break in order to test how changes in the pension system in 2010 might have affected results. In this case the results were unchanged, suggesting that they could not be ascribed to the changes in the pensions system. 5. What factors explain spatial disparities at the household-level? 48. At the center of the policy debate over addressing spatial disparities in living standards, is the question of whether policies should prioritize investing in poor people or in poor places. These represent two contrasting views on what are the main drivers of spatial disparities. These two explanations have been dubbed the ‘concentration’ and the ‘geography’ perspectives, respectively (Skoufias and Lopez- Acevedo, 2009). The “concentration” hypothesis posits that poverty arises from a concentration of individuals with lower levels of endowments, specifically human capital. Under this hypothesis, individuals with the same levels of endowments would achieve the same living standards regardless of their place of residence, and therefore policies should focus on improving human capital outcomes. Examples of such policy interventions include the conditional cash transfer Oportunidades program in Mexico and Bolsa Familia program in Brazil. On the other hand, the “geography” hypothesis suggests that regions lagging behind in living standards are primarily due to low returns to individual characteristics. Geographic areas that are better endowed with public goods such as infrastructure and basic services (electricity, water, sanitation) yield higher productivity levels and economic returns to endowments. According to this view, individuals with the same endowments may still achieve different living standards as influenced by the geographic characteristics of their place of residence, and therefore policies that focus on reducing the gap in endowment levels are not sufficient. In reality, however, spatial differences are a result of both concentration and geography drivers. While these views advocate different interventions, they are not incompatible in that prioritizing one does not preclude the other. 49. Agglomeration economies that arise from external economies of scale can lead migration to widen rather than narrow spatial income differentials. Unlike the human capital theory of migration where migration equalizes returns to labor, the New Economic Geography literature posits that as labor 37 migrates in response to wage differentials, the size of the market grows in the destination region and through a variety of mechanisms related to scale economies, the real wage in the destination region increases rather than decreases (Kanbur and Rapoport, 2005). Well-developed infrastructure, a high degree of market specialization, competition, information exchange and more efficient matching in the labor market create an environment that is conducive to lower costs and higher returns and can help households escape poverty (Skoufias and Lopez-Azevedo, 2009). Agglomeration effects can be substantive; the extent of which differs across industries (Ellison and Glaeser, 1997). 50. Empirical evidence across countries and subnational regions suggests that both endowments and returns to endowments play a role in driving spatial disparities. A number of studies have explored how differences in endowments vis-à-vis differences in returns explain variances in welfare in countries such as Brazil, Mexico, Ecuador, Peru (Skoufias and Lopez-Azevedo, 2014), Indonesia (Skoufias and Olivieri 2013a), Tanzania (Zeufack and Hassine 2015), Ghana (Agyire-Tettey et al. 2017), Cambodia, Laos, Vietnam, Thailand (Lozano-Gracia et al 2016; Nguyen et al. 2007; Thu Le and Booth 2014), and China (Sicular et al. 2005). A common element across most refers to both the endowment and returns to education and the role of occupation as it reflects differences in economic activities. Data, methodology, and regional groupings 51. Standard decomposition methods allow us to understand household-level drivers of regional income inequality. 9 This methodology allows welfare disparities to be decomposed into endowment effects and return effects—which reveal whether disparities in welfare are attributable to differences in portable characteristics of the population (i.e. human capital) or to differences in returns to those characteristics that are determined by location-specific characteristics (i.e. physical infrastructure). This analysis uses real per capita income as the measure for standard of living, and explores education level, occupation, demographics and household composition as the set of “covariates” which summarize the portable characteristics (see Box 8 for a more in-depth description of the methodology).10 Data for the analysis comes from the 2014 round of the Survey of Income and Program Participation and includes households in 79 regions.11 9 The traditional decomposition method is Oaxaca-Blinder (1973). An important limitation of the Oaxaca decomposition, however, is that it is not performed on groups of individuals with the same characteristics: for example, there may be skilled agricultural workers with less than secondary education living in region A but none in region B. The Nopo decomposition (Nopo 2004) is a nonparametric matching method to decompose outcome differences into components that are analogous to the Oaxaca decomposition. The Nopo decomposition, however, accounts for differences in support (comparable characteristics) which can help correct for the Oaxaca’s potential over-estimation of differences due to rewards for characteristics. Alternative results for this analysis using the Nopo decomposition method, however, were not very different in a qualitative way. 10 The analysis is repeated with consumption expenditure as measure of living standard using 10 years of HBS data to confirm that the results do not vary much from year to year. Indeed the analysis reveals that the patterns observed from the Oaxaca decomposition are relatively stable across time. Detailed results are shown in the annex. 11 Prices are deflated using a constructed indicator that combines the regional CPI (to account for temporal price differences) and the value of a fixed basket of consumer products and services for each region (to account for spatial price differences). 38 Box 8: Using the Oaxaca Blinder decomposition to understand sources of regional inequality in Russia The Oaxaca-Blinder decomposition estimates the relative contribution of endowments and returns to endowments to differences in welfare. The determinants of welfare can be broadly classified into a set of covariates that summarize the portable characteristics of the household and a set of structural parameters that estimate the marginal effects (“returns”) to these endowments. In this analysis, the set of covariates is represented by household head’s demographic characteristics (age, age squared and sex), household head’s education, occupation and household composition variables (number of babies aged 0-2, number of teens aged 3 to 15, number of adults aged 16 to 64, number of elderly, the square of each, and the number of working adults in the household). For two regions A and B, the log of the welfare measure can be estimated as a linear function of the above portable characteristics. Specifically, for each region the following regression is estimated separately: = + , = + where ε is a random term with the usual properties. The mean difference in living standards between regions A and B is expressed as: − ̅̅̅̅̅̅̅ ̅̅̅̅̅̅̅ − ̅̅ =̅̅̅ ̅̅ and ̅̅̅̅̅̅̅ where ̅̅̅̅̅̅̅ indicate sample means of covariates in region A and B, respectively, and and are the associated structural parameters to be estimated. After adding and subtracting the term ̅̅̅ , and following Jann (2008) the above difference can be written as: ̅̅̅̅̅̅̅ − ̅̅̅̅̅̅̅ =(̅̅̅ − ̅̅ ̅̅ ̅̅̅ ̅̅̅̅ )( + (1 − ) ) + ( − )((1 − ) + ) where 1 is an identity matrix and D is a matrix of weights. While the original Oaxaca decomposition uses either D=0 or D=1, the analysis in this paper uses matrix D with diagonal elements equal to 0.5 so that the average coefficients and average characteristics are used. It should be noted that since these decompositions are performed at the mean of the distribution, they may or may not hold at other parts of the distribution. Source: Adapted from Skoufias and Olivieri (2013b) 52. While there is an advanced Russian scholarship on regional classification, this analysis chooses to rely on a simpler approach and classifies regions based on the share of extractive activities in GDP,12 supplemented by a special category for metropolitan areas. GDP data for the last 10 years is used to average out the impact of the commodity cycle. Among these regions, Tyumen oblast has the highest average share of mining, comprising approximately 54% of economic activity. Regions are classified into 5 different regional groupings based on their economic structure and urban/rural delineation in order to analyze differences between regions and within regions. The groupings include Metro, Resource (Urban/Rural), and Rest (Urban/Rural). Regions in “Metro” grouping include Moscow city and St Petersburg city and is therefore 100 percent urban. Regions in “Resource” grouping include 12 regions where mining accounts for at least 20 percent of the regional GDP: these are Arkhangelsk oblast, Chukotka autonomous okrug, Kemerovo oblast, Komi republic, Magadan oblast, Orenburg oblast, Sakha (Yakutia) republic, Sakhalin oblast, Tatarstan republic, Tomsk oblast, Tyumen oblast, and Udmurtia republic. 12A commonly used synthetic classification of regions developed by Grigoryev and Urozhaeva (2011) for example, relies on four broader categories, and nine subcategories. This approach classifies regions into: 1. Highly developed regions (1a. Financial and economic centers, capitals; 1b raw materials- and export-oriented regions), 2. Developed regions (2a. regions with diversified economy; 2b. Industrial, manufacturing regions; 2c. Extractive regions relying on natural resources), 3. Average development (3a. Industrial and agrarian; 3b. Agrarian industry), 4. Less developed regions (4a. Less developed raw-materials- focused regions, 4b. Less developed agrarian regions) 39 Regions in “Rest” grouping include all other regions. Table 6 provides a more detailed description of the regional groupings and their characteristics.13 Table 6: Regional groupings Metro Resource Rest Urban Urban | Rural Urban | Rural Share of GDP 26% 22% 52% Mostly services, Mostly services, Economic structure manufacturing, Mining 20%+ manufacturing, <1% mining <1% mining Pop share (%) 12% 15% 73% Urban 100% 76% 71% Gini 0.420 0.394 0.377 Theil (GE(1)) 0.348 0.307 0.269 Number of regions 2 12 65 Understanding welfare gaps across and within regions: Endowments and returns to endowments 53. There is a large gap in living standards between regions, as measured by income per capita. Metro areas have the highest standard of living, followed by Urban Resource, Urban Rest, Rural Resource, and Rural Rest (Figure 17, Panel a). In particular, rural regions lag significantly behind, attaining less than half the level of welfare achieved in Moscow and St. Petersburg. To understand these differences, this analysis estimates how much of the spatial differences in income is accounted for by differences in observed household characteristics and differences in returns (or welfare gains) to those characteristics. 13While there are advantages to the simplicity of this approach, there are also key limitations. In particular, by relying only on indicators of primary economic activity the approach does not consider its compatibility with different mixes of secondary activities – such as agriculture or manufacturing. For example, resource rich regions in which there is heavy manufacturing or production of military items have seen a high government demand for these goods in recent years, despite the key role played by extractive activities. 40 Figure 17: Welfare and endowment gaps by region a. Income per capita (rubles) b. Share of household heads with c. Share of household heads in higher education highly-skilled occupation 700000 80% 80% 600000 60% 60% 500000 400000 40% 40% 300000 200000 20% 20% 100000 0 0% 0% Metro Resource Urban Metro Resource Urban Metro Resource Urban Resource Rural Rest Urban Resource Rural Rest Urban Resource Rural Rest Urban Rest Rural Rest Rural Rest Rural Source: Staff estimation using Income Survey 2014. 54. The results suggest that roughly half of the difference in living standards between Metro and other urban regions is accounted for by differences in household characteristics (endowments, such as education levels) and the other half by differences in the returns to those characteristics(Figure 18, panel a). This result is underpinned by the fact that the Metro region has significantly better endowments than all other regions. For example, over 60% of the household heads in the Metro region have higher education and work in highly-skilled occupations, compared to under 40 percent in all other regions (Figure 17, Panels b and c).14 In comparison, the disparities in welfare between either urban or rural non-Metro regions, while also much smaller in levels, is driven almost entirely by differences in returns (Figure 18, panels a and b). Only about 20 percent of the welfare difference between urban Resource regions and urban Rest regions is accounted for by endowment differences. The corresponding estimate for the difference between rural regions is even smaller at about 5%. This can be explained by the fact that households in non- metro urban areas and households in non-metro rural areas tend to be very similar in terms of characteristics (Figure 17, Panels b and c). 14Highly-skilled occupations refer to occupation levels 1-3 out the 10 occupation levels defined in the survey. Occupation levels 1-3 correspond to managers and specialists with high and medium level of qualifications. 41 Figure 18: Inequality across regions Panel a. Oaxaca decomposition across urban regions: Metro vs Urban Resource, Metro vs Urban Rest, Urban Resource vs Urban Rest 0.5 100% 0.45 90% 0.4 80% 0.35 70% 0.3 60% 0.25 50% 0.2 40% 30% 0.15 20% 0.1 10% 0.05 0% 0 Metro/Resource Metro/Rest Resource/Rest Metro/Resource Metro/Rest Resource/Rest Characteristics (%) Returns (%) Mean diff Characteristics Returns Panel b. Oaxaca decomposition across rural regions: Rural Resource vs Rural Rest 0.16 100% 0.14 90% 80% 0.12 70% 0.1 60% 0.08 50% 0.06 40% 30% 0.04 20% 0.02 10% 0 0% Resource/Rest Resource/Rest Mean diff Characteristics Returns Characteristics (%) Returns (%) Source: Staff estimation using Income Survey 2014. Note: The left hand figures show the size of the welfare gap and how it is attributed to differences in endowments or returns. The right hand figures show the same results in terms of shares. 55. Urban-rural differences in living standards within the same region are large and attributed almost equally to endowments and the returns to those endowments (Figure 19). This urban-rural gap is comparable in magnitude to the large gap between the Metro and other urban areas. Large urban-rural differences in productive characteristics contribute to the wide gap in living standards. The share of household heads with higher education or working in highly-skilled occupations drops by approximately 10 to 20 percent when comparing urban and rural areas in both Resource and Rest regions (Figure 17, Panels b and c). While returns are overall greater in urban areas and therefore play a role in further widening the welfare gap, somewhat surprisingly, the returns to occupation are comparably smaller in urban areas holding all else constant, effectively offsetting the welfare advantage attained through better endowments. 42 Figure 19: Inequality within regions: Oaxaca decomposition of urban-rural areas within region 0.45 100% 0.4 90% 0.35 80% 70% 0.3 60% 0.25 50% 0.2 40% 0.15 30% 0.1 20% 0.05 10% 0 0% Urban vs Rural Urban vs Rural (Rest) Urban vs Rural Urban vs Rural (Rest) (Resource) (Resource) Mean diff Characteristics Returns Characteristics (%) Returns (%) Source: Staff estimation using Income Survey 2014. Note: The left hand figures show the size of the welfare gap and how it is attributed to differences in endowments or returns. The right hand figures show the same results in terms of shares 56. These results partly reflect the different sectoral distribution of economic activity across and within regions. In fact, wage differentials across sectors can be significant with employment in mining or finance on average paying more than twice as much as jobs in sectors such as manufacturing or those that are dominated by low-end services such as construction, education, trade, hotels and restaurants. The employment share in non-market service sectors such as education and health services is notable. Given that wages constitute a large share of household income, wage differentials across sectors holding education and occupation constant, account for some of the gap in household income. Table 7 shows the sectoral distribution of similarly educated household heads working in similar highly-skilled occupations: in Moscow and St. Petersburg, approximately one quarter are engaged in finance and real estate, followed by almost 20% in trade, hotels and restaurants, and transport and communications. On the other hand, industry is the dominant sector in other urban areas and education—one of the lowest paying sectors—is the biggest employer of highly educated individuals in rural areas (30-40%). To make an illustrative point: on average, an investment banker in Moscow would get paid significantly more than an accountant working for a telecoms company in Krasnoyarsk, who in turn would make significantly more than a lecturer in Kalmyk republic—even if all three had attained the same level of formal tertiary education. However, a separate analysis (results not shown) suggests that the differences in welfare gaps due to returns persist even after formally accounting for sector of activity in the decomposition. 43 Figure 20: Average wages by sector, 2015 Financial activities Mining Fishing, fish farming Public administration Real estate services Transport and communications Production and distribution of utilities Manufacturing Other community, social and personal services Construction Health and provision social services Wholesale and retail trade Education Hotels and restaurants Agriculture, hunting and forestry 0 10000 20000 30000 40000 50000 60000 70000 80000 Source: Rosstat. Table 7: Sectors of activity for household heads in highly-skilled occupations (manager or specialist) by education level a. Household heads with higher education Moscow/St. Urban Urban Rural Rural Petersburg Resource Rest Resource Rest Agriculture, forestry, fishing 0.002 0.008 0.011 0.049 0.063 Industry 0.138 0.287 0.221 0.178 0.114 Construction 0.102 0.067 0.085 0.015 0.050 Trade, hotels, restaurants, transport, communication 0.196 0.148 0.151 0.054 0.089 Finance and real estate 0.252 0.117 0.156 0.053 0.074 Public administration 0.063 0.124 0.135 0.125 0.178 Education 0.096 0.154 0.130 0.411 0.311 Health and social services 0.096 0.068 0.085 0.066 0.099 Other services 0.053 0.028 0.026 0.050 0.024 Other 0.003 0.000 0.002 0.000 0.000 b. Household heads with secondary professional education Moscow/St. Urban Urban Rural Rural Petersburg Resource Rest Resource Rest Agriculture, forestry, fishing 0.002 0.023 0.007 0.047 0.117 Industry 0.163 0.272 0.233 0.050 0.137 Construction 0.120 0.029 0.083 0.014 0.043 Trade, hotels, restaurants, transport, communication 0.254 0.229 0.182 0.286 0.101 Finance and real estate 0.186 0.100 0.092 0.011 0.063 Public administration 0.049 0.087 0.098 0.054 0.104 Education 0.049 0.093 0.084 0.125 0.212 Health and social services 0.136 0.152 0.189 0.210 0.198 Other services 0.042 0.013 0.031 0.203 0.026 Other 0.000 0.002 0.001 0.001 0.000 Source: Staff estimation using Income Survey 2014. 44 57. Among the set of covariates, the age of the household head determines a large part of the differences in welfare gaps. The underlying reason for that may differ across regions, however. In the Metro region households with an older household head have higher income levels, all else equal, whereas the opposite appears to be the case in other regions. The former is intuitive: being older is generally associated with a higher labor income (due to a rising age-earnings profile) and more asset ownership. The reason for the latter is not quite clear, although it should be noted that the estimated coefficient that indicates that relationship is sometimes insignificant or on the smaller side. The urban-rural gap within regions is largely explained by the household composition, likely reflecting the younger demographics and higher number of working adults in urban areas, and household head’s education. In decompositions that compare urban areas, the household head’s education plays a relatively minor role and occupation appears to drive the difference between the Resource regions and the Rest of the country (Table 8).15 Moreover, the advantage of greater educational attainment for households in the Metro region is slightly offset by lower “returns,” the reason for which is not immediately clear. Table 8: Contribution of Covariates to welfare gap within and across regions Across: Across: Within: Urban areas Rural areas Resource and Rest Metro vs Metro vs Urban Rural Urban vs Urban vs Urban Urban Resource vs Resource vs Rural Rural Resource Rest Urban Rest Rural Rest (Resource) (Rest) Head's demographics 0.2131 0.3856 0.1725 0.1840 -0.0491 -0.0376 Head's education 0.0400 0.0483 0.0082 0.0865 0.0217 0.1000 Head's occupation -0.1442 -0.0574 0.0868 0.1102 -0.1037 -0.0802 HH composition -0.2233 -0.2284 -0.0051 -0.1561 0.1716 0.0206 Constant 0.4644 0.3485 -0.1159 -0.0671 0.3421 0.3909 Mean diff 0.3500 0.4965 0.1465 0.1576 0.3827 0.3938 Source: Staff estimation using Income Survey 2014 A closer look at metropolitan areas 58. As predicted by the literature on agglomeration, city size matters: different-sized cities tend to have different economic structures and therefore different welfare and inequality dynamics. Urbanization happens invariably as part of economic development which accompanies the structural transformation from a predominantly agrarian economy to an economy dominated by the production of non-agricultural goods and services. Increased city size brings greater industrial diversification (Marshall, 1975) and consequently, living standards tend to be positively correlated with city size – this also holds for Russia, although cities in the resource-rich regions seem to exhibit a slightly different pattern (Figure 21). In particular, Moscow and St. Petersburg represent the richest and largest urban centers within Russia which is distinctive from other large urban areas in the country. These two cities attract the most internal migrants, 15 One exception is the decomposition of spatial differences between rural regions, but the welfare difference is relatively small. 45 have the highest population density, highest per capita income, very low poverty rates and a young, active and highly skilled workforce (Table 9). Figure 21: Mean income by region and city size (rubles) 1600000 1400000 1200000 1000000 800000 600000 400000 200000 0 Less than 50k 1 mill+ 100-249.9k 250-499.9k 500-999.9k < 200 5000+ 201 - 1000 1001-5000 50-99.9k Urban Rural Metro Resource-rich Rest Source: Staff estimation using Income Survey 2014 Table 9: Demographic and Socioeconomic Characteristics of Households in Metro vs Other Large Urban Areas Metro Other large urban areas Age of head (years) 43.5 47.5 Sex of head (=1 if male) 0.645 0.597 Household size (# people) 2.546 2.516 Education (Household head, %) University and higher 0.608 0.443 Secondary professional 0.221 0.257 Primary professional 0.164 0.249 Less than secondary 0.006 0.051 Sector of activity (Household head, %) Not working 0.099 0.232 Agriculture, forestry, fishing 0.001 0.002 Industry 0.125 0.183 Construction 0.098 0.093 Trade, hotels and restaurants 0.263 0.207 Finance and real estate 0.179 0.101 Public administration 0.053 0.052 Education 0.056 0.046 Health and social services 0.068 0.058 Other services 0.056 0.025 Other 0.001 0.000 Occupation (Household head, %) Managers 0.120 0.104 Professionals 0.337 0.198 Professionals with medium qual. 0.152 0.121 Clerical support workers 0.025 0.019 Service workers 0.063 0.059 46 Skilled agricultural workers 0.068 0.110 Plant and machine operators and assemblers 0.061 0.069 Non-skilled workers 0.029 0.055 Source: Staff estimation using Income Survey 2014 59. The decomposition can also be used in order to better understand welfare disparities across large and small urban areas. This distinction based on size further allows for comparison of welfare disparities between: (i) Metro and other large urban areas and (ii) large vs small urban areas within each region type. Using available information on the size of Russian cities and towns from the 2014 Survey on Income and Program Participation, large urban areas are defined as areas with one million or more residents, and small urban areas are defined as those with less than one million residents. Urban areas from the original regional groupings (Urban Resource and Urban Rest) were each further disaggregated into large urban areas and small urban areas. Only a small share of the population lives in major urban centers with more than one million residents (Figure 22). It is notable that especially in the resource-rich regions there is a large concentration of cities in the 500,000-999,900 range and only very few large urban cities.16 Figure 22: Distribution of cities and towns by region (%) 30 25 20 15 10 5 0 < 200 100-249.9k 250-499.9k 500-999.9k 5000+ 1 mill+ 201 - 1000 1001-5000 50-99.9k Less than 50k Urban Rural Resource Rest Source: Staff estimation using Income Survey 2014 Note: Moscow and St. Petersburg have a population of greater than 1 million (not shown here). 60. Differences in living standards across urban areas of different sizes are accounted for by a relatively large role for endowments (Figure 23). This is especially the case for the comparison of Metro 16 It is well known that Russian cities do not follow the “rank-size” rule (Zipf’s law), a robust empirical regularity found between the log of city size and log of rank across various countries. According to this rule, Moscow, St. Petersburg and other large Russian cities would have populations larger than their current sizes. These cities would need to grow further to realize the benefits of agglomeration (World Bank, 2011). The relative lack of major urban agglomerations is largely a legacy of administrative limitations imposed on the growth of major cities during Soviet times. The distribution of cities and towns across regions shown in Figure 22 reiterates this point. 47 vs other large urban areas and large vs small urban areas in the rest of the country. Here again it appears that the demographic characteristics of the household head are driving the advantage in terms of living standards. Within resource-rich regions, a similar pattern as before emerges in that returns to endowments appear to play a more important role, although it should be noted that the difference between other urban areas is considerably smaller and the contribution of various covariates is more mixed (results not shown). Figure 23. Oaxaca decomposition: Metro vs Other large urban, Large vs small urban areas 0.4 100% 90% 80% 0.3 70% 60% 0.2 50% 40% 30% 0.1 20% 10% 0.0 0% Metro vs Other Large vs Small Large vs Small Metro vs Other Large vs Small Large vs Small Large Urban Urban Urban (Rest) Large Urban Urban Urban (Rest) (Resource) (Resource) Mean diff Characteristics Returns Characteristics (%) Returns (%) Source: Staff estimation using Income Survey 2014. Note: The left hand figures show the size of the welfare gap and how it is attributed to differences in endowments or returns. The right hand figures show the same results in terms of shares 61. The results point to the importance of endowments as policy entry points in addressing spatial disparities. Differences in endowments play a bigger role especially when welfare disparities between areas are greater. This implication is also consistent with the policy framework of the WDR 2009, which identifies spatial equity in health and education as the bedrock of policy packages for successful territorial integration. Investments in education and health care not only improve the productivity of people but they also help people in case they decide to seek opportunities in other places. 6. Unbalanced growth and inclusive development in Russia: Lessons from the spatial analysis 62. Against the background of the legacy of a dispersed economic structure and the largest world land masse, Russia since transition has been following policies “more equalizing than most.” Arguably over the last few years the model worked – we find evidence of declining regional disparities and of conditional convergence between regions. Transfers appear to be effectively targeting poorer regions. Yet poorer regions have been bypassed by the big investment flows, that have gone to oil and gas producing regions and big metropolitan areas, thereby reinforcing the dependence of poorer regions from federal transfers. 63. This model might not be sustainable any longer and different regions are differently placed to deal with these policy developments. The widening fiscal deficit might result in large cuts in transfers, 48 from which two thirds of the regions depend on for at least 15 percent of their budgets. The nature of the transfer system should ensure that transfer will continue flowing to the poorer regions. Heavily populated metropolitan areas and resource-rich mining regions, despite having been most adversely affected by the recent crisis as indicated by evidence of decreasing inequalities due to downward convergence, will be able to rely on own resources. It can therefore be expected that regions which are not among the poorest but do not have natural resources are likely to be most affected. 64. Given these prospects, how can policies better support regional development in Russia? Addressing spatial disparities does not necessary imply “balancing” growth across a geographic territory – but rather focusing on creating opportunities for all people, regardless of where they live. From a policy perspective this may imply investing in spatially targeted interventions with concentrated economic activity in few places, and spatially connective infrastructure to connect more those living in more distant regions with these areas. Critically underpinning both of these strategies are investments in spatially blind institutions, which ensure equal provision of basic services – such as education, health, or utilities - to all, regardless of their geographic location (World Bank 2009). 65. Our analysis suggests two major priorities. First, leveling the playing field in terms of opportunities. Our micro level decompositions show that differences in endowments continue to play a major role in explaining differences between regions: in comparisons which involve large welfare gaps (such as metropolitan areas versus others) they account for almost half of that gap. Addressing disparities in terms of access to services, and in particular to health and education, needs therefore to remain at the heart of policies seeking to improve both efficiency and equity. 66. In this regard, regional governments may seek to prioritize different policies as determined by their conditions in terms of endowments, access to markets and fiscal constraints. For example, rich and poor regions face very different fiscal constraints--with the latter in particular having limited fiscal space which would manifest especially during economic downturns. Rich regions, in particular the large metropolitan cities of Moscow and St. Petersburg and the resource-rich regions, largely drive poverty and inequality and have more resources to spend on innovative approaches that could help address the challenges of more inclusive growth (see Box 9 for an example of an innovative approach). For poor regions, investing in assets (such as education) that matter the most for people should be a priority and an effort should be made to preserve social spending even during times when fiscal space is limited. An in-depth assessment into why their endowments are still lagging and identify effective policy actions could be useful in this context. Combined with investments in large, spatially connective infrastructure, human capital investments would help people in poor regions gain greater access to markets and opportunities from agglomeration economies. Here policies that support the movement of people would be instrumental, especially in areas where there are limited economic opportunities. Policy pillars to enhance labor mobility include 1) establishing institutions to facilitate the transaction of real estate that would otherwise tie people to locations, 2) removing administrative obstacles (related to registration of permanent residence) and 3) investing in human capital by ensuring uniform access to services regardless of location of residence. The lack of connective infrastructure has also been pointed out as a constraining factor for migration (World Bank, 2011). 49 Box 9. New York City’s Conditional Cash Transfer Program As an example of an experimental approach to poverty, New York City’s conditional cash transfer program (CCT) was the first comprehensive CCT program in a developed country. Launched in 2007 by New York City’s Center for Economic Opportunity, it offered cash assistance to low-income families to reduce immediate hardship, but conditioned that assistance on fa milies’ efforts to build up their “human capital” to reduce the risk of longer-term and second-generation poverty. The program thus tied a broad array of cash rewards (financial incentives) to prespecified activities and outcomes in the areas of children’s education, families’ preventive health care, and parents’ employment. Six community-based organizations, in partnership with a lead nonprofit agency, ran Family Rewards in six of New York City’s highest-poverty communities. Family Rewards transferred over $8,700, on average, to families during the three-year period in which it operated. By the end of the study, it had produced some positive effects on some outcomes, but left many other outcomes unchanged. For example, the program 1) reduced current poverty and material hardship, including hunger and some housing-related hardships (especially for families in severe poverty), although those effects weakened after the cash transfers ended, 2) substantially increased graduation rates and other school outcomes for ninth-graders who entered high school as proficient readers, and increased their likelihood of subsequently enrolling full time in four-year colleges, 3) substantially increased families’ receipt of preventive dental care but 4) did not increase parents’ employment in or earnings from jobs covered by the unemployment insurance system (thus impeding sustained reductions in poverty), and led to some small earnings reductions for certain more disadvantaged subgroups. Source: http://www.mdrc.org/publication/new-york-city-s-first-conditional-cash-transfer-program (accessed May 26, 2017) 67. Strengthening local institutions would also help to address the wide within-region inequalities that drive disparities at the national level. Here institutions refer to spatially blind investments to ensure that basic services are uniformly provided across territory. Our analysis suggests that the impact of differences in endowments in driving inequalities between rural and urban areas within regions is comparable to the one it has between metropolitan areas and non-resource rich regions. Ensuring effectiveness in delivering services at the local level despite the challenges of remoteness and depopulation that many of these regions face requires again good management and innovative solutions. 68. Second, fostering greater agglomeration effects. Our macro analysis shows how persistent spatial effects affect both growth and poverty reduction. With regions apparently clustered in “successful” and “unsuccessful” spots, investing in connective infrastructure which can break down these barriers is clearly a priority. Connective infrastructure facilitates labor mobility, which would improve the returns to household endowments with greater access to markets, jobs and opportunities. The aim should be however to create an enabling environment for private resources to flow where they can be most productive, rather than trying to channel and disperse those to new “agglomerations”. These are also the economic centers where people and economic activities continue to concentrate and where the benefits from “thick” markets are likely to be the greatest. Agglomeration should be driven by market forces, instead of dispersing investment towards widespread agglomerations. While concentration can also bring with it negative externalities such as congestion and crime--the key for agglomeration economies lies in the role of policies to efficiently manage and control them. Similarly, addressing implicit barriers to mobility might help support this agenda, for example by addressing constraints on the housing market which might curtail mobility. 50 References Agyire-Tettey, F., C. Godfred Ackah, and D. Asuman. 2017. “An Unconditional Quantile Regression Based Decomposition of Spatial Welfare Inequalities in Ghana”. The Journal of Development Studies Vol. 0, Iss. 0,0. Arellano, Manuel; Bond, Stephen. 1991. "Some tests of specification for panel data: Monte Carlo evidence and an application to employment equations". Review of Economic Studies. 58 (2): 277. Barro R.J., Sala-i-Martin X. 1991. Convergence across States and Regions. Brookings Papers on Economic Activity, 1991(1), pp. 107–182. Blinder, AS. 1973. “Wage Discrimination: Reduced Form and Structural Estimates.” Journal of Human Resources, 8(4), 436–455. Blundell, R., and S. Bond. 1998 “Initial conditions and moment restrictions in dynamic panel data models.” Journal of Econometrics, 87, 115-143. Ceriani, Lidia and Luis F. Lopez-Calva. 2016. “State Discontinuity.” Background Note, WDR 2017, World Bank. Washington, DC. Commander, S., Tolstopiatenko, A. and Yemtsov, R. 1999. “Channels of redistribution: Inequality and poverty in the Russian transition.” Economics of Transition, 7: 411–447. doi:10.1111/1468- 0351.00019 Credit Sussie. 2016. Global Wealth Report 2016. November, 2016. Credit Sussie AGE Research Institute. Zurich, Switzerland. EBRD. 2016. Transition for all: Equal opportunities in an unequal world. Transition Report 2016-17. Ellison, G. and E.L. Glaeser. 1997. “Geographic Concentration in U.S. Manufacturing Industries: A Dartboard Approach,” Journal of Political Economy 105(5): 889–927 Ericson, Ricahrd E. 2013. “Command Economy and its Legacy.” The Oxford Handbook of the Russian Economy. Eds. Michael Alexeev and Shlomo Weber. Oxford University Press. Esteban, J. and D. Ray. 2006. “Inequality, Lobbying, and Resource Allocation.” American Economic Review 96 (1): 257–79. Federov, Leonid. 2002. “Regional Inequality and Regional Polarization in Russia, 1990–99.” World Development Vol. 30, No. 3, pp. 443–456, 2002. Gluschenko, Konstantin. 2010. “Methodologies of analyzing interregional income inequality and application to Russia” MPRA Working paper No. 66824. 51 Gorina, E. 2016. “General description of the fiscal position and social spending in Russia during 2008- 2015.” Mimeo, World Bank. Grigoryev L., Urozhaeva V., Ivanov D. 2011. “Synthetic classification of regions: Basis for Reg ional Policy.” In Grigoryev, Zubarevich, Khasayev. “Russian Regions: Economic Crisis and Problems of Modernization”. Pp. 35-56. Guriev, Sergei and Elena Vakulenko. 2012. “Convergence between Russian Regions.” Working Paper No 180. Centre for Economic and Financial Research at New Economic School. Jann, B. 2008. “A Stata implementation of the Oaxaca-Blinder decomposition for linear regression models” The Stata Journal 8(4): 453-479 Kaliuzhnova, N. 2011. “Institutes of Regional Development and Competitiveness in the Conditions of Modernization.” Экономика Региона № 2/2011 Kanbur, R. and A. J. Venables. 2005. “Spatial Inequality and Development.” In Spatial Inequality and Development, edited by Ravi Kanbur and Anthony J. Venables, 3–11. Oxford, U.K.: Oxford University Press. Kanbur, R., and H. Rapoport. 2005. “Migration selectivity and the evolution of spatial inequality.” Journal of Economic Geography, 5(1), 43–57. Kanbur, Ravi. 2006. “The policy significance of inequality decompositions.” Journal of Economic Inequality 4: 367–374. Kholodilin, Konstantin, Aleksey Oshchepkov, and Boriss Siliverstovs. 2009. “The Russian regional convergence process: Where does it go?” Paper presented at the Higher School of Economics (Moscow) and at the 10th Bi-annual conference of the European Association of Comparative Economic Studies. Kolenikov, S., and A.F. Shorrocks. 2005. “A decomposition analysis of regional poverty in Russia.” Review of Development Economics, 9: 25-46. Krugman, P. 1991. “Increasing Returns and Economic Geography.” Journal of Political Economy, 99, 483–99. Kukenova M., Monteiro J.-A. 2008. Spatial Dynamic Panel Model and System GMM: A Monte Carlo Investigation, MPRA Paper 13404, University Library of Munich, Germany. Limonov, Leonid. Undated. “Regional Research and Regional Policy in Russia.” . Accessed April 2017. LeSage J. P., Pace R. K. (2009). Introduction to spatial econometrics. Taylor & Francis / CRC Press Boca Raton, FL. 52 Lozano-Gracia, Nancy, Daniel Walker and Sarah E. Antos (2016) “Spatial Disparities in EAP”, mimeo, the World Bank Marshall, John U. 1975. City size, economic diversity, and functional type: The Canadian case, Economic Geography, Vol. 51. No. 1. Matytsin M., D. Popova, and E. Sinnott, 2017. Distributional impact of fiscal policy in Russia over the boom and downturn, mimeo. Nguyen, B. T., Albrecht, J. W., Vroman, S. B., and Westbrook, M. D., 2007. “A Quantile Regression Decomposition of Urban-Rural Inequality in Vietnam.” Journal of Development Economics 83(2), 466-490. Oaxaca, R., 1973. “Male–female wage differentials in urban labor markets.” International Economic Review 14 (3), 693–709. OECD, 2010. “Regional Development Policies in OECD Countries”, OECD Publishing, Paris. Ravallion, Martin and Quentin Wodon 1999. “Poor Areas, or Only Poor People?” Journal of Regional Science, VOL. 39, NO. 4, 1999, pp. 689–711. Sargan, J. D. 1975. "Testing for misspecification after estimating using instrumental variables". Mimeo. London School of Economics. Shorrocks, Anthony and Guanghua Wan. 2004. “Spatial Decomposition of Inequality.” Discussion Paper No. 2004/01. UNU/Wider. Sicular, T., Y. Ximing, B. Gustafsson, and L. Shi. 2005. “The Urban-Rural Gap and Income Inequality in China”. Paper prepared for UNU-WIDER Project Meeting “Inequality and Poverty in China,” 26- 27 August 2005, Helsinki, Finland. Skoufias, E. and G. Lopez-Acevedo. 2009. “Determinants of regional welfare disparities within Latin American Countries.” Volumes 1 & 2. Washington, DC: World Bank. Skoufias, E. and R. Katayama. 2008. "Sources of Welfare Disparities across and within Regions of Brazil: Evidence from the 2002-03 Household Budget Survey," Policy Research Working Paper Series 4803, The World Bank. Skoufias, E. and S. Olivieri. 2013a. "Sources of spatial welfare disparities in Indonesia: Household endowments or returns?" Journal of Asian Economics, Elsevier, vol. 29(C), pages 62-79. Skoufias, E. and S. Olivieri. 2013b. “Spatial Disparities in East Asia.” Washington, DC: World Bank. Thu Le, H. and Booth, A. L. 2014. “Inequality in Vietnamese Urban–Rural Living Standards, 1993–2006.” Review of Income and Wealth, 60: 862–886 53 UNDP. 2011. National Human Development Report for the Russian Federation 2011: Modernization and Human Development. Ed. Alexander Auzan and Sergey N. Bobylev Vakulenko E. S 2016. “Does migration lead to regional convergence in Russia? International Journal of Economic Policy in Emerging Economies.” 2016, 9(1), 1-25. Venables, A. 2003. “Spatial disparities in developing countries: cities, regions, and international trade.” Journal of Economic Geography 5(1): 3-21. World Bank. 2009. World Development Report 2009 Reshaping Economic Geography. Washington, DC: World Bank. World Bank. 2011. “Russia: Reshaping Economic Geography.” Washington, DC: World Bank. World Bank. 2014. Back to work: Growing with jobs in Europe and Central Asia. Washington, DC: World Bank. World Bank. 2015. “From demographic dividend to demographic burden? Regional trnds of population aging in Russia” Washington, D.C. World Bank. 2016a. “Pathways to Inclusive Growth: The Russian Federation Systematic Country Diagnostic.” Washington, DC: World Bank. World Bank. 2016b. Russia: Subnational governments' fiscal response to the economic downturn. Washington, D.C.: World Bank Group. World Bank. 2017. World Development Report 2017 Governance and the Law. Washington, DC: World Bank. Yemtsov, Ruslan. 2003. “Quo Vadis? Inequality and Poverty Dynamics across Russian Regions .” Discussion Paper No. 2003/6. UNU/Wider. Zeufack, A.G and N.B. Hassine. 2015. Inequality of Outcomes and Inequality of Opportunity in Tanzania. Policy Research Working Paper No. 7260. World Bank, Washington, DC. Zubarevich Natalia. 2009. “Regional Development and Regional Policy in Russia during Ten Years of Economic Growth.” Journal of the New Economic Association, (1-2), 161-174. Zubarevich N. V. 2012. «Crafty figure» on the map of the Motherland. ECO, 4, 74–85 (in Russian). 54 Annex A. Inequality Trends Year Theil Within Between %Within %Between Gini p90/p10 p75/p25 2005 0.280 0.258 0.022 0.92 0.08 0.395 5.87 2.55 2006 0.274 0.255 0.019 0.93 0.07 0.391 5.77 2.54 2007 0.316 0.295 0.022 0.93 0.07 0.411 6.01 2.56 2008 0.319 0.302 0.017 0.95 0.05 0.407 5.71 2.48 2009 0.258 0.238 0.020 0.92 0.08 0.379 5.34 2.40 2010 0.270 0.247 0.023 0.91 0.09 0.382 5.30 2.40 2011 0.265 0.248 0.017 0.94 0.06 0.380 5.23 2.34 2012 0.295 0.274 0.021 0.93 0.07 0.393 5.31 2.34 2013 0.291 0.272 0.019 0.93 0.07 0.393 5.30 2.32 2014 0.299 0.280 0.018 0.94 0.06 0.392 5.15 2.29 2015 0.245 0.231 0.014 0.94 0.06 0.364 4.79 2.25 55 B. Convergence Results (Real income) (1) (2) (3) (4) (5) (6) (7) Effects ML (4) GMM + VARIABLES FE GMM Spatial ML Direct Indirect Total lag Time lag 0.816*** 0.618*** 0.665*** (0.051) (0.035) (0.027) Spatial lag (rho) 0.358*** 0.224*** (0.061) (0.039) Net internal migration rate 0.018 0.006 0.023** 0.013* 0.014 0.004 0.018 (t-1) (0.023) (0.010) (0.011) (0.008) (0.009) (0.003) (0.011) - Unemployment rate (t-1) -0.003 0 -0.004** -0.004** -0.001** -0.005** 0.010*** (0.003) (0.002) (0.002) (0.001) (0.002) 0.000 (0.002) Share of transfers in region - - - - - - -0.001 budget income 0.002*** 0.003*** 0.002*** 0.002*** 0.000*** 0.002*** (0.001) (0.001) (0.001) 0.000 0.000 0.000 0.000 Real investments per capita 0.092*** 0.044** 0.034* 0.039*** 0.042*** 0.012*** 0.054*** (log) (0.024) (0.019) (0.019) (0.012) (0.011) (0.004) -0.014 - - - - - Share of mining in GDP -0.001 -0.001** 0.004*** 0.003*** 0.002*** 0.002*** 0.002*** (0.001) (0.001) (0.001) (0.001) (0.001) 0.000 (0.001) Share of manufacture in - - - - - -0.004 -0.002** GDP 0.012*** 0.012*** 0.007*** 0.007*** 0.009*** (0.004) (0.003) (0.004) (0.002) (0.002) (0.001) (0.003) Share of construction in - - - - -0.006** -0.002 -0.003* GDP 0.003*** 0.004*** 0.001*** 0.005*** (0.002) (0.002) (0.002) (0.001) (0.001) 0.000 (0.001) Share of young(0-15) -0.026** -0.012 -0.008** 0.004 0.004 0.001 0.005 (0.012) (0.009) (0.004) (0.004) (0.004) (0.001) (0.005) - - Share of pensioners (65+) 0 -0.010** -0.010** -0.003* -0.013** 0.055*** 0.016*** (0.014) (0.006) (0.003) (0.005) (0.004) (0.002) (0.006) Population growth -0.005** 0 -0.001 -0.002 -0.002 0 -0.002 (0.002) (0.001) (0.002) (0.002) (0.001) 0.000 (0.002) Observations 936 858 858 858 858 858 858 R-squared 0.899 0.944 0.944 0.944 0.944 Number of code 78 78 78 78 78 78 78 Number of instruments - 58 69 - - - - AR(2), p-value 0.64 0.41 AR(3), p-value 0.21 0.07 Sargan, p-value 0.16 0.2 Notes: Robust standard errors in parentheses and cluster standard errors in (4) *** p<0.01, ** p<0.05, * p<0.1 Dependent variable – log of per capita real income (with respect to modified fixed basket). Spatial lag – weighted real income per capita in other regions by inverse distance between regions by railway. The methods of estimation are OLS with FE, GMM (Blundell-Bond system gmm) and ML. Endogenous variables: time and spatial lags. Exogenous instruments: time dummies (year dummies 2005-2014). Time period: 2004-2014. 56 C. Convergence Results (Poverty) (1) (2) (3) (4) (5) (6) (7) Effects ML (4) GMM + VARIABLES FE GMM ML Direct Indirect Total Spatial lag Time lag 0.764*** 0.640*** 0.641*** (0.039) (0.045) (0.027) Spatial lag (rho) 0.486*** 0.323*** (0.047) (0.043) Net internal migration rate (t-1) 0.399 0.433 -0.171 -0.169 -0.150 -0.074 -0.224 (0.670) (1.961) (0.313) (0.246) (0.274) (0.134) (0.406) Unemployment rate (t-1) 0.13 -0.001 0.056 -0.003 0 -0.001 -0.001 (0.126) (0.085) (0.055) (0.063) (0.068) (0.033) (0.100) Share of transfers in region budget 0.064** 0.002 0.085*** 0.022** 0.022** 0.010** 0.032** income (0.030) (0.018) (0.019) (0.010) (0.010) (0.005) (0.014) Real investments per capita (log) -1.973*** -0.346 1.644*** -0.181 -0.099 -0.05 -0.149 (0.733) (0.565) (0.515) (0.431) (0.402) (0.186) -0.584 Share of mining in GDP 0.055 -0.008 0.001 0.001 0.006 0.003 0.01 (0.046) (0.022) (0.030) (0.021) (0.022) (0.011) (0.032) Share of manufacture in GDP 0.501*** 0.13 0.117 0.272*** 0.290*** 0.137** 0.427*** (0.168) (0.085) (0.108) (0.101) (0.101) (0.060) (0.155) Share of construction in GDP 0.178** -0.02 -0.185*** -0.002 -0.002 -0.001 -0.003 (0.083) (0.054) (0.062) (0.036) (0.040) (0.018) (0.058) Share of young(0-15) 0.036 0.471** 0.592*** 0.072 0.086 0.035 0.121 (0.583) (0.213) (0.088) (0.249) (0.254) (0.118) (0.371) Share of pensioners (65+) 1.011* 0.001 0.420*** 0.413** 0.394** 0.193* 0.586** (0.534) (0.180) (0.095) (0.178) (0.172) (0.110) (0.278) Population growth -0.006 0.009 0.097* 0.056 0.060* 0.028 0.088 (0.056) (0.045) (0.057) (0.038) (0.036) (0.018) (0.053) Observations 936 858 858 858 858 858 858 R-squared 0.714 0.79 0.79 0.79 0.79 Number of code 78 78 78 78 78 78 78 Number of instruments - 58 73 - - - - AR(2), p-value 0.1 0.02 AR(3), p-value 0.64 0.03 Sargan, p-value 0.12 0.12 Notes: Robust standard errors in parentheses and cluster standard errors in (4) *** p<0.01, ** p<0.05, * p<0.1 Dependent variable – poverty rate. Spatial lag – weighted poverty rate in other regions by inverse distance between regions by railway. The methods of estimation are OLS with FE, GMM (Blundell-Bond system gmm) and ML. Endogenous variables: time and spatial lags. Exogenous instruments: time dummies (year dummies 2005-2014). Time period: 2004-2014. 57 D. Oaxaca (Means of Covariates) URBAN RURAL National Variable Metro Resource Rest Resource Rest Rural Urban HH head is male 0.58 0.54 0.53 0.51 0.52 0.56 0.60 HH head's age 45.49 49.03 51.97 52.78 53.92 51.68 48.15 HH head's age squared 2282.45 2662.81 2986.44 3049.51 3167.37 2922.10 2571.82 HH head's education Tertiary 0.51 0.27 0.29 0.17 0.15 Secondary professional 0.27 0.30 0.29 0.25 0.25 0.26 0.27 Primary professional 0.21 0.36 0.33 0.42 0.41 0.40 0.27 Less than secondary education 0.01 0.07 0.10 0.16 0.19 0.15 0.06 HH head's occupation Not working 0.18 0.32 0.39 0.41 0.47 0.05 0.10 Managers 0.08 0.07 0.06 0.05 0.04 0.10 0.19 Professionals 0.29 0.12 0.13 0.09 0.08 0.08 0.11 Professionals with medium level qualification 0.16 0.09 0.09 0.08 0.07 0.01 0.02 Clerical support workers 0.03 0.02 0.01 0.02 0.01 0.05 0.06 Service workers 0.08 0.07 0.06 0.06 0.05 0.11 0.11 Skilled agricultural workers 0.07 0.13 0.11 0.08 0.09 0.12 0.09 Plant and machine operators and assemblers 0.06 0.12 0.09 0.13 0.10 0.08 0.06 # babies (age0-2) 0.04 0.08 0.07 0.09 0.08 0.10 0.08 # babies sq 0.04 0.09 0.08 0.10 0.09 0.11 0.09 # teens (age3-15) 0.33 0.36 0.31 0.39 0.38 0.49 0.38 # teens sq 0.46 0.54 0.47 0.65 0.70 0.91 0.59 # adults (age 16-64) 1.72 1.61 1.52 1.65 1.61 1.82 1.76 # adults sq 3.88 3.52 3.37 3.92 3.83 4.63 4.19 # elderly (age 65+) 0.25 0.27 0.38 0.32 0.38 0.32 0.28 # elderly sq 0.33 0.37 0.52 0.45 0.50 0.43 0.39 Number of working adults 1.43 1.20 1.11 1.07 0.99 1.21 1.41 Source: Staff estimation using Income Survey 2014. 58 E. Oaxaca Results: Household budget survey (2005-2014) Inequality across regions 0.35 120% Urban: Metro vs Resource 0.3 100% 0.25 80% 0.2 60% 0.15 0.1 40% 0.05 20% 0 0% -0.05 -20% Characteristics Returns Mean diff Characteristics (%) Returns (%) a. 0.4 100% Urban: Metro vs Rest 0.3 80% 60% 0.2 40% 0.1 20% 0 0% b. Characteristics Returns Mean diff Characteristics (%) Returns (%) 0.15 120% Urban: Resource vs Rest 100% 0.1 80% 60% 0.05 40% 20% 0 0% -20% -0.05 Characteristics (%) Returns (%) c. Characteristics Returns Mean diff 0.14 160% Rural: Resource vs Rest 0.12 140% 0.1 120% 0.08 100% 80% 0.06 60% 0.04 40% 0.02 20% 0 0% -0.02 -20% -0.04 -40% Characteristics Returns Mean diff -60% Characteristics (%) Returns (%) d. 59 Oaxaca Results: Household budget survey (2005-2014) (Continued) Inequality within regions 0.5 Resource: Rural vs Urban 100% 0.4 80% 0.3 60% 0.2 40% 0.1 20% 0 0% Characteristics Returns Mean diff Characteristics (%) Returns (%) a. 0.4 100% 0.35 Rest: Rural vs Urban 0.3 80% 0.25 60% 0.2 0.15 40% 0.1 20% 0.05 0 0% b. Characteristics Returns Mean diff Characteristics (%) Returns (%) Source: Staff estimation using HBS 2005-2014. Note: The yellow trend line shows the difference between the two regions being compared over the period 2005-2014. 60