The Fragmented Welfare State: Explaining Local Variations in Services for Older People

Abstract

Much research focusing on the welfare state is based on the assumption that welfare regimes are homogenous entities. This idea is supported by studies analysing cash benefits. In the area of welfare services, however, local governments in most countries have some autonomy regarding policy formation as well as the design and implementation of policies. In practice, substantial local differences exist with regard to the provision of welfare services, which in turn challenge our conception of nation-wide homogenous welfare state regimes. This paper examines the factors causing marked differences in local government spending in the provision of care for older people in Denmark. The conclusion is that the wealth of the municipality, local demographics and privatisation can explain about 48 per cent of the differences in local government spending. Political factors such as the ‘colour’ of local government have no explanatory power, while a high percentage of women in municipal councils appears to have a slightly negative effect on spending.

Introduction

Which factors influence the qualitative and quantitative development of the welfare state? This is one of the central questions in welfare research, and one of the most well-known responses to this question was provided by Gøsta Esping-Andersen in his seminal book, The Three Worlds of Welfare Capitalism, which was published in 1990. According to Esping-Andersen, party politics and ideology are epitomised as ‘politics matter’, a decisive driving force behind the development of the welfare state. This leads Esping-Andersen to draw a distinction between liberal, conservative and social democratic welfare regimes. Esping-Andersen's theoretical insights were anchored in analyses of the cash benefits provided by the welfare state, and his approach has been criticised for neglecting the provision of services (see for example Bambra, 2005; Jordan, 2005), including care for older people, which poses a major challenge to most late-modern societies due to ageing populations (e.g. Ungerson, 2005).

More recently, the welfare regime approach has been reconfirmed in the area of welfare services. Jensen (2008) and Bambra (2005) have thus argued that social care services conform to Esping-Andersen's regime typology, albeit with a number of minor modifications. This reconfirmation of Esping-Andersen's welfare regime approach within the area of welfare services is based on an analysis of aggregate (national) public expenditures on social care, and one relevant question is whether such an analytical strategy is appropriate. In contrast to cash benefits, the provision of services by the welfare state largely depends on local political decisions and priorities. In most European countries, welfare services are under regional or municipal authority, and in some countries local authorities have considerable autonomy with respect to determining the quality and quantity of their services, which allows for a considerable scope of inequalities and possible inequities across municipal boundaries (Davey et al., 2006; Rummery, 2009; Trydegård and Thorslund, 2001).

In Denmark, strong local autonomy has given rise to substantial variations in the distribution of welfare services from municipality to municipality. In some instances, such intra-national variations may be even greater than the variations between countries.¹ In the case of care for older people in Denmark, for instance, municipal spending in 2005 varied from DKK 25,148 to DKK 61,651 per older adult (sixty-five-plus) per year (between about €3,350 and €8,220), while the number of hours of long-term home care delivered varied from 2.0 to 13.6 hours per older adult per week. Such differences in the provision of care for older adults undermine our notion of the existence of homogenous and uniform welfare state regimes. It may actually be more appropriate to talk about a multitude of different ‘welfare municipalities’ (Kröger, 1997, 2009) as opposed to a single, uniform welfare state.

The variations in care for older people from municipality to municipality might perhaps be explained by the responsiveness of decentralised service states to the wants and needs of the population, just as they might be the result of different local traditions and political priorities. But we do not know. Knowledge of the actual reasons for municipal differences in care for older people is very limited. Very few studies have actually investigated why such major local differences exist in the provision of care for seniors. Some Swedish studies have attempted to cast light on the issue, but with varying results. Davey et al. (2006) argue that local variations in the delivery of public home help can be ascribed to differences in needs across municipalities, while Savla et al. (2008) argue that the coverage rates of home help and institutional care depends on the proportion of the population age sixty-five and older, the proportion of unmarried elders, the median income in the municipality and level of municipal spending. Berg et al. (1993) failed to explain the differences in the municipal coverage of care and spending per citizen, simply concluding that local taxes play a role and that local differences are longstanding. Yet another Swedish study (see e.g. Trydegaard and Thorslund, 2001, 2010), which analysed the degree of coverage for residents over age eighty, were only able to account for 15 per cent of the variation in spending using multivariate models and suggested that the explanation for variations in the municipal provision of care for the older adult lies in historical patterns of care provision. Hansen and Hjort (1998), using Danish data, selected spending on older people as their dependent variable, and found that the following factors have a significant effect on spending on older adults: the share of single older adults, the share of over seventies in relation to the entire municipal population, the labour force activity among women aged fifty to sixty-six, growth in the population of older people and the share of over seventies who were admitted to hospital in 1995. More specifically, Nielsen and Andersen (2006) have carried out a study of the ten Danish municipalities with the highest and lowest spending, respectively. They found that the spending among high-spending municipalities is a deliberate political priority, which can be justified with service improvements in home help activities, such as, for example, cleaning. All in all, however, one is left with the impression, as Mouritzen (1991: 106) noted in 1991, that municipal spending on care for older adults is difficult to explain, and no conclusive evidence has been established about what drives such variations.

The aim of this paper is to contribute to a better, more comprehensive understanding of the factors which explain variations in cross-municipal spending on elder care in a social policy perspective. Drawing on assumptions embedded in the social policy tradition, we will analyse demographic, political, structural, institutional and path-dependency explanatory factors using linear regression as well as multi-level analyses. How local welfare states meet the need for the care of older citizens has been an under-researched field, and the reason for this is most likely that the data, including expenditure data on elder care in local welfare states, have been difficult to get or does not exist. In this study, however, we have been able to draw on comprehensive Danish register data suited for the purpose of this study.

The paper is structured as follows: first, we provide a brief account of how care for older adults is organised in Denmark. A theoretically grounded account of the choice of the dependent and independent variables will then be presented and the empirical data will be accounted for. We then present our analysis of the data, before offering a final summary and conclusion.

Care for older adults in Denmark

In Denmark, the municipalities play a central role in the provision of care to older citizens. The relevant legislation – the Social Services Act – stipulates that municipalities bear the responsibility for care for frail older people as the municipality is legally obliged to facilitate the help and support that older people need. Two forms of care for older adults exist: (1) institutional care in care homes owned and operated by the municipalities,² and (2) home help in the form of practical (e.g. vacuum cleaning) as well as personal (e.g. showering, getting dressed, assistance with the eating of meals) care services delivered in the private homes of older citizens.

Although care for older people is a citizen's right delivered by the municipality at no charge, the Social Services Act emphasises that the ‘[h]elp provided according to this law is premised on the principle that the individual in need and his or her family takes responsibility for their own situation’. This basic principle has been interpreted to mean that, as a rule, practical home care will not be extended to older adults living with healthy and fit spouses, partners or relatives living under the same roof, thereby de facto shifting the practical care responsibility onto the immediate family. This has been controlled for in the empirical analysis by including a variable of the number of older people living alone. It should be noted, however, that personal care will be provided by the municipality if needed, despite the presence of a capable and healthy partner.

It must also be emphasised that the practical care responsibility does not extend to the immediate family members if they are not cohabitating with the individual in need of practical home help. In effect, there has been a strong tendency in Denmark towards the formalisation of care for older adults and towards such care becoming a public matter. Most older adults not living alone live together with a partner; intergenerational cohabitation is atypical in Denmark, and family members (daughters, sons or other relatives) rarely provide personal care to frail older adults in Denmark (cf. Rauch, 2008; Rostgaard et al., 2011; Rostgaard and Szebehely, 2012). That is, the informal care sector is small relative to other countries with less developed formal care sectors. In 2005, around 3.7 per cent of the total (formal) labour force were employed in caring for older adults (Jensen and Rathlev, 2009).

Although the municipalities bear the primary responsibility for running residential homes, nursing homes and providing personal and practical care, they actually have considerable autonomy when it comes to the actual extent and substance of care as well as the situations in which one is entitled to receive care (cf. Jensen et al., 2004). This means that the municipal council has extensive authority to define the level of municipal services (e.g. quality, coverage) for older people. Still, the municipalities are obliged to account for their political priorities to the municipal residents, and the municipalities have been required to work out quality standards for municipal services since 1999. These standards must state the level of service that the municipal council has set for the municipality, just as the municipalities are required to install particular control mechanisms aimed at ensuring the fulfilment of the quality standards.

More generally, a strong local autonomy in the provision of welfare services, such as child and elder care, schools etc., intersects with a high degree of political decentralisation. In Denmark there is a solid connection between responsibility for decision-making and financing service provisions. Thus, most service provisions are financed entirely by the municipalities. Danish municipalities impose taxes, which, in combination with user fees, are the most significant source of municipal revenue. Taxes and user fees accounted for 56 per cent of the combined revenues for municipalities and counties in 2002 (Strukturkommissionen, 2004). Other sources of revenue include operating and installation revenues, state refunds, borrowing and conditional grants (from the state). Conditional grants aim to level out the conditions between municipalities associated with spending needs, tax bases and so forth. On the face of it, there are rather considerable differences between municipalities with regard to their respective tax bases and spending levels. In 2002, the average tax base per capita which is subject to taxation in Danish municipalities fluctuated between DKK 94,413 (approx. €12,693) and DKK 279,303 (approx. €3,755), while average spending per capita fluctuated between DKK 25,506 (approx. €3,429) and DKK 36,503 (approx. €4,908) (Strukturkommissionen, 2004).

To increase efficiency in the production of welfare services and thereby lowering costs municipalities have several times been merged to achieve economies of scale. The number of municipalities in Denmark was reduced from about 1,100 to 275 in 1970 and 271 in 2003. The number of residents varied considerably from municipality to municipality. As of 1 January 2001, the number of inhabitants ranged between 2,266 and 499,148, the average size being around 19,000 (Indenrigs- og Sundhedsministeriet, 2002; Mouritzen, 2003). A major reform carried out in 2007 reduced the number of municipalities to ninety-eight. In 2007, the total population in Denmark was 5.4 million.

Recent reforms in elder care

Trends towards formalisation of elder care have been strong in Denmark. In the formative years of the Danish elder care sector, emphasis was on the provision of residential and old people's homes. By the end of the 1980s, however, the public sector and formal care for older people was restructured. Since then, Denmark has been marked by a trend towards ‘de-institutionalisation’, decentralisation, flexibility – and even individualisation – in the provision of public services for older adults who are no longer able to fully care for themselves (Rostgaard, 2006); where possible, home care − that is personal assistance and care and assistance with practical tasks in the home and residential homes where conditions resemble ‘normal’ housing − has to the extent possible replaced institutional care in care homes. This shift towards de-institutionalisation has been accompanied by a trend towards flexibility and privatisation in the provision of care for older people. Until 1 January 2003, municipalities could freely choose whether they wanted to opt for a private contractor or whether all of the staff in the ‘home care sector’ should be municipally employed. Since 1 January 2003, however, individuals requiring home care have been free to choose from among different home care providers (public or private). However, not just anyone can establish a private company and provide home care. Such companies must be recognised by the municipality, and such recognition is based on quality requirements outlined by the municipality. Private companies must thus be able to deliver the level of service that the municipal council has set for the municipality; that is, they must be able to live up to the quality standards worked out by the municipality. As such, there is no difference in the quality standards between publicly or privately provided home help, and private contractors receive payment corresponding to what it costs the municipality to provide the same service; that is, private contractors cannot compete on the price for care. The municipality pays directly to the private contractor for services delivered – not to the care receiver. In other words, the scheme is not based on cash-for-care principles.

Other things being equal, one might expect private home care to be more expensive for the municipality than public home care. As already mentioned, the municipality pays the same to the private entity as it would cost the municipality to provide the same service. But in addition to this, the municipality must bear costs in connection with the documents for tender, just as the municipality is also responsible for some of the paperwork associated with the delivery of for-profit private home care. However, it is worth noting that the rationale behind the introduction of the free-choice arrangements under the Liberal-led, right-of-centre government in 2003 was not to achieve savings. Instead, free-choice arrangements were aimed at constructing a private market for the provision of home care, as the reform was established on liberal and libertarian values, epitomised as the ‘freedom to choose’, which was believed to empower older people by providing them with opportunities to choose. The advantage for older people to be able to choose a private contractor is that private contractors are allowed to offer additional services, such as gardening or dog-walking, to the individual receiving the care. The care receivers themselves must pay directly for such additional services to the private contractor.

Variables: theory and data

This section presents the dependent and independent variables and their theoretical bases. The basis for the data material will also be accounted for.

The dependent variable

The dependent variable in this paper is the annual municipal care expenses per older adult, with 2005 selected as the base year. This is owing to an interest in applying data pre-dating the 2007 municipal reform, at which time many municipalities were amalgamated. As an outcome of the municipal reform, probable relationships between municipal spending and our independent variables, such as the share of the labour mandates, are possibly erased. For instance, if three different political coloured municipalities are merged, for example a social democratic, conservative and liberal one, the distinct impact of different political parties may disappear. However, it is worth mentioning that the spending differences between the municipalities remain considerable after 2007. In 2010, for instance, spending in the lowest spending municipality was DKK 34,631 per older adult per year, while spending in the highest spending municipality was DKK 78,407 (De Kommunale Nøgletal). Average spending was DKK 47,330 and the standard deviation DKK 6,564. Still, differences in spending tend to decrease. The standard deviation was greater in 1993, 1999 and 2005 as compared with 2010 (De Kommunale Nøgletal).³

Spending on older people is officially recorded as ‘Older adults expenses (net) per 65+/67+ year-old’, which in 2005 ranged from DKK 25,148 to DKK 61,651 (De Kommunale Nøgletal). ‘Net’ means that it is the gross expenses minus income, including the conditional grants from the state, rent payments and so forth. This spending covers the expenses for care for older adults (including home care) and disabled younger people, rehabilitation, housing for older people, contact persons and an accompanying scheme, preventive measures, nursing homes and sheltered housing as well as aids and other equipment for older adults. As the retirement age was reduced from sixty-seven to sixty-five as of 1 July 2004, the figures since 2004 are for expenses for those over age sixty-five, whereas they were for those over age sixty-seven prior to 2004.

As Nielsen and Andersen (2006) and others have pointed out, some of the differences from municipality to municipality in the spending on older people are due to variations in accounting practices. As regards the overall accounts, however, only a limited percentage of these variations can be explained by accounting differences due to state instructions regarding municipal accounting.

The net expenses of municipalities on older people include, as indicated, the expenses for the disabled and seriously ill persons under age 65, which might undermine the accuracy of the expenditures data on elder care. However, the number of younger disabled persons requiring care makes up a relative small proportion of the total number of older adults subject to care,⁴ and the impact on such spending is further limited as 50 per cent of the costs for care to younger people are reimbursed by the state (Hansen and Hjorth, 1998: 92).

Independent variables

In the following analysis, we will examine a number of potentially relevant explanatory factors. We have included political, structural, institutional and demographic factors as independent variables. These factors are summarised in Table 1 at the end of this section.

TABLE 1. Independent variables (2005)

Note: ¹Farum has had a massive debt due to extraordinary conditions.

Municipal structure and composition of the population

Thus far, studies of local variations in the provision of care for older people in Denmark indicate that the structure of the municipality and composition of the population effect the spending on care (see e.g. Hansen and Hjort, 1998; Lolle, 2000; Mouritzen, 1991). There is, however, no straight line between the composition of the population and spending. A large share of older people in a given municipality may thus help reduce the spending per older person for at least two reasons: first, spending on older people is a major component in the budget for most municipalities, and the greater the share of older people, the more pressure on the budget, which will often lead to reducing the spending on each older adult. Second, if older people make up a large share of the population in the municipality, then there may be economies of scale in such provision, which in turn may help reduce the spending per older adult. One factor which tends to bring about increasing expenditures is a high mean age of older people, as older seniors require more care than younger seniors. We expect that there will be an effect on spending levels per older person in the municipality related to the share of eighty-five-plus of total population aged sixty-five-plus.

The denser the population in a municipality, the less costly one would expect each older person to be. For instance, in a densely populated area, transportation costs in relation to items such as home care and food delivery may be expected to be lower than in less densely populated areas. We therefore analyse how factors such as (1) the overall number of people living in the municipality, (2) the municipal area in square kilometres and (3) population density effect on the spending on elder care. The degree of urbanisation could also serve as a measure of the ‘urban culture’, e.g. a more demanding population, which could give rise to other spending patterns than in smaller rural municipalities (cf. Birgersson, 1971).

Several studies indicate a dynamic link between wealth and health (see, for example, Feinstein, 1993; Smith, 1999); that is, wealth is associated with positive health outcomes in terms of mortality or morbidity, which in turn may reduce the need for care. In as much as wealth is associated with owning one's own home as opposed to renting, it is plausible to assume that the greater the number of homeowners in a given municipality, the greater the wealth and health of that population. One would therefore expect that the need for and expenses related to care for older people are less in municipalities with a high percentage of homeowners.

One might also assume that the share of single older adults has significance for care costs, as singles require more public care than if spouses or common-law partners can help one another (Savla et al., 2008). It is therefore to be expected that the larger the share of older adults living alone in a given municipality, the higher the spending per older person.

Moreover, one might also expect a trade-off between spending on care and female labour force participation. Extensive coverage of high-quality public care services reduce the necessity of informal care, which in turn facilitates the entry of women into paid employment (Jensen and Pfau-Effinger, 2005), and evidence seems to indicate (cf. Viitanen, 2005) that formal public care for older people – which is expensive for the public – is associated with high female labour force participation rates, while informal family care giving – which saves the public money – may be predominant in areas with low female labour force participation. It has even been observed that within countries; that is, between different regions and cities a trade-off exists between public care provision and the labour market integration of women (cf. European Commission, 2009).

Council composition

According to Briggs (1961), an institutional welfare state is characterised by that ‘all citizens without distinction of status or class are offered the best standards available in relation to a certain agreed range of services’. In effect, the degree of institutionalisation of the local welfare state is bound to have an effect on spending. Furthermore, as the class mobilisation thesis (see e.g. Korpi, 1983) hypothesises that the labour movement is the most significant force driving the development of an institutional welfare state, it must be reasonable to assume that the strength of the local labour movement, measured in terms of the percentage of labour mandates in the municipal council, is bound to have an impact on spending.

Welfare state research has increasingly focused on gender as a mechanism behind welfare state development. ‘Critical mass’ theories have thus raised the question as to whether gender makes a difference in politics – whether a ‘critical mass’ of women within the legislature can affect state policies in women-friendly directions (see e.g. Beckwith and Cowell-Meyers, 2007; Berkman and O'Connor, 1993). As large and comprehensive welfare states are usually considered to be women-friendly (see e.g. Hernes, 1987), it is worth noting that cross-national OECD data indicate that women's political representation has an impact on levels of social spending (Bolzendahl and Brooks, 2007). Whether this is actually also the case at the local level remains to be proven.

Municipal financial resources

The financial resources of municipalities vary greatly. Some are in a strong financial position, while others are more or less under permanent fiscal stress. As permanent austerity (cf. Pierson, 2001) is very strong in some municipalities, one might expect these municipalities to display low levels of spending. One might therefore assume that factors causing fiscal stress, such as long-term municipal debt and a weak tax base, have an impact on the capacity to organise extensive, high-quality care. Denmark does have a conditional grant scheme partly adjusting the differences in both tax base and needs in such a way that wealthy municipalities with low needs make transfer payments to poor municipalities with high needs, but conditional grants do not level out financial resources among municipalities. The measure that we use for tax base is adjusted for the conditional grant mechanism which helps equalise the actual tax base in the municipality.

Municipalities are often in a situation in which they must prioritise between various areas of welfare. Due to inadequate financial resources, the municipalities can thus be forced to prioritise between childcare, schools, care for older people, cultural affairs and so on. As such, a municipality with extensive financial resources may actually not be able to spend as much on elder care if schools and childcare require funding; or the municipality can spend a relatively greater amount of money on older people if spending is less in other areas such as schools, culture, etc. It is therefore relevant to account for the extent to which spending on other welfare services has an impact on spending on elder care. We will therefore analyse how a measure of the overall socio-economic and demographic pressure per capita is associated with spending on elder care.

Privatisation and flexibility

As already discussed, there have been tendencies since the turn of the millennium towards increased flexibility and privatisation of the care provided for older adults. Normally, privatisation is assumed to lead to productivity increases and lower costs, which could help explain the differences in spending, not least because tendencies towards privatisation of eldercare have not had the same direction or strength in all municipalities (cf. Fersch and Jensen, 2011; Szebehely, 2005). Privatisation, however, does not always lead to reduced spending (Hansen, 2010), and this has special bearing in the care sector in Denmark. The private provision of care in Denmark is based on quality standards and fixed prices set by the municipality, and privatisation and increased flexibility was primarily justified with reference to the ‘freedom to choose’; that is, home care recipients should be free to choose between different home care providers (public or private, for-profit). This in turn may result in increased costs for the municipality, partly due to the costs involved in drafting tender materials and such documents and partly because the municipality must have a response that makes it possible to fulfil the obligation to re-assume the provision of care if a private contractor were to choose to leave the business, and the question becomes whether the construction of clients as the consumers of welfare services (see e.g. Rostgaard, 2006) has increased the expenses related to the care for older adults and may help explain differences in municipal spending. The analysis draws a distinction between the for-profit, private provision of practical assistance (such as cleaning) and personal care.

Several studies have indicated how variations between municipalities appear to be stable over time (Daatland, 1997; Hansen and Hjort, 1998; Trydegård and Thorslund, 2001, 2010). This continuity in the variations in the care for older adults fits the path-dependency argument, according to which past political decisions take on a dynamic of their own or that welfare programmes are marked by vested interests and feedback mechanisms (Pierson, 1994, 2001). The path-dependency argument will therefore be tested, after which attempt is made to explain the 2005 expense levels with reference to spending levels six years earlier.

The independent variables included in the analysis are presented in Table 1.

Data

The analyses in this paper are not directed by theoretically interesting factors alone. They have also been shaped by the accessibility of data. The data set serving as the basis for the regression analysis below consists of multiple sources of data. All of the data are from 2005, with the exception of the path-dependency measure and occasions where data from 2005 were unavailable, in which case 2006 data are used instead. All of the variables are downloaded from De Kommunale Nøgletal (‘the municipal key figures’ – www.noegletal.dk), Statistikbanken from Statistics Denmark (www.statistikbanken.dk) and data dealing with the labour mandates and the proportion of women in municipal councils have been provided by Kurt Houlberg from ECO-nøgletal (www.ecoanalyse.dk). The figures from 1993 and 1999 together with De Kommunale Nøgletal are also used in the subsequent path-dependency analysis.

Results from the regression analyses of the effects on spending on older adults

The following is an investigation of the extent to which the major differences in spending on older people can be explained by objective characteristics. These explanatory variables are introduced using four models on the basis of theories and existing empirical material regarding the causes for the differences in spending. We try to structure the models as a recursive block system where the effect from lower block variables is possibly mediated through higher block variables. Table 2 presents the results of the regression analyses.

TABLE 2. Explaining municipal spending on elder care in Denmark 2005. Linear regression. Regression coefficients and explained variance. N = 270

Notes: Backward selection of variables in regression equation for each model. Minimum accepted p-value: 0.10. Significance level based on robust standard errors.

^NS Not statistically significant (left out of model); † p < 0.10, *p < 0.05, **p < 0.01, ***p < 0.001.

^a Numbers correspond to Table 1.

Model 1 includes variables describing the municipal structure and composition of the population. This includes variables pertaining to the area and population density of the municipality, the municipal housing structure (owned homes versus dwellings), the size of the population and age of the municipal residents (percentage of sixty-five-plus of total population and percentage of eighty-five-plus of total population sixty-five-plus), including the share of single older adults, the labour force participation of women and total spending on municipal welfare services, including schools, childcare and so forth. Controlling for these factors accounts for 36 per cent of the variation in spending on the care for older people. However, only four of the ten variables in this model were found to be statistically significant at a 10 per cent level or less. Health (owned homes) and the percentage of the population aged sixty-five-plus (economy of scale) tend to reduce spending, while the percentage of single older adults as well as the percentage of the population aged eighty-five-plus of total population sixty-five-plus tend to increase spending. It is somewhat surprising that the variables for area, population and especially the degree of urbanisation, do not have an impact on spending, since higher service levels can be expected in larger urban municipalities (Birgersson, 1971).

Model 2 adds political mobilisation variables in relation to the council composition (the share of female and labour mandates) as well as the financial resources of the municipality (tax base and depth). When controlling for these factors, 45 per cent of the differences in spending can be accounted for. Data show that a greater tax base typically means higher spending on older adults, while depth has a negative impact on spending. It is rather surprising that no effects can be found from the share of labour mandates in the municipal council, indicating that the municipalities cannot be divided in high-spending social democratic municipalities and low-spending liberal or conservative municipalities. One plausible reason for this may be that ideas and ideology possibly play a far more important role at the national level than at the local level, where local politicians are confronted with the challenge of solving practical problems in citizen's daily lives. It is also surprising that the stronger the representation of women in the municipal council, the less the municipality tends to spend on elder care. The effect is rather weak, though, and not easily interpretable. But even after adjusting for the effects from the variables in blocks 1 and 2, most of the variation in the expenditures remains to be explained.

Model 3 also includes financial resources in the form of tax levels in addition to trends towards privatisation and flexibility, comprising the degree of free choice in the care for older people. Unsurprisingly, higher tax levels typically mean higher spending on care for older people, whereas the coverage of free choice in relation to practical and personal home help has no impact on expenditures. However, it seems as though the use of private suppliers in the home help sector, as expected, increases spending on the care for older adults. All in all, the variables in the first three blocks explain 48 per cent of the variation in expenses on care for older people.

Having reached a level of explanation of 48 per cent, there is one important question to answer: are differences in expenditures mostly due to ad hoc adjustments of expenses in the municipalities to maintain the service quality at status quo? If this is the case, differences represent random fluctuations. Or are variations an indication of what might be termed ‘real differences’, referring to a more stable pattern of substantial differences over time? Should this be the case, we would expect past expenditures to have a strong effect on present and future expenditures – what might be labelled past dependency. The question is, then, whether past spending can serve as a good predictor for spending in 2005, adjusted for relevant control factors, like the percentage of older people, the percentage of single older adults, etc., from Model 1.

Model 4 provides an indication of this. Here, we add a variable to Model 1 indicating spending on the care for older people six years prior to 2005, that is in 1999.⁵ This time span includes one municipal election, held every fourth year in Denmark. We see a highly statistically significant effect from past expenditures and the R² increase from 0.36 in Model 1 to 0.52 in Model 4. Including the path-dependency argument allows us to explain 52 per cent of the difference in spending among municipalities. This also indicates that the differences between municipalities do not occur due to year-to-year fluctuations in spending.

Path-dependency may occur because municipalities employ marginal budgeting/incremental budgeting. In other words, when deciding the next year's budget, the past year serves as the point of departure. To dig into this perspective, we have also analysed a model (not shown) where we add past expenditures to Model 3 (see Table 2). This leads to an increase in the R² from 0.48 to 0.61. All of these findings tell us that there seem to be some real and rather stable differences in municipal spending on older adults. Taking the analysis of path-dependency back six more years to 1993, the effect from past spending decreases considerably, although it remains highly statistically significant. That the effect of past spending falls over time probably owes more or less to random fluctuations and policy changes. So there are limits to the path-dependency argument, as path-dependency seems to be declining, as Trydegård and Thorslund (2001, 2010) also found in the care for older people in Sweden.

In addition to examining path-dependency in the municipalities, we have also examined whether the spread in spending between municipalities is above the same level over the years. One could hypothesise that the municipalities would converge over time. Nevertheless, the variation seems rather stable from 1993 to 2005. The coefficients of variation in the years 1993, 1999 and 2005 are 0.14, 0.12 and 0.13, respectively.

Finally, we have carried out a longitudinal analysis of the changes over time for the period 1993 to 2005, as we have carried out so-called repeated measurement analysis in the form of a multi-level linear regression. If the interest focuses on the investigation of causal relationships – effects on the dependent variable – the opportunities to draw conclusions in such analyses are improved in relation to ordinary cross-sectional analyses. Here, the question is what the changes over time within the individual municipality typically mean for the dependent variable. For example, does an increase in the percentage of seniors aged sixty-five-plus lead to a fall in the spending per senior, as the conclusion from the cross-sectional analyses would lead us to believe?

Unfortunately, we are not able to conduct the overall analysis in this manner, as many of the important independent variables are only found for the last years of the period under investigation. But we can carry out the analysis with repeated measurement for the relevant variables in Model 1; that is, the variables that can possibly change significantly over time in the individual municipality. If the results from such an analysis confirm the image from the cross-sectional analyses, they would support the results from Model 1 as well as the conclusions from the other models.

As seen in Table 3, three out of four variables have a statistically significant effect with the same sign as in the cross-sectional analyses in Table 2. A rise in the percentage of people aged sixty-five-plus tends to lower the expenditures on each senior, and a rise in the share of seniors aged eighty-five-plus and single elderly households, respectively, tends to increase the spending on each senior. However, the effect of change in the proportion of homeowners has an opposite sign as the coefficient in the cross-sectional analysis. Here, it is worth adding that the hypothesis regarding the effect of this variable is not as strong as the others, so it is not particularly surprising that the ‘expected’ effect is not found here. All told, this longitudinal investigation supports our results and conclusions.

TABLE 3. Multi-level analysis of intra-municipal change in expenditure levels (repeated measurement in 1993, 1999 and 2005

Notes: The statistical model includes variables that are not shown in the table. These variables are, first, a variable measuring time in three categories, among other things adjusting for rising prices over the years, and, second, group means for all four variables shown in the table, controlling for between municipality effects (Rabe-Hesketh and Skrondal, 2008). These variables are not shown.

*p < 0.05, **p < 0.01, ***p < 0.001.

What is the money spent on?

Thus far, our main focus has been on trying to explain differences in municipal spending on elder care. We have not investigated how and for what purposes money is actually spent. It is therefore possible that high-spending municipalities simply misuse taxpayer's money due to organisational slack and less value for money. It is also possible that high-spending municipalities employ more staff in order to deliver high-quality provisions, given that there is a connection between quality and the number of care staff. Yet another option could be for municipalities to choose different institutional solutions for the care challenge; some municipalities may prioritise institutional care, which is also more expensive than home help. Table 4 shows how municipal spending co-varies with (a) the share of older adults enrolled in institutional care and (b) the number of full-time care workers employed in the elder care sector.

TABLE 4. Spending on elder care: bivariate correlation (Pearson's r)

Note: ***p < 0.001.

Table 4 shows a rather strong correlation (0.29) between spending levels and the share of older people enrolled in institutional care. This seems to indicate that differences in spending are founded in different institutional priorities at the local level. While some municipalities are inclined to make use of institutional care, others favour home care. The table also reveals a very strong correlation (0.60) between spending and the number of staff in the elder care sector. This could either signify that some municipalities are over-staffed – and that there are gains to be achieved via improved efficiency in the high-spending municipalities – or that differences in municipal spending are founded in different priorities with regard to the quality of social provisions delivered to elder care recipients.

Whatever the case, it should be noted how existing research has demonstrated that there is not necessarily a link between expenses, quality and user satisfaction (Lolle, 1999), nor is it possible to say anything about whether those with the greatest needs are actually those receiving the most public services.

Conclusion

In Denmark, services provided by the welfare state are generally a municipal responsibility. In many countries, municipalities are responsible for the organisation, financing and provision of the services to the citizens, and municipal autonomy has meant that the extent and quality of social services may vary enormously among municipalities, meaning that the welfare state regime approach has exploded from within, so to speak. Using Denmark as our test case, the paper has aimed to explain the enormous variations in municipal spending on care for older adults. It has been rather difficult to identify the factors constituting differences in municipal spending. We have, however, been able to explain about around 52 per cent of the differences in spending, which is a relatively high figure compared to other studies in this field. The insights from the paper are twofold.

First, different factors have different effects on spending. Factors such as the share of population aged sixty-five-plus (economy of scale), the share of population owning one's own home (a proxy for health) and municipal depth reduces spending, whereas factors such as the share of older adults living alone, the wealth of the municipality (high tax base and tax levels), percentage of population eighty-five-plus of total population sixty-five-plus and the proportion of private suppliers tend to increase spending. The analysis also supports the path-dependency argument, as previous spending is a good predictor of present spending levels.

Second, from the point of view of political mobilisation, it is interesting to note that traditional forms of politics do not matter. The percentage of labour mandates in local municipal councils has no effect on spending levels, whereas a strong representation of women in local municipal councils has a weak negative effect on spending. This seems to indicate that in the case of welfare services and welfare municipalities, conventional typologies and labels such as liberalism, conservatism and social democratism are not applicable.

The fact that policy does not have great significance on the local level is rather remarkable, as the municipalities have great autonomy and freedom to choose how the local authority will solve the challenges and problems they are facing. The marginal role of policy at the local level is presumably owing to different ways of governing and that different political dynamics are at play at the national and local levels. At the local level, there is broad support among the politicians and population alike that it is the municipality's responsibility to provide the best possible services to the citizens, including ensuring high-quality elder care, but that the financial framework in particular (spending levels and possible sources of income) sets limitations and conditions in this regard. To put it bluntly, one might say that ideas or ideology have primacy over problems at the national level, whereas problems have primacy over ideas and ideology at the local level. From the perspective of political citizenship, however, one must remember that quality in the provision of services at the local level is transparent (quality standards must be communicated to citizens in local communities), just as local decision makers (the municipal council) can be held accountable for the quality of the care provided together with other kinds of services, such as schools, childcare and so forth.

Elder care in Denmark is highly formalised, and the public provision of elder care has crowded out informal care. From a social citizenship perspective, however, the massive local variations in elder care make topical a debate about equity and fairness in social policy and welfare systems; not least because a number of municipalities in recent years have made cut-backs in practical home care in the wake of the global financial crisis. However, the regional state administration issued guidelines in January 2012 for minimum standards in the elder care area. One can therefore regard elder care in Denmark as a universal welfare service. Elder care in Denmark is characterised by citizenship-based universalism, minimum standards, services financed by individual and progressive taxation and high degrees of user satisfaction. As such, local differences in elder care may not challenge social citizenship. But municipal differences remain an intellectual challenge. The major question is how to integrate interactions between national and local governments, including local variations in welfare services, into the welfare regime modelling business.