Read Microsoft Word - For MWIEDC.docx text version


Abstract: Using rich panel data on all of Ghana's districts' local public finances over 11 years, this paper investigates the way that intergovernmental and other transfers to local governments affect local governments' incentives to collect internally generated revenues and funds (IGF). We find that despite an incentive scheme built into one of the major intergovernmental grants, the flow of all grants taken together discourages, rather than encourages, own-sourced revenue effort. This is reflected both in the depressing effect of transfers on IGF levels, as well as on IGF growth. Keywords: Decentralization; Local public finance; Internally generated revenues; Intergovernmental transfers; Africa; Ghana.

______________________________________________________ a

Corresponding author. Email: [email protected] International Food Policy Research Institute c Care Quality Commission.


1. INTRODUCTION Since the late 1980s, many developing countries have started the process of devolving political, administrative, and fiscal responsibilities from central to provincial and local jurisdictions. The motivations of countries for undergoing such a governance change are varied and include a range of political as well as social and economic factors. Even where political factors have been the main drivers in implementing decentralization, strong financial backing from the donor community has usually been founded on economic arguments, many of which have been laid down in the canonical literature on decentralization. Subnational governments may have better information (or can obtain information more cheaply) about the local needs for and requirements of public services, and thus a decentralized system would generate greater allocative efficiency in public service provision (Hayek 1945). Improved public service provision would also emerge in a politically and fiscally decentralized system through greater political competition at the local level, leading to more accountable local governments (Crook and Manor

1998), and through greater jurisdictional competition arising from citizens choosing their location on the basis of the quality of services (Tiebout 1956).i These and other arguments for decentralization rest strongly on the assumption that the fiscal aspect of decentralization in fact results in local governments gaining substantial discretion in allocating public resources to competing economic uses in their jurisdictions. This assumption is germane to the realization of the benefits from decentralization mentioned above, since achieving these benefits are predicated on the notion of local governments as capable decisionmakers, able to act upon information and upon the pressures of political and jurisdictional competition. In practice, in many developing countries--as in the case of Ghana, which this paper focuses on--local fiscal discretion is highly restricted in the sense that local authorities may have relatively little control even over their own budgets. In Ghana, a substantial share of their revenues is made up of transfers from upper tiers of government or from donors, and these funds tend to be tied to particular activities, projects, or sectors.ii Revenues that local governments generate themselves, through the tax and fee bases assigned to them, can, in contrast, be used in a completely flexible manner. In this sense, local governments' fiscal autonomy is intimately tied to their ability to generate own resources. Therefore, an important part of the policy debate around decentralization concerns the question of how local governments can expand their fiscal autonomy by increasing their internally generated funds (IGF) and realize the hypothesized equity and efficiency gains in the local provision of public services. In Ghana, there are a range of potential constraints affecting the ability of district assemblies--the term for Ghana's district-level governing bodies--to expand their IGF. These may include the scope of local governments' revenue assignments, revenue collection capacity, discretion in setting rates on their tax and fee bases, and enforcement of honest revenue-collection practices. This paper focuses on a specific potential driver of IGF, by investigating what impact the flow and size of external grants (i.e., central government and donor funds) have had on the incentives of local governments to generate own revenues. While this question has received research attention in the developed country context, there is hardly any literature in developing regions empirically examining the effect of intergovernmental and other external fiscal transfers on local governments' internal revenue effort. With grants comprising the bulk of local governments' total revenues, the incentive effect that grants have on local governments' own revenues is a critical policy concern in Ghana. The recent Decentralization Policy Review, conducted jointly by the government and donors in Ghana to inform the development of a new decentralization policy by the cabinet, states that although the DACF [District Assemblies Common Fund] formulae contains [sic] a small incentive to improve on IGF (very small criteria weight for the so-called `responsiveness' factor), this is not perceived sufficient to promote improvements in the MMDA

[Metropolitan, Municipal, and District Assembly] revenue mobilization. As indicated in a report from 2000, the incentives to collect revenues may be impacted negatively by the increase in grants. Further studies of this and of the real MMDA revenue potential within the existing legal framework is urgently required. (GoG 2007, p. 52) This paper sets out to investigate this very question: How does the flow of intergovernmental grants to local governments affect their incentives to raise their own revenues? In the next section, we present our framework and a review of the empirical literature on how external grants, as well as other factors, may affect local revenues. iii This is followed in Section 3 by an outline of the policy and empirical context for this study, a discussion of district governments' revenue structure and degree of fiscal autonomy, and the criteria used in determining how one major grant facility is allocated across districts, particularly the own revenue incentivization criterion. Section 4 describes the empirical model for determining the effect intergovernmental and external transfers have on districts' incentives for generating own revenues. In so doing, we describe the potential confounding factors which may introduce bias in the estimation of this effect, and outline an estimation and instrumentation strategy to address these factors. The results are discussed in Section 5, and the final section offers concluding thoughts.

2. DETERMINANTS OF LOCAL GOVERNMENT REVENUE GENERATION The interest in the question of how local tax-generation efforts and revenues respond to intergovernmental transfers is variably motivated by policy concerns with regard to how local government fiscal behavior contributes to evolving decentralization, by the question of the efficiency of local tax rates, by that of the benefit incidence of local taxes, or by other factors. iv Nearly all quantitative empirical research on this topic, however, involves developed economies, likely due in large part to the scarcity of data on local public finances of developing countries. This paper may contribute useful empirical findings particularly in light of this lacuna of research in the developing country context. A frequently used conceptual foundation for the pathway through which external grants affect local revenues is the median-voter model. Applied in this context, the hypothesis is, simply put, that grants from upper-tier governments crowd out revenues from local taxes. Assuming an initial optimal balance between local public consumption and private consumption, additional intergovernmental grants would be passed on by local governments to local residents as reductions in local taxes and fees, other factors remaining unchanged (Bradford and Oates 1971). Wildasin (1984) theoretically sketches out in a general equilibrium framework the pathway of impact that different types of intergovernmental transfers (matching grants versus lump-sum grants) have on welfare, via their effect on local taxes, finding that

matching grants are superior to lump-sum grants when transfers can be optimally tailored to each local government. Other studies theoretically examine the efficiency and equity effects of intergovernmental transfers that are designed for particular policy purposes. For example, Bucovetsky and Smart (2006) and Smart (1998), examine the effect of transfers specifically designed to increase equity in revenues across local jurisdictions, and draw different conclusions based on their different assumptions about the mobility of tax bases: when tax bases are immobile, equalization grants can induce excessive local tax rates and thus increase the dead-weight loss from distortionary taxation; this effect is mitigated when tax bases are mobile. The literature on the relevance of taxation for political accountability (recent studies on developing countries include Moore 2008 and Ross 2004) may suggest yet another type of welfare effect of intergovernmental transfers, through their impact on local tax effort. The potential fiscal effort disincentives of higher tier grants may be detrimental in light of the argument that local governments are, all else equal, more likely to spend efficiently resources they have to raise themselves from the populace than resources they don't (Bird and Smart 2002). Several studies have empirically tested the effect that higher-tier grants have on locally raised revenues. Zhuravskaya (2000) establishes a crowding-out effect in a transitional country (Russia), where each monetary unit raised in own revenues by a local government is offset by a reduction by 0.9 units in revenue sources from the higher-tier government, which strongly implies that local governments will have nearly no incentive to exert any tax-generating effort when transfers increase. Similar findings of reductions of locally raised revenue being greater than the grants that induce them emerge in one of the few studies of a developing country context, Rajaraman and Vasishta (2000) in their analysis of data from Kerala, India. Buettner and Wildasin (2006) take an integrated approach in which all interrelationships between various local public finance variables--general expenditure, debt service expenditure, intergovernmental grants, and own revenues--are assessed, with minimal imposition of structure on the empirical model. They find that the adjustment of local governments to an increase in external grants results in reduced subsequent own revenue generation--a finding broadly consistent with the above hypothesis. In a similarly empirically oriented study, Dahlberg et al. (2007), who focused on econometrically addressing the potential endogeneity of grants, does not find a crowding-out effect of intergovernmental transfers on local tax rates, nor on local tax revenues. Only a few studies empirically refute the above hypothesis and identify a positive (crowding-in) effect of higher-tier government aid to local governments on locally generated revenues (an example is Skidmore 1999). Local revenues can also be affected by several factors other than intergovernmental grants. These include other fiscal variables, such as tax limitation measures imposed by higher-tier government (Skidmore 1999; Dye and McGuire 1997), pressures exerted by local expenditures on subsequent local

revenue raising (Buettner and Wildasin 2006; Holtz-Eakin et al. 1989; de Mello 2002), and the impact of federal deductibility of local taxes on the latter (Holtz-Eakin 1992).v Political economy factors can have a potentially important influence on locally raised revenues, through the effects of ideology on fiscal policy, the structure of government, or the pressures exerted by the level of electoral competitiveness (Allers et al. 2001; Borge and Rattsø 1997; Solé-Ollé 2006; Brennan and Buchanan 1980; Boyne 1998). Natural, socioeconomic, and demographic factors in the jurisdiction of the local government are additional potential determinants of locally generated revenues, especially factors that determine the tax revenue base and the capacity of the local government to collect taxes. Although local tax and fee assignments are likely to be the same across different jurisdictions, differences in the above factors will have different effects on the amounts of revenue generated by local governments. For example, jurisdictions with larger deposits of natural resources are more likely to generate greater local revenues through royalties from mining or other natural resource extraction. The same is likely to apply to jurisdictions with greater nonfarm economic activity and more private residences, through more collection of business license fees, which are mostly levied on firms, and of property taxes, which are levied on private residences. As local residents may to some extent have a say in the level and variety of taxes instituted, the social composition of the jurisdiction along ethnic, religious, or other social lines is another potential factor, although the direction of impact on local revenue generation is ambiguous and likely highly context-specific.

3. DATA AND EMPIRICAL CONTEXT The analysis draws mainly on three sources of data. At the core is a relatively long panel of data on Ghanaian local governments' public finances, obtained from the Ministry of Local Government and Rural Development (MLGRD). Districts' revenues are disaggregated by source (revenues from licenses, property taxes, external grants, etc.) and districts' expenditures by economic classification (personnel, travel and transport, capital, etc.). The panel spans 11 years (1994­2004) and covers all of Ghana's 110 districts existent at the time of the The other two main sources of data are Ghana's 2003 Core Welfare Indicators Questionnaire (CWIQ) survey and the 2000 Population and Housing Census, which are used primarily to identify district socioeconomic and demographic variables. The CWIQ has a sample size of over 49,000 households and the data are representative at the district level, enabling aggregation of household information to the district level. The Population and Housing Census data, which are representative even below the district level, also allow aggregation to districts. In addition to data from these three primary sources, this paper also draws on data on: central government aggregate public expenditure accounts from Ghana's Ministry of Finance and Economic Development and from the

Controller and Accountant General's Department; the District Assemblies Common Fund (DACF) from the DACF Administrator's Office; and district-level rainfall from the Meteorological Services Department. In per capita and absolute terms, the magnitude of average local government-generated revenue is modest. During the time period under study, the average district collected 65,000 GHC of its own revenues annually. vii In per capita terms, IGF amounted to 0.29 GHC. Locally generated revenues make up only a small, but not negligible, fraction of district assemblies' (DAs') total sources of funds. On average, they constituted 16 percent of total district government revenues. This share has not been uniform over time, and the direction of change is surprising. There has been a gradual decline in the share of IGF in local budgets over a decade. In the initial period (i.e., 1994), internal revenue sources were nearly equal to external funds, 46 percent of the total local budget. This high rate, compared to the decade average of 16 percent, may be a feature of the fact that the DACF, which is the main vehicle for intergovernmental transfers, commenced in 1994, and although the constitution stipulated that five percent of central government revenues be allocated to the districts in the form of the DACF, it is likely that the financial management infrastructure to disburse the DACF was not yet fully in place in its first year. By 2004 however, the IGF share of the district revenues declined to 10 percent. This is not what one would expect in light of consistent, albeit only partially successful, efforts to deepen decentralization in Ghana, including efforts to strengthen the fiscal position of district governments. Figure 1 outlines the fluctuation of IGF and external funds over time, for all districts on average. Both revenue sources follow an approximately similar pattern, including an accelerated increase of both revenue sources in the later years. However, as the above information on budget shares of the two types of revenue sources suggests, the increase in external funds was faster, leading to the decline in IGF share. <FIGURE 1 HERE> District governments have seven key sources for generating own revenues: property taxes; a head tax levied in a fixed amount on each district resident; royalties from natural resources such as timber and minerals; a range of fees and tolls, such as those charged on market stalls; business licensing; returns from financial investments (e.g., dividends and interest rate earnings on financial capital); and rents collected on DA-owned properties. In contrast, central government has control over all income taxes. As seen in the top panel of Figure 2, local governments' revenue portfolio has generally not changed substantially over time. While it is, in the aggregate, relatively diversified, there are some sources that tend to dominate on average, such as fees (e.g., market stall user fees and parking tolls) and property taxes. Business licenses and permits, and royalties on natural resources, generate a somewhat smaller portion of revenues. Similarly, revenue from financial capital and rents collected on DA-owned properties play a minor role. <FIGURE 2 HERE>

The nationwide relative diversity of local revenue portfolios hides the stronger concentrations visible at the regional levels. The bottom panel of Figure 2 shows the example of three regions in which one revenue source constitutes over one-third of the districts' IGF, with the dominant source differing across these regions. For example, the Western region, which is very rich in timber and mining resources, relies on royalties from natural resources to generate 44% of local revenues (well above the national average of 15%). An important component of the funds which make up local governments' budget and do not emanate from internally generated revenues is the DACF, which makes up slightly more than half of all external sources of revenue. Intergovernmental transfers also include "ceded revenues," which are revenues collected by the central tax agency on behalf of DAs. The transfer of ceded revenues, terminated in 2005, was a more important source of funding in the early years of decentralization, especially before the institution of the DACF. After the introduction of the DACF, ceded revenues played a smaller role in local governments' budgets, and their disbursement was often unreliable in timing and magnitude (GoG 2007). A more recently instituted (and by 2006 discontinued) grant is the Heavily Indebted Poor Countries (HIPC) fund, commenced in 2002 and allocated both to central government ministries as well as to DAs. As Table 1 shows, it constitutes a non-negligible share of DAs' revenue. The HIPC funds to the districts have allocation criteria, earmarks, and reporting formats that are distinct from those of other sources. There is a lack of clarity on the allocation criteria, however, and therefore it is not clear whether there are explicit incentive mechanisms in the HIPC funds for districts to develop own sources of revenue. Donors also have district-specific projects, the funds for which they either channel through the DAs' budgets or maintain off budget. Several districts may not have donor projects for which the funds go through the DAs, but for those that do, donor funds can be a substantial share of revenues. Table 1 captures only such on-budget district donor funds, as data are not available for off-budget external aid to districts. <TABLE 1 HERE> Local fiscal autonomy is typically associated with the size of local governments' budgets. That is, the larger (relative to total public spending) the local governments' expenditures, the more one can speak of local fiscal autonomy. While this is true in Ghana as well, in this country's context the issue of fiscal autonomy is just as intimately tied to the relative size of internally generated revenues in districts' revenue portfolios, as alluded to earlier in the introduction. In addition to the general expenditure assignments of district assemblies, DAs' expenditures out of budget sources originating from central government or donors are further governed by guidelines laid down by the respective sources (Crook 2003; Kokor and Kroës 2000; ISODEC 2005). For example, the DACF administrator's office establishes guidelines on how the DACF funds are to be used. Detailed information about the guidelines obtained for

five years (1999 and 2001­2004) are characterized not only by a high degree of detail in the specification of the areas for which the funds can be used, and the percentage of the funds to be used in each area, but also by uniformity across time. For the five years for which the information was available, the guidelines and the mandated shares for each spending area are identical (GoG 2007, IMF 2006).viii Similarly, the use of resources that district assemblies receive from donor agencies is guided by parameters set by these agencies; most commonly this is a direct consequence of donor funds being disbursed in connection with a particular project (for example, a water and sanitation project, a microcredit project, etc.). The one revenue source over which districts exercise full discretion--subject, naturally, to limitations arising from specific expenditure assignments to the local level--are the internally generated revenues. Therefore, the capacity, legal framework, and incentives that govern own revenue generation also determine the extent of local governments' fiscal autonomy in the prevailing fiscal constellation in Ghana. The effectiveness of the incentives in central government and donor grants --whether these incentives are by design or an unintended consequence of the nature of external budgetary flows to districts--depends not only on the strength of these incentives but also on the extent to which there is any potential to increase internal revenues. Three factors play a role in this potential: first, the extent to which, given the revenue assignments, DAs have discretion in setting rates for taxes and fees; second, the extent to which the revenue-collection effort is constrained by administrative capacity and internal incentives (e.g., compensation design to improve local tax officials' efforts and prevent leakages); and finally, the extent to which extending tax and fee collection either through greater efficiency or higher tax and fee rates encounters the resistance of residents. As to the first, the discretion to set tax rates and fees on DAs' revenue bases is large: With the exception of royalties, for which the central government determines the rates, DAs are permitted to establish the rates and fees on all other revenue sources assigned to them, albeit given ceilings set by central government (primarily through the MLGRD).ix However, even these ceilings have de facto not always been treated as binding: Given that ceilings are for the most part over a decade old, there are several instances in which DAs have in practice exceeded them. While autonomy in rate setting is thus not a binding constraint to the potential impact of external incentives for increasing revenues, district assemblies' ability to tap into certain revenue sources does constitute a limitation, especially given capacity at the level of the revenue collectors. GoG (2007) even suggests that the districts' revenuemobilization capacity is one of the weakest among the various aspects of local financial management. Finally, political resistance to local tax increases is likely to be an issue in Ghana, where local revenue assignments make IGF a relatively visible form of tax, in contrast to income taxes deducted at the source. Such resistance, combined with the fact that the local nature of the spending of these revenues creates

greater transparency about the quality of these expenditures, may however contribute to more judicious use of these public resources, even where it will likely pose a constraint on the size of local revenues (ISODEC 2004). In light of the attention in the fiscal decentralization policy debate that is given to increasing local fiscal autonomy, the framework for the allocation of intergovernmental transfers and donor funds to the DAs gives some consideration to incentives for own revenue generation. The DACF, however, is the only outside revenue source that has a built-in own revenue generation incentive scheme, through the formula for its allocation to districts. Given that information about the DACF allocation criteria will be employed in the subsequent empirical analysis, it is useful to provide an adequate description of the formula. The allocation formula gives weight to four criteria, with the weights adding up to 100% (Table 2a). The first criterion is referred to by the DACF administration as the "equality factor", which refers simply to a base sum distributed in equal absolute amounts to each district. The second is the "needs factor", which provides greater funds to poorer districts. As Table 2a shows, the neediness of districts has been proxied by different indicators over time. In earlier years of the DACF starting from 1994, an estimate of districts' 1992 per capita GDP, along with districts' 1984 population size, was used to determine district need. In later years, availability of different public services was used instead to proxy for district poverty (see Table 2a for details). <TABLE 2a HERE> The third "responsiveness factor" is the dimension which is intended to incentivize own revenue generation. For the first two years of the study period, this factor included only IGF levels as an indicator. Starting 1996 an additional indicator, improvement in IGF, and the level of IGF-indicator was dropped as of 2002. Table 2b provides further details on the indicators used in the responsiveness factor. The level indicator is simply the per capita IGF, and the change indicator is the percentage increase in IGF. The formula only considers increases in IGF, not awarding any amount on the basis of the change-indicator to districts which saw decreases or no change in their IGF). As can be seen in the table, there is usually a lag in the variables used, i.e. the formula relies on two to three years older IGF data, and in the earlier years, even a decade old population data. <TABLE 2b HERE> Finally, the "service pressure factor", seeks to account for the way that, all else equal, greater numbers of residents in a given space require more services. It is simply proxied by the population density of a district. Based on the indicators used and the weights given each indicator, a formula is used to derive the amount each district receives. Both the specific aspects of the underlying indicators, as well as the weights across the criteria, vary from year to year, based on decisions made by the DACF administrator. Table 2 shows how the weights have changed over time.

The formula broadly works as follows: First, the total (that is, for all districts taken together) Common Fund of a given year is apportioned into four parts based on the weights of each of the factors. Then, based on district-specific characteristics, each of the four parts is allocated across districts. The straightforward case is that of the "equality factor". If its weight for a given year is, say, B, with B [0, 1], the total DACF is X, and the number of districts is 110, then each district receives the equal amount of

B · X b / 110, where b (0, 1) is the share of the DACF that is not apportioned for the contingency fund.

The allocation on the basis of the other three factors is somewhat more involved. The appendix provides the details of the formula used for one sample year.

4. EMPIRICAL APPROACH The long panel dataset on local governments' finances in Ghana permits us to assess empirically the extent to which revenues received externally spur or disincentivize local governments' internal revenue generation, with internal revenues being the sole source of fully discretionary expenditures by districts. The first model examines the drivers of the magnitude, or level, of IGF, which is hypothesized to be affected by both public financial variables and core district characteristics: ln = ln + ln + + +

EXTit-1 represents lagged (at time t-1) external transfers to district i. The vector EXPit-1 consists of different types of past expenditures of local governments. The vector xit is made up of basic economic, demographic, social, and climatic characteristics of district i. Specifically, economic features, reflecting economic welfare and development, are the district poverty rate, the literacy rate, and average access to roads by households in the district; demographic district attributes include the degree of urbanization and population size and density; variables defined as the shares of Ghana's two largest ethnic groups (Akan and Ewe) and Christians in the district represent core social characteristics; and climatic conditions are measured by district rainfall. Ri and Dt control for region-specific effects and time effects, respectively. Finally, the composite error term i + it consists of a district unobserved effect and an idiosyncratic disturbance. See Table 3 for a more detailed description of the district characteristics and their summary statistics. By including the district characteristics in the estimation, we are also considering the factors--not the central question in this study, but important to take into account--that may influence the size of district IGF through mechanisms other than the incentive effects of external grants and transfers. These mechanisms were discussed in Section 3. For example, local governments' tax and fee bases define an approximate upper bound to how much revenue they can collect. The tax bases in turn are strongly determined by the income levels of the district population, and by district governments' revenue




assignments. While data are not available on income levels, districts' poverty levels, included in the model, offer a proxy to this. Revenue assignments are naturally the same across districts; but districts with different attributes will have narrower or more expanded tax and fee bases, for the given (universal) revenue assignments. For example, licenses, which are mostly levied on businesses, and property taxes, which are levied on private residences, are predominantly generated in urban areas. Because the extent of urbanization of a district is thus strongly correlated with the relative importance of these revenue sources in the overall IGF portfolio, inclusion of urbanization in the model captures the effect of this dimension of revenue assignments. <TABLE 3 HERE> The vector of public expenditure variables, EXPit-1, consists of an economic decomposition of local government spending into personnel, nonpersonnel recurrent, and capital spending. The inclusion of local expenditures is motivated by the "spend-tax" hypothesis. Applied to local governments, as has been done for example in Dahlberg and Johansson (1998), this hypothesis suggests that a need to increase local expenditures in a given time period may exert pressure to expand revenues in subsequent time periods, especially in the local government context where room to incur budget deficits is very narrow or absent. x In this study, the use of a vector of expenditures reflecting an economic decomposition of local spending has an a priori rationale: while internally generated revenue in Ghana's districts can be spent at the discretion of local governments, it is more often than not used to cover maintenance and other (nonpersonnel) recurrent costs. Since the DACF, an important part of intergovernmental transfers, is supposed to be used primarily for capital investment spending, IGF is frequently not allocated for capital investments, in order to maintain a desirable balance between capital and recurrent expenditures. Some share of salaries is paid through IGF, but as the administrative dimension is perhaps the least complete of the three dimensions of decentralization in Ghana, most civil servants, including those operating at the district level, are still under the management of and compensated through the budget of the central government. This places a check on the extent to which district government revenues, especially IGF, are used for salary expenditures. All this suggests that changes in different types of expenditures may be likely to exert quite different influences on subsequent efforts in local own revenue generation, and that this differentiation in influence should be accommodated in the empirical model. We estimate equation (1)--and variations of it in terms of specification and measurement of the dependent variable--using as the primary method the Hausman-Taylor estimation (Hausman and Taylor 1981). This approach addresses limitations that would arise with both random-effects and fixed-effects estimations in the context of this model. Within-group estimation is not able to identify the time-invariant variables xi, the effects of which, while not central to the core research question of this paper, are nevertheless of interest. Random-effects models on the other hand assume that all covariates are

uncorrelated with the unobserved effect i, which, if false, leads to inconsistent estimates, and such orthogonality is a stronger assumption than we are willing to impose. In particular, one should be concerned that the overall amount of external grants and transfers local governments are able to attract may plausibly be influenced by a range of district (and district leadership) features that we are not able to capture in our model. This is likely to include a district's political history, especially features that contribute to political connectedness with the central government, as well as the ability and capacity of local governments to draw resources from donors and other sources (see for example Banful 2011). We therefore treat the magnitude of external grants and transfers as endogenous in the model in the sense that it may be correlated with district unobserved effects. The Hausman-Taylor estimation approach builds on the within-estimators of the time-varying variables, and uses those (included) time-varying and time-fixed variables that are assumed to be uncorrelated with the unobserved effect to instrument for those variables that are potentially correlated with the district effect. The Hausman-Taylor estimates will be compared with those estimates that use the methods the Hausman-Taylor estimation is partially built on--random effects and fixed effects--by way of gaining additional insights into how the respective underlying assumptions of the different methods affect the key results. One other extension in the empirical analysis is undertaken to further explore potential unmitigated endogeneity. As mentioned above, the Hausman-Taylor estimation method is able to correct estimates' inconsistency arising from potential correlation of unobserved characteristics with districts' ability to both leverage greater amounts of intergovernmental transfers as well as to generate relatively higher local revenues. However, although we assess the effect of past transfers on subsequent local revenues, this does not fully mitigate potential simultaneity. Thus, we employ an analytical extension that consists of a fixed-effects two-stage least squares (2SLS) estimation approach, which takes advantage of the information available on DACF's resource allocation formula, described in some detail in Section 3 above. The weights for each factor and indicator are determined on a yearly basis by the DACF administrators' office, with the recommended weights by this office usually approved by parliament as is. The annual reports by the DACF administrator to parliament detail the rationale for each year's changes made in the weighting scheme. The reasonings, while clearly affecting the transfers districts receive, are not related to the levels of districts' local revenues, and thus serve as an optimal building block for an instrument for the external grants variable.xi The instrumental variable is an index which multiplies the weights used in each year's formula with the corresponding indicators' values for each district and year, excluding the IGF-related weights and indicators.xii A second instrument in the fixed-effects 2SLS estimation is the central government's overall budget for each year. Since the DACF is set to be a certain

percentage of the central government's budget, this variable will be correlated with the amount of intergovernmental and external grants, and affects locally generated revenues solely through the transfers provided from central to local government. We test the overidentifying restrictions implied by this model to ascertain the validity of the instruments and specification. The second model investigates whether external grants also have a growth effect on local revenue generation. Compared to the level model, this investigation is more indicative of the longer-term trajectory local revenues may see as a consequence of the external grant flows into district budgets. In the growth model, we account for a possible path-dependency of IGF growth due to the slow-changing nature of districts' tax bases (for example due to low mobility of taxable entities, e.g. businesses, and the stationarity of natural resources which generate royalties) by including districts' past IGF growth as an explanatory variable. We also allow for the possibility that low-local-revenue districts may have a different growth trajectory than high-local-revenue districts, by including the lagged level of IGF; this parameter speaks to the nature of convergence (or divergence) of districts' local revenues over time. ln The model of IGF growth is then: = ln + + + ln + ln + ln +



While the level of a district's per capita IGF is expected to depend importantly on the district attributes captured in the level model, as illustrated above, the growth rate of IGF is less obviously reliant on such factors, and we can concentrate our analysis on local public financial factors. In econometric terms, the focus on public financial explanatory variables, with the inclusion of past IGF growth to account for possible path-dependency in local revenue trajectories, also permits us to apply estimation methods that are appropriate for the case at hand but that are not able to estimate time-invariant factors. We estimate the model using the Arellano-Bond (1991) and the Arellano-Bover (1995) / Blundell-Bond (1998) dynamic system-GMM estimators. Blundell and Bond (1998) show in Monte Carlo simulation that when the true parameter value of the lagged dependent variable is relatively large, the number of time periods in the panel moderately small, and the number of panels not too large, the system-generalized method of moments (GMM) estimator can reduce both finite sample bias that would obtain in the difference-GMM estimation, as well as improve on the precision of the estimator. Although in our case the latter two conditions do not raise immediate concerns (both T and n are not small), a comparison between the results from these two estimators will be presented and results discussed in this light. The estimations treat the past public expenditure variables as endogenous, using both GMM type instruments as well as the external instruments employed in the 2SLS estimation of the level model. As before, overidentifying restrictions are tested.

5. ESTIMATION RESULTS Tables 4a and 4b present the results of model (1). The fixed-effects estimation results, including those of the fixed-effects instrumental variables model, naturally do not identify the time-invariant control variables. The results are estimated, in the case of random effects, both with and without these variables, to examine the robustness of the core results to their exclusion. In Table 4a, the results regressing the most recent year's (2004) log per capita own revenues on district characteristics are contrasted with a three year moving average of this measure, to consider a formulation of the variable with possible yearon-year instabilities in revenue collection smoothed out. We undertake four main diagnostic tests of the estimation methods under consideration. First, the Hausman test rejects the GLS random effects approach as delivering the correct assumptions about the individual-level effects being random and uncorrelated with the covariates (Hausman 1978). Secondly, a Sargan-Hansen test of overidentifying restrictions in the Hausman-Taylor model fails to reject the null hypothesis that the instruments used in this model are valid (Hayashi 2000, Schaffer and Stillman 2010). The remaining two tests examine the instrumentation strategy for the 2SLS estimation method. The Anderson canonical correlations test result strengthens confidence that the instrumental variables model is identified and the instruments relevant (the null hypothesis that the equation is underidentified is rejected), and the Sargan overidentification test statistic, with the null hypothesis that the instruments are correctly included, suggest that the instruments satisfy exclusion restrictions (Schaffer 2007). Overall, the coefficient estimates are quite robust across the estimation methods, and robust to the exclusion of the control variables, however with some differentiation between the models with the dependent variable specified as a year's value versus as a three-year moving average. The results suggest quite consistently across the various estimation methods, which reflect different empirical assumptions, that external grants and transfers taken as a whole do not lead to greater internal revenue generation; on the contrary, we see statistically significant evidence of a negative relationship. Transfers appear to discourage rather than encourage internal revenue generation. The magnitude of the negative effect is, however, quite small in the Hausman-Taylor estimation as well as GLS results: A 10 percent increase in external transfers to local governments is associated with an approximately one-third of a percent drop in own-revenue generation. When considering the smoothed measure of own-revenues, the effect is less than half that and is only close to being statistically significant (with a p-value of 0.107 for the core regression result in Column 1 of Table 4a). <TABLE 4a HERE>

The results also show that, as hypothesized, there is strong differentiation across the types of public spending in the effect they have on subsequent revenue collection. There is no statistically significant response of own revenues to past capital and personnel expenditures in the case of the main specification and estimation. In contrast, the response is large with regard to past nonpersonnel recurrent spending, in which a 10 percent increase is associated with a 3.9 percent increase in local revenues in the subsequent year. This strong response differentiation by spending category is not surprising, and is in fact quite consistent with the way that revenue sources are linked to expenditure types. As discussed in Section 3, local governments undertake capital investments primarily using the DACF, donor grants, and other external transfers, rather than through IGF. The latter are mostly used for maintenance and operational expenses, and only to a limited extent for personnel expenditures. Thus, higher past expenditures in the nonpersonnel recurrent category are more likely to invoke greater subsequent local revenue mobilization than are increases in past capital or salary spending. Finally, the Hausman-Taylor results show that with the public financial variables accounted for, only few of the district characteristics emerge as significantly affecting own revenue size. Poorer districts tend to generate lower per capita own revenues, as is seen in the statistically significant results across both ways of measuring own revenue and across estimation methods. This is not particularly surprising, as the income base for taxation is likely lower in districts with a greater share of residents in poverty. Similarly intuitively, more urbanized districts have greater per capita own revenues. With businesses being an important source of revenue for local governments, through license fees and other fees levied on enterprises, more urbanized districts with their greater prevalence of businesses offer more opportunities for local revenue generation. The first column in Table 4b presents the two-stage least squares estimation results, in which potential simultaneity bias is corrected for by instrumenting the external grants. The GLS and standard fixed-effects results are included for comparison. Qualitatively, the coefficient estimates lie in the same direction as the Hausman-Taylor results: the effect of external grants on own-revenue effort is negative. However, here, a 10 percent increase in external grants is associated with a 6.2 percent decrease in own revenue effort, an important departure from the much more modest magnitudes in the previous regressions. One likely reason for the Hausman-Taylor results--and for that matter, the results of the alternative estimations--being less negative (that is, smaller in absolute terms) may precisely be the efforts through one of the funding facilities to incentivize local revenue generation. The possible effect of higher IGF "lifting" the disbursement of grants from central government to the districts would not necessarily make itself substantially felt, since this reverse effect would run from past IGF to future grants, while in our equation (1) we are modeling the effect of lagged grants on IGF. Nevertheless, this remedy alone mitigates, but may not fully eliminate, potential simultaneity (Wooldridge 2002, p.301),

and the remnant traces of such reverse effect may be at play in the first regressions. What is worth highlighting, however, is that even without clearing out the potential boosting effects of incentive schemes, external grants still exercise a negative, albeit small, effect on own revenues. <TABLE 4b HERE> The analysis thus far has considered how the flow of external grants to districts may influence the magnitude of per capita internal revenue funds generated by the local governments themselves. For the reasons discussed in Section 4, it is useful to examine as well how financial factors affect the growth, or changes, in IGF. The results of this set of estimations are presented in Table 5. In all the estimations, GMM-type instruments are used for the external grants variable as well as the expenditure variables, all of which are treated as endogenous. In addition, the same excluded instruments as in the 2SLS estimation of the level model are used in Cols. 1, 2, 4 and 5, and further robustness of the main results are examined by considering a specification without the year effects (Cols. 2 and 5) and without the excluded instruments (Cols. 3 and 6). <TABLE 5 HERE> The results show evidence that external grants inhibit own revenue growth. Robustly in statistical significance and magnitude across nearly all of the alternative models, a 10 percent increase in external grants effects an increase in the local revenue growth rate of approximately half a percentage point, which corresponds to a nontrivial decrease from the mean local revenue growth rate of 4.4 percent to a growth rate of about 3.9 percent. Consistently with our findings in the level equation, among the expenditure types, the more pronounced drivers of IGF growth are past increases in public spending on nonpersonnel recurrent spending and, to a somewhat lesser extent, salary expenditures. Capital expenditure increases do not induce subsequent greater growth in local revenues.

6. SUMMARY AND CONCLUSIONS The main motivation for decentralization, namely, that subnational governments are better placed to allocate public resources more efficiently and effectively, is often supported by the argument that subnational governments have better information about the needs for and requirements of public services in their jurisdictions. This argument in favor of decentralization rests strongly on the assumption that local governments have a substantial degree of fiscal autonomy and are able to use local discretion in resource allocation. However, the fiscal responsibilities of local governments often remain quite circumscribed, and their budgets are dominated by external transfers that are tied to specific investments, which may or may not match the priorities of local governments. Given such a constellation, local governments' fiscal autonomy is exercised predominantly through their internally generated funds (IGF),

which are allocated to services and public goods without central government or other external spending criteria imposed. Using 1994­2004 panel data on all 110 district governments' public finances and other districtlevel data, this paper examined the impact of the flow and size of externally generated revenues on the incentives these induce on the part of local governments to undertake own revenue raising effort. The impact of fiscal transfers to local governments on their internal revenue generation incentives has been analyzed in developing countries, but has hardly received attention in the developing country context. In this respect, research lags behind policy attention, as concern about a potential disincentive effect has been in the decentralization policy discourse in Ghana for some years. We find that greater past external transfers to district governments do not encourage internal revenue generation, but instead have a slightly depressing effect on own revenues. Despite inbuilt incentives in the allocation formula for one of the key intergovernmental grants, increases in grants and transfers lead to a modest decline in local revenue effort, not the increase hoped for by policy makers in the decentralization community. When more explicitly accounting for the potential simultaneity emerging from the incentive schemes, the negative effects of external grants on local revenues becomes much more pronounced, suggesting that the incentives may mitigate, but by no means eliminate, the negative impact transfers have on locally raised revenues. In addition to examining a level effect on own revenues, we assess how transfers affect subsequent growth of IGF, finding again a depressing effect. One can argue that a necessary (albeit not sufficient) condition for the gains of decentralization to be realized is that local governments have substantive autonomy over how resources are allocated across competing needs in their jurisdiction. Ghana's decentralization landscape is characterized by local budgets which are very small relative to central government spending in districts, own-sourced local revenues which are small relative to other sources of local budgets (and limited local autonomy over the latter), and, as discussed in this paper, a prevailing negative impact of transfers on IGF. In this context, serious consideration should be given to the alternative mechanisms to increase local governments' fiscal autonomy, and perhaps just as importantly, to the political economy factors which may affect feasibility of implementing these mechanisms.

APPENDIX This is an illustration of how the amount of funds each district receives out of a given year's DACF pool is calculated. The following formula, derived from information from the DACF reports to parliament, was used for the 2003 allocation: = + + + 110 + + + + +

where Yi is the DACF amount district i receives in this year, b is the share of the DACF pool X that is not going toward the contingency fund, N is the population size, and indicates the weights for the different factors/indicators. S, E, P, H, T, D, W refer to the number of schools, educational facilities that require major repair, pupils, health facilities, teachers, doctors, and people with access to potable water, respectively; A is the land area in km2, and C is the percentage growth, or change, in IGF.

_____________________________________ Acknowledgements: The authors are grateful for financial support from the Ghana Strategy Support Programme (GSSP) of IFPRI. We thank an anonymous reviewer, the late Leah Horowitz, and Nethra Palaniswamy for remarks on an earlier working paper version of this paper. Thanks also goes to comments provided by participants at the University of Oxford's Centre for the Study of African Economies (CSAE) conference, participants at a local government workshop at IFPRI, and discussant Ian Coxhead and participants at the Allied Social Science Association (ASSA) meetings, all conferences at which this work was presented.

REFERENCES Alemán, E., & Treisman, D. (2005). Fiscal politics in "ethnically mined," developing, federal states: Central strategies and secessionist violence. In P. Roeder, & D. Rothschild (Eds.), Sustainable peace: Power and democracy after civil wars (pp. 173-216). Ithaca, NY: Cornell University Press. Allers, M., de Haan, J., & Sterks, C. (2001). Partisan influence on the local tax burden in the Netherlands. Public Choice 106(3­4): 351­363. Arellano, M., & Bond, S. (1991). Some tests of specification for panel data: Monte Carlo evidence and an application to employment equations. Review of Economic Studies 58(2): 277­297. Arellano, M., & Bover, O. (1995). Another look at the instrumental variable estimation of errorcomponents models. Journal of Econometrics 68(1): 29­51.

Bahl, R. (2010). Conditional vs. unconditional grants: The case of developing countries. In J. Kim, J. Lotz, & J. Mau (Eds.), General grants versus earmarked grants: Theory and practice (pp. 126-148). Albertslund, Denmark, and Seoul, Korea: Korea Institute of Public Finance and Danish Ministry of Interior and Health. Banful, A. B. (2011). Do formula-based intergovernmental transfer mechanisms eliminate politically motivated targeting? Evidence from Ghana. Journal of Development Economics (forthcoming). Bardhan, P. (2002). Decentralization of governance and development. Journal of Economic Perspective 16(4): 185­205. Bardhan, P., & Mookherjee, D. (2006). Decentralization and accountability in infrastructure delivery in developing countries. Economic Journal 116(508): 101­127. Bird, R. M., & Smart, M. (2002). Intergovernmental fiscal transfers: International lessons for developing countries. World Development 30(6): 899­912. Blundell, R., & Bond, S. (1998). Initial conditions and moment restrictions in dynamic panel data models. Journal of Econometrics 87(1): 115­143. Borge, L.-E., & Rattsø, J. (1997). Local government grants and income tax revenue: Redistributive politics in Norway 1900­1990. Public Choice 92(1­2): 181­197. Boyne, G. A. (1998). Party competition and local spending decisions. British Journal of Political Science 28(1): 210­222. Bradford, D. F., & Oates W. E. (1971). Towards a predictive theory of intergovernmental grants. American Economic Review 61(2): 440­448. Brennan, G., & Buchanan, J. (1980). The power to tax: Analytical foundations of a fiscal constitution. Cambridge, MA: Cambridge University Press. Bucovetsky, S., & Smart, M. (2006). The efficiency consequences of local revenue equalization: Tax competition and tax distortions. Journal of Public Economic Theory 8(1): 119d­144. Buettner, T., & Wildasin, D. E. (2006). The dynamics of municipal fiscal adjustment. Journal of Public Economics 90(6-7): 1115­1132. Coulombe, H. (2005). Ghana census-based poverty map: District and sub-district level results. Accra, Ghana: Ghana Statistical Service. Crook, R. C. (2003). Decentralization and poverty reduction in Africa: The politics of local-central relations. Public Administration and Development 23(1): 77­88. Crook, R. C., & Manor, J. (1998). Democracy and decentralization in South Asia and West Africa. Cambridge, UK: Cambridge University Press. DACF. (District Assemblies Common Fund). Annual report to parliament from District Assemblies' Common Fund. Reports for the years 1993/94, 1994, 1995, 1997, 1999, 2000, 2001/02, 2003, 2004. Accra, Ghana.xiii Dahlberg, M., & Johansson, E. (1998). The revenues-expenditures nexus: Panel data evidence from Swedish municipalities. Applied Economics 30(10): 1379­1386. Dahlberg, M., Mörk, E., Rattsø, J., & Ågren, H. (2007). Using a discontinuous grant rule to identify the effect of grants on local taxes and spending. Journal of Public Economics 92(12): 2320­2335. de Mello Jr., L. R. (2002). Public finance, government spending and economic growth: The case of local governments in Brazil. Applied Economics 34(15): 1871­1883. Dye, R. F., & McGuire, T. J. (1997). The effect of property tax limitation measures on local government fiscal behavior. Journal of Public Economics 66(3): 469­487. GoG (Government of Ghana). (2007). Decentralization policy review. NCG and DEGE Consult, Accra, Ghana. Hausman, J. A. (1978). Specification tests in econometrics. Econometrica 46(6): 1251­1271. Hausman, J. A., & Taylor, W. E. (1981). Panel data and unobservable individual effects. Econometrica 49(6): 1377­1398. Hayashi, F. (2000). Econometrics. Princeton NJ: Princeton University Press. Hayek, F. (1945). The use of knowledge in society. American Economic Review 35(4): 519­530.

Holtz-Eakin, D. (1992). Elections and aggregation: Interpreting econometric analyses of local governments. Public Choice 74(1): 17­42. Holtz-Eakin, D., W. Newey, & Rosen, H. S. (1989). The revenues-expenditure nexus: Evidence from local government data. International Economic Review 30(2): 415­429. IMF. (2006). Ghana poverty reduction strategy paper annual progress report. IMF Country Report No. 06/226. International Monetary Fund, Washington DC. ISODEC. (Integrated Social Development Centre). (2004). Budgeting and accounting structures of district assemblies. Accra, Ghana: Centre for Budget Advocacy, Integrated Social Development Centre. ISODEC. (2005). Financing decentralized development: How well does the DACF work? Accra, Ghana: Centre for Budget Advocacy, Integrated Social Development Centre. Kokor, J. Y., & Kroës, G. (2000). Central grants for local development in a decentralized planning system. Dortmund, Germany: SPRING Centre, University of Dortmund. Moore, M. (2008). Between coercion and contract: Competing narratives on taxation and governance. In D. Bräutigam, O.-H. Fjeldstad, & M. Moore (Eds.), Taxation and state-building in developing countries (pp. 34-63). Cambridge, UK: Cambridge University Press. Payne, J. E. (2003). A survey of the international empirical evidence on the tax-spend debate. Public Finance Review 31(3): 302­324. Prud'homme, R. (1995). The dangers of decentralization. World Bank Research Observer 10(2): 201­ 220. Rajaraman, I., & Vasishtha, G. (2000). Impact of grants on tax effort of local government. Economic and Political Weekly 35(33): 2943­2948. Ross, M. L. (2004). Does taxation lead to representation? British Journal of Political Science 34(2): 229­ 249. Schaffer, M. E. (2007). xtivreg2: Stata module to perform extended IV/2SLS, GMM and AC/HAC, LIML and k-class regression for panel data models. Schaffer, M. E., & Stillman, S. (2010). xtoverid: Stata module to calculate tests of overidentifying restrictions after xtreg, xtivreg, xtivreg2 and xthtaylor. Skidmore, M. (1999). Tax and expenditure limitations and the fiscal relationship between state and local governments. Public Choice 99(1­2): 77­102. Smart, M. (1998). Taxation and deadweight loss in a system of intergovernmental transfers. Canadian Journal of Economics 31(1): 189­206. Solé-Ollé, A. (2006). The effects of party competition on budget outcomes: Empirical evidence from local governments in Spain. Public Choice 126(1): 145­176. Tiebout, C. (1956). A pure theory of local expenditures. Journal of Political Economy 64(5): 416­424. Wildasin, D. E. (1984). The welfare effects of intergovernmental grants in an economy with distortionary local taxes. Journal of Public Economics 25(1­2): 103­125. Wooldridge, J. M. (2002). Econometric analysis of cross section and panel data. Cambridge, MA: MIT Press. Zhang, T., & Zou, H. (1996). Fiscal decentralization, public spending, and economic growth in China. Policy Research Working Paper 1608, World Bank, Washington, DC. Zhuravskaya, E. V. (2000). Incentives to provide local public goods: Fiscal federalism, Russian style. Journal of Public Economics 76(3): 337­368.

TABLES Table 1. Average district revenue in 2004, by revenue source

External sources (% of total) Central government Salary HIPC DACF transfers 7.6 18.8 38.4 13.1 21.6 47.1 8.1 27.7 50.4 11.1 17.9 51.3 11.1 18.1 20.9 5.1 20.0 65.7 6.6 3.3 6.7 6.8 8.1 15.6 30.9 34.3 26.9 22.6 41.9 36.9 44.9 41.6 43.5 Donor funds a 19.5 7.5 6.0 6.2 0.7 6.5 27.7 22.8 6.3 2.1 9.8 Total external 84.3 89.3 92.2 86.5 50.9 97.2 91.8 93.9 92.1 77.4 84.0 Internal (% of total) IGF 15.7 10.7 7.8 13.5 49.1 2.8 8.2 6.1 7.9 22.6 16.0 Total (2000 constant GHC) 653,975 471,750 491,149 365,634 1,604,321 600,811 591,765 759,999 489,934 633,756 593,781

Region Ashanti Brong Ahafo Central Eastern Greater Accra Northern Upper East Upper West Volta Western Ghana

DACF = District Assemblies Common Fund; HIPC = Highly Indebted Poor Countries; IGF = internally generated funds a While HIPC funds also originate from donor resources, they are disseminated by the central government and thus here reported separately from other donor funds. This category refers to resources from donors going directly to districts.

Table 2. Formula weights for criteria for District Assemblies Common Fund (DACF) allocation (%)

Factor/Indicator Equality: Flat rate Need: Population size in 1984 GDP in 1992 No. of health facilities Doctor/population ratio Nurse/population ratio No. of schools Teacher/pupil ratio % with access to potable water km of tarred roads No. of schools needing repair* Responsiveness: Per capita IGF level % increase in IGF Service pressure: Population density Total: Reserve fund:

Source: DACF, various years.

1994 1995 1996 1997 1998 1999 2000 2001 2002 2003 2004 30 30 30 30 35 35 35 35 35 35 60 35 35 35 35 35 35 40 40 50 55 35 5 5 0 0 0 0 0 0 0 0 0 30 30 15 10 5 10 0 0 0 0 0 0 0 10 12.5 15 12.5 12.5 12.5 12.5 12.5 5 0 0 0 0 0 0 7.5 7.5 7.5 7.5 5 0 0 0 0 0 0 0 0 0 0 5 0 0 10 12.5 15 12.5 12.5 12.5 12.5 12.5 5 0 0 0 0 0 0 7.5 7.5 7.5 7.5 5 0 0 0 0 0 0 0 0 10 10 5 0 0 0 0 0 0 0 0 0 0 5 0 0 0 0 0 0 0 0 0 5 0 20 20 20 20 20 20 15 15 5 5 2 20 20 15 15 15 15 10 10 0 0 0 0 0 5 5 5 5 5 5 5 5 2 15 15 15 15 10 10 10 10 10 5 3 100 100 100 100 100 100 100 100 100 100 100 5 10 10 10 10 10 10 10 10 10 10

Table 2b. Indicators used for the responsiveness factor


Year 1994 1995 1996 1997 1998 1999 2000 2001 2002 2003 2004

Level indicator '92 IGF / '84 pop '92 IGF / '84 pop '92a IGF / '84a pop '94 IGF / '95a pop '95a IGF / '96a pop '96a IGF / '97 pop '97a IGF / '98 pop '99a IGF / '00 pop

Change indicator

('92 IGF - '93 IGF) / '92 IGF ('94 IGF - '95 IGF) / '94 IGF ('95 IGF - '96 IGF) / '95 IGF ('96 IGF - '97 IGF) / '96 IGF ('97 IGF - '98 IGF) / '97 IGF ('99 IGF - '00 IGF) / '99 IGF ('99 IGF - '01 IGF) / '99 IGF ('99 IGF - '01 IGF) / '99 IGF ('99 IGF - '02 IGF) / '99 IGF

The exact years for these variables were not available from the DACF reports obtained, thus these constitute our best guess on the specific variable year, based on other information in the report (for example, for the 1997 formula the 1994 IGF was used in the level indicator, which is also the starting year for the change indicator. We thus assume that the starting year's IGF in the change indicator of the formulas for 1996 and 1998-2001 were used in the respective level indicators.

Table 3. Variable definition and district-average descriptives Variable sh. urban pop. density pop. size poverty rain no road literacy Christian ethn. Akan ethn. Ewe R


Description Proportion of urban residents Persons per km2 land area Population size in 1,000s Headcount poverty rate b Long-term annual rainfall, in mm Share of households with road not / not easily accessible all year round Literacy rate Share of Christian residents Share belonging to the Akan ethnic group Share belonging to the Ewe ethnic group Vector of region dummies

Mean a

Standard deviation 0.30 0.22 391.84 188.76 0.17 225.83 0.16 0.19 0.24 0.35 0.23

155.61 171.93 0.47 1304.81 0.19 0.50 0.66 0.47 0.13

Unweighted averages across districts. b Based on data from the 2003 Core Welfare Indicators Questionnaire and the 2000 Population and Housing Census, drawn from Coulombe (2005).

Table 4a. External grants and level of internally generated funds a


Standard errors are in parentheses. p-values are in square brackets. Levels of statistical significance: *** 1%, ** 5%,* 10%. All local public finance variables are measured as the natural log of real per capita GHC (10,000s).

Hausman-Taylor GLS random effects Dep. var as: Single-year 3-year average Single-year 3-year average ** *** EXTit-1 -0.031 -0.015 -0.032 -0.015 * (0.012) (0.009) (0.012) (0.009) * Personnel-Eit-1 0.023 -0.006 0.047 0.002 (0.031) (0.02) (0.028) (0.019) Recurrent-Eit-1 0.388 *** 0.418 *** 0.454 *** 0.454 *** (0.035) (0.025) (0.032) (0.024) Capital-Eit-1 0.017 0.021 * 0.025 * 0.025 ** (0.015) (0.011) (0.015) (0.011) pop. size -0.001 -0.001 -0.001 -0.001 (0.001) (0.001) (0.001) (0.001) sh. urban 0.235 * 0.263 * 0.210 0.260 ** (0.141) (0.138) (0.133) (0.133) 0.013 0.000 0.006 pop. density 0.028 (0.036) (0.033) (0.033) (0.032) poverty -1.302 *** -1.177 *** -1.163 *** -1.086 *** (0.281) (0.273) (0.266) (0.263) * 0.309 0.222 0.268 rain 0.611 (0.347) (0.264) (0.251) (0.252) -0.221 -0.196 -0.205 no road -0.227 (0.169) (0.165) (0.16) (0.159) literacy 0.292 0.366 0.440 0.414 (0.429) (0.413) (0.4) (0.397) -0.363 -0.431 -0.363 Christian -0.467 (0.291) (0.283) (0.276) (0.272) ethn. Akan -0.151 -0.105 -0.095 -0.098 (0.17) (0.165) (0.159) (0.159) 0.221 0.211 0.193 ethn. Ewe 0.335 (0.226) (0.216) (0.206) (0.207) 3.527 * 3.635 ** 3.379 * Constant 1.746 (2.441) (1.845) (1.763) (1.767) No. of obs.: 1,028 951 1,028 951 Sargan-Hansen Hausman 2.17 4.317 29.21 23.83 (2): [0.989] [0.889] test (2): [0.0098] *** [0.033] **


Table 4b. External grants and level of internally generated funds a


Standard errors are in parentheses. p-values are in square brackets. Levels of statistical significance: *** 1%, ** 5%,* 10%. All local public finance variables are measured as the natural log of real per capita GHC (10,000s).

IV EXTit-1 -0.621 * (0.318) Personnel-Eit-1 0.342 *** (0.126) Recurrent-Eit-1 0.383 *** (0.067) Capital-Eit-1 0.283 ** (0.138) Constant -- -- Anderson 4.96 * LM stat (2): [0.06] Sargan 1.74 overid. (2): [0.187]

GLS -0.042 *** (0.013) 0.093 *** (0.03) 0.593 *** (0.032) 0.014 (0.016) 3.042 *** (0.302) Hausman test (2):

FE -0.032 *** (0.012) 0.023 (0.031) 0.382 *** (0.035) 0.017 (0.015) 5.098 *** (0.346)

335.10 [0.000]***

Table 5. External grants and growth of internally generated funds a

Blundell-Bond est. EXTit-1 Personnel-Eit-1 Recurrent-Eit-1 Capital-Eit-1 Lagged dep. var. past IGF level Constant Year-effects External instruments GMM type instruments AB Test, order 1: 2: (1) -0.051 ** (0.024) 0.110 *** (0.042) 0.327 *** (0.083) 0.018 (0.016) -0.026 (0.041) -0.752 *** (0.1) 3.004 *** (0.533) yes yes yes -6.4459 .37312


Arellano-Bond est. (3) -0.050 ** (0.024) 0.109 *** (0.042) 0.327 *** (0.084) 0.017 (0.016) -0.027 (0.041) -0.751 *** (0.101) 2.998 *** (0.533) yes no yes (4) -0.042 * (0.023) 0.144 *** (0.042) 0.313 *** (0.072) 0.021 (0.016) 0.067 * (0.036) -0.998 *** (0.08) 4.654 *** (0.516) yes yes yes


(2) -0.044 * (0.024) 0.098 *** (0.037) 0.336 *** (0.08) -0.002 (0.016) -0.083 * (0.043) -0.648 *** (0.097) 2.260 *** (0.468) no yes yes -6.7445 .29458


(5) -0.039 (0.026) 0.145 *** (0.037) 0.311 *** (0.067) 0.011 (0.016) 0.001 (0.041) -0.871 *** (0.081) 3.630 *** (0.452) no yes yes

(6) -0.040 * (0.024) 0.142 *** (0.042) 0.313 *** (0.072) 0.020 (0.016) 0.066 * (0.036) -0.995 *** (0.081) 4.644 *** (0.518) yes no yes


-6.4475 0.36907

-6.0058 1.2963


-6.4428 1.1651

-6.0073 *** 1.2949

Number of observations: 893. a Robust standard errors are in parentheses; levels of statistical significance: *** 1%, ** 5%,* 10%. All local public finance variables are measured as the natural log of real per capita GHC (10,000s).



Figure 1. Components of districts' revenue: internally generated funds (IGF) and external funds

0.45 0.40 0.35 Per capita GHC 0.30 0.25 0.20 0.15 0.10 0.05 0.00 '94 '95 '96 '97 '98 '99 '00 '01 '02 '03 '04 4.0 3.5 3.0 2.5 2.0 1.5 1.0 0.5 0.0 Per capita GHC

IGF (axis l.)

Ext. funds (axis r.)

Figure 2. Components of internally generated funds


1994­2004 average:



There are a range of realities in many developing countries which render some of these classic arguments empirically tentative, including low citizen mobility and nondemocratic systems. Scholars have also elaborated on the potential downsides of decentralization, such as efficiency loss, greater geographic inequality, elite capture, and social unrest. See for example Bardhan and Mookherjee (2006), Prud'homme (1995), and Alemán and Treisman (2005). ii Tied grants are not necessarily a priori "better" or "worse" than untied grants. Arguments marshalled for the former include the need to align local expenditure patterns to national priorities; achieve equity or minimum standards in key public services across local jurisdictions; and prevent under-provision of certain public goods which have externalities. Arguments in favour of untied grants include better alignment of expenditures with locally determined needs and preferences, and the stimulation of political voice in articulating local priorities and holding local governments accountable for adherence to these priorities. See Bahl (2010) for a think piece on conditional versus unconditional grants in developing countries. iii The terms `internally generated funds (IGF)', `own revenues', `local revenues', `locally generated funds', etc. are used interchangeably to refer to the revenues local governments levy through their local tax and fee assignments. iv The theoretical and empirical literature on the consequences of decentralization for development outcomes is vast, and this section does not have the ambition to provide an adequate background to the wider literature on fiscal decentralization in developing countries. Prud'homme (1995) and Bardhan (2002) provide good overviews. v de Mello (2002) uses Brazilian municipal revenues as a proxy for local GDP, and thus interprets his findings in this light. vi After 2004 an additional 28 districts were created by splitting some larger districts, and further splits in 2007 resulted in the current 168 districts. vii In the midpoint of the study period, 1999, the exchange rate was US$1 = 0.30 GHC. The Ghanaian Cedi (GHC) was introduced in 2007, by removing four zeros from the old cedi which prevailed until that time, that is, as of 2007, 10,000 cedis were converted to 1 GHC. Although the study period is before this recalibration, we express all monetary units in GHC instead of cedis. viii The guidelines for these years were: 2% of the funds to be used for `Capacity and Human Resource Improvement', 20% for `Productivity Improvement and Employment Generation (Poverty Alleviation)', 10% for `Self-help projects', 2% for the `District Education Fund to support scholarships and needy students', 5% for `Establishing and strengthening of sub-district structures, 1% for the `District response initiative on HIV/AIDS, 1% on `Malaria control', and 59% on `Other projects: economic, social administration, environment'. ix While DAs can set property tax rates, the revaluation of taxable properties--proper implementation of which would be likely to increase property taxes accruing to DAs--is undertaken by the Land Valuation Board, which is not under the control of the DAs and itself lacks the capacity to execute this task. x Payne (2003) reviews the literature that tests this and other hypotheses concerning the intertemporal dynamics between revenues and expenditures, separately presenting studies at the national and at the subnational levels. xi The following are some examples of the reasonings driving changes in the formula, namely data limitations, and the creation of new districts leading to a greater fall back on the equal-allocation-per-district factor. Specifically, the 2000 annual report states: "With the elimination of GDP per capita and additional indicators introduced for Education and Health, the weight for the Need Factor is proposed to be increased to 40% from 35%" (DACF 2000, Executive Summary, section IV). And the 2004 annual report describes the rationale for changes in that year's weighing scheme as follows: "This year 28 new Districts are to be created [...] The Equality Factor has been given prominence over the other factors. The reason being that there are no specific data on the other factors pertaining to the new districts. Therefore given prominence to the equality factor gives a reasonable minimum to each district" (DACF 2000, p.10). xii For cases in which the allocation to a district increases with the decrease of the indicator, the corresponding weight is multiplied by the reciprocal of that indicator. xiii There are slight variations in the name for the various annual reports.



Microsoft Word - For MWIEDC.docx

26 pages

Find more like this

Report File (DMCA)

Our content is added by our users. We aim to remove reported files within 1 working day. Please use this link to notify us:

Report this file as copyright or inappropriate


You might also be interested in

Microsoft Word - Mar 10 FINAL
city revenue sources
Microsoft Word - BABcover.doc