Decentralization measures revisited.

Author:Martinez-Vazquez, Jorge

    For the last half of a century, public policy analysts and economists have been trying to measure decentralization to use as either the dependent or an independent variable in their models. From a theoretical perspective, the suggested measures essentially boil down to a few concepts: locally raised revenues, locally decided expenditures, locally spent national grants, or the number and relative size of local units. However, these basic concepts enter actual decentralization measures in various functional forms: additive, multiplicative, exponential, or various combinations thereof.

    The first goal of this paper is to develop a taxonomy of decentralization measures showing their relationship to the primary theoretical aspects of decentralization and how the different measures are related to each other.

    The second goal of the paper is to argue convincingly that from an empirical analysis perspective the aggregation of those distinct dimensions of decentralization into a single indicator inevitably leads to a loss of information and that a preferable approach in multivariate frameworks is to enter these distinct aspects of decentralization in the regression analysis separately in the most flexible functional form possible.

    In terms of decentralization measures, besides the shares of different government levels in public revenues and expenditures, which are commonly used in the literature, we also take another look at the measure proposed by Breton and Scott (1978): the territorial fragmentation of different government tiers weighted by their relative roles in the public finance. This measure has been all but forgotten in the literature but we argue that, conceptually, it has considerable relevance for assessing the extent of decentralized decision-making.

    While no single indicator can capture all aspects of decentralization, some indicators are more inclusive and informative than others. Therefore, if for some reason we have to use a single indicator, our study suggests which indicator would be more appropriate for the hypothesis at hand.

    The rest of the paper is organized as follows. In section 2 we identify various dimensions of decentralization. In Section 3 we review approaches to measuring different dimensions of decentralization. In Section 4 we propose a single indicator capturing multiple dimensions of decentralization. In Section 5 we present empirical evidence on the relative explanatory powers of various indicators of fiscal decentralization and their combinations. Section 6 concludes.


    Decentralization is commonly defined as the process of transferring decision-making powers to sub-national tiers of government. Three forms of this process have been distinguished in the policy literature: deconcentration, delegation and devolution (Bird, 1993). Through deconcentration the central government gives some autonomy to its territorial branches that are appointed by, and are accountable to, the higher hierarchy. Under delegation, locally elected government bodies assume new responsibilities subject to strict regulations (mostly concerning service outputs) by the upper-level government. The process of devolution establishes complete autonomy of locally elected government bodies in their exclusive spheres of responsibility. The definition of decentralization suggests that it is generally a multifaceted process. In particular, at least three different dimensions jointly constitute this concept: the scope of authority, the degree of autonomy, and the direction of accountability.

    In practice, we can observe varying extents of decentralization for each of the distinct dimensions of transfer of government power. Therefore there is a multidimensional universe of decentralized systems representing different combinations of power-sharing chosen for each of the dimensions. Using this paradigm, a particular decentralization process would represent a transition from one point in this universe to another, such that local powers for at least one of the dimensions increases (but leaving open the possibly that there are decreases for other dimensions that in combination we weight less). Thus, for example, a central government might offload responsibility for the provision of certain public goods so that it can refocus and centralize some aspects of regulation concerning the same public goods (MacLeod and Goodwin, 1999).

    We illustrate this point in Figure 1, by focusing on three dimensions of decentralization of expenditure competencies commonly considered by public finance economists: regulation, financing and delivery of public services (Philip, 1954). The three graphs show how different forms of decentralization increase powers of local government along those three dimensions. Thus deconcentration increases local government powers only on one dimension--delivery of services--while reserving to the central government the powers to regulate and raise finances (bottom graph). Deconcentrated services can be further decentralized by relaxing central government regulation, in particular by allowing local communities to make (some or all of) those decisions via elected local officials; what is known as "political decentralization." Under "delegation," local governments are given responsibilities to deliver and (partially) finance public services while regulation of these services remains centralized, as depicted in the middle graph. Finally, through "devolution" local governments receive powers along all the three dimensions (regulation, financing, and delivery) with respect to services that fall in their exclusive spheres of responsibility.


    Devolution can also be achieved in two steps by first undertaking delegation and then following with "administrative decentralization," or giving local governments more discretion with respect to the different modes of delivery of the formerly delegated functions.

    While this three-dimensional picture is helpful for visualizing the multi-dimensionality of decentralization, it still misses important information about the entire universe of decentralized systems, which in fact has higher dimensionality. There have been several attempts in the decentralization literature to try to capture and quantify this multi-dimensionality. For example, very recently Blume and Voigt (2008) attempted to reduce the dimensionality of their dataset through factor analysis of 25 indicators of decentralization and federalism, including political, fiscal, and administrative aspects (with the number of countries varying from 33 to 136 depending on the variable). Their factor analysis suggests that the information captured by these indicators cannot be condensed to a dataset of less than seven dimensions. Three of these seven principal components capture fiscal aspects: local role in spending, reliance on own-source revenues, and unconditional sharing of national revenue. The four non-fiscal principal components capture elected local executives, elected local councils, veto power of the house of regional representatives, and political fractionalization. However, the achieved reduction of dimensionality does not imply that conceptually all aspects of decentralization can be defined in terms of those seven principal components. Rather, the results mean that, in the currently observed practice, various manifestations of decentralization would appear to be driven by the seven identified driving forces, which Blume and Voigt (2008) call 'latent variables'

    The main conclusion from the discussion above is that, even in the fiscal realm alone, a single decentralization indicator cannot capture the entirety of powers assigned to the subnational level; simply put, the different aspects of public finances (regulation, financing, administration, and delivery) cannot be captured with the same indicator.


    The right approach to measuring the extent of fiscal decentralization has been long debated in theoretical works and, more recently, in empirical studies focusing on the link between fiscal decentralization and various fiscal and economic outcomes. (1) The economics literature has focused heavily on the number and size of local jurisdictions and their role in raising and spending public finances. The common caveat is that regulation, while being the most common form of government intervention, cannot be measured by any indicator constructed from fiscal data. Setting aside the regulation aspect, we can use fiscal data to separately approximate other aspects of fiscal decentralization (Levin 1991). Thus, responsibility for financing public services can be measured as total expenditures of a given government net of received grants. (2) The power of administration can be measured as total expenditures of a given government net of grants provided to other governments. Finally, the direct responsibility for service delivery can be measured as total expenditures of a government net of grants and contracts awarded to private contractors.

    Even in the studies narrowly focused on fiscal aspects of decentralization, there has been diversity of the specific approaches to measuring decentralization. A recent meta-analysis of 26 studies in this field finds that about one third of them measure decentralization as the share of local governments in total government expenditures, while another quarter measure the share of local spending financed from own resources (Feld et al., 2008). Other, less frequently used, measures include the share of total government revenues raised by local governments and the gap between revenues and expenditures of local governments. In what follows we attempt to develop a taxonomy of measures, first, for those dealing with fiscal aspects only, and second for measures focusing on the vertical and territorial structure of sub-national governments.


To continue reading