Welfare Stateness as an Explanatory Concept

Abstract

After the preceding systematisation of the main assumptions about the influence of the welfare state on individual-level outcomes, this chapter turns to the derivation of specific measurements of welfare stateness that capture the mechanisms underlying the explanations and the corresponding hypotheses.

After systematising main assumptions about the influence of the welfare state on individual-level outcomes, I turn to deriving specific measurements of welfare stateness that capture the mechanisms underlying the explanations and the corresponding hypotheses. According to Adcock and Collier, “[v]alid measurement is achieved when scores […] meaningfully capture the ideas contained in the corresponding concept” (Adcock & Collier 2001: 530). Achieving such meaningfulness is a considerable challenge and, to my knowledge, there has not yet been a comprehensive attempt to derive measurements for specific concepts of the welfare state that serve as explanans.

A very good and recent example for a similar undertaking is given in a publication by Andersson (2018). She proposes a theoretically grounded way to operationalise the specific examples of social investment policies. Relying on Sartori’s (1970) ladder of abstraction, she first deduces relevant dimensions of the broad concept of social investment, before suggesting specific variables capturing these dimensions. Although her aim is to provide a conceptual framework leading to the proposal of a typology of social investment approaches, the main objective of her study—to improve and standardise the measurement—is related to the aim of my project. In this chapter, I pursue a similar approach of isolating indicators by gradually narrowing down the underlying concepts. To this end, I draw on an analytical framework proposed by Adcock and Collier (2001), which shows resemblance to Sartori’s ladder but focuses more on deriving indicators for empirical operationalisations.

Figure 5.1

(Own figure based on Adcock & Collier (2001: 531))

Conceptualisation and Measurement.

Adcock and Collier’s framework starts at an abstract level. The background concept is very broad and includes all meanings associated with a research object. This most general way of looking at a concept is usually not the one that is actually addressed in specific research questions. Instead, the focus is on only one specific meaning or understanding. This selection of specific parts within a broader background concept is guided by the research question and it involves the task of conceptualising a given phenomenon in depth. This leads to the second step in the framework, which is the formulation of a systematised concept. Up to this point, the main objective has been a conceptual one, aimed at identifying the relevant understanding of a given phenomenon. As soon as the systematised concept is deduced and defined, the question of empirical operationalisation arises. The third step within the framework is thus to select indicators that allow the systematised concept to be captured empirically. The last step consists of the actual attribution of scores for cases based on these indicators. After implementing the measurement and putting it to the test, the model also includes the possibility to revise the measurement based on empirical observations (Adcock & Collier 2001: 530). Figure 5.1 summarises these steps and tasks.

Their framework fits my objective for several reasons. First, it is possible to combine their proposal with the criteria derived at the end of chapter 3. Comparability, clarity, and availability have been identified as key factors to consider when selecting an operationalisation. These three criteria can be well integrated into the argumentation when it comes to selecting indicators (third step in Figure 5.1). Second, their framework allows for continuous improvement as it implies a circular process of approximation. Since I hope that my systematisation of mechanisms, concepts and indicators will be critically discussed, applied and expanded in future research, it seems very useful to review each step and modify it if necessary. Third, it enables a differentiation of perspectives within a broader phenomenon through a specific distinction between a background concept and systematised concepts. In this case, this refers to different conceptualisations of welfare stateness depending on the respective dependent variable, research question, pursued explanation and hypotheses. Fourth, and perhaps most importantly, it emphasises the importance of separating conceptual issues from issues related to measurement. Since I argue that the conceptualisation of welfare stateness is lacking prior to its operationalisation as an independent variable, it is particularly important to distinguish between the concept and the resulting possible measurements. This additional step holds the potential to narrow down and standardise distinct perspectives on welfare stateness as an intermediary step.

5.1 From Background Concept to Systematised Concepts

As a background concept, the welfare state embodies various different meanings, influences different societal levels and affects all the functions and mechanisms discussed in relation to social policy-making. As such, the welfare state is tied to institutions and institutional developments on the macro-level and to individuals as beneficiaries, contributors and/or judges on the micro-level. Likewise, it can be an explanatory factor and/or the phenomenon to be explained as it is interrelated with structures and dynamics on both levels. It causes outcomes and is shaped by feedback processes. Industrialisation and democratisation were certainly relevant drivers of welfare state development and change (chapter 2), but conversely, social policy arrangements have also shaped macroeconomic and societal outcomes. Not least, because they are strongly linked to economic cycles and demographic developments. Moreover, there are also reciprocal relationships with the micro-level. Since welfare states are shaped by governments, individuals do in fact possess the opportunity to influence the direction of social policy-making in elections (this idea of simultaneous feedback is discussed in more detail by Breznau 2017, 2018).

With the elaborate conclusions from previous chapter in mind, it is apparent that conceptualising ‘the welfare state’ as one all-encompassing construct is not conducive if it is to serve as an independent variable. Therefore, if we want to approach a more precise, standardised and comparable manner of looking at distinct facets of welfare stateness, the idea of deriving systematised concepts is quite promising.

In the broadest sense, the previous chapter revealed two ways of approaching the welfare state: (1) top-down, as a responsive and not necessarily visible actor, and (2) bottom-up, perceived as a visible and salient institution. These two perspectives frame the welfare state in different ways. Within these two main perspectives, several different mechanisms are at work, depending on the research question, explanation and hypotheses.

The top-down perspective takes up mechanisms that explain why and how different welfare states in general and different social policies in particular affect individuals. As illustrated in the review of hypotheses in the previous chapter, those mechanisms correspond to the functions of the welfare state highlighted in the second chapter of this book. In this sense, the welfare state can constitute a source of security against risks. Moreover, it redistributes resources, shapes patterns of social inequality, enables or incentivises specific behaviours and shapes a normative reference frame. Some of these functions overlap quite a lot, while others offer a more isolated view.

In contrast, the bottom-up perspective explores why and how individuals perceive and evaluate welfare states and why and how these assessments shape outcomes—especially related to behaviours and attitudes. Two mechanisms may be in effect, which both entail an evaluation of social policies by the individual. On the one hand, the evaluation may follow self-interest. On the other hand, other explanations—such as an activation of individual values—are possible. Since both processes are based on the same premise, the systematised concept of welfare stateness is similar: a welfare state that is being assessed by individuals. Since mechanisms do not yet reveal systematised conceptualisations of the welfare state, I would like to use and further distil the insights gained so far on theoretical premises, mechanisms and hypotheses.

Looking at the welfare state from the broadest perspective, all seven discussed mechanisms are related to each other—even transcending the two main perspectives. Modes of redistribution may mould political support for the welfare state and public support—if reaching a critical mass—can change the logic behind redistribution of resources. Equally, the way in which the welfare state stratifies societies is linked to how it socialises respondents and this may again be subject to individual evaluation. Incentives may be shaped by patterns of inequality instead of intended (or unintended) activation through specific policies, which in turn can determine how effective a welfare state responds to risks. One could further combine these different mechanisms and it would become clear that all these different processes related to social policy-making are interconnected in one way or another. Considering that they are tied to the same background concept—the welfare state—this is no surprise. In deriving systematised concepts based on these mechanisms, the question is therefore not which mechanisms are at play, but which mechanisms are most characteristic. Here, an important distinction lies in the key actor within an explanation– welfare state or individual—and the mechanism (or mechanisms) highlighted.

Following this approach, there are at least four systematised concepts that emerge from the discussion so far. Three of these concepts refer to the top-down perspective; one addresses the bottom-up perspective. They are summarised in Figure 5.2. This figure also highlights the most relevant mechanism(s) for each concept. It should be reiterated that other combinations are possible.

Figure 5.2

Conceptualisations of the welfare state

The Responsive Welfare State emphasises those mechanisms that determine how effective a welfare state operates. Especially characteristic mechanisms are the efficiency of reducing risks and uncertainties (security) and reducing social inequality (stratification). This happens particularly through redistributive processes. Evidently, the responsiveness of welfare states could also be determined by its potential to activate individuals and promote specific norms and justice principles (socialisation). These mechanisms are however of secondary importance in this case.

The Enabling Welfare State is characterised above all by the mechanism of activation. The name is based on the term enabling state used in the literature, which is introduced in the second chapter of this book. This systematised concept highlights processes in which the welfare state incentivises and enables individuals to take responsibility for their own welfare. As discussed, one way this is achieved is through labour market integration through investment in human capital. Again, there is overlap with other mechanisms, but the focus of the conceptualisation is on activation.

The Normative Welfare State represents the last systematised concept within the top-down perspective. Socialisation is particularly characteristic here, while other mechanisms are possibly at play but are only relevant in terms of their socialisation potential (e.g. when it comes to justice principles underlying redistributive logic).

In contrast to these three conceptualisations, the Assessed Welfare State represents a bottom-up perspective. Here, individual evaluation in general and activation of self-interest in particular represent the most relevant mechanisms. Even if individuals form attitudes and behaviours based on an assessment of the five top-down mechanisms, it is their perception of those processes that matters. Moreover, these individual perceptions can deviate substantially from how policies are designed and performing in reality.

At this point, it is important to reiterate that this summary does not claim exhaustiveness. Its purpose is to illustrate a possible approach to conceptualising the welfare state in a way that is explicitly intended to treat it as an explanans. In this contribution, a selection of dependent variables is explored and adding others may reveal further conceptualisations or adjustments of my proposal. Following the selected measurement framework, revisiting the proposal critically is not just possible but highly encouraged. While this is important to recognise, the central message is that when adressing any of the concepts of welfare stateness in a hypothesis, it is important to clarify which perspective and mechanism is chosen (and why). Committing as clearly as possible to a systematised concept, will allow the most targeted selection of indicators. Moreover, by stating, which perspective is adopted, comparability should be increased considerably. Finally, a combination of different concepts can be beneficial to many research objectives. This includes combinations of top-down and bottom-up perspectives. The strength of adding systematised concepts of welfare stateness as an intermediate step to empirical operationalisations lies in the limitation to specific perspectives. By not attempting to cover all facets and meanings of the welfare state, theoretical arguments on the one hand and empirical decisions on the other are sharpened.

Overall, this proposal resembles a typology. Unlike the typologies discussed so far, however, this is not a typology of welfare state regimes, but of conceptual perspectives that we encounter when we assume that social policies affect individual outcomes—be they risks or attitudes or behaviour or more. Intended as an auxiliary tool, it still has to prove its usefulness in this contribution and hopefully further research on the matter.

5.2 Indicators and Scores for Cases: Criteria and Measurement Validity

While the previous section sketched a possible framework for conceptualising distinct perspectives on the welfare state, the following part discusses how to get from concept to operationalisation aiming at ultimately deriving specific indicators. The analysis of the problems associated with common operational approaches in the third chapter of this book led to the identification of three criteria that are often not satisfactorily met: (1) comparability, (2) clarity, and (3) availability. These three criteria are discussed in detail in this section, with emphasis on how systematised concepts of welfare stateness may help to solve underlying issues at least partly. This is followed by a discussion of how measurement validity can be determined.

5.2.1 Criteria for the Selection of Indicators

The first criterion, comparability, relates to four issues: (1) different kinds of indicators, (2) different sources for the same indicator, (3) different points in time, and (4) different countries. Turning to the first, it seems almost trivial to say that comparable indicators need to be used. However, as explained in chapter three, this is often not the case. Instead, different types of indicators are compared in the review of literature as well as in the selection of empirical operationalisations. This happens for a number of reasons already outlined—not least because of the limited availability of many indicators.

Comparability can be affected not only by the use of different indicators but also by using the same indicator based on different data sources. In the third chapter, I referred to the case of net replacement rates and illustrated different effects obtained by using two different versions of this indicator. Such deviance in results seems highly problematic—especially since someone less familiar with the surrounding debates would probably not expect it. As discussed in detail by Ferrarini and colleagues (2013) and others (such as Scruggs 2013; Wenzelburger et al. 2013), different values result from different theoretical premises and operational decisions. The discussion shows that comparability is not just a matter of comparing different indicators but also different operationalisations within one indicator. Both issues can be solved at least partly by being more aware of the issue as well as discussing and explicating the selection. Thus, differing results in literature using different indicators for testing similar hypotheses, should not come as a surprise but should instead be expected. In my opinion, much can be achieved by providing an overview of possible indicators and results obtained from different approaches.

The situation is more difficult with the other two issues that could affect comparability. The same score for an indicator can have a different meaning depending on its context (Adcock & Collier 2001: 534). This can refer to different times, in which specific indicators have different meanings (e.g. due to external shocks or reforms). Furthermore, it can refer to different countries—the latter being of particular importance. As the case of CEE countries frequently shows, there may be severe differences between de jure and de facto benefits and other issues, which may complicate the use of established indicators in these cases. Recent contributions dealing with the greater inclusion of CEE countries in comparative welfare state research are a first step towards addressing the issue (e.g. Kuitto 2018).

So, how can comparability be maximised in the process of operationalising distinct systematised concepts of welfare stateness? First of all, explaining which concept is adopted and why draws attention to how the concept should be operationalised. By reducing the welfare state to narrowed and specific concepts instead of a comprehensive and multidimensional background concept, it should also be easier to determine the comparability of different operational decisions as they are limited to a clearly defined part of the welfare state. This especially helps solving issues associated with the selection of indicators but it should also help to assess the context-spanning comparability of operationalisations.

The criterion of clarity is related to comparability but emphasises a particular concern. Indicators should not only be comparable; they also have to be sufficiently meaningful and interpretable. In other words, it should be as obvious and clear as possible what an indicator is measuring. In the case of social policy indicators, this is often not the case. Several examples for this were used repeatedly throughout this book. Indicators on social expenditure and regime approaches are the most affected. Both kinds of operationalisations give too little indication of what they actually measure. As the case of social expenditure as an indicator of welfare effort shows, it cannot be ruled out that high spending in one country is a sign of high generosity, while equally high spending in another country may indicate a high number of beneficiaries. Moreover, even controlling for benefit receipt (e.g. by adding unemployment rates) still does not reveal redistributional principles, eligibility rules and many other important aspects. Regime approaches, on the other hand, incorporate too many facets of welfare stateness making it impossible to determine, which one is responsible for an observed effect. In addition, cross-cultural differences are marginalised by reducing variation between countries to a handful of types. Such approaches, which aim at grasping either distinct types of welfare states or overall measures of redistributive budget in a broad manner, have their place in comparative welfare state research. However, they are rarely suitable for testing hypotheses. Too little is known about what they capture to test whether specific mechanisms behind causal assumptions prove to be valid. The same can be said about composite indicators (Kvist et al. 2013: 332).

Much of the literature and my own analysis in chapter 3 are consistent with this assessment of the regime and expenditure approaches,¹ but other operationalisations receive less criticism. However, popular indicators such as replacement rates may be affected by similar problems—albeit to a lesser extent. For instance, addressing average benefits (e.g. in the case of replacement rates) may not adequately reflect the situations of different income groups (Kvist et al. 2013: 329).

The use of specific concepts rather than broad conceptualisations of welfare stateness helps to provide a frame for discussions on clarity. Like in the case of comparability, it reduces the complexity of potential operationalisations by limiting them to those relevant for the addressed mechanisms. The main agenda here is to find indicators, which meaningfully and comprehensibly capture the specific aspect of social policy-making relevant to test respective hypotheses.

Availability is a rather pragmatic criterion and it is closely tied to data sources. As such, it is the most strongly affected by a need for compromise. As previously discussed, availability of indicators is often limited to country samples and periods. When selecting indicators, it is thus important to make sure that operationalisations can be replicated as easily as possible. This entails compromise, because focussing on accessible indicators may restrict the selection. While it cannot be ruled out that even better measurements than the ones available exist, the advantage of relying on easily accessible data sources is its high comparability with other research using the same sources and indicators. This issue will be discussed in more detail in a later section of this chapter, where data sources are explored (cf. Section 5.4). For now, besides comparability and clarity, the availability of indicators ought to guide the selection of meaningful operationalisations of welfare stateness in a given systematised concept.

The three criteria discussed so far, highlight important issues, which cannot be solved entirely at this point. Instead, they require compromise at first and more extensive conceptual and empirical work in the long run. Hereinafter, I am going to substantiate the required proceeding further and I will illustrate its potential application in the next chapter. The fact that compromises are necessary (to meet the three criteria) should in no way discourage further research into improved measurements to separate the wheat from the chaff (in the words of Bambra 2007). After all, at this point of the discussion there seems to be evidence for quite a bit of chaff out there. Applying these three criteria, the use of typologies and overarching composite measures should be ruled out in many cases. The same could be said for measures of social expenditure. What remains and calls for a more detailed discussion are other single indicators that—precisely selected—can do exactly what is intended: addressing distinct perspectives on the welfare state and clear-cut concepts within the broad spectrum of welfare stateness to test hypotheses about the impact of specific social policies on individuals.

5.2.2 Determining Measurement Validity

Once indicators have been selected, it must be clarified whether they meet the criteria for measurement validity. Adcock & Collier (2001: 538–543) highlight three types of measurement validation: content validation, convergent (or discriminant) validation and nomological (or construct) validation, which are summarised in Table 5.1. The table also points out the main disadvantages of each approach that need to be taken into account.

Table 5.1 Types of measurement validity

The first type of validation (content) occurs on a conceptual level. It requires a certain amount of agreement on the nature of the examined phenomenon and involves conceptual reasoning. In contrast, the other two types (convergent and nomological) address empirical aspects and applications. In line with Adcock and Collier, I believe that distinguishing between these types of validation helps to assess which indicators are a better or worse fit for a concept. Moreover, the distinction between conceptual and empirical issues is particularly important in the specific case of welfare state indicators.

After discussing criteria that potential indicators should meet and the types of validity against which they should be measured, the following section is devoted to identifying concrete indicators for each of the four systematised concepts.

5.3 Measurement of the Systematised Concepts

While the dependent variables assumed to be influenced by social policies are diverse, the underlying assumptions are not. In the course of this chapter, four systematised concepts of welfare stateness have been highlighted: the Responsive, Enabling, Normative and Assessed Welfare State. While the first three capture the top-down perspective, the latter represents the bottom-up perspective. These concepts help to narrow down the role of the welfare state for specific research objectives and hypotheses, thereby making its operationalisation simpler and more standardised. After the rather abstract discussion so far, concrete indicators for the individual concepts are mentioned below. Since I am only approaching a solution for the independent variable problem discussed in this book, these recommendations are phrased relative to possible alternatives. A deterministic proposal of operationalisations seems inadvisable for two reasons. First, I do not expect my proposal to be an exhaustive solution to all problems related to the use of the welfare state as an independent variable. Rather, it is intended as a blueprint for a possible procedure. Secondly, some of the recommendations cannot (yet) be based on empirical evidence. Rather, they represent expectations arising from the preceding discussions. I will empirically test some of them in the next chapter to determine their validity.

Even before going into the individual concepts in more detail and proposing indicators, two very specific recommendations can already be made at this point, as they relate to issues that have arisen repeatedly throughout this book. One addresses the shortcomings of composite approaches, the other those of expenditure approaches. Neither seems to be well suited as independent variable.

Recommendation 1: The use of single indicators instead of composite measures such as regime typologies and indices allows a more targeted operationalisation of systematised concepts.

This recommendation is supported by the discussed lack of clarity, comparability and applicability of composite measures. If one assumes that a certain area of the welfare state is effective and follows a certain mechanism, composite measures cannot be used to determine which part of the multidimensional operationalisation is responsible for the observed effects. Drawing conclusions about mechanisms from such measures is therefore highly problematic in most cases. This view is supported by several voices in the relevant literature, highlighting the importance of distinguishing between important aspects of welfare stateness in order to accurately assess effects (Palme 2006: 400; Kvist et al. 2013). It is crucial to add that there can be cases in which regime approaches are a feasible option, since the conceptualisation does not imply mechanisms addressing distinct policies. This can occur if a somewhat latent welfare culture is addressed—for instance in the case of the Normative Welfare State, which will be discussed later in this section.

The second recommendation addresses expenditure measures. Even though they possess the great advantage of being easily available and frequently updated, they are no panacea to the problem of operationalising welfare states as independent variables. While they represent the overall redistributive budget or the “concept of ‘welfare resources’” according to Dahl & van der Wel (2013: 61), they only partly reveal information about which of their components is responsible for an observed effect. Therefore, their applicability and potential to reveal in detail how and why welfare states shape outcomes is limited. This view is supported by various scholars, raising concerns about the validity of expenditure data (e.g. Kvist et al. 2013).

Recommendation 2: Social expenditure should not be chosen as a targeted operationalisation of systematised concepts. Only hypotheses addressing the redistributive budget can be more adequately modelled by expenditure indicators than by others (e.g. institutional).

While these two recommendations address general issues, the following sections elaborate on a possible measurement for each of the four systematised concepts. Two general questions will guide the selection of indicators for each of the four concepts. (1) Which policy field is addressed in the hypotheses? (2) Is an immediate effect of policy arrangement assumed or one that entails some delay? These two issues provide an overarching frame, relevant to all concepts and they address important aspects of the operationalisation. The first issue relates foremost to the policy area covered by an indicator. While the general nature of an indicator is the same within each concept (e.g. eligibility, expenditure, universalism et cetera), the policy area determines which risk is highlighted (such as old age or sickness). The second issue adds a temporal perspective. In some cases, recent policies are relevant, while in others, long-term policy developments or policies at a particular point in the respondent’s life are important. Since these two aspects apply in all cases, they will not be discussed separately for each concept of the welfare state. Instead, they are highlighted only when they are particularly relevant and are taken up again in the summary discussion at the end of this chapter.

5.3.1 The Responsive Welfare State

The Responsive Welfare State is arguably the most obvious systematised concept within the welfare state and is characterised by how it fulfils its essential task of preventing risk and meeting need. Furthermore, since reducing social inequality receives growing attention in political agendas, shaping social stratification can be partly attributed to such main tasks as well. As the preceding literature review and discussion reveals, the Responsive Welfare State is embedded most frequently in research examining how and why different welfare states show varying degrees of success in securing individuals and reducing social inequalities and inequalities in outcomes. Exemplary dependent variables, which are closely related to the Responsive Welfare State are various risks (such as poverty), as well as manifestations of well-being (such as health).

As discussed in the previous chapter, the literature dealing with methodological issues related to the selection of an independent variable in this context emphasises the role of the institutional approach (e.g. Ferrarini et al. 2014a). Particular attention is paid to indicators measuring eligibility, generosity of benefits and population coverage. These indicators capture the social rights perspective and address different areas within it. Indicators of eligibility include who is entitled to benefits, while generosity refers to the type of benefits and coverage measures how widespread access to such benefits is in a society. All of these issues refer to immediate responses to risk: who is covered, who is entitled and how much does a person at risk receive? In contrast to other aspects of welfare stateness such as redistributive budget or overall welfare culture, indicators of social rights represent features of welfare states that capture characteristic responses to risk.

Recommendation 3: The Responsive Welfare State should be operationalised using indicators for social rights (e.g. benefit generosity, insurance coverage, eligibility criteria).

It seems plausible to expect that, depending on the dependent variable and the research question, not all indicators of the social rights perspective are equally suitable. The previous chapters have also discussed evidence that some indicators produce unexpected results when others are not controlled for in the same model (e.g. in case of net replacement rates and benefit coverage). This will be taken into account in the empirical test of various potential indicators in the next chapter.

5.3.2 The Enabling Welfare State

Highlighting the mechanism of activation, the Enabling Welfare State is characterised by policies, which incentivise a specific behaviour. This can happen in an intended as well as in an unintended way—though the former is more relevant in this case than the latter. Indicators capturing this conceptualisation of welfare stateness should thus highlight policies, which are intended to incentivise a specific outcome. Operationalising the Enabling Welfare State in the field of family policies presents a very tangible example for such intended incentives. Here the main question is how strongly equality is emphasised when it comes to the relationship between care responsibilities, housework and employment. In this context, the Enabling Welfare State can thus be operationalised by features tied to incentives for equal division of labour and female labour market participation. For instance, this includes public childcare provision (Brewer & Shaw 2018).

The case of enabling in the field of employment is similarly palpable. Here, active labour market policies are the most obvious choice. However, their measurement is not as obvious. Apart from expenditure in the area of active labour market policy, there do not seem to be many alternative indicators. The same can be said about other indicators of social investment. As mentioned before, the empirical operationalisation of social investment is an ongoing issue.

Regardless of the emphasised policy field, indicators capturing the Enabling Welfare State should model an activating component. This can relate to a specific field (such as childcare provision or ALMP) or can refer to social investment in a more general sense.

Recommendation 4: The Enabling Welfare State should be operationalised with the help of indicators for efforts to activate and provide incentives (e.g. active labour market policies).

As long as it is assumed that the indicators capture enabling components of the welfare state, they are suitable for measuring the Enabling Welfare State.

5.3.3 The Normative Welfare State

While responsiveness and activation are linked to specific policy goals such as risk reduction and labour market integration, the normative welfare state is characterised by its potential to socialise the individual, which is more subtle. There are only a few examples where conveying justice principles and egalitarianism are seen as actual policy agenda. Indeed, welfare states can create normative environmental pressure on companies, leading (for instance) to more work-family support within them (Beham et al. 2014: 34–35). Such pressure can be politically stirred and such active shaping of normative frames by political actors may be found in other examples as well. Still, in the majority of cases—including those related to attitudes formation—the socialising influence of welfare states is more likely to be unintentional. This makes the proposal of indicators more difficult. The Normative Welfare State has to be captured through policy measures, which are tied to dominant justice principles, like egalitarianism—without this being necessarily part of political agendas. In principle, various kinds of social policies can thus have a socialising effect, since they all shape what is perceived as legitimised and institutionalised norm and serves as a reference frame for individuals. Narrowing down specific indicators therefore means identifying which policies are particularly tied to cultural principles.

According to the literature discussed, policies that determine the principles of stratification and eligibility are particularly relevant. Thus, the degree of equality and universalism in benefit access and provision appears to be especially important. This excludes, for example, the use of social expenditure, as it may reflect the welfare commitment, but does not reveal principles guiding social policy-making. In contrast, it is however worth discussing, whether the regime approach, which has received criticism so far, is somewhat suited in this particular case. Given that many typologies capture the overall design of welfare states and their dominant characteristics, they come closest to what is sometimes referred to as different “welfare cultures” (e.g. Vrooman 2013). While this does not help to overcome the problem of the approach’s lack of clarity, this fuzziness does not appear to be as problematic in this case as in other cases where mechanisms address clear implications of policy-making (such as lowering specific risks). This should be kept in mind and it seems worth discussing, whether indicators of universal access (e.g. benefit coverage) are better suited as proxies for a normative framework than typologies. Regardless, the recommendation at this point is to focus more on indicators of organisational principles than expenditure or other indicators of budget size.

Recommendation 5: The Normative Welfare State should be operationalised using indicators linked to principles that guide the provision of social services, such as universalism or egalitarianism (e.g. coverage, eligibility criteria).

In this recommendation, welfare regime approaches are intentionally left aside. As mentioned, it may be worth discussing, whether they could aid in capturing the Normative Welfare State. Since however, I focus particularly on defining characteristics of each systematised concept of welfare stateness and intend to compare different operational choices, using typologies does not fit the agenda of this project.

5.3.4 The Assessed Welfare State

While the three conceptualisations discussed so far follow a top-down perspective, the Assessed Welfare State represents a bottom-up approach. It particularly emphasises individual evaluation and self-interest as mechanisms. Even though self-interest implies a (cost-benefit) evaluation of what is perceived to be the social policy arrangement in a relevant context, not all hypotheses addressing evaluation, necessarily include a rational-choice argument for the link between evaluation and outcome. As discussed previously, other determinants—such as value orientations—can be more relevant than cost-benefit considerations. Thus, evaluation and self-interest represent different mechanisms within the Assessed Welfare State.

Various dependent variables are closely related to the Assessed Welfare State. Attitudes and behaviours are the most obvious ones discussed in the previous chapters. In addition, subjective manifestations of well-being (such as happiness) may entail this kind of conceptualisation as well.

The operationalisation of the Assessed Welfare State is more difficult than the previous types and the literature offers less guidance in this case. The most challenging aspect is that evaluation requires a certain amount of knowledge about relevant processes. However, as previously discussed (cf. chapter 4), citizens’ knowledge about spending and complex regulations appears to be rather unreliable. It thus seems very important to choose indicators that represent aspects that are likely to be familiar to most citizens. Literature treating the welfare state as an independent variable does not offer much insight in this case. We can still derive some hints from the mechanisms assumed to be in effect. For instance, if self-interest (or even rational choice) is assumed to be responsible for a given outcome (e.g. a specific attitude or behaviour), indicators that capture potential benefits appear more relevant, as it seems likely that there is at least vague knowledge about the average amount of replacement benefit one is entitled to and about the eligibility criteria. The same applies to explanations that do not rely on self-interest but on other aspects guiding evaluation of social policies. Salient issues, which are measured up against individual ideology, value orientation and other determinants of the formation of attitudes and behaviours, are likely those that are the most tangible for people who are not themselves beneficiaries of social security measures. It seems plausible to expect that this again relates to replacement rates and eligibility criteria, as most adult citizens should have at least some knowledge about what they could expect in case of need.

In contrast, indicators of procedural aspects—such as waiting days or benefit duration—are unlikely to be salient for individuals who have never been in any contact with social security measures. These considerations are derived from the previous discussions, but they lack empirical evidence.Testing citizens’ knowledge about specific policy aspects beyond expenditure or very particular examples (such as unemployment) and on a comparative scale seems to be a desideratum, which is worth exploring in detail in the future. Determining to what extent the indicators commonly used in comparative welfare state research correspond to individual knowledge would greatly facilitate the operationalisation of the Assessed Welfare State. For now, two recommendations can be stated:

Recommendation 6: If the mechanism highlights the individual evaluation of performance, the Assessed Welfare State should be operationalised using indicators representing salient evidence of the performance of a welfare state.

Recommendation 7: If the mechanism entails an activation of self-interest, the Assessed Welfare State should be operationalised using indicators representing salient potential benefits for the individual and their generosity.

While both recommendations take up the premise of salience, the latter relates especially to those research questions and hypotheses that highlight the mechanism of self-interest. Here, potential benefit seems particularly relevant. Additionally, it may also be worth exploring whether an operationalisation of costs could also provide valuable insights. Again, this goes beyond the issues examined in this book—yet individual (private) contributions to insurance schemes, for example, could be a measure of the perceived costs of the welfare state.

5.4 Data Sources

Four different conceptualisations of the welfare state emerge from the discussion in the previous sections of this chapter, which are used to identify possible empirical operationalisations for each concept. In addition, several criteria are reviewed that should be met by such operationalisations. These criteria include clarity, availability and comparability. Before turning to an empirical test and illustration of the proposed framework, I would like to take a closer look at the possible data sources that can be used for the indicators discussed. Although there are many sources of comparative data on social policies, many of them require some kind of compromise because at least one of the above criteria is not met. Therefore, this section briefly presents a selection of particularly useful data sources, including a brief assessment of their respective advantages and disadvantages.

Among the most commonly used datasets are those provided by international organisations such as the OECD and the World Bank, or by official statistical agencies such as Eurostat. These sources provide information on popular variables such as social spending (e.g. European Commission 2016; Adema & Fron 2019).

In addition to such sources of specific indicators, there are also comprehensive comparative datasets that contain a variety of variables on welfare states. The Comparative Welfare Entitlement Dataset (CWED, Scruggs et al. 2014), the Social Citizenship Indicator Project (SCIP, SPIN 2017) and the Social Insurance Entitlement Dataset (SIED, SPIN 2019a) were mentioned and/or used previously in this contribution. Other datasets with very general information on welfare policies include the Quality of Government—Social Policy Dataset (QoG, Teorell et al. 2019). These different sources are discussed and compared in detail for instance by Grünewald (2014), and they all have advantages and disadvantages. Some are limited in the size of their country sample, others in terms of years for which data are provided. Furthermore, the selection of indicators varies. Still, the listed examples share that they all emphasise the “old risks”. Thus, they are focussed mainly on decommodification and unemployment, pensions, sickness and disability policies. Family policies or social investment are not widely covered. Information on such ‘newer’ risks and policies are included in more specified datasets. For instance, the Multilinks Database (Keck & Saraceno 2012) offers information on family policies; so does the Child Benefit Dataset (CBD, SPIN 2019b).

One could add many other examples to this list but I would like to refer to existing comprehensive overviews instead (e.g. Grünewald 2014; Lohmann & Zagel 2018). For the purpose of this contribution, it is mainly important that information on many aspects of welfare stateness does exist—even though it is scattered over different sources and not always available for all countries and periods. One additional deficit has to be noted: information on social investment is indeed scarce (Andersson 2018)—at least when aiming at an operationalisation that goes beyond social expenditure.

At least two aspects need to be considered when choosing data. First, there are a number of sources from which to choose. The available indicators, the country sample, and the years for which data are available should guide the selection. Second, especially the last aspect (the reference year or period) requires additional attention. Here we may have to take into account a certain time lag in the impact of social policy on a variety of issues. Especially drastic changes in social policy-making—for instance as a result of external shocks such as the ‘Eurocrisis’ or reforms such as the ‘Agenda 2010’ in Germany—will not immediately impact outcomes. This refers especially to hypotheses referring to individual perception and evaluation of social policies (the Assessed Welfare State) as well as socialisation (the Normative Welfare State), where policy changes have to sink into the fabric of societies first before they will shape individual outcomes. This raises entirely different issues regarding the age (difference) of macro and micro data depending on hypotheses. Although this problem is not the focus of this project, it will be kept in mind in the next chapter, where the conceptual considerations of this chapter will be discussed and empirically implemented.

5.5 Discussion

In this chapter, I set out to explore how to improve and standardise the measurement of social policy arrangements as independent variables in multilevel analytical frameworks. This is done by deriving different conceptualisations of the welfare state based on the existing literature and its demand. This introduces an intermediate step to narrow down different perspectives before an empirical operationalisation takes place. This proposal is supposed to serve as an input to a—hopefully—larger debate on how to standardise the selection of operationalisations of welfare stateness in multilevel frameworks. At this point it is important to point out again that there are probably other conceptualisations in the literature that are not covered in this book. Should the approach I have outlined in this project prove useful, it would be necessary to further test the proposed conceptualisations to explore possible sub-variants and additional types.

In summary, I argue that in order to standardise the selection of indicators, to improve the comparability of the methods and to increase the robustness of the empirical results, the explanatory concept—the welfare state—must first be considered in more detail. By determining conceptualisations of welfare stateness embedded in the mechanisms highlighted by research objectives and hypotheses, we fill in a piece missing so far. The main concern of this project is to show that the lack of specification of the perspective on the welfare state when it is studied as an independent variable results in empirical operationalisations that lack theoretical guidance. As a result, approaches used in the relevant literature are too heterogeneous (cf. chapter 3) and lack comparability. By narrowing down the perspective on the welfare state, the selection of indicators can be standardised within explicitly defined research objectives. Furthermore, combining several conceptualisations may even enrich debates by adding new perspectives.

Table 5.2 Guiding questions

In the course of this chapter, four systematised concepts of welfare stateness were discussed in detail, derived from different strands of literature. They highlight specific perspectives on the welfare state—all of which are supported by previous work on the matter, albeit with varying specificity.² For each of these concepts, I have discussed possible sets of indicators which—from a theoretical point of view—seem more suitable for an empirical test than others. In the selection, I tried to meet the criteria of comparability, clarity and availability as much and with as little compromise as possible. Still, there is certainly room for improvement in many cases. For the different steps and considerations necessary to move from the research question to the systematisation of concepts to specific indicators, it is helpful to ask several questions (cf. Table 5.2). The entire selection process is summarised in Figure 5.3.

Figure 5.3

Conceptualising and operationalising welfare stateness (proposal)

In order to test the applicability of the proposal in detail, two exemplary analyses are carried out in the following chapter, which on the one hand illustrate the practical implementation of the proposed framework for the selection of indicators and on the other hand help to assess the measurement validity.