Should Large Spanish Municipalities be Financially Compensated? Costs and Benefits of Being a Capital/Central Municipality

This paper analyses the costs and benefits of being a capital or central municipality, where central costs are understood to be incurred specifically as a result of the problems large municipalities located at the centre of an urban agglomeration face and capital costs are understood to result from the presence of regional and/or central government institutions in the municipality. However, these two qualities might also be beneficial to municipalities, resulting in a direct increase in their fiscal capacity. Here, by estimating an equation of the expenditure needs and the fiscal capacity of Spanish municipalities with more than 75,000 inhabitants, we find that the central costs incurred by large municipalities are offset by their greater fiscal capacity, but that the same is not true for municipalities that serve as political/administrative capitals.


Introduction
Local public finances are currently under considerable stress and the financial crisis serves only to highlight the need to rethink both expenditure and revenue policies at all lev 68 NÚRIA BOSCH, MARTA ESPASA AND DANIEL MONTOLIO els of government. The situation is particularly critical among Spanish municipalities, which as the level of government closest to the citizens (voters) are responsible for delivering the most highly demanded public services, at a time when the Spanish financing system is fail ing to provide these authorities with adequate revenue tools to meet such demands 1 . The outlook is even worse for large municipalities, which are beset by a series of specific problems that have a direct impact on their local budget, but which are not contemplated by the local financing system. In the light of this current situation, the aim of this article is to estimate the impact on the expenditure needs and fiscal capacity of being a large capital or central municipality.
In terms of costs, the expenditure needs of large municipalities differ from those of other municipalities as their specific socioeconomic characteristics generate both central (spillover effects, concentration of social problems and/or congestion costs) and capital costs 2 (including, the substitution of productive activities with less productive administra tive activities or the loss of fiscal revenue because of the tax exemption of administrative ac tivities).
Outside Spain, Washington D.C. is a good illustration of a city whose central and capi tal costs have not been solved by granting it a special financing system, but rather by increas ing the percentage of financing of some of the federal transfer programmes and requiring the federal government to accept the provision of certain services (prisons and civil service pen sions). A further illustration is provided by the case of the United Kingdom, where a system of unconditional transfers to municipalities, based on sophisticated methods of calculation, seeks to take into account certain factors related to central costs (McLean and McMillan, 2003;Midwinter, 2002).
In terms of revenue, the capital and/or central nature of a municipality may affect the de cisions of individuals and firms when choosing where to locate in the territory and, there fore, have an impact on the size of the tax base for each municipality (fiscal capacity). Ac cording to Ladd and Bradbury (1988), the fiscal capacity of a sub-central government depends on the decisions adopted by individuals and enterprises, which in turn depend on the perceived costs and benefits from the localisation in the territory of this sub-central gov ernment.
The main contribution of this paper is to analyse the factors that affect large municipal ities' expenditure needs and to determine whether these same factors also have a positive im pact on their fiscal capacity. Specifically, we analyse the impact of being a capital or central municipality, factors which have not received much attention in the academic literature. However, Greene et al. (1974), Ladd and Yinger (1989), Chernick and Tkacheva (2002) do analyse whether commuter costs are offset by the revenues commuters generate, concluding that commuter costs (reflecting centrality) tend to be higher than any additional revenues. This paper provides further empirical evidence on this question, analysing from a wider per spective than that taken by these earlier studies (that is, by taking into account other factors in addition to commuting) whether the various costs borne by large municipalities are offset by the revenue generated by the economic activity, for instance, that the large municipalities can attract. To do so we draw on a dataset for Spanish municipalities with more than 75,000 inhabitants 3 , and estimate the determinants of both expenditure needs (Bosch and Solé-Ollé, 2005) and fiscal capacity. Note that, given that we focus on a certain type of mu nicipalities (those with more than 75,000 inhabitants), we are studying the differential effects of some capital/central factors among those large municipalities; we do not analyse the im pact of such factors with respect small size municipalities. This point is important to inter pret and understand properly the results we obtain.
The present study comprises six sections, the first being this introduction. Section 2 de scribes the specific traits of large municipalities, in terms of their costs and revenues. Sec tion 3 reports the empirical analysis and Section 4 describes the data used in the estimation of the equations of the expenditure needs and the fiscal capacity of large municipalities. Sec tion 5 describes the main results obtained. Finally, section 6 concludes.

C Central costs
Central costs are those generated by the specific problems faced by large municipalities situated at the centre of an urban agglomeration. The main central cost is attributable to the flow of non-residents (those residing at a certain distance from the municipality in question) and which central cities absorb each day. Here, the main cost is caused by the spillover ef fect that municipal services are put under. Other central costs are attributable to the concen tration of social problems (immigration, poverty, etc.) and those related to diseconomies of scale, density or congestion. Each of these costs is briefly described below.

Spillover effects and congestion costs of municipal services
A central municipality typically receives a large flow of non-residents: for reasons of work (commuters), education, shopping, administrative activities and leisure. These non-resident visitors use the services provided by the municipality for its inhabitants (and tax payers). The main costs for the municipality generated by non-residents can be identified as traffic manage ment, cleaning, urban waste management, the maintenance of parks and public spaces as well as the provision of cultural and sports facilities and citizen safety and security. These services are provided by municipalities and (also) benefit non-residents that generally do not pay for them and, hence, they generate spillover (or external) effects and congestion costs.
For the Spanish case, the literature here is not very extensive and comprises basically the studies undertaken by Solé-Ollé (2001, 2006 and Bosch and Solé-Ollé (2005) examining the impact on the cost of the provision of municipal services of daily visitors who come to the central city for reasons of work and study. More specifically, Solé-Ollé (2001) quantifies the effects of a range of variables on the cost of providing municipal services in the municipal ities of the Barcelona province with more than 5,000 inhabitants in 1996, concluding that commuters have an impact on expenditure on citizen safety and security, culture and sports, housing and urban development and public welfare. Solé-Ollé (2001) excludes, for reasons of comparability with the rest of the municipalities, the City of Barcelona; however, this in convenience is solved in Bosch and Solé-Ollé (2005) and Solé-Ollé (2006) which draw on a sample of more than 3,000 Spanish municipalities with more than 1,000 inhabitants in the year 1999. In these studies the impact of non-residents on municipal public services is meas ured using the size of the population (i.e., potential non-resident users of the central city's public services) residing in locations within a 30-kilometre radius of the main central city. Solé-Ollé (2006) reports that a non-resident generates an average increase of 9.4% in the mu nicipal expenditure generated by a resident, a figure that rises to 24% in the case of residents in municipalities located in large urban areas.

Attractivity cost
Another type of non-resident with quite specific characteristics is the tourist. Here the costs will be either direct (promotion paid for by the city council) or indirect (greater inten sity of use of public services such as transport, cleaning, cultural activity and security).

Cost of the concentration of social problems
Central cities typically attract social problems associated with immigration, unemploy ment, poverty (the medium-high income population having abandoned the core areas) and delinquency. Solé-Ollé (2001) reports, for example, that rising poverty levels increase the cost of citizen safety and security, social services, culture and sports, and housing and urban development. Bosch and Solé-Ollé (2005) further report that immigrant population is poten tially related with poverty and with a positive and statistically significant impact on munic ipal expenditure.

Other cost factors
The level of population has an impact on public service unit costs, either reducing (economies of scale) or raising them (costs of congestion). The costs of congestion can be related to the effects of commuting, as we have described above, and also to population den sity, which is normally higher in large municipalities. A high population density can, for ex ample, increase the costs related to traffic, delinquency and the destruction of urban furni ture. Likewise, in some cases, a low population density can also generate higher costs in the provision of public services, including refuse collection, public lighting or security (costs re 71 Should Large Spanish Municipalities Be Financially Compensated? Costs and Benefits of Being... lated to disperse populations). For a detailed discussion of urban agglomerations seen from the perspective of public finance see Cadaval (2004).

C Capital costs
Capital costs are related to the presence in the municipality of State and Autonomous Community government institutions. The main costs in this case can be grouped into two categories: i) loss of income due to activity substitution, and ii) an increase in expenditure due to activity substitution. Each of these factors is described briefly below.

Loss of income due to activity substitution
A municipality that serves as a political capital specialises in activities of an administra tive nature and as a result the resources of the municipality (some of which are scarce, such as land) are used for administrative activities as opposed to other economic (primarily com merce and services) or residential activities. Activity substitution is not a neutral factor for the municipal treasury when the administrative activities of the Autonomous Communities and the State are exempt from payment of municipal taxes. So, for example, if a municipal ity contains the headquarters of government institutions it no longer receives the property tax corresponding to residences and/or businesses that hypothetically could occupy the area used for administrative purposes 4 . This argument can be extended to other taxes in the municipal domain, for example the Motor Vehicle Tax and some fees and public prices. Moreover, ac tivity substitution with the increase of civil servants may lead to a lower average productiv ity in the municipality and, hence, weaken its fiscal bases.

Increase in expenditure due to activity substitution
It might also be the case that the substitution of residential, commercial and service ac tivities with administrative activities generates an increase in municipal expenditure, for in stance, in activities concerned with security and public safety (public events), in urban de velopment and the maintenance of public spaces (parks and gardens) and the improvement and promotion of the image of the capital. These impacts are difficult to quantify although Bosch and Solé-Ollé (2005) propose a methodology that uses public sector employees (au tonomous and central governments) per inhabitant as an approximation of municipal expen diture needs.

Revenue factors: benefits from being a capital/central municipality
The centrality or capital status of a large municipality can also potentially generate ad ditional revenue. For instance, economies of agglomeration and infrastructures of communi 72 NÚRIA BOSCH, MARTA ESPASA AND DANIEL MONTOLIO cation can attract the localisation of individuals and firms to large municipalities, and so have a positive effect on their fiscal capacity (Ladd and Bradbury, 1988). In this sense, Brueckner et al. (1999) report in a theoretical analysis conducted in Paris that individuals with higher levels of income tend to localise in zones with the best amenities (which in many cases are in, or close to, large urban areas). Therefore, we also analyse the effect that all the aforementioned factors might have on the fiscal capacity of large municipalities. For instance, it is not clear whether the benefits of the economic activity generated by visitors (commuters or tourists) offset their costs (given the limited revenue instruments available to the municipalities). Empirical evidence from the United States seems to indicate that the costs derived from commuting are not recovered by such large cities as New York and Philadelphia (Chernik and Tkacheva, 2002). Greene et al. (1974) also evaluate the additional costs and revenues generated by non-residents for the US city of Washington.
Tourist enclaves exhibit many characteristics that are highly attractive to visitors and which can affect the fiscal capacity of the municipality through, for instance, a higher de mand for housing. In such cases, real estate prices rise and, as a result, the value of the prop erty tax base also rises. While only a few studies have sought to analyse whether tourism de termines the tax base volume of a jurisdiction, Glaeser et al. (2001) are able to show that American cities with the highest amenity levels (variety of private services, adequate clima tology, etc.) have higher housing prices, which in turn means a higher value of the property tax base. Likewise, Hawkins and Murray (2004) show that the number of hotels per capita, as a proxy of the number of visitors for leisure purposes, has a markedly positive effect on the per capita bases of a city's sales taxes.
In some ways, the previously presented costs could be seen as potential revenue raisers. Apart from the obvious cases of commuting (work, studies, shopping, administrative activ ities and leisure) and tourism, which are easily identified as potential sources of revenue for recipient municipalities via a variety of channels, other variables, including immigration, unit costs (wages) or activity substitution, could have a positive impact on a municipality's fiscal capacity. Indeed, all of them could attract more economic activity and, hence, increase directly and indirectly local revenues such as property tax or the tax on economic activities.
Few studies to date have tested whether these factors also affect the fiscal capacity and, so, the analysis undertaken here also represents a new contribution to the testing of this relation ship.

Estimation of an equation of expenditure needs
The quantification of expenditure needs is based on econometric estimations of the av erage estimated impact of the cost variables (both, central and capital) on municipal expen diture per inhabitant 5 . In this way we can quantify the additional expenditure (measured with regard to the average) that these variables represent given the actual value that they take for each municipality under consideration (Solé-Ollé, 2001, 2006. In line with Bosch and Solé-Ollé (2005), the determinants of municipal expenditure can be expressed as: where g/pop is the per capita expenditure (current or total) 6 , Z is a vector of cost variables, F is a vector of fiscal variables that account for the resources available to the municipality, X is a vector of other variables that are taken into account and ε is the error term.

Estimation of an index of fiscal capacity
Estimating the impact of different variables on the fiscal capacity of a municipality re quires that this index is first calculated. This is by no means straightforward as there is more than one way of calculating it, and there are various difficulties that must be overcome.
An index of fiscal capacity must capture as closely as possible the fiscal capacity of the municipality. In order to do this various methods have been proposed to measure the poten tial revenues of a municipality: i) tax collection; ii) macroeconomic indicators, including mu nicipal GDP or municipal income; and iii) microeconomic indicators, including the Represen tative Tax System (RTS) 7 . More specifically, the RTS draws on the available information for the tax base (B) of any type of municipal revenue and the standard tax rate (t). With this in formation, and using the RTS approach, the index of fiscal capacity can be constructed as: where n is the number of different revenue sources considered and N is the total number of municipalities considered. Eq.
(2) can also be expressed as: where α j is the share (summing up to 1) of each type of revenue in a representative budget and b is the tax base expressed in per capita terms 8 . Given the budgetary structure of rev enues for large municipalities, for our calculations we opt to use just the property tax and the tax on economic activity. In 2008 these local taxes represent 22.9% of current revenues and 66.7% of tax revenues 9 .
In the case of the property tax (the main source of municipal revenue), there is a major shortcoming to our obtaining a true measure of the fiscal capacity. The measure of the tax base is the cadastral value of the properties located in the municipality; however, the cadas tral values are not homogenously updated for all municipalities. In other words, local gov ernments can endogenously decide when to revise cadastral values and, hence, to increase (or otherwise) the tax base of the property tax 10 . There are two ways in which we can over come this problem: by correcting the cadastral value or by implementing a correction via the tax rates. The first option is in practice particularly difficult to implement because we would require data on the evolution of the market prices of properties in all the municipalities and for all years or, alternatively, the evolution of data on housing prices (assuming that all prop erties evolve in a similar fashion to housing).
The second option is the correction of the tax base via tax rates based on the assumption that nominal tax rates differ between municipalities according to the time elapsed since the last updating of cadastral values. More specifically, we expect lower tax rates in those mu nicipalities that have most recently revised their cadastral values 11 . Indeed, in the appendix we provide evidence supporting this assumption 12 . In practical terms, we use tax rate infor mation to modify the property tax base: to each municipality i, and depending if it has re vised cadastral values in 2001 or not, we apply the average tax rate of all those municipali ties that have revised their cadastral values in the same period of time (before or after 2001).
In the case of the tax on economic activity, we use the municipal tax base (the so-called Cuotas mínimas), which is the estimated contents of each business, professional and artistic activity weighted by coefficients depending on the location and turnover of the activities present in the municipality. Therefore, we construct an index of fiscal capacity for large municipalities in Spain as presented in Eq. (3), with the shares (α) of the property tax and the tax on economic activi ty being equal to 84 and 16%, respectively. Once the measure has been constructed we can then estimate its possible determinants in the same way as for the expenditure needs present ed in Eq. (1).
where fc is the index of fiscal capacity of municipality i, F is a vector of fiscal variables, S is a vector of economic and socio-demographic variables (some of which may coincide with the expenditure needs equation) and u is the error term.

Simultaneous determination of needs and capacity
Finally, an important point to take into account to properly interpret the results obtained for the estimation of both the expenditure needs and the fiscal capacity equations, is that it might also be the case that the fiscal capacity determines the expenditure needs of a munic ipality or vice versa: i.e. a municipality might make its spending decisions in line with the fiscal capacity of its tax bases or (productive) expenditure might have a positive impact on the tax bases of the municipality and, hence, increase its fiscal capacity. Moreover, the afore mentioned endogeneity problem might also be seen as simultaneous decisions being taken regarding property tax rate and fiscal capacity (and spending). In order to address this issue we estimate a simultaneous equation model using 3-SLS as follows: In Eq. (5) municipal expenditure per inhabitant (g/pop), current and total, is both a de pendent variable and enters as an explanatory variable of the fiscal capacity. Similarly, fc is also introduced as a determinant of expenditures needs. Finally, fiscal capacity and the prop erty tax rate are also assumed to be decided simultaneously.

Data issues 4.1. Variables and data used for the estimation of expenditure needs
For the empirical estimation, all the municipalities that had over 75,000 inhabitants in 2008 are available, except Granada 13 ; hence, we have information for 91 large municipali ties in Spain. However, for reasons of data availability, we perform some of the estimations for a subsample of 86 municipalities, that is, excluding the municipalities of the Basque Country and Navarra. Table 1 presents summary statistics of all the variables used in the em pirical estimations and which are described below.

Dependent variable
The econometric estimations take the current municipal expenditure per inhabitant (gc/pop) as a dependent variable, as well as the total non-financial municipal expenditure per inhabitant (gt/pop). All municipal budgetary data come from the Spanish Ministry of Fi nance and Public Administration.

Explanatory variables Spillover effects, congestion and attractivity costs (central costs)
We compute the population resident within a 30-km radius of the municipality to proxy the potential spillover effect on its public services resulting from the (daily) use of non-res ident visitors as well as the congestion costs derived from these potential commuters, as pre viously explained. The municipal population data are obtained from the Spanish National Statistics Institute (INE in its Spanish acronym) for the year 2008 14 . More specifically, the variable constructed (potential_users) is a measure of the potential users (population within 30-km radius) in relation to the population of the central city interacted with a dummy vari able that takes a value of 1 if the municipality is considered a central city and 0 if the mu nicipality is peripheral and close to a central city 15 . By interacting potential users with this dummy variable means we only take into account those municipalities that have a power of attraction over their neighbouring municipalities.
To estimate the attractiveness cost on municipal expenditure of the number of tourists we use a municipal tourist index (i_tourism) calculated from the data provided by the "Spain Economic Yearbook" edited by La Caixa savings bank. The index is expressed as a percent age of the Spanish total and in relation to the resident population in the municipality (also expressed as a percentage of the Spanish total). This variable seeks to capture the intensive use of the resources of central municipalities that attract a large tourist population (because of such services as museums, cultural activities, beaches, etc.).
To approximate the congestion costs from commuters for consumption of private com mercial services we use the employ_serv variable that represents the number of employees in the municipality working in the service sector in relation to the municipality's total pop ulation. This variable is created using data from the Institute of Economic Research of Va lencia (IVIE in its Spanish acronym) and approximates the intensive use of the service sec tor in central municipalities that can attract the neighbouring population for consumer activities.

Concentration of social problems (central costs)
The number of immigrants allows us to capture the influence of social factors on munic ipal expenditure. The data are obtained from the official register of residents published by the INE. Specifically, we use the percentage of foreign nationals from outside the EU in re lation to the total population of the municipality (immg_non_eu). To account for other social factors we use the municipal unemployment rate in 2008 (unemp), which should also cap ture the level of economic activity in the municipality, and the illiterate population and the population without schooling in relation to the total population of the municipality (pop_il literate).

Costs of the factors (capital costs)
The variation in public sector salary costs between municipalities is introduced as a cap ital cost. This variable, w_cost 16 , is calculated as the total amount of wages in relation to the total number of workers in the local administration of the municipality 17 . Another variable, which likewise approximates the capital costs that a municipality incurs, is the number of workers employed in the central and autonomous administration (employ_public) that are present in the municipality. This variable is obtained by subtracting the number of workers in the local public administration (available from the official figures published by the Span ish Ministry of Public Administration -the Statistical Bulletin of Personnel Serving in the Public Administration) from the total number of workers in the municipality employed in the "public administration, defence and obligatory social security" (employment data taken from the two digit National Classification of Economic Activities published by IVIE). As with the other employment variables, they are calculated for the year 2006 18 .

Other expenditure needs (control variables)
In addition to the above variables that capture the features of the municipalities' capital and central statuses, we use other potential determinants of their expenditure needs as con trol variables. The urban area in relation to the population in 2008 (urban_area/pop) is taken into account to capture the effect of density on the cost of providing municipal services (for instance, refuse collection services). This variable is constructed with the area of buildings and building lots in each municipality (in square meters) in relation to its total population (data taken from the Cadastral Register).
Other variables included are those that refer to the characteristics of the municipal pop ulation that might have an impact on public expenditure: specifically, the population under 16 years of age in relation to the total population of the municipality (pop_under_16) 19 and the population over 65 years of age in relation to the total population of the municipality (pop_over_65). These variables, calculated for the year 2008, measure the impact of the presence in the municipality of two groups of individuals who are likely to be the main re ceivers of municipal public services. Another control variable is the population with higher education in relation to the total population of the municipality (pop_higher_educ). In this case, people with higher education seem to tend to return to central municipalities, with the consequence of increases in the housing prices, and also higher demand for cultural servic es and leisure amenities from the municipality. In this way we are able to control the level of municipal expenditure in relation to the characteristics of the resident population, as these variables capture resident needs in a particular municipality as regards one level of expendi ture or another. In this case, given that census data from the INE are required for their con struction, the variables refer to values for the year 2001.

Availability of revenue
The estimations of the equations for expenditure needs also require the use of variables associated with the revenues obtained by the municipality. The revenue from current trans fers in relation to the population of the municipality in 2008 (current_transf/pop) is used in the regressions to quantify the resources received by the municipality from other levels of government 20 . We assume that the municipalities that receive most transfers can spend more. Likewise, income from capital transfers in relation to the population of the municipal ity for 2008 (capital_transf/pop) as well as the revenue from their own taxes per capita (own_taxes/pop) are calculated.

Variables and data used for the estimation of fiscal capacity
In general, the variables that might affect the fiscal capacity of a municipality are prima rily fiscal variables and economic and socio-demographic variables. Below, we provide a de tailed description of the variables included in the estimation of Eq. (4).

Dependent variable
The dependent variable is the fiscal capacity index (fc) as described in detail in section 4.2 and in the appendix. We use data from the Cadastral Register which includes detailed informa tion for all municipalities (except those in the Basque Country and Navarra). The urban land registry files include variables that accurately identify the municipality, province and Au tonomous Community to which the municipality belongs. They also include information for the year of the last cadastral revision, the number of urban units, the assessed value (in thou sands of euros), the number of property tax receipts and taxable income (in thousands of euros). Finally, the dataset includes the tax rate applied by each municipality and the tax deductions.

Fiscal variables
In the short run, the main determinant of the fiscal capacity of a municipality is the tax rate on property, the main source of revenue for local governments with a fairly fixed tax base. Thus, as a determinant of fiscal capacity we include the property tax rate in the munic ipality in 2008 (property_tax_08). However, this rate is not exogenous to the fiscal capacity itself and, as a result, we might face problems of endogeneity. To overcome this problem we estimate the fiscal capacity equation using a 2-SLS procedure, using as instruments of the property tax rate in the municipality in 2008 the property tax rate in the previous year (prop erty_tax_07) and the other sources of revenue current_transf/pop, capital_transf/pop and own_taxes/pop 21 .

Economic and socio-demographic variables
To analyse other possible determinants of the fiscal capacity of a given municipality, we adhere to a strategy founded on the fact that, a priori, many of the variables presented as 80 NÚRIA BOSCH, MARTA ESPASA AND DANIEL MONTOLIO costs factors might also well have a positive impact on a municipality's revenues. As dis cussed above, certain variables might also attract economic activity to the municipality, such as potential users (potential_users), tourists (i_tourism), a predominant presence of the serv ice sector (employ_serv), immigration (immg_non_eu), higher wages (w_cost) or the pres ence of public servants from other administrations (employ_public).
Note that for some of the socio-demographic variables, we have no a priori expectations regarding their impact (if any) on the municipality's fiscal capacity. This is the case for those variables that refer to the characteristics of the municipal population (pop_under_16 and pop_over_65) 22 and those that capture the educational level of the population (pop_illiter ate and pop_higher_educ). In the case of unemployment (unemp), we expect a negative sign with respect to the fiscal capacity index, whereas we expect a positive impact of the variable capturing the urban surface area in relation to the population in 2008 (urban_area/pop).

Main results
The results of the estimation of the equation for expenditure needs are presented in Table  2 for the sample of 91 municipalities (columns 1 and 2) and for the sample of 85 municipal ities with more than 75,000 inhabitants (columns 3 and 4). As discussed above, we do not have access to cadastral data for the municipalities of the Basque Country or those of Navar ra and, hence, we are unable to use variables such as the urban surface area in relation to the population as a determinant of expenditure needs. The estimations have been carried out using two definitions of the dependent variable: current expenditure per capita and total non financial expenditure per capita 23 .   The results in Table 2 indicate that, in the case of factors related to central costs (i.e., spillover effects), the number of tourists and the number of employees in the service sector are positive, statistically significant determinants. Moreover, given that these variables are introduced in log format, they can be interpreted as elasticities. Hence, a one per cent increase in the tourist index increases total non-financial expenditure per capita by 0.042%, while a one percent increase in the share of employees engaged in the service sector increas es total non-financial expenditure per capita between 0.43 and 0.54%. By contrast, the po tential number of users of the municipality's public services (measured in terms of the pop ulation resident within a 30 km radius of the central city in relation to the total population of the city interacted with a dummy variable indicating the municipality's central status) is not statistically significant. The fact that this factor is not statistically significant is somehow ex pected given that, as previously pointed out, we do not analyze differential costs of large mu nicipalities in front of small municipalities. We study the effect of some factors across cen tral and capital municipalities, all of them above 75,000 inhabitants. Therefore, it seems that commuters are not a factor that differentiates the cost among those cities. Most probably, and as pointed out in Solé-Ollé (2001), commuters are a differential cost with respect small mu nicipalities, but we are not analyzing that effect. Similarly, immigration and unemployment variables do not seem to be significant determinants of expenditure needs either across large municipalities.
As for the factors representing capital costs, the two variables tested were found to be positive and statistically significant factors with an estimated impact of around 0.30%. These results indicate that the number of workers employed by the central and autonomous admin istrations located in the municipality, in addition to the salaries paid to the employees of the local administration, have a positive impact on municipal expenditure.
The results obtained for the variables capturing the municipalities' expenditure needs, including the share of population under 16, are statistically significant and positive with a very high estimated elasticity (0.8%) 24 . Similarly, the share of population over 65 is also positive, but this factor presents a lower level of significance and a lower point estimate (0.16%). Both variables reflecting the educational level of the population and the population density variable were found not to be statistically significant. In the case of the fiscal vari ables, the estimations conducted confirm that only revenue derived from current transfers has a positive impact on expenditure needs. Note that these results are fairly robust accord ing to the four models analysed (two dependent variables and two samples) and do not pres ent significant differences across the estimated models.
The results of the estimation of the equation for fiscal capacity using instrumental vari ables are shown in Table 3. First, the Sargan test reported at the bottom of Table 3 deter mines the validity of the instruments used. Second, as expected, the property tax rate has a negative impact on the fiscal capacity index (equal, remember, to 0.84 and 0.16% of the tax bases of the property tax and the tax on economic activity, respectively). Third, in the case of the impact of the variables that account for the centrality costs of municipalities we find, in line with the expenditure needs estimation, that the variables capturing the number of po tential users and the level of immigration in the municipality are not statistically significant, whereas tourism and the number of employees in the service sector are positive and signifi cant determinants of the fiscal capacity index of the municipality. Moreover, unemployment is found to be significant and to present a negative sign; that is, the higher the level of un employment in the municipality, the lower is its fiscal capacity -a result that, in principle, makes economic sense. Fourth, an inspection of the variables that capture a municipality's capital status shows that the results are not as significant as they are in the expenditure needs equation. Here, only the variable that approximates the salaries paid to employees in the local public sector is pos itive and slightly significant (0.17-0.20%). Fifth, as regards our findings for the other poten tial determinants of a municipality's fiscal capacity (in the main, the socio-demographic variables), the most notable result is the positive and very significant impact of population density (which presumably captures the positive impact on the tax base of the property tax) and the fact that, in general, the variables capturing the age structure and the level of educa tion of the population do not have a significant effect on the fiscal capacity index.
Finally, and given the (more than) likely simultaneity of expenditure needs, fiscal capac ity and property tax, we estimate the three equations simultaneously by means of 3-SLS (see, Eq. 5) to confirm, or not, the results obtained up to this juncture. The results, reported in Table 4, confirm our previous findings. Note that the results suggest a clear simultaneity be tween fiscal capacity and property tax rate, but this is by no means so clear when analysing needs and fiscal capacity. In the latter case, note that the fiscal capacity index is not a signif icant determinant of expenditure needs and, likewise, the current (or total non-financial) ex penditure per capita does not seem to affect the fiscal capacity index. Therefore, the results presented above can be considered valid, particularly given that the results for the fiscal ca pacity index (Table 4) were obtained by introducing the property tax rate as an explanatory variable and using 2-SLS to instrument the possible endogenous determination of tax rates and tax bases, especially in the case of the property tax.  In short, the results in Table 4 show that the centrality variables of tourism and employ ment in the service sector affect both expenditure needs and fiscal capacity. At the same time, the number of potential users and immigration are not significant factors, although the unemployment rate does have a negative impact on the fiscal capacity of municipalities. The capital status variables seem to have a greater impact on expenditure needs than they do on fiscal capacity, a finding that is similar for the socio-demographic variables which, as ex pected, seem to have a greater effect on expenditure needs.

Conclusions
This study has determined the costs and benefits for Spain's large municipalities (de fined as those with a population over 75,000 inhabitants) of being a capital or central munic ipality. We bring a series of rigorously tested results to what remains an on-going and high ly ideological debate in Spain regarding the financing of local governments, and especially the governments of large municipalities. While it is indisputable that these municipalities face higher costs given their particular nature, it is equally true that they also obtain addi tional economic benefits that need to be evaluated with these costs to obtain a complete pic ture of the public finances of these governments, especially if these questions are to be taken into account in possible reforms of the local financing system.
Our results indicate that there are indeed significant costs factors associated with the central/capital nature of a municipality that influence the local per capita expenditure of these municipalities. These factors are essentially related to the existence of spillover effects (central costs) in central cities, measured by the number of tourists and the weight of the service sector, and the existence of capital costs associated with the amount of employment in other levels of government and the salaries paid in the local public sector. Our results also show that the socioeconomic structure can influence expenditure needs (primarily, the share of the population under 16 and over 65). Interestingly, we also find that the variables captur ing the central status of municipalities result in an increase in their fiscal capacity, but the variables capturing their status as a capital do not seem to be determinants of this fiscal ca pacity.
Thus, the main conclusion to be drawn from this study is that the centrality characteris tics of large municipalities in Spain increase their expenditure needs; yet, they have a posi tive impact on their fiscal capacity. As such, centrality implies certain additional costs, but it also brings some benefits. By contrast, the capital status of a municipality seems to gener ate mainly costs; that is, being a capital increases expenditure needs. Consequently, the local financing system for the large municipalities in Spain needs to give greater consideration to factors linked to capital characteristics, since the factors associated with the centrality char acteristics of municipalities appear to be offset by their greater fiscal capacity and, hence, do not, in principle, need to be explicitly recognised by the financing system.

Notes
1. The Spanish Law on Local Authorities draws a distinction between municipalities on the basis of population size; however, this distinction differs according to the specific area being regulated (organizational, compe tences or financing). Referring to competences Article 26 of this Law lists the minimum competences that can be required to all municipalities (compulsory competences). These are public lighting, cemeteries, waste col lection, street cleaning, the domestic supply of drinking water, drains and sewers, access roads to centres of population, paving of public roads and control of food and beverages. The same Law divides municipalities into population ranges, the highest of which is municipalities with a population over 50,000 inhabitants and, based on those ranges, municipalities are designated additional competences. These are the so-called delegat ed competences. Moreover, the municipalities can also provide other additional services if they decide to do it (discretional competences). As for financing, the same model is applied to all the municipalities, with some adjustments being made according to population size and delegated competences, affecting above all tax rates and unconditional transfers. Nevertheless, the assignment of delegated responsibilities is not properly covered by adequate concomitant resources. The municipalities do not receive specific resources for the additional/dis cretional competences provided to citizens.
2. Throughout the paper we use, for our convenience, the term "capital cost" as a synonym of "costs of being a capital".