Differences That Matter: Overcoming Methodological Nationalism in Comparative Social Policy Research

Abstract

Welfare states are often discussed as if they were territorially homogeneous state-wide institutions measurable by state-wide expenditure averages and explained by country-level variables. It is rare in comparative policy studies to investigate the role of territorial politics in the outcomes of even federal countries. This article argues, using social policy examples in the UK and US, that the impact of intergovernmental finance and division of labour profoundly shapes social investment and redistribution – producing almost as much expenditure variance within the US as within the OECD. The findings show the importance of incorporating territorial politics and intergovernmental arrangements into comparative welfare state and policy analysis.

Keywords:

• comparative welfare states
• comparative federalism
• fiscal indicators

1. Introduction

Students of welfare state and public policy are well aware of the importance of territorial allocation of authority. No serious textbook of public policy in a federal or decentralised country fails to mention the roles of regional governments and their policy activities; in some countries, such topics merit their own professional journals and undergraduate courses. Yet the study of comparative public policy is strangely uncomfortable with the role of regional governments in explaining policy outcomes (Agnew 2013). Students of the detail of the policy sometimes fail to pay much attention to the constraints and politics of the precise government bodies formulating interesting policies. Comparativists, especially those focused on the welfare state, often become methodological nationalists (Chernilo 2006). Methodological nationalism is “the naturalisation of the equation of society, state, and nation” – the assumption that states are the best or only units of observation or analysis (Jeffery and Wincott 2010, p. 170). In comparative politics, policy, and welfare state studies, it amounts to the use of country-level data to characterise and explain welfare decisions.

Methodological nationalism poses problems for the analysis of welfare states. First, it is problematic because it defines distinctive regional welfare arrangements out of existence. Second, it substitutes state-level averages for more telling regional data, reducing the number of cases available. Causal analysis about the effects of a given variable might be better served by avoiding a once-removed analysis of aggregate data, and, instead, pursuing an inspection of the actual political units where the decision making takes place. A study of the effects of leftist government on education spending, for example, would be poorly executed through an analysis of aggregate spending data and political representation at the federal level if the actual spending levels were determined locally. Third, it oversimplifies the complexity revealed by the historical study of federalism and welfare (Obinger et al. 2005). Fourth and finally, it is a problem because it makes it much harder to study the interaction of nationalism, territorial politics, and the welfare state (McEwen and Moreno 2005; Wincott 2006; Béland and Lecours 2008; Greer 2009).

Methodological nationalism, we posit, is particularly likely to be a problem in the analysis of “new” social policy debates, such as those about the ways welfare states are handling demographic changes such as ageing populations, the decline of male-breadwinner households, and the challenges of long-term unemployment – broadly speaking, the debates surrounding new social risks. Unlike pension policies, the diverse policies aimed at women, youth, and the long-term unemployed all tend to be decentralised and depend on regional or local government finances, collaboration, and implementation. The scale of social investment or adaptation of the welfare state in these areas, with the large “footprints” of staff and programmes found in areas such as health or activation policies and the large role of regional governments, is particularly unlikely to be captured by state-wide averages.

Working with the cases of the United Kingdom and United States, in the perspective of data from the OECD countries, this article argues that analyses of the politics of important social policy developments are illuminated if we incorporate regions into comparative welfare state analysis. The following sections sketch a method for understanding regional welfare state effort and develop an example in the development of social policy in the United Kingdom and United States. We focus on two kinds of social policy, often associated with the discussion of new social risks: the relative age alignment of spending (Lynch 2001, 2006) and active labour market spending. The analysis finds substantial variation within the United States, especially if placed in juxtaposition with the heavily used OECD data, which stands as the basis for a large portion of current understandings of different welfare states. As a result we can see the effects of the different forms of decentralisation in the United States and United Kingdom on two key kinds of distributional concern, namely the extent to which risks associated with age are addressed by policy, and the extent to which the regional governments charged with much active labour market policy (ALMP) are acting to address the risks of long-term unemployment.

2. Regions in Comparative Welfare State Literature

There are two broad ways in which the existing literature handles the role of regions in comparative welfare state studies (Greer 2009). One emerges from the literature in comparative welfare state politics and macrosociology and treats decentralisation as a property of states. It regularly finds that decentralisation correlates with lower welfare state effort, argues that the problem is veto points, and is usually hostile to decentralised allocations of authority. This body of work converges with conservative public choice scholarship, which formulates similar arguments but applauds smaller welfare states and posits intergovernmental competition as a more important mechanism. The other broad way of handling territorial welfare states emerges from studies of individual regions, and often praises their policy experiments and higher welfare state effort.

2.1 Macro-Comparative Approaches

The traditional macro-comparative approach, practised equally in sociology and political science on both sides of the Atlantic, is to treat any decentralised allocation of authority as an attribute of the state. The units of analysis are states, and the allocation of authority is either a variable or part of a larger variable such as “fragmentation” (Hicks and Swank 1992; Hicks 1999; Lancaster and Hicks 2000; Huber and Stephens 2001; Swank 2001, 2002; Brooks and Manza 2007). Either way, decentralisation, like bicameralism or referenda, is an attribute found to a variable degree in countries. Almost every study with this premise finds that centralisation promotes welfare state growth, and decentralisation the reverse (Castles 1999; Stepan and Linz 2011). They tend to explain this pattern with a form of veto-players argument: federalism increases the number of obstacles to welfare-expanding legislation. Such a veto-players argument also seems like a plausible explanation for what appears to be relatively slower retrenchment of welfare states in decentralised countries; they are just slower to change (Simeon 2006).

Public choice scholars come to many of the same conclusions while employing a different analytical framework. They argue that mechanisms of competition keep governments from “overproviding” desirable things and thereby raising taxes too much or incurring too much debt, while competition and shorter principal–agency connections (e.g. greater proximity to voters) reduce waste and local rent-seeking (Weingast 1995, 2009; also discussed in Costa-i-Font and Greer 2013). While the logic focuses on competition rather than the veto points that interest macro-sociologists, and the normative commitment is to small government, it concurs with the comparative sociologists on one basic point: more decentralisation yields less welfare state expenditure.

Neither school of thought explores the possibility that variation within the state can produce higher or lower levels of expenditure on priorities in different places; both are narrowly focused on using stylised state-level institutional structures to explain state-level averages. But treating decentralisation only as a property of sovereign states reduces regional governments’ activities to their participation in state averages. There are several problems with this approach.

First, regional governments may or may not play any role in determining welfare spending. If they lack decisive influence over the central government, why should they matter? Decentralisation is logically and practically independent of regional government vetoes over central government (Greer 2009) – self-rule is not the same thing as shared rule (Elazar 1987).

Second, state-level measures of welfare effort should decompose into state and regional welfare state effort. If the regions are dominant welfare providers and the central state plays little role then there might be considerable variation. Overall welfare expenditure will be an unrevealing average, as it often is in the United States.

Third, there is no theory of regional or state welfare politics in these accounts; it is not at all clear that even territorial decentralisation means anything on its own. It might be an unrevealing aggregate of more focused variables such as the structure of intergovernmental finance or the specific extent of regional government influence over central government actions. This is because the basic idea of “fragmentation” has no meaningfully specified mechanism associated with it.

Fourth, the endogeneity problem that afflicts all studies of territorial politics means that it is hard to see, at the state level, whether welfare state effort is a property of institutions, or whether both welfare state effort and institutions are effects of deeper social structures (Rodden 2004; Erk and Koning 2009). Insofar as the fragmentation runs deeper than the institutional level, there is a good chance that these structures are territorially differentiated and would be revealed in a territorially sensitive analysis, as studies have indeed found (Linz and de Miguel 1966; Snyder 2001). It is certainly the case that territory shapes the intense arguments about the distribution of government resources that are found in many countries (e.g. Beramendi 2011).

Fifth, the narrow concept of “decentralisation” obscures the very complex ways in which territorial politics and institutions shape welfare state priorities and expenditure. For example, areas of the Spanish and US welfare states that are the responsibilities of autonomous communities and states are under more fiscal pressure than areas in the hands of the central governments (Greer 2009). Gais (2009), accounting for the complex system of US welfare programmes, presents an intricate comparison of US state and local expenditures on cash assistance, medical assistance, and social services and discerns a decline in assistance and growing divergence in state spending depending on fiscal capacity. The historic roots of this differentiated American welfare state, other major scholars argue (Lieberman and Lapinski 2001; Katznelson 2013; Mickey 2013), involves precisely the exploitation of federalism in social policy design to preserve inequality in undemocratic racial enclaves in the South.

These problems more or less automatically flow from treating decentralisation as a property of countries in regression analyses. As Benz and Broschek (2013) conclude, “although many scholars have emphasized the dynamic character of federalism, there obviously is a shortage of informed studies on the topic”. The next section reviews the main alternative.

2.2 Regional Agency

The alternative in a great deal of the literature is to focus on the activities of individual regional governments. This means writing about regional protagonism: what did a given region do, and why? This genre of work is predominantly case-based. It focuses on small numbers of regions, or regions in a given country, and explains what they did and what they were capable of doing in a theoretically informed way (McEwen and Moreno 2005; Béland and Lecours 2005, 2008). It encompasses studies that discuss, in varying degrees, the use of welfare policy as a nation- or region-building exercise as well as the development of distinctive welfare politics and policy in any given region. Work in this vein, as often as not, celebrates regional welfare states; it is fairly clear that the desire to defend a Scottish welfare state was a major driver of devolution (regardless of what scholars might say about the Scottishness of that desire, or that welfare state). In principle, this literature is complemented by the general public policy literature that exists for some regions (Keating 2005; McGarvey and Cairney 2008; Birrell 2009; Lodge and Schmuecker 2010) as well as by qualitative comparative studies which tend to find that the territorial politics of welfare is very complicated (Obinger et al. 2005).

This second approach has advantages over the standard macro-comparative account. It is more precise, connects better with relevant areas of comparative politics such as the study of parties, institutions, and intergovernmental relations, and is more theoretically fertile because it is not based on rough and nearly uninterpretable correlations. The problem is that it has been pursued through small-N studies. This creates difficulties in comparing specific studies as well as problems of case bias: many discussions of Catalan, Quebec, and Scottish policy, few studies of Valencian, Manitoban, and Northern Ireland policy.¹ There is a risk that a focus on the experiences of a handful of regions shapes the theory.

2.3 Regional Welfare State Effort

So long as comparative findings are derived from studies that use states as units and “federalism” or “fragmentation” as a variable, macro-comparative welfare state studies will continue to write regional governments out, fail to connect with case study and intra-country comparative literature, or slight decentralisation by lumping it into a single variable with other presumed enemies of the welfare state. Qualitative comparisons of policy variations within countries and case studies of divergent social policy approaches both have great merits but fail to straightforwardly respond to the macro-sociological or public choice claim that decentralisation impedes the development of the welfare state.

We propose that it is possible to compare the extent of divergence within countries, showing how different degrees of financial centralisation interact and shape policy (Agranoff 2004). We argue that decentralisation can allow or create substantial internal variation through greater or lesser redistributive spending across territory and financial integration. If this is true, then it should be possible to choose two countries with very different institutional forms and see substantially different levels of expenditure within them, comparing the role of regions within their systems, identifying the scale of territorial divergence within countries, differentiating divergence between countries. We undertake this endeavour in our study of the cases of United States and Great Britain, elaborating a cross-national comparative analysis of social policy questions that places territorial governance at the centre of the investigation.

3. Decentralisation and New Social Risks in the Federal United States and the Devolved United Kingdom

This section compares regional and welfare state spending in two countries with substantially different arrangements for allocating money, focusing on the way territorial politics shape two key aspects of the welfare state: the distribution of spending between young and old, and the extent and vigour of active labour market programmes. These are preoccupations of the literature on new social risks (Taylor-Gooby 2004; Armingeon 2007). As Bonoli (2005) summarises it, “new social risks” are risks that fall onto the young, long-term unemployed, single mothers, and others who are structurally vulnerable but not priorities of welfare states built on supporting full-time male wage earners. They are also preoccupations of the newer work on social investment (Morel et al. 2012). Simply put, ALMP is a key form of social investment. Spending on the young is more of an investment than spending on the old because of the longer term expected payoff due to longevity, and the actual spending on the young tends to be investment such as education.

Comparative analyses of new social risks, while folding new elements within the comparative cross-national analyses, share the common ground with previous studies in this vein comparing and grouping cases of social policy spending levels (Cameron 1978), degrees of de-commodification (Esping-Andersen 1990; Scruggs and Allan 2006), or tendency towards relatively high spending on the elderly (Lynch 2006).

3.1 Case Selection

Our cases were selected to provide a basic test of the thesis that the structure of territorial politics shapes the welfare state; they are therefore two countries with substantially different territorial politics and welfare states that are generally categorised in the same way, as liberal. The United Kingdom and United States are frequently classified as similar countries: English-speaking, liberal in Esping-Andersen’s terms (1990), held up as paragons of neoliberalism, closely linked in politics and many policy debates, and with (broadly) residualist welfare states and low levels of employment protection. While the politics and policies of their welfare states still differ, their welfare states overall resemble each other much more than they do those of corporatist or Southern European countries. In terms of territorial politics, however, they are very different. The United States has long been a very decentralised federation (the states preceded and had to agree to the creation of the federal government), while the UK’s modern experience as a politically decentralised state began in 1998² and the UK remains asymmetric, with England directly governed by the UK government while Northern Ireland, Scotland, and Wales have their own autonomous parliaments with responsibilities concentrated in the welfare state. The central government of the UK still substantially controls finances within the country, despite some recent liberalisation of the Scottish financial mechanisms (if not the underlying distributional formula).

This makes the two countries excellent cases to assess whether the given political institutions produce greater or lesser territorial variation in the welfare state. In countries such as the UK, where finances are substantially centralised, we should expect relatively limited variation. In the United States, the institutional and financial arrangements permit considerable variation and put much less effort into inter-territorial equalisation, and we should see it in the variation between states. The structure of territorial politics – centralised finance and responsibility in the UK, decentralised finance and responsibility in the US – should produce varying levels and kinds of inequality among residents of different territories.

3.2 Methodological Approach and Data

Conceptually, our approach reconstructs a standard international framework for measuring comparative welfare state effort by incorporating the division of governmental powers with regard to a specific function. Data on state-level spending in the United States was extracted from the United States Census of State and Local Governments, year 2009, and where available 2010, which reports expenditure data for the federal and state governments, and survey-based returns for local government. This data presents numerous advantages in terms of degree of detail but comes from a data series that has since been discontinued.³ The United Kingdom data is from the Public Expenditure Statistical Analysis (PESA) series, produced by HM Treasury, for 2009 as well.

Coding is one of the biggest challenges for cross-national quantitative comparisons. Our approach was to rely on broad, largely comparable, budget categories: health, retirement income, primary and secondary education, higher education, and public assistance. We then translated the categories reported by each government into the larger categories. While it might be interesting to have more specific data on more specific policy areas, it becomes difficult to code coherently and comparatively below the level of those large categories. The least self-explanatory category, public assistance, consists of any form of benefits or income support excluding pensions. This includes cash benefits such as unemployment insurance as well as more distinctive benefits such as the Supplementary Nutritional Assistance Program (SNAP) in the United States.⁴ The categories we use essentially follow the same categories as the UK PESA statistics.

The age-alignment spending ratio was calculated closely following Lynch (2006). This measure is calculated by dividing the total spending on pension and survival benefits by the number of individuals over 65 and then dividing this figure by the per capita spending on social expenditures for the population under 65.⁵ With regard to the percentage of spending on active labour market programmes or close equivalents, the denominator, the calculated gross national product of each state or region, was calculated based on data from the US census data, Eurostat, and the OECD (Adema et al. 2011).

Calculations for the OECD countries drawn from the OECD Social Expenditures Dataset were made based on data from the OECD. As with Lynch (2001, 2006) we are obliged to exclude several types of direct expenditures from this measure given the problem with ascertaining the age of the recipients: incapacity-related benefits (notably, disability pensions), housing, and health spending. Spending included in the elderly category are the following: old-age pensions, survivors’ pensions, services for the elderly, and early retirement pensions. Non-elderly expenditures include sums for unemployment compensation, severance pay, early retirement, family cash benefits and services (child allowances and credits, childcare support, income support during leave), ALMPs (OECD 2011), and other social assistance expenditures such as income maintenance as well as spending on education.

4. The United States

It is a commonplace of American politics that states vary widely in their wealth, taxes, welfare benefits, and political commitment to welfare. What is less widely studied is the extent to which this produces a distinctive welfare state in each state as a result of interaction between state politics and federal programmes. What is the interaction between state welfare policies and federal welfare policies and how does it shape the overall US welfare state?

Governments in the United States have many financial options and trade-offs. As a legacy of the country’s age and long history of intergovernmental relations, practically every form of intergovernmental arrangement – and gaming – may be located in some place at a given point in time. State and federal spending are both conceptually simple: the state or federal government raises a sum of money and spends it in a given state. The more complex problems arise from federal grants and co-financing, which meld state and federal money by inviting states to participate in programmes that allow the federal government to match their expenditures for some reason. While the state governments have no formal representation in the government (they serve as electoral circumscriptions for the directly elected Senate), members of Congress are able to direct spending to the state governments in a variety of ways (Richardson 2009). When the federal government co-finances services, it often uses different matching formulae for different states; a higher matching rate means that a state gets more federal dollars for each dollar it spends (e.g. Thompson 2012). Negotiating this rate can have important consequences for state budgets.

Figure 1 shows the 2008 returns for spending numbers for state and federal governments from the Census. For simplicity, it only shows three categories: state spending out of own revenue; direct federal spending; and conditional grants – federal spending via grants to the states such as Medicaid. This excludes local government spending, which is especially important in education. Figure 1 uses the measure of per capita expenditure, which shows substantial variance, summarised in Table 1. A look at the relative weight of state, federal, and federal grant expenditure reveals that not only are there wide differences in American states’ welfare expenditure but also divergences in the degree to which they affect federal grant outlays (e.g. Medicaid) but not direct federal expenditure (e.g. Medicare). For instance, the coefficient of variation for federal spending is 0.20 and for state spending it is 0.31, while the variation is much greater for federal via state per capita spending (0.82).

Table 1. Variation in US state, federal, and conditional spending ($ thousands per capita)

	State	Federal	Conditional grants from centre
Average	2.29	6.15	1.54
SDD	0.72	1.23	1.26
Coeff. of variation	0.31	0.20	0.82

Figure 1. State, federal, federal via state per capita ($ Thousands) social spending in the United States (2009)

Notes: Per capita spending in thousands includes spending in the following areas: primary, secondary and higher education, old-age pension, public assistance with TANF and public assistance without TANF, health, housing, and local amenities. Source: Census Bureau, authors’ calculations.

Based on a more detailed breakdown of per capita spending data (Data Appendix, Table A), there is substantial variation in states’ share of welfare spending, less variation in federal spending via states (co-financing, i.e. Medicaid, Temporary Assistance for Needy Families (TANF), and the Children’s Health Insurance Program (CHIP)), and still less in federal spending, as shown in Figure 1. Including covariates such as age and poverty rates would lessen the variation between expenditure by federal programmes in different states. The variation is therefore not just between states; it is between programmes. Programmatic areas such as higher education that have more state money have more variation in per capita expenditure than the paradigmatic example of a federal programme, old-age pensions. Where the federal government leaves space for states to vary (Béland et al. 2014), states will; when the federal government creates incentives for states to take, they vary greatly and the states take them at very different rates, as the coefficient of variation in Table 1 shows.

Figure 2. Age-alignment spending ratios in the United States (2009)

Further cross-state comparisons emerge based on measures of state spending presented in Figures 2 and 3. The first is the elderly–nonelderly spending ratio (ENSR), introduced in Section 3.2. A higher value reflects more spending directed at the elderly, controlling for populations. Crudely, in the US this maps onto a ratio between federal spending, directed at the elderly, and state spending, directed at the non-elderly. The considerable variation in ENSR shown in Figure 2 suggests substantial variation in state spending against a background of a federal welfare state more focused on the elderly.

The second dimension of social risks under analysis, active labour market programmes, are difficultly captured in the United States. We use one of the only schemes with an active job training programme, TANF, a federal block grant programme that provides “assistance and work opportunities to needy families by granting States the Federal funds and wide flexibility to develop and implement their own welfare programs”.⁶ It might not be all that advocates of ALMPs would recommend, but it is a labour market policy with elements of activation in the American context,⁷ and is classified as a component of ALMPs by the OECD (OECD 2011). We report TANF as percent of state GDP to match OECD data.

Figure 3. TANF spending in United States as a percentage of State GDP (2009)

Spending on TANF varies to a greater extent than the variation observed in the age-alignment ratio. As illustrated in Figure 3, the coefficient of variation for the TANF data indicates relatively high variance across states for this measure – 0.58, compared to that of the ENSR 0.17. In the case of the TANF data, 12 states are outside the range of the median plus or minus 50 per cent. With regard to the age-alignment orientation of spending in a given state (ENSR), there is some overlap between the top 15 states in terms of degree of age spending ratio and the lowest proportion of state-calculated GDP dispensed on TANF (New York, Ohio, Vermont).

For comparative welfare state studies, overall low US average welfare state effort is not very revealing. There are deep divergences between states that reveal different priorities, expenditure levels, and corresponding welfare states and effects on citizens’ lives. The overall US programmatic bias towards age and medical care is in large part a function of the structure of federal government spending, though it seems help for the aged is largely parallel (run without reference to the states, except for Medicaid’s role in long-term care for the elderly) and healthcare is largely additive, with states having the capacity and sometimes the will to expand healthcare spending beyond the apparent minimum of Medicaid and CHIP participation. The age profile of the US federal welfare state is biased towards the elderly (Lynch 2001); the extent to which states reduce that bias is in large part driven by state politics.

5. United Kingdom

If the decentralised finances and division of labour in the US seems to encourage interstate variation in some services (loosely those aimed at under-65s) and federalise benefits in others, the UK’s centralised finances at devolution were geared to produce rough equality between residents regardless of their territories. UK debates tend to focus on the variation in per capita spending between England, Northern Ireland, Scotland, and Wales, and more sophisticated versions incorporate differences between spending in English regions as well (McLean 2005; Trench 2008; McLean et al. 2009). But compared to the United States, what stands out is the lack of variation and lack of scope for variation in welfare spending so far.

Devolution finance in the UK has two components. One is a block grant that allows the three devolved administrations (Northern Ireland, Scotland, and Wales) large sums of unrestricted money. The money given to devolved governments is calculated from a historic baseline, which traditionally gave them all higher per capita spending than England. Changes in the amounts budgeted then come via something known as the “Barnett formula”. The Barnett formula stipulates that for each change in spending in England on services that are devolved in Northern Ireland, Scotland, and Wales, the devolved administrations each receive a pro-rata per capita share; so if an extra pound is spent on English services, an additional per capita share of that is given to the devolved administrations (about 11p for Scotland). There are a variety of significant loopholes (classifying some spending as UK rather than comparable, or manipulating local government taxation), but the system is relatively transparent. Devolution of specific tax bases could produce more variation in expenditure as the share of Barnett block grants shrinks relative to taxes that devolved administrations can vary.

Atop the block grant comes an arrangement being introduced in Scotland, which implements the recommendations of the Calman Commission (issued in 2009). The new Scottish system goes beyond the (unused) Scottish power to raise income tax slightly, and instead devolves taxes totalling approximately 30 per cent of its expenditure. The devolved taxes include some entire individual taxes, such as landfill tax and the “stamp tax” on real estate transactions, and a slice of the income tax. The taxes gained will be matched with a cut in the block grant to Scotland (that will match the revenue raised in Scotland each year). There are discussions of parallel reforms for financing in Northern Ireland and Wales. These arrangements are new and still being implemented as of 2015.

The devolved governments are then free to spend that money within their legal powers, which vary in form and extent but basically make them responsible for health, education, housing, and community amenities, and those aspects of public assistance involving personal social services. The bulk of devolved budgets go on the “comparable” welfare state services, i.e. ones that are overwhelmingly the responsibility of devolved governments or England-only departments in England. The UK government’s all-UK social welfare responsibilities are mostly in old-age pensions and those areas of public assistance that involve cash transfers. Formally, the social security system in Northern Ireland is devolved, but it historically moves in lock-step with UK benefits and is paid by the UK for that purpose (meaning that, properly understood, it should cause no variation in spending per capita on the elderly).

Figure 4 shows the territorial breakdown of welfare state spending in the UK. It displays the well-established inequalities in spending by territory, inequalities that are actually much more substantial if we consider the large inequalities between English regions (McLean et al. 2009). Such comparisons often appear in discussions of whether the allocation of funding to one region or another is “fair”. In the variation in the percentage of spending that is allocated by regional governments, we can see some effects of differences in formula funding, as well as different eligibility for UK benefits.

Figure 4. UK social spending by region and area, 2009, per capita (£ thousands)

Source: Public Expenditure Statistical Analysis (PESA) series. See notes to Table 2 for discussion of Northern Ireland.

Table 2 presents the percentages of expenditure that are regional. For a Scot, about 56 per cent of the welfare state effort is regional. And for the UK overall, about the same percentage of the welfare state is “regional” – meaning spent by either a devolved administration or an England-only department such as Health. Apart from Northern Ireland, where the formally regional social security field is essentially part of the UK system, the variations are probably due to a mixture of devolved finance and the age profile of different parts of the UK. Compared to the US, what stands out is the uniformity of the percentages; the two categories of mechanism (central block grants to devolved regions and central programmes operated centrally) are in rough balance.

Table 2. Percentage of social expenditure by regional government (relative to total including regional and UK government)

	% Regional
England	48.40%
Northern Ireland	79.20%
Scotland	56.81%
Wales	50.82%

Source and notes: The United Kingdom data is from the Public Expenditure Statistical Analysis (PESA) series. Northern Ireland is formally responsible for social security, which accounts for its higher regional percentage, but as benefit levels are set in lock-step with overall UK social security and it receives a subsidy to maintain those levels the regional role is essentially administrative rather than policy making.

Table 1 and Figure 4 both show that the impact of regional welfare states depends substantially on the financing mechanism and degree of fiscal equalisation. Devolved governments in the UK do not have much freedom to change overall welfare effort. They have a block budget calculated primarily based on three major English services (health, education, and social care) and spend most of it on those three services in their jurisdictions. They have limited trade-offs between sectors and no meaningful trade-off with taxation levels, which also means their incentive to reorganise might be limited. It should be no surprise that the divergence in devolved policy is more in organisation than in levels of expenditure (Greer 2004; Birrell 2009; Lodge and Schmuecker 2010). For the UK, the future, with at least some greater degree of fiscal autonomy for Scotland, might allow greater variation.

6. Cross-National and Cross-Regional Comparisons

Standard analyses often mask the composition of the welfare state by hiding the substantial role of regional politics. Approximately one-third of US welfare spending is by state governments out of their own resources, with the effect that Connecticut and Kansas, or Texas and New York, can run welfare states with different priorities, eligibility, and services. To some extent they must; Connecticut is a much richer state than Kansas. To some extent they choose; Connecticut is also much more left-leaning than Kansas. Approximately another 12 per cent of expenditure in the US is state spending of federal grants, in which they have some autonomy to treat their populations differently. Approximately half of UK welfare spending is by devolved administrations or England departments, with the devolved governments highly autonomous; the formula financing system means that they do not vary their spending much but the half of welfare that they expend does explain why the experience of the same welfare services in different parts of the UK might be quite substantially different.

This divergence raises an obvious question: is the divergence within at least some countries perhaps greater than the much more studied divergence between countries? Comparing Figures 2, 5, and 6 addresses this question. The coefficient of variance is designed to allow for the comparison of the degree of variance across samples of data without being sensitive to the measure utilised (calculated as the standard deviation divided by the mean). With regard to the age-alignment spending ratio, the coefficient of variation is 10.6 per cent greater for the sample of OECD countries when compared to the state ratios for the United States and 63 per cent greater than in the case of regions within the United Kingdom. Given the volumes of critical comparisons and theoretical groupings of OECD countries, it is surprising to consider that such a high degree of variation may be found within a single case. Moreover, in TANF, which the OECD considers to be the key American ALMP programme, the coefficient of variation is actually greater in the USA (0.58) than in the OECD overall (0.53) (compare Figures 3 and 7). The comparison is imperfect because TANF is a mixture of ALMP and straightforward income support, but as a statement about variable commitment to the labour market and life prospects of the poor, the extent of variance within the US is noteworthy. American states, many of which rival OECD member states in size, seem also to rival them in the divergence of their approaches to, and spending on, new social risks.

Figure 5. Age-alignment spending ratios in the United Kingdom (2009) (ENSR)

Source: The United Kingdom data is from the Public Expenditure Statistical Analysis (PESA) series.

Figure 6. Age-alignment spending ratio in the OECD

Source: The OECD countries in question are limited to those within a 33 per cent range of GDP per capita relative to the US. Data from OECD Social Expenditures and Population from OECD and OECD.stat, education and skills (primary, secondary, and tertiary education (excluding post-secondary non-tertiary education so as not to double-count spending on ALMPs and educational programmes)).

Figure 7. OECD countries’ ALMPs as a percentage of GDP (2009)

Source: Social expenditures OECD; employment and labour markets: Key tables from OECD (OECD 2011).

Second, the different patterns of spending have effects on the sustainability and nature of the policy. American states, for example, operate under balanced budget constraints. This makes them procyclical, along with any services they provide – they have more money to spend in good times and less to spend in bad times (when demands on the welfare state are likely to be higher). Thus, for example, it is no surprise that US higher education spending is procyclical: much of it is by states. Likewise, the large share of public assistance entrusted to states means that these are less effective as automatic stabilisers than might be thought. The United Kingdom, by contrast, achieves very similar welfare states because of its simple division of powers (98 per cent comparability is quite strikingly high compared to the US and probably most other states, where funding streams are much more mixed) and block budgets.

Third, the effect of breaking down spending by both programme and territorial level highlights the importance of institutions. The US federal government could in theory legislate universal single-payer healthcare or massively increase its higher education spending, so in that sense the variation and averages in state expenditure are a federal decision (though one attributable to fragmentation in Washington, not the existence of states). But given the well-established difficulties of passing comprehensive and coherent legislation in Washington, it seems more useful to appreciate the areas of state discretion and the effects of state discretion. This is especially so at the time of writing, when an historically unusually number of states are under one-party government are unsurprisingly diverging rapidly in many kinds of public policies including implementation of the 2010 Patient Protection and Affordable Care Act.

The United Kingdom has a party system and legal framework seemingly designed to promote divergence, but the financial system has substantially constrained it and produces what we have seen so far: divergence primarily in how things are done, not how much of them are done. Greater financial autonomy might produce variation not just in organisation but also in the priorities and levels of welfare spending.

7. Conclusion

If the groupings of liberal, conservative-corporatist, and egalitarian welfare states matter, and more literature than can possibly be contained in a bibliography argues that it does, then the finding, albeit modest, that in issues of social risk there is practically as much difference within a single country as there is between groups of countries falling into these pre-established ideal types is unnerving. Moreover, as Lynch (2006) showed, there is little doubt that the age orientation of a welfare state has tremendous distributional and policy implications. By this same token, the breadth of variation within a given country reveals important dynamics not only about the type of territorial politics but also about the differentiation of policy debates. In this article we showed how two superficially similar liberal states turned out to have very different territorial social politics with regard to both age and unemployment. Further, we hope we showed the extent of variation and its potential usefulness not just in domestic but in cross-national research. The variation permits a renewed evaluation of the usefulness of averages in comparative research as well as suggesting the need for a more sophisticated treatment of territorial politics in theories of welfare state politics.

Methodological nationalism and “mean-spirited thinking” (Snyder 2001, p. 98) afflict welfare state studies as much as any other part of comparative politics, and it probably explains why it has been so difficult to move beyond the old arguments that first divided centralisers and decentralisers at the dawn of the twentieth century. Decentralised states have lower welfare state efforts but we know neither why nor how; it is possible that the correlation is basically spurious. The main line of attack on such macro-comparative arguments has been to point out cases of regional mobilisation in defence of welfare states, or regions with higher welfare spending, or regions with distinctive welfare policies.

We went a step further and investigated the share of regional and central expenditure in the two countries by expenditure category. We have presented quantitative evidence of the scale of differentiation that is possible within countries, thereby questioning the basic value of methodological nationalist analyses that, for example, handle an average value for the United States and seek to explain it without examining its subnational territorial structure. The United Kingdom has much less variance in expenditure, reflecting its institutional structures, which means that the usefulness of averages is in itself a reflection of the structure and extent of decentralisation within the country.

The United States has almost as much internal variation as the OECD has between countries, which is a remarkable finding in itself. The causes are clear enough since in the US state spending is focused on social investment and protection against new social risk, while federal policies are mostly targeted towards the elderly. What is striking is how this division of labour, combined with the diversity of states, produces radically different age and ALMP orientations. The contrast with the United Kingdom illustrates the extent to which institutional variables can support a very different approach; while the division of labour between regions and the central state is quite similar, the financing of regions in our time period gave them almost no leeway to change the variance between their expenditure levels. Wales or Northern Ireland, simply put, had a guaranteed budget to spend on social services that US federal states cannot, or choose not to, have. Divergence happened either because of financial formulae or in-service organisation, not due to the financial decisions of the three devolved administrations (Birrell 2009). The result is that the United States has 50 different approaches to new social risks for the under-65s, as varied in their age profile as the whole OECD, while the UK jurisdictions vary in policies rather than expenditure levels.

At the present, the approach adopted in this article has some limitations. First, at this stage, the availability of data does not permit longitudinal comparisons. Moreover, our data shares the drawbacks of a spending variable in comparative welfare state studies. The internationally comparable data for any aspect of welfare state policy typically focuses on inputs – hence the phrase “welfare effort” – and therefore ignores all questions of the organisation and effectiveness of provision (see Jochen and Siegel 2007). The focus on expenditures distinguishes it from approaches that focus on the issue of benefits eligibility and its measurement (Scruggs 2006; Scruggs and Allan 2006). The argument for eligibility, as against expenditure, data has two legs. One is its political realism: as Esping-Andersen famously wrote, “it is difficult to imagine anyone struggled for spending per se” (Esping-Andersen 1990, p. 21). The other is the risk of spurious results – governments can cut back eligibility and benefit levels without decreasing expenditure if they do so when need is increasing (lower unemployment benefits can coexist with higher unemployment replacement expenditures in a recession).

Nevertheless, we opted for expenditure data. First, in territorially decentralised countries people do struggle for spending per se. Intergovernmental finance is the stuff of much intergovernmental relations (Rodden 2005; Trench 2006; Bednar 2009; McLean et al. 2009; Weingast 2009; Beramendi 2011). Second, we believe that many of the pitfalls (such as the unemployment insurance example) could be handled with covariates such as unemployment. Third, we see no way to collect region-level data on eligibility: not only would it be a very long and expensive process, it would also be nearly impossible in the case of eligibility standards for complex services such as healthcare or education.⁸

For scholars, the conclusion is that methodological nationalism obscures a great deal of important variation, and that disaggregating decentralised states reveals a much more complex and powerful set of explanations than the mere concept of fragmentation. For policy, the conclusion is that the design of intergovernmental relations, law, and finance is very important for the welfare state. Party politics in the UK diverge more than party politics in the United States, with explicit nationalists in government in Northern Ireland, Scotland, and Wales, and very different partisan configurations, but the financial structure limits their ability and need to vary overall expenditure.

In our specific cases, of policies to address “new social risks” in the United Kingdom and United States, we found that the structure of intergovernmental responsibilities and finance qualify for generalisations about both the role of decentralisation and about their approaches to new social risk. The United States federal welfare state’s substantial bias towards the elderly means that states’ expenditure decisions have dramatic effects on the distribution of resources between the under- and over-65s, contributing to substantial territorial variation in welfare state strategies. The programmatic design of the US welfare state focuses variation on services for the non-elderly, which depend more on state politics. By contrast, the UK legally permits much policy variation and, like the United States, focuses the activity of the central state on pensions, but the centralisation of finance and provision of under-65 income replacement at the level of the UK government combine to substantially harmonise expenditure on new social risks across the UK. In neither case has there been a simple federalism or decentralisation effect as public choice and macro-sociological accounts might suggest; rather, the specific allocation of responsibilities and funding responsibilities has deeply shaped the meaning of territory and federalism for citizens.

Supplemental data

Supplemental data for this article can be accessed at http://dx.doi.org/10.1080/13876988.2015.1060713.

Additional information

Notes on contributors

Scott Greer

Scott L. Greer is Associate Professor of Health Management and Policy at the University of Michigan and Senior Expert Advisor on Health Governance at the European Observatory on Health Systems and Policies. His research focuses on federalism, decentralisation, and Europeanisation in social policy.

Heather Elliott

Rebecca Oliver is currently a Lecturer in the Department of Political Science at the University of Southern California. Her research examines the politics of inequality with respect to labour markets and social policy in Europe. Substantive topics of her work include labour union strategies, collective bargaining institutions, childcare policy and territorial inequalities in social spending.

Rebecca Oliver

Heather Elliott is a joint PhD candidate in the departments of Political Science and Health Management and Policy at the University of Michigan. Her research focuses on the comparative political economy of health workers in Europe and the United States, health politics and policy as well as territorial politics.

Notes

1. The list of Spanish regions with distinctive social policies does not look much like the list of stateless nations (Gallego 2001; Gallego et al. 2002).

2. “State of Unions” that the UK is, the phrase must be qualified. It previously had an autonomous Northern Ireland government in its territory from 1921 until it was suspended in 1972 (see Mitchell 2009).

3. The PESA series is comparable across short periods, but periodic recalculations of the data mean that it is not comparable over more than about five years.

4. Data and detailed coding notes are posted at http://sph.umich.edu/faculty-profiles/greer-scott.html.

5. In line with Lynch (2006), housing is excluded from the equation.

6. http://www.hhs.gov/recovery/programs/tanf/tanf-overview.html (accessed May 2013). Under the welfare reform legislation of 1996, TANF has replaced the Aid to Families with Dependent Children (AFDC) programme, the Job Opportunities and Basic Skills Training (JOBS) programme, and the Emergency Assistance (EA) programme. More recent details on the changes to TANF in Schott and Pavetti (2011).

7. In line with categorisation of programmes within the OECD Employment Outlook (2011), Statistical Index, p. 273.

8. With regard to the TANF data, we note Gais’ (2009) discussion of the pros and cons of using “State Spending on Cash Assistance, Per Poor Person” and decided against it since this method appears to pose greater difficulties when between-state comparisons are embedded within international comparisons.