WPS4774 Policy ReseaRch WoRking PaPeR 4774 Centralization, Decentralization, and Conflict in the Middle East and North Africa Mehmet Serkan Tosun Serdar Yilmaz The World Bank Sustainable Development Network Social Development Department November 2008 Policy ReseaRch WoRking PaPeR 4774 Abstract This paper examines broadly the intergovernmental discuss the current intergovernmental structure by structure in the Middle East and North Africa region, examining the type and degree of decentralization in which has one of the most centralized government five countries representative of the region: Egypt, Iran, structures in the world. The authors address the reasons West Bank/Gaza, Tunisia, and Yemen. Cross-country behind this centralized structure by looking first at regression analysis using panel data for a broader set of the history behind the tax systems of the region. They countries leads to better understanding of the factors review the Ottoman taxation system, which has been behind heavy centralization in the region. The findings predominantly influential as a model, and discuss its show that external conflicts constitute a major roadblock impact on current government structure. They also to decentralization in the region. This paper--a product of the Social Development Department, Sustainable Development Network--is part of a larger effort in the department to study local governance systems and decentralization in the client countries. Policy Research Working Papers are also posted on the Web at http://econ.worldbank.org. The author may be contacted at syilmaz@ worldbank.org. The Policy Research Working Paper Series disseminates the findings of work in progress to encourage the exchange of ideas about development issues. An objective of the series is to get the findings out quickly, even if the presentations are less than fully polished. The papers carry the names of the authors and should be cited accordingly. The findings, interpretations, and conclusions expressed in this paper are entirely those of the authors. They do not necessarily represent the views of the International Bank for Reconstruction and Development/World Bank and its affiliated organizations, or those of the Executive Directors of the World Bank or the governments they represent. Produced by the Research Support Team Centralization, Decentralization, and Conflict in the Middle East and North Africa Mehmet Serkan Tosun* University of Nevada, Reno and Serdar Yilmaz** World Bank JEL classification: H77, H87, N45, O53 Key words: Fiscal decentralization, intergovernmental relations, Middle East and North Africa The authors thank Lyn Squire, Blanca Moreno-Dodson Jennifer Keller and Robert Beschel for helpful comments and suggestions, Dilek Uz and Omid Harraf for excellent research assistance. The findings, interpretations, and conclusions are entirely those of authors, and do not represent the views of the World Bank, its executive directors, or the countries they represent. Corresponding author: Serdar Yilmaz. *Department of Economics, College of Business Administration, University of Nevada-Reno, Reno, Mail Stop 0030. Nevada 89557; tosun@unr.edu; phone: +1-775-7846678. **Social Development Department, World Bank, 1818 H Street, NW, Washington D.C. 20433; Syilmaz@worldbank.org; phone: +1-202-473-9350. Centralization, Democracy, and Conflict in the Middle East and North Africa 1. Introduction There have been significant decentralization efforts in developing countries in recent decades. These efforts led to an extensive literature on the causes and consequences of both centralization and decentralization. In this paper, we examine decentralization and intergovernmental relations in the Middle East and North Africa (MENA), a region that has lagged behind other developing regions in decentralization. The MENA region has unique socio-economic and political characteristics. First, there is economic dependency on oil revenues particularly in Gulf Cooperation Council (GCC) countries. Second, the MENA region has one of the highest population growth rates in the world. This demographic explosion has led to young population structures throughout the region. Third, the region has been impacted by both internal and external conflicts that have affected economic performance of the countries. Despite these good reasons to study government structure in the MENA, there is surprisingly sparse literature on the region. While recent studies followed a comparative perspective and showed similarities and differences between the decentralization efforts in a variety of developing countries, the MENA countries are left out of those comparisons (Bardhan and Mookherjee, 2006). This is mainly due to lack of appropriate data and information on decentralization in the countries of the region. Our goal in this paper is to combine descriptive and empirical analyses to provide a comprehensive picture of decentralization and the reasons behind its slow progress. We also discuss some reform options. 2 The paper is structured as follows. In the next section we look at the history of decentralization and centralization in the region by examining the taxation system of the Ottoman Empire. We provide a comparative description of the intergovernmental structure in five MENA countries, Iran, Egypt, Tunisia, West Bank/Gaza and Yemen in Section 3. In Section 4 we explore the determinants of centralization and decentralization in the region in a regression analysis. We discuss the intergovernmental reform options in the final section. 2. Centralization in the MENA Countries: A Historical Perspective MENA countries have relatively more centralized government structures compared to other developing countries. It is argued that many MENA countries have based their government organization on the administration structure of the Ottoman Empire. 2.1. Ottoman Taxation System Overall, the Ottoman taxation was a centralized system of taxation. Lewis (1979) argues that Ottoman taxation included elements from Islamic taxation as well as from taxation law and practices of Roman, Byzantine, Mongolian, Turkish and pre-Islamic Persian civilizations. Cosgel (2005) examines the evolution of the tax system in the Ottoman Empire as a clash between competition and rigidity in institutions. For example he argues that while the Ottomans changed the tax system in the conquered lands to collect revenue in the most efficient way, they were faced with local institutional constraints. Hence, this explained the substantial regional variation in taxes throughout the empire. 3 Lewis also asserts that Ottoman tax administration changed from a relatively decentralized system with strong center to a weaker but more centralized system. This change started around the end of the sixteenth century. There had been a consolidation of tax collection after the sixteenth century, particularly during the eighteenth and nineteenth centuries.1 While many argue that this coincides with the stagnation and decline in the Empire, Barkey (2008) argues that this "adaptation was a sign of flexibility and pragmatism, not a sign of decline."2 In places such as Egypt more government centralization took place in accordance with heavier European influence (Hanna, 1995). In this centralization trend, we see a change from the Timar system (strong center with decentralized military expenditures) to Iltizam or Tax-farming system (weaker center with more centralization of military and more decentralization of other government services at the province level) towards the end of the sixteenth century. Timar is a land- tenure system where the land was allocated to Sipahis (feudal cavalry) in return for military service in Ottoman provinces (Barkey, 1994: 36).3 In the Iltizam system the state receives an initial monetary sum from private interests in return for rights to collect taxes (Barkey, 2008: 229). Some of the reasons for the system change from Timar to Iltizam were: Sipahi cavalry becoming less important due to introduction of new war technologies, higher demand for full-time regular troops, changing patterns of trade, influx of Spanish silver from America and subsequent inflation leading to a sharp 1See Barkey (2008, Part 2) and Inalcik (1977) for excellent accounts of transformation in the Ottoman administration towards centralization in the eighteenth and nineteenth centuries. 2Barkey, Empire of Difference: The Ottomans in Comparative Perspective, 194. 3Inalcik (1994: 114-118) gives a detailed description of the Ottoman Timar system. As Inalcik notes, an interesting characteristic of the Ottoman Timar system was "the lack of inheritance rights on land and the frequent dismissal from timars" (Inalcik, 1994: 115). With the Timar system, Ottoman state was able to maintain provincial armies without direct centralized expenditures. In addition, the lack of inheritance rights prevented timar holders from forming land-based hereditary nobility that could become a threat to the central authority. 4 decrease in real income (Lewis 1979, Barkey 1994). Barkey (1994) argues that Ottoman state was able to overcome (global) economic distress without any significant rebellions throughout the empire. She explains that this was mainly due to success of the state in aligning officials at different levels of government with the objectives and success of the central government. Similarly, Hess (1977) describes what can be called Ottoman central administration's success in keeping the North African provinces of the empire in the midst of decentralization in the eighteenth century. There have been few sporadic efforts to change the centralized government structure in the Ottoman Empire. For example, Saliba (1978) explains in detail efforts and achievements of Midhat Pasha to decentralize political and fiscal powers when he served as Governor of the Syrian province. Such efforts faced strong opposition from the central government in Istanbul and were mainly approached with suspicion as the Empire was experiencing significant loss of territories. Reform efforts during the Tanzimat or Reorganization period in the nineteenth century aimed at creating a more centralized government structure (Ortayli, 2006). This was thought to be a way to save the Empire from total collapse. McLure (2001) argues that tax assignment over different levels of government depends on history, and has been subject to change through economic evolution. Hence, he asserts there is no one-size-fits-all formula for tax assignment. Cosgel and Miceli (2005) examine the tax assignment in the Ottoman Empire in the mid-sixteenth century. They provide evidence that the tax assignment was done according to a transaction cost hypothesis rather than risk hypothesis. They classify taxes into categories of fixed and variable taxes. Fixed taxes category includes personal taxes based on marital and 5 economic status, input taxes based on production inputs such as land, trees and animals, and enterprise taxes. These taxes are thought to be less risky due to absence of factors that would cause variation and they also have lower transaction costs. Variable taxes category includes trade taxes from market exchange of goods in towns and output taxes such as taxes on grains, legumes and fibers. These taxes are riskier due to high variability and also carry greater transaction costs for the same reason. Using data from Ottoman tax registers (tahrir defterleri) the authors conduct an ordered probit analysis for the influences on tax assignments and find that the proportion of variable taxes are negatively associated with the higher levels of government lending support to the transaction cost hypothesis which asserts that variable taxes can be more efficiently collected at the local level due to lower transaction costs. The key underlying assumption here is that transaction costs are higher for higher levels of government. The authors conclude that transaction costs were more important in tax assignment in the Ottoman Empire. However, they do not model fixed taxes specifically. In summary, we see that the taxation system was not as centralized in the early period (until the end of the sixteenth century) of the Ottoman Empire as it was in the later period. Centralization gained momentum particularly during the Tanzimat (Reorganization) period in the nineteenth century when the empire was declining rapidly and consolidation of power at the center was seen as a solution to prevent total collapse. Centralization that started in the sixteenth century is, by no means, unique to the Ottoman Empire. In fact, Europe started its transformation from small, decentralized city-states to large and centralized territorial states at the end of the fifteenth century.4 It was also 4Alesina and Spolaore (2003) note that French invasion of Italy in 1494 marks the beginning of the period of absolutism in Europe. They argue that main reasons behind the centralization trend were the need for 6 argued that European influence in the Middle East and North Africa fueled more government centralization (Hanna, 1995). Aside from few courageous attempts at decentralization, the region inherited a heavily centralized system of taxation and public administration from the Ottoman Empire and other European States. 3. Comparison of Intergovernmental Structure in Selected MENA Countries In this section, we examine the intergovernmental structure in a selected number of MENA countries to demonstrate the centralized features of local government systems in the region. Our selection includes five countries representing both Maghreb and Mashreq regions. These are Egypt, Iran, Tunisia, West Bank/Gaza and Yemen. The central government in these countries is the senior partner in the intergovernmental relationship. The share of subnational expenditures in GDP is very low compared to OECD average of 17 percent5 (see Table 1). In many countries in the MENA region, such as Yemen and Iran, a large share of the expenditures is disbursed through subnational governments that act as agents of central government ministries and departments. In these countries locally elected representatives have little decision-making power over expenditures. 3.1. Deconcentration and Decentralization in MENA Overall, local government systems in the MENA region can be characterized as a form of deconcentration rather than one of devolved local self-government. In general, the public administration system is highly centralized, equipped with an elaborate system of deconcentrated field offices of line agencies. Decisions for the most part, especially institutions and larger domestic markets caused by growing economies in Europe, and rising costs of war and public administration. 5http://siteresources.worldbank.org/INTDSRE/Resources/WBFDISummaryNote17Aug2004.doc 7 service delivery decisions, are made by the central government and the role of subnational authorities is largely confined to carrying these out. In the region, the general trend is to have two distinct types of local government units: deconcentrated and decentralized (municipal) units (See Table 2). The operations of these two types of local governments are subject to totally different rules. In all countries, the deconcentrated units of the central government provide a big chunk of public services, including health and education, under strict guidance of the central government. Whereas, decentralized units (generally municipalities) perform limited number of functions such as street paving and maintenance, construction of local roads, street lighting, garbage collection, library and park services, and issuing permits for constructions. Table 2 shows that our five selection countries differ in their deconcentrated systems. While West Bank/Gaza and Tunisia have only provincial or governorate level governments, Iran, Yemen and Egypt have a variety of sub-provincial level governments. In all five countries, the central government and/or the President appoint the heads of provincial governments. On the other hand, elected officials in all of them, except Egypt, run municipalities. In Iran, representation of people is partial and indirect. First, there is involvement of the Ministry of Interior (MOI) in both urban and rural municipalities. Second, people elect city or village councils that in turn appoint mayors jointly with the MOI. 3.2. Expenditure Assignment Although the degree of centralization differs among MENA countries, they generally have heavily centralized expenditure and revenue systems. In almost all 8 countries in the MENA region the assignment of expenditure responsibilities between central and local governments does not conform to generally accepted principles for setting the right incentives for efficient and equitable delivery of public services. In contrast, intergovernmental expenditure responsibilities have been assigned largely according to static bureaucratic and administrative considerations. In the region, subnational deconcentrated governments have a very limited number of "own" responsibilities. Most of the local expenditure responsibilities can be classified as "delegated" expenditures as opposed to "own" expenditure responsibilities.6 Central government ministries make decisions on most services that are traditionally provided by local governments in other countries (See Table 3). Table 3 shows the expenditure assignments in five MENA countries and also includes the worldwide practice. Expenditure assignments are indicated as central government (C), provincial government (P), and municipal government (M). Assignments are shown for Social Services, Transportation, Utility Services and Other Services, and are also broken down by Macro Policy and Oversight, Financing, and Provision. MENA countries have significantly more centralized expenditure functions compared to the worldwide practice. To compare MENA countries among themselves, we assigned points to central, provincial and municipal government involvement in expenditures. Accordingly, we assigned 1 point to C, 2 points to P and 3 points to M.7 Total expenditure assignment scores and average scores are listed in Table 4. The average 6Delegated responsibilities are those transferred to the deconcentrated branch units of the central government for delivery of services while the actual decisions on budgeting and financing are carried out at the central level. Some of the delegated responsibilities, such as primary education, health, and public security have high local public good characteristics and in many countries are not delivered by locally elected governments. 7In addition we assigned 1.5 points to C, P; 2.5 points to P, M and 2 points to C, P, M. We gave 1 instead of 0 to central government assignments since the worldwide practice shows that some central government involvement is probably desirable to facilitate decentralization. 9 scores for countries across government services and government involvement show that West Bank/Gaza has the most decentralized expenditure system among the group followed, in order, by Egypt, Iran, Tunisia and Yemen. 3.3. Revenue Assignment Table 5 shows revenue assignment characteristics in the region. West Bank/Gaza is the only country with a truly decentralized revenue system. However, municipal revenue autonomy in West Bank and Gaza is not something granted by the laws. Due to the circumstances, municipalities invent their own ways of raising revenues and they often find ways of rationalizing such practices (World Bank, 2006). Other than West Bank/Gaza, all other countries in the region have largely an ad- hoc local revenue system. The only exception might be property tax system in Iran. We see a glimpse of autonomy in land-based taxation in Iran. 3.4. Intergovernmental Transfer Characteristics Table 6 shows that allocation rules for transfers to local governments are also mainly ad-hoc, except in Tunisia and West Bank/Gaza. In both countries, the transfer system is based on a formula. In West Bank and Gaza municipalities have been historically less dependent on central government transfers compared to other countries in the region. Both due to historical reasons and current political and security problems, West Bank and Gaza municipalities rely little on central government transfers in providing services. In fact, this situation has not been affected significantly by the Intifada period, as there is no difference today in the amount of central government 10 transfers as a share of total local revenues, which remains constant at about 5 percent8 (World Bank, 2006). 3.5. Discussion In this section, we presented an overview of the deconcentration and decentralization systems in selected MENA countries and compared these countries in terms of three important aspect of decentralization: expenditure assignment, revenue assignment and intergovernmental transfers. We also compared these countries to the worldwide practice in expenditure assignments. First, we see a good potential for a rich deconcentrated and decentralized government structure in MENA countries. Most of the countries analyzed have a good number of lower tier governments such as governorates, districts and a variety of municipalities (e.g. 1000 urban municipalities and 68,000 rural municipalities in Iran). Second, MENA countries are significantly behind the worldwide practice in decentralization regarding expenditure assignments. We see in Table 4 that this is particularly true in social services, which is one of the most important government expenditures that directly affect the welfare of residents. Third, West Bank/Gaza seems to have the greatest degree of decentralization from the perspective of both expenditure and revenue assignments. Among the countries analyzed, West Bank/Gaza gives the greatest autonomy to its local governments. While this highlights West Bank/Gaza as an interesting case to consider for other countries in the region, one should approach the decentralization efforts cautiously as the observed decentralization seems to spring from special political and security circumstances of that country. Finally, there is need for improvement in intergovernmental relations in MENA countries. Intergovernmental transfer rules are largely ad-hoc. Our comparative analysis leads us to conclude that 8This is the average figure for West Bank and Gaza municipalities in the Intifada period. 11 MENA countries have highly centralized government fiscal structures. This is despite our observation that there is a potentially rich deconcentrated and decentralized government systems. In the next section, we will examine the reasons behind this heavy centralization using different regression analyses. 4. Determinants of Centralization and Decentralization 4.1. Measurement Issues and Review of the Literature Decentralization is difficult to measure. Despite the popularity of decentralization as a research topic in recent literature, there are serious problems with the measurement of decentralization. Ebel and Yilmaz (2003) note that this is mainly due to data imperfections, particularly in the IMF's Government Finance Statistics (GFS). Problems with data availability and quality get worse in the case of the MENA countries. Hence an empirical investigation of decentralization in the MENA region is difficult due to aforementioned data problems. As so many other studies did before us, we started with data from the GFS. We have realized however that data on conventional fiscal decentralization measures such as share of local governments in total government expenditures or revenues are simply nonexistent. That left us with using data on total central government expenditures relative to an economic base such as gross domestic product (GDP). Thus, we first used total budgetary central government expenditures as share of GDP of the country. While this measure captures mainly the size of the central government, it will also be closely related to the degree of delegation of central government powers to lower levels of government. A reduction in the size of the central government could be one avenue for decentralization. Indeed, we get a negative correlation (-0.47) between the share of local government spending in GDP for the 12 countries shown in the first column of Table 1 and the average share of central government expenditures from our panel dataset. In the absence of better government finance data on fiscal decentralization, we use central government expenditures to get a feel for the weight of central government in economic activity. We also use a non-financial measure of decentralization, which is derived from the World Bank's Database of Political Institutions (DPI). This decentralization indicator is created by taking the sum of three other indicator variables from DPI for local elections in municipal governments (Muni), local elections in state/province governments (State) and authority over taxing, spending or legislating in state/provinces (Author).9 Hence, higher numbers in our decentralization indicator mean greater degree of decentralization. There is considerable literature on the effects of fiscal decentralization on economic growth. Determinants of fiscal decentralization received less attention. Two recent papers provide theory and empirical evidence on the determinants of the decentralization process in a wide variety of countries. Panizza (1999) builds on the work by Alesina and Spolaore (1997) on the number and size of nations by introducing different levels of government. Panizza's theoretical model shows negative correlations between fiscal centralization and country size, income per capita, taste differentiation and level of democracy. He finds empirical evidence supporting these theoretical correlations. Panizza also compares his results to two other cross-sectional or panel studies by Oates (1972) and Wallis and Oates (1988) and finds similarities particularly in regards to country size and income per capita. Arzaghi and Henderson (2005) examine the determinants of fiscal decentralization using panel data on 48 countries over the period 9Muni, State and Author are the names used in DPI under the sub-heading "Federalism." See Beck et al. (2001) and Keefer (2007) for detailed descriptions of these variables. Higher numbers indicate greater degree of federalism, and thus decentralization of central government authority. 13 1960 to 1995. They show that the Middle East and North Africa region has the highest share of central government and the lowest federalism index.10 They also show that the Middle East and North Africa lags behind all other regions in the democracy index. Arzaghi and Henderson follow Panizza's approach to model decentralization first and then empirically test hypotheses derived from the theoretical model. Their empirical findings mainly confirm previous findings on the determinants of decentralization (or centralization). They find that past colonial experience and initial constitutional provisions matter in the decentralization process. 4.2. Empirical Analysis We are using an unbalanced panel for fourteen countries with a study period broadly from 1975 to 2004.11 As mentioned above, we are examining centralization and decentralization in MENA in two parts. In the first set of regressions we use total budgetary central government expenditures as share of GDP as our dependent variable, which we see as a rough indicator of centralization. To address the shortcomings of this variable as a measure of centralization (or decentralization), we run a second set of regressions with a decentralization indicator derived from DPI as the dependent variable. Given our key dependent variable measuring degrees of decentralization, which is thus inherently ordered, our second set of regressions are based on the probabilistic ordinal dependent variable regression model of the Logit type P Y = j X = j - X - j - X , ( ) ( ) ( -1 ) 10Their evidence on MENA is based on only few countries in the region due to problems with available data on fiscal decentralization. 11These fourteen countries are Bahrain, Djibouti, Egypt, Iran, Jordan, Kuwait, Lebanon, Libya, Morocco, Oman, Saudi Arabia, Syria, Tunisia and Yemen. We had to reduce our sample to nine countries in the regressions with central government expenditures due to lack of government expenditure data for Djibouti, Lebanon, Libya, Saudi Arabia and Yemen. 14 where P Y = j X ( ) is the probability of observing j {0,.., J} outcome of the dependent variable Y conditional on the vector X of individual country characteristics, is the standard logistic distribution function, and is the vector of regression coefficients to be estimated by the Maximum Likelihood method. Denoting Y* the latent variable driving the observed outcomes of Y and j the J -1 cutoff points such that j < j , the model assumes that Y = 0 if Y* < 0, Y = j if j