Andalusia's social service system aims to guarantee social protection. Community Social Service Centers (CSSCs) are crucial access points, managed by local and supra-local entities. Funding is distributed based on population (90%), dependent population (2%), surface area (2%), scattered population (3%), and large cities (3%), with a relative poverty index correction (33.34%). This distribution focuses on population size and geography, considering social disadvantage directly (ad hoc poverty index) and indirectly (age structure). However, the poverty index is limited, not fully capturing capability deprivation. The geographical distribution of CSSC funding is expected to be heterogeneous due to varying population, geographic characteristics, and uneven deprivation distribution. A spillover effect in spatial distribution of public expenditure is also anticipated. Spatial data analysis, while common in fields like epidemiology and economics, is less prevalent in public service financing. This study uses spatial data analysis to examine the geographical distribution of CSSC funding per inhabitant in Andalusia in 2019 and explores its relationship with demographic and socioeconomic indicators.

Existing research has examined spatial aspects of public service financing in various contexts, including school funding in China (Xiao and Liu 2014), disability expenditures in Italy (Agovino and Parodi 2016), social expenditure in the EU (Miśkiewicz-Nawrocka and Zeug-Žebro 2019), welfare and housing benefits in the UK (Hamnett 2009), and public expenditures and social services spending in Spain (López et al. 2017; Gallego Valadés et al. 2023). These studies demonstrate the utility of spatial analysis in understanding the geographical patterns and determinants of public spending.

The study used the catchment areas of Andalusia's 250 CSSCs (Social Service Zones or SSZs) as analysis units (n=184 after data cleaning and aggregation). Data on funding (from the Andalusian Equality, Social Policies and Reconciliation Department) and demographic and socioeconomic indicators (from the Institute of Statistics and Cartography of Andalusia) were collected for 2019. Data were aggregated at the municipal level for smaller SSZs. The dependent variable was per capita CSSC funding (€), and explanatory variables included sex ratio, ageing index, dependency index, emigration rate, immigration rate, unemployment rate, population density, and employment rate in the primary sector. Population density and the dependent variable were logarithmically transformed to meet regression assumptions. Global Moran's I was used to assess spatial autocorrelation, followed by Local Moran's I and Local Getis & Ord's G* to identify spatial clusters (hot and cold spots). Ordinary Least Squares (OLS) regression was initially used, followed by spatial lag and error regressions to address spatial autocorrelation using Anselin's model selection rule. A distance matrix with second-order queen contiguity weights defined spatial relationships.

Global Moran's I indicated positive spatial autocorrelation (0.261, p<0.01) in CSSC funding. Local Moran's I and Local Getis & Ord's G* consistently identified similar hot spots (northern Huelva, central Cordova, northwestern Almeria, and northern Granada) and cold spots (around Seville, Malaga, Granada, and Cadiz). These patterns suggest a rural/urban dichotomy. Model 1 (OLS regression with all variables) showed collinearity issues and some non-significant variables. Model 2 (OLS regression with significant variables) included ageing index, primary sector employment, immigration rate, and population density. Spatial lag regression models (Models 3 and 4) addressed spatial autocorrelation. Model 4 (spatial lag regression without non-significant variables) showed that a one-unit increase in the ageing index and primary sector employment led to a 0.14% and 0.05% increase, respectively, in per capita funding, while a one-unit increase in foreign population and population density led to decreases of 0.31% and 0.06%, respectively. The dependency index, a criterion in funding distribution regulations, was not a significant predictor.

The findings challenge the assumption that funding distribution aligns perfectly with the needs of high-population and high-dependency areas. The urban/rural pattern in hot and cold spots suggests potential organizational or practice differences. Comparing spatial clusters with a comprehensive deprivation index (Rodero-Cosano et al. 2014) revealed a correspondence, particularly for unemployment, income, and housing deprivation domains. Spatial lag regression, implying spillover effects, supports the idea that funding in one area influences neighboring areas. The significant association of funding with primary sector employment highlights the impact of rurality, while the negative association with foreign population suggests the presence of populations outside the formal social safety net, particularly illegal migrants and perhaps some wealthier foreign residents who don’t use social services. The study provides valuable information for evidence-informed social service policy and planning, specifically highlighting areas needing further assessment (e.g., the cold spots in disadvantaged Cadiz Province).

The geographical distribution of social service funding in Andalusia is not random. Hot spots are concentrated in rural areas, while cold spots are predominantly urban. Funding distribution criteria defined by regulations are not entirely sufficient predictors of funding levels. Other variables like primary sector employment and foreign population are significant factors. Spatial analysis offers a valuable tool for assessing and improving social service resource allocation. Future research should address the limitations, particularly by using longitudinal data and sub-municipal level data to capture greater detail, and include other relevant variables for a comprehensive understanding.

The study's limitations include the use of 2019 data only (limiting longitudinal analysis), the inability to analyze sub-municipal areas due to data limitations, and the lack of financial data for smaller municipalities (precluding the inclusion of tax revenue and transfer variables). The economic context has changed significantly since 2019 due to the COVID-19 pandemic and the Ukraine war, though these impacts on social service funding have been relatively less pronounced than in prior crises.