Mental health expenditure in England: A spatial panel approach

https://doi.org/10.1016/j.jhealeco.2006.12.008Get rights and content

Abstract

We empirically investigate the determinants of local authority mental health expenditure in England. We adopt a reduced form demand and supply model, extended to incorporate possible interaction among authorities, as well as unobserved heterogeneity. The model is estimated using an annual panel dataset that allows us to explore both time-series and cross-municipality variation in mental health expenditure. Results are consistent with some degree of interdependence between neighbouring municipalities in spending decisions. This first attempt to apply spatial panels in investigating health expenditure offers insights and raises new questions.

Introduction

Recent years have witnessed a gradual rebalancing of responsibility from National Health Service (NHS) to local authority budgets in financing mental health services in the United Kingdom. Local authorities have assumed responsibility for funding placements, such as in nursing homes, residential care homes and supported accommodation, that previously would have been paid for by the NHS or covered by social security transfers. As a consequence, in the last few years, expenditure on mental health services by municipalities has grown faster than NHS spending, accounting today for roughly one fifth of its size (Knapp et al., in press). This proportion varies substantially across the country. For instance, social services spending on mental health by local authority in London ranges from 23% to 79% of health service spending (Aziz et al., 2003).

Two empirical strands of health economics literature have been developed to explore variations in mental health expenditure and costs. One strand uses data on individual people, perhaps collected in controlled experiments, to study how mental health costs vary in relation to characteristics such as diagnosis, illness severity, physical health, and socio-economic characteristics. A second strand focuses on area-level characteristics, including some that are recognized as proxies for mental health needs, such as ethnic mix and deprivation, in examining variations in mental health spending. The tendency to use risk factors rather then direct measures of needs follows from the lack of information on the prevalence of mental illness at area level, and from recognition that measures of treated prevalence depend on service supply as well as need. The work in this paper fits within the second strand of literature, taking as statistical unit the administrative area (local authority or council), and trying to explain the geographical disparity in mental health spending in relation to a set of risk factors. The approach taken is to extend the traditional demand and supply framework to allow for possible interaction among authorities as an additional source of expenditure variation. If it exists, interaction of this kind could have implications for the methods used by central government to monitor and influence local performance across the country and over time, and perhaps also to allocate resources across municipalities (cf. Carr-Hill et al., 1999, Jarman et al., 1992, Carr-Hill et al., 1994, Kelly and Jones, 1995, Smith et al., 1996, Lelliot et al., 1996).

A substantial proportion of English local authorities’ revenue, roughly two-thirds, comes from central government grant and uniform business rates; the residual part is locally generated from council tax, sales, fees and charges. Central government grants are distributed via a formula that tries to identify the spending requirement of an authority if it were to adopt a common level of services, given differences in needs and exogenously determined input prices. The allocation is further adjusted by an equalization factor, to account for possible distortions generated by disparities in the council tax revenue. Central government assesses personal social services needs separately for each client group, and a number of formulae are used to estimate the (adjusted) spending requirements. However, local authorities have considerable autonomy in allocating these resources from central government, both to and among the various local services (education, housing, leisure, community resources as well as social services), prioritizing particular areas and client groups in line with local interpretations of need, local preferences and so on. Therefore, authorities’ actual spending will at least partly, and perhaps predominantly, reflect local policies rather than the standard spending assessment (SSA) by central government. In particular, expenditure is likely to depend on the risk factors perceived by the authority as well as local preference and income, competition for other services, and local policies (Carr-Hill et al., 1999).

In this study we adopt a general approach where expenditure choices of municipalities are examined by reference not only to needs and supply determinants, but also to policy interdependence and geographical concentration of unobservable risk factors. A recent strand of literature in public economics has analyzed the role of interaction among policy makers. In particular, several studies from a range of countries have investigated interdependency between local authorities in setting tax rates Ladd, 1992, Besley and Case, 1995, Revelli, 2001, in deciding standard and regulatory measures Brueckner, 2000, Fredriksson and Millimet, 2002, and in determining expenditure levels Case et al., 1993, Kelejian and Robinson, 1993, Bivand and Szymanski, 1997, Revelli, 2002. In contrast, the effect on local decisions of expenditure behaviour in neighbouring areas has been almost completely neglected in health economics. In the next section we therefore provide a series of arguments that suggest why there might be strategic interactions among municipalities. Our line of reasoning will build on recent economic theories on interaction among individuals Manski, 1993, Manski, 2000, the literature on public economics (Brueckner, 2000), and intuition from studying the English health and social care systems (e.g. Moscone and Knapp, 2005). We provide an econometric framework that models mental health expenditure as a function of risk factors, supply determinants, spending decisions of neighbouring municipalities, and time trends. This model is then estimated using a yearly panel dataset that allows both time-series and cross-municipality variation to be explored in per-capita mental health expenditure by English local authorities.

The rest of the paper is organized as follows. Section 2 discusses some possible reasons why local policy makers may interact. Section 3 synthesizes existing empirical evidences on interaction between municipalities provided by a recent strand of literature on public economics. Section 4 introduces the empirical models of health expenditure. Section 5 presents the data. Section 6 focuses on the estimation of our models, and emphasizes some limitations of the study. Finally, Section 7 closes with some concluding remarks.

Section snippets

Interaction and reaction in mental health policy

Local interaction models in economics can be defined as models in which each individual’s behaviour is affected by the characteristics or behaviour of other people (Manski, 1993). A major assumption of these models is that individuals interact locally with a finite subset of the population defined by a social, economic or geographical distance metric. Their interaction generates spillovers and externalities, and can lead to an emergent collective behaviour and aggregate pattern that empirically

Evidence of interaction in public spending

There exists an ample literature studying cross-sectional variation of public spending in relation to various economic, political, and demographic factors. Differences in public expenditure across municipalities are traditionally explained by differences in local characteristics such as per-capita income, taxes, socio-economic attributes and political structure Foster et al., 1980, Wildasin, 1986. Recent literature in public economics has pointed out that an important element explaining

Empirical models

We can suggest two econometric specifications that allow the representation and testing of the hypotheses introduced earlier.

The demonstrative or exemplar effects, explained by the benchmarking of each municipality performance against the performance of other authorities, might induce endogeneity of spending choices. The statistical framework for this effect is a generalization of models usually presented in the public economics literature, where the value of the dependent variable for one

Data

Our empirical study follows the 150 English local authorities (also referred to as councils with social services responsibility) over a period of 6 years, from 1998 to 2003. The dataset is drawn from several sources. The dependent variable is net personal social services expenditure for adults aged under 65 years with mental health problems,3

Empirical analysis

Fig. 2 maps the geographical distribution of per-capita expenditure averaged over the 6 years. The higher level of mental health spending, indicated in the map by darker colour, shows important concentrations across space. It is evident that the variable tends to distribute in clusters, with the highest concentrations in metropolitan areas such as Greater London, Greater Manchester and Birmingham. Per-capita expenditure is also concentrated in the areas contiguous to these clusters, though with

Conclusions

In this paper we have explored the possibility that policy makers’ choices concerning mental health expenditure are interdependent. In particular, we offered a range of motivations underlying possible interactions between local authorities when deciding their own level of social care spending on mental health. This led us to suggest two econometric specifications to reflect these hypothesis. Using data on English personal social service mental health expenditure by local authority, we explored

Acknowledgments

We would like to thank participants at the 14th Workshop in Econometrics and Health Economics in Dublin, Mario Tosetti and two anonymous referees for valuable comments and suggestions.

References (58)

  • F. Revelli

    Performance rating and yardstick competition in social service provision

    Journal of Public Economics

    (2006)
  • J. Tavares

    Does the right or left matter? Cabinets, credibility and adjustments

    Journal of Public Economics

    (2004)
  • L. Anselin

    Spatial Econometrics: Methods and Models

    (1988)
  • L. Anselin

    Spatial econometrics

  • G. Arbia

    Introductory Spatial Econometrics with Applications to Regional Convergence

    (2006)
  • F. Aziz et al.

    Financing Mental Health Services in London: Central Funding and Local Expenditure

    (2003)
  • T.J. Besley et al.

    Incumbent behavior: vote seeking, tax setting and yardstick competition

    American Economic Review

    (1995)
  • Brock, W., Durlauf, S., 2000. Interaction Based Models. National Bureau of Economic Research, Technical Working...
  • J.K. Brueckner

    Welfare reform and the race to the bottom: theory and evidence

    Southern Economic Journal

    (2000)
  • J.K. Brueckner

    Strategic interaction among governments: an overview of empirical studies

    International Regional Science Review

    (2003)
  • R.A. Carr-Hill et al.

    Allocating resources to health authorities: development of method for small area analysis of use of inpatient services

    British Journal of Psychiatry

    (1994)
  • R.A. Carr-Hill et al.

    The determinants of expenditure on childrens’ personal social services

    British Journal of Social Work

    (1999)
  • A.C. Case

    Spatial pattern in household demand

    Econometrica

    (1991)
  • A.C. Case et al.

    Budget spillovers and fiscal policy interdependence: evidence from the states

    Journal of Public Economics

    (1993)
  • T.G. Conley et al.

    Socio-economic distance and spatial patterns in unemployment

    Journal of Applied Econometrics

    (2002)
  • L. Crivelli et al.

    Federalism and regional health expenditures: an empirical analysis of Swiss cantons

    Health Economics

    (2006)
  • J.P. Elhorst

    Specification and estimation of spatial panel data models

    International Regional Science Review

    (2003)
  • C. Foster et al.

    Local Government Finance in a Unitary State

    (1980)
  • G.R. Glover et al.

    A needs index for mental health care

    Social Psychiatry and Psychiatric Epidemiology

    (1998)
  • Cited by (65)

    • Spatial interaction effect in renewable energy consumption in sub-Saharan Africa

      2022, Renewable Energy
      Citation Excerpt :

      In the SSA context, there various factors can contribute to spatial spill-over effects in REC. Due to the “yardstick competition” governments in SSA may tend to match the renewable energy promotion policies of close countries [38], or due to demonstration effects, some governments may follow the policies of other countries [39]. Second, in SSA, there is high concentration of hydroelectric resources near-borders and this may have an important bearing on spatial spill-over effects [40].

    • The fiscal impact of health care expenditure: Evidence from the OECD countries

      2020, Economic Analysis and Policy
      Citation Excerpt :

      On the other hand, factors such as the environment, diet and lifestyle of the populations might have raised the prevalence of certain pathologies which will subsequently create considerable geographical concentrations of these pathologies at the national level (Haining and Haining, 2003). The existence of unobservable risk factors of HCE has also been described in the spatial econometrics literature (Moscone and Knapp, 2005; Revelli, 2006; Moscone et al., 2007). Few are the studies though that combine health and economic variables in a spatial context.

    View all citing articles on Scopus
    View full text