W hy C oun tries Are F iscally DecentralizingMoh am mad ArzaghiVernon HendersonDepartm ent of E conom ics, Bro wn Universit y64 Waterman St.Pro vidence, RI 02912December, 2002We thank Enrico Spolaore for early advice and Rajiv Vohra for helpful discussions on the topic. We are alsograteful to James C. Davis for his help throughout the project, Christine Kearney for preparation of thefederalism index, and Fumiko Tamura for her help as a research assistant. Finally we thank two anonymousreferees and an editor for helpful detailed comments.1IntroductionOver the last twen ty-fiv e years, there has been an international movement towards both institutional andfiscal decentralization. Fiscally, the role of regional governments has expanded, in the sense that the shareof state/provincial and local governments in total government consumption has risen. Sometimes the riseis preceded, accompanied, or followed by a change in formal institutional structures, where countries mayadopt a federal constitution or embark on local democratization with elected local or regional governments.In the paper we model the economic forces that would promote formal institutional change towards a greaterdegree of federalism, and test the empirical implications of the model. As suc h, in some of t he empirical workand in the early discussion of world trends we will look at development of formal institutions. However acoun try which in the data has a formal federal structure may still be de facto highly centralized, in that thelocal or regional governments have little fiscal responsibilities; and a country with a formal unitary structuremay be de facto highly decentralized fiscally. Constitutional changes are discrete events which in certaincontexts may be difficult culturally and politically to achieve. Thus we will also examine the actual extent offiscal decentralization, which may be more revealing about underlying forces for decen tralization in certaincontexts.In the model, we will argue that income and population growth and shifts in relative population acrossregions of countries, whic h occur during economic development, explain the mo ve toward greater federalismand greater fiscal decentralization, both in theory and in our empirical results. We start by examining someof these patterns of change, as well as related changes in the extent of democratization and we review theliterature on federalism and fiscal decentralization.1.1 Patterns of Institutional and Fiscal Decen tralizationFor a sample of 48 countries with populations over 10 million in 1990, we construct a federalism index everyfive years from 1960-1995. The index varies from zero to four, with higher values meaning a greater degreeof institutional regional autonomy. The description of the index is in the data Appendix, but it is based1on official governmental structure (unitary versus federal) and the degree of autonomy and democratizationof state/provincial and municipal governments. Later in the paper w e will break the index into some of itscomponen t parts in the empirical work. Figure 1 shows how the index varies worldwide and by regions of theworld. After 1975 or 1980, both worldwide and in all regions the index rises. It rises from a world averageof 1.03 in 1975 to 1.94 by 1995. There are significant differences among the regions. Developed countriestend to be more federalized to begin with. Latin American countries reached the same level of federalism asdeveloped countries in 1995, with former Soviet bloc countries not far behind. Only countries in the MiddleEast and North Africa have little local autonomy, with minor changes over time.Federalism and democratization seem to go hand-in-hand, as expected. First democratization involvesrepresen tative government and a voice for regional interests. Arguably that leads to a greater call for regionalautonomy. In addition one third of our federalism index reflects local and regional democratization, whichmay be correlated with national democratization. Indeed the simple correlation between our federalismindex and a national democracy index is 0.63 for five years intervals over the period of 1960-1995, and thecorrelation between changes in those indices is 0.39 over the same time period. The democratization indexis based on Jaggers and Gar (1995), who used Polity III indicators on autocracy and democracy, and itranges from -10 (least democratic and most autocratic) to +10. Figure 2 shows democratization trends forthe w orld. The pattern is similar to that for federalism.Indices of changes in formal government structures have the limitation that the “de facto” situation maydiffer from the formal situation, with formal structures changing infrequently. For example between 1975and 1985, federalism only changes in 9 of 48 countries; and between 1985 and 1995 it c h anges in 22 countries.One can argue that changes in formal structures towards enhanced local autonomy and democratization arethe key elements, since they represent a structural change with some degree of permanence and guarantees.However more subtle changes in de facto situations are more likely to be reflected in a continuous measure,such as the share of state and local governments in government expenditures. Using IMF data, we constructeda variety of measures representing the degree of centralization. However, we tend to rely on one, which is the2share of the central government in total governmen t current consumption. This measure, as will be discussedbelow, excludes transfer payments.The time coverage for these data is 1975, 1985, and 1995, involving respectively 39, 47 and 48 coun-tries. The average central government share is graphedinFigure3. Consistentwiththechangesinformalstructures, the world average for fiscal centralization declines from 0.74 in 1975 to 0.64 in 1995. An averageweighted by country population drops from 0.62 to 0.52. Again there are significant regional differences.Developed countries and countries in the former So viet bloc and Latin America have the lowest centralizationwhile Middle Eastern and Sub-Saharan African countries have the highest centralization. The share of thecentral go vernment drops in all regions, with an exception of former Soviet bloc countries in the period of1985 to 1995.1The overlap in country co verage between our centralization measure and the federalism measure is limited,a total of 65 country-years over 1975, 1985, and 1995. Overall the correlation in levels for those years