Article Text

Download PDFPDF

Welfare regimes, population health and health inequalities: a research synthesis
  1. Sarah Brennenstuhl1,
  2. Amélie Quesnel-Vallée2,3,
  3. Peggy McDonough1
  1. 1Dalla Lana School of Public Health, University of Toronto, Toronto, Ontario, Canada
  2. 2Department of Epidemiology, Biostatistics and Occupational Health, McGill University, Montreal, Quebec, Canada
  3. 3Department of Sociology, McGill University, Montreal, Quebec, Canada
  1. Correspondence to Dr Peggy McDonough, Dalla Lana School of Public Health, University of Toronto, 155 College Street 6th floor, Toronto M5T 3M7, Ontario, Canada; peggy.mcdonough{at}utoronto.ca

Abstract

Background Research on the social determinants of health is increasingly using welfare regime theory. Although a key argument is that population health will be better and health inequalities lower in social democratic regimes than in others, this research has not been subjected to a systematic review. This paper identifies and assesses empirical studies that explicitly use a welfare regime typology in comparative health research.

Methods 15 electronic databases and relevant bibliographies were searched to identify empirical studies published in English-language journals from January 1970 to February 2011. Thirty-three studies appearing in 14 peer-reviewed journals between 1994 and 2011 met the inclusion criteria.

Results Three welfare regime typologies and their variants dominated existing work, which consisted of two broad study types: One compared population health and health inequalities across welfare regimes; the other considered relationships between health and the political determinants and policies of welfare regimes. Studies were further distinguished by the presence or absence of statistical significance testing of relationships of interest. Just under one half of studies comparing outcomes by regime found at least some evidence that health inequalities were lowest or population health was the best in social democratic countries. Studies analysing the relationship between health (mortality) and the political determinants or policies of welfare states were more likely to report results consistent with welfare regime theory.

Conclusions Health differences by regime were not always consistent with welfare regime theory. Measurement of policy instruments or outcomes of welfare regimes may be more promising for public health research than the use of typologies alone.

  • Gender
  • inequalities
  • medical sociology FQ
  • poverty
  • public health
  • review
  • social epidemiology
  • social factors in
  • socioeconomic inequalities
  • welfare regimes
View Full Text

Statistics from Altmetric.com

The positive association between health and socioeconomic resources is well established in social epidemiology.1–4 In trying to understand the causal links, most work focuses on pathways involving important ‘proximal’ markers of inequality, such as occupational class, education, employment status and income.5–7 Recently, however, research on social inequalities in health has drawn increasingly from the social policy literature in recognition of the importance of ‘distal’ (upstream) contexts in generating socioeconomic inequalities in the first place.8 9 As one such context, the state plays a critical role; life chances are shaped—sometimes profoundly—through various policies concerning education, taxation, family support, unemployment, childcare and healthcare, among others.10 11

The social programmes that reflect the policy contexts of western states have been developed to organise labour markets and buffer certain risks to economic and mental/physical wellbeing encountered in modern competitive industrial economies.12 However, as Esping-Andersen13 14 demonstrates, there is considerable cross-national variation in the extent and type of protections offered. His well-known typology of ‘worlds of welfare capitalism’ groups countries partly on the basis of how much they ‘decommodify’ labour—or free citizens from reliance on the labour market to maintain a decent standard of living—and on the accompanying extent of class-based inequality. According to his schema, liberal welfare regimes, such as Britain and the USA, represent the least decommodified class of nations, while in social democratic regimes, such as Sweden, the level of decommodification is high. Conservative or corporatist welfare regimes, such as Germany, fall somewhere in between these two types. Because decommodification is inversely related to socioeconomic hardship,15 it follows that population health should be better and health inequalities smaller in welfare regimes that have higher levels of social protections than in others.16 17

Despite its widespread use, Esping-Andersen's framework has encountered much criticism in the social sciences,18 leading some to seek alternative typologies. For example, Huber and colleagues19 20 classify countries on the basis of prevailing political traditions; Ferrera21 and Korpi and Palme22 developed typologies that group countries according to how social benefits are organised and awarded; and feminist critics argue for the need to consider state approaches to reducing women's reliance on the family for survival and promoting their economic independence.23–26 To date, however, health research based on these various welfare regime models, including that of Esping-Andersen, has not been subjected to a systematic review. Therefore, we do not yet know which approaches, if any, are most informative for understanding cross-national differences in population health and social inequalities in health. Indeed, as governments increasingly turn their attention to tackling the social determinants of health,27 28 international comparisons become more significant than ever for building an evidence base that can inform relevant policy development.29

The purpose of this paper is to identify and critically review empirical studies that explicitly use a welfare regime typology to compare population health and social inequalities in health across regime types. We describe the methods used to select relevant studies, outline their findings, and discuss the research and policy implications of this literature as they pertain to population health, health inequalities and welfare regime theory.

Methods

Search strategy

Fifteen electronic databases were searched from January 1970 to February 2011. These included ASSIA, EconLit, IBSS, MEDLINE, PAIS International, Political Science, psycINFO, Public Administration Abstracts, Social Science Abstracts, Social Service Abstracts, Social Science Citation Index, Sociological Abstracts, Sociology, Web of Science and World Wide Political Science Abstracts. An initial search combined welfare state-related terms (‘welfare state*’ or ‘social welfare*’ or ‘welfare theor*’ or ‘welfare typ*’ or ‘welfare regime*’ or ‘welfare polic*’ or ‘welfare nation*’) with health terms (eg, ‘health status’ or ‘health inequal*’ or ‘health inequit*’ or ‘health disparit*’). This strategy yielded 1230 articles. A second search used the broader term ‘health’, combined with identical welfare state terms to ensure that no relevant articles had been missed in the first search. A total of 3373 articles was found in the second search. In a final step, the bibliographies of all articles that met our selection criteria (see below) were also examined for relevant articles.

Study selection and data extraction

Articles selected for the review had to: (1) be an empirical study published in an English-language, peer-review journal; (2) examine individual or aggregate health at the national level as a main outcome (ie, articles about healthcare or the health of populations of cities were excluded); and (3) explicitly compare health outcomes (including health inequalities) on the basis of welfare state theory as a primary objective. Based on these criteria, 82 articles were originally selected by one author (SB) using the article title and abstract. Selected articles were re-reviewed by the other authors using information from the full article, reducing the number of relevant articles to 32. One article was also added after reviewing bibliographies, bringing the total to 33 articles for analysis. Relevant data on the study population, data source(s), welfare regime theory, welfare regime types, health outcomes, findings and other study characteristics were collected from the selected papers by one author (SB) and validated by the other two authors.

Analysis

We were particularly interested in whether or not a study supported a central hypothesis that can be drawn from welfare regime theory, namely, that social democratic regime countries have the best population health and lowest health inequalities because of their higher levels of social protections compared with other regimes.15 When studies lacked statistical tests of this hypothesis, we extracted quantitative data from fully adjusted models (when provided) and determined whether or not between-regime estimates were significantly different at the p<0.05 level by adapting the approach of Cumming and colleagues30 31 to assess the extent of overlap between CI. The details of this analysis are provided in the supplementary appendix (available online only).

Results

Selected articles were published between 1994 and 2011, with nearly four-fifths appearing in the past 5 years. Our discussion of the studies focuses on the basis of their approaches to operationalising welfare regime typologies and assessing the association between regime types and health.

Operationalisation

Two features of operationalisation stand out. First, despite the fact that three main groups of researchers, one based in Scandinavia,32–35 another centred primarily in North America and Spain36–42 and a third in the UK, Norway and The Netherlands,43–50 were responsible for 18 of the 33 articles, we found substantial variability in the measurement of welfare regime typology. Three regime typologies dominated (Esping-Andersen,13 Huber and colleagues19 20 and Ferrera)21 (table 1), but they were often modified. Sometimes this took the form of adding another typology to the original classification scheme. For example, several papers36–40 built on the work of Huber's group19 20 to consider ‘ex-fascist’ or ‘ex-dictatorship’ regimes, while others added ‘eastern European/post-Soviet’43–48 51–53 or ‘east Asian’45 regimes to their schema. At other times, typologies were modified by excluding a welfare regime type. For example, slightly more than one in five studies excluded either the conservative/Christian democratic/Bismarckian/corporatist regime54–58 or the liberal/Anglo-Saxon/basic security (liberal) regime.38 58 Moreover, at least five of the seven studies that used Huber and colleagues' typology did not analyse data from one of the original four key types, the wage earner cluster,36–40 a system of social provision rooted in arbitrated wages and benefits.19 Approximately one quarter of studies used one country only to represent at least one regime.

Table 1

Dominant welfare regime typologies in selected studies

A second notable feature of the operationalisation of typologies is that a sizeable number of studies included what we have termed ‘contextual explanatory variables’ in their analyses. These variables are of two types: political determinants of welfare states and social policies or their output. Political determinants include the percentage of the population who voted for left parties,40 42 voter participation rates,39 40 42 and the duration of time that political parties with a strong re-distributive orientation held power in national governments.39 40 This group of measures was unique to the operationalisation of Huber and colleagues'19 20 approach to welfare regime theory and its modifications.39 40 42 The second type of contextual explanatory variables (social policies) were used in studies based on either Esping-Anderson's typology13 or that of Korpi and Palme.22 Examples of such variables are the levels of family benefits32 33 and decommodification,49 the levels and coverage of income replacement benefits,34 57 the proportion of the population with medical care coverage,39 40 42 total social expenditure37 40 42 and total healthcare spending.39 40 59

Welfare regimes and health

In addition to their operationalisation approaches, we organised the studies on the basis of their assessment of between-regime differences in health and health inequalities. We distinguish between those that undertook statistical testing of health differences by welfare regime and/or reported the health point estimates and their 95% CI by welfare regime and those that did not. Based on these two considerations, the selected studies are presented in three tables. Table 2 shows those in which statistical testing of between-regime differences in health estimates (as derived from fully adjusted models, when relevant) was undertaken by either the original study authors or by us. Table 3 is composed of studies that, with one exception,49 use multivariate models to test for statistically significant associations between health and contextually explanatory variables characterising welfare state regimes (ie, political determinants or policy instruments). Finally, table 4 presents studies that left us unable to determine whether or not health or health inequalities varied statistically by welfare regime.

Table 2

Population health and health inequalities studies with statistical testing of between-regime health differences

Table 3

Population health and health inequalities studies examining association between contextual explanatory variables and health

Table 4

Population health and health inequalities studies with no statistical testing of the health–welfare regime association

Among the 16 studies in which conclusions could be reached about between-regime health differences, six examined population health and the remainder socioeconomic inequalities in health (table 2). Among the former group, three (50%) supported the hypothesis that social democratic regimes would have higher levels of population health than other regime types,41 44 49 one refuted it by finding better self-assessed health among liberal regime countries,52 and two others either found no significant differences in health by welfare regime type34 or failed to provide the data needed to determine what regimes did better than others.45 However, two of these population health studies examined self-assessed general health rather than an objective health measure.44 52 Their findings must be interpreted with caution because of concerns about the validity of cross-national comparisons using subjective health indicators of this nature.65 The 10 studies of health inequalities all examined morbidity, with self-rated health and limiting long-standing illness being the most commonly used indicators. One study reported no differences by regime type,60 while the remainder either countered the hypothesis that inequalities would be lowest in social democratic regimes38 46 47 54 61 or reported equivocal results.37 48 50 53 Notably, support (and refutation) varied according to measures of socioeconomic inequality,47 48 health outcomes,48 50 gender,37 38 46–48 54 birth cohort50 and the specific regime types being compared.53

Of the 11 studies in which contextual explanatory variables were examined, 10 focused on population health (table 3). Among the latter, results from two investigations33 35 were also presented in a third,32 thus reducing the total count of studies with unique findings of interest to nine. They showed that at least some political determinants of welfare states and policies were associated with health outcomes that, with two exceptions,37 57 were ‘objective’ in nature (eg, life expectancy at birth, infant mortality rate, low birth weight, etc). Among political variables, the percentage of the population voting for left parties and voter turnout were related inversely to low birth weight and infant mortality and positively to life expectancy at birth,40 42 but this was not always the case.39 42 Results were more consistent among the studies that examined policy instruments: decommodification,49 the generosity of various types of public expenditures32–34 40 57 59 and their degree of coverage34 40 42 fostered population health and reduced health inequalities.37

Finally, the results of studies listed in table 4 differ in their assessment of the association between welfare regimes and health. Although some reported that social democratic welfare regimes enhanced population health or reduced inequalities compared with other regime types,36 56 64 other studies found the opposite relationship55 or described inconsistent patterns.43 51 58 62 63 In any case, even if their findings had been less equivocal, no conclusive statements can be made about them because tests of statistical significance for observed between-regime differences were not undertaken, or the data that would have allowed us to perform this assessment (ie, difference measures and their CI) were not provided.

Discussion

This paper set out to identify existing empirical studies (published in peer-reviewed journals up to February 2011) that compared population health and socioeconomic inequalities in health on the basis of applied welfare regime theory. Thirty-three met our inclusion criteria, and 24 of them employed tests of statistical significance or provided data that enabled us to draw tentative conclusions about health patterns across regime types. Based on this latter subgroup, we reach two conclusions. First, there is some evidence supporting the hypothesis that the populations of social democratic regimes are in better health than their counterparts in other welfare state regimes. However, this indication comes largely from studies that examined mortality measures (eg, infant mortality rate, life expectancy at birth, etc) and included specific policy instruments in analytical models (eg, extent of public healthcare coverage, public health expenditure, dual family earner policies, benefit generosity, etc). The findings of this latter group increase our confidence in saying that, with respect to public health and social policy spending, ‘bigger is better’.34 It also allows us to begin to disentangle the causal mechanisms that link ‘distal’ social contexts to the life chances of a nation's inhabitants. For example, showing that population health is better when generosity and universalism, rather than means testing, define social programmes32 66 is more useful for creating relevant policy instruments than simply comparing health across the ‘black box’ of welfare regime typologies. Needless to say, such broad categorisations cannot do justice to the myriad of complex issues that they encompass, including considerable within-regime variation in enacted social policies and change over time in key dimensions of welfare states and, thus, in their policies.34

A second conclusion of our study is that here is little support for the hypothesis that socioeconomic inequalities in health are smaller in social democratic regimes than they are in other regime types. This finding has been noted elsewhere and attributed to a variety of factors, including artefact, health behaviours, ‘culture’, immigration and temporal change in the nature of welfare regimes.50 67–69 However, just because health inequalities may not conform to expectations, it cannot be assumed that the pathways and policy influences leading from relative disadvantage to ill health are the same in different welfare regimes.70 Nonetheless, some argue that a more fruitful approach would be to change the focus from relative to absolute inequalities in health.66 71 This is because relative inequalities depend on the circumstances of more privileged members of society who, like their more disadvantaged counterparts, are eligible for universal social programmes in social democratic regimes. As all social groups benefit from such programmes, comparing absolute, rather than relative, health may be a better reflection of the extent to which the welfare state buffers and reduces market-generated inequalities. However, careful attention must be paid to ensuring that the health indicators used can be appropriately compared in absolute terms across different national contexts (see below).

Our findings suggest at least three directions for future research. First, when available, measures of actual policies and policy outputs should be used instead of welfare regime typologies. The studies reviewed suggest several appropriate data sources, including the Social Citizenship Indicator Programme.34 Also relevant is the Comparative Welfare Entitlements Dataset,72 which, like the Social Citizenship Indicator Programme, provides information on a broad range of social policies for 18 countries from the latter half of the 20th century to the beginning of the 21st century. A more in-depth analysis of family policies can also be found in the Comparative Family Policy Database, which provides information on family allowances and parental leave regulations for 22 Organisation for Economic Co-ooperation and Development countries from 1960 to 2010.73 In our view, the detailed accounting of main social insurance programmes and family policies makes these harmonised data ideal for investigating both population health and health inequalities, especially when hypothesised pathways foreground income and gender.

Second, we need to determine whether the findings from studies linking policy instruments to mortality generalise to morbidity. In this, a key challenge is the need for health indicators that minimise the problem of measurement bias. This is always an issue when the research focus is between-country absolute differences in health, given cross-cultural variation in the meaning attributed to various morbidity measures.74 Moreover, although it is assumed that such bias is minimised in comparative studies of health inequalities, within-country differences in reporting by social group may also lead to erroneous conclusions regarding the extent of the social gradient in health.75 Clinical, biological and physical performance measures permit the most valid comparisons,76 and are becoming increasingly available in comparative social surveys (eg, SHARE). A recent, promising innovation to make self-reported measures more comparable is the use of anchoring vignettes to adjust the response scales of populations that are being compared, be they national populations or groups within a nation, but this approach awaits further development.77 78

A third direction for future research is to focus more closely on individuals and the ways in which their lives are shaped over time by their institutional contexts. Indeed, the tendency of comparative policy research to rely on outputs of state instrumentalities (eg, generosity of public pension benefits) to the relative exclusion of direct measurement of individual behaviours, has meant that we have to infer what policies signify for individual outcomes.79 Combining longitudinal microdata with institutional measures80 will allow us to consider the complexities of people's lives and the evolving policy contexts in which they live.81 Moreover, the advent of internationally harmonised microdata sets on ageing populations78 will strengthen our ability to compare these processes cross-nationally. Nonetheless, doing so means that we must theorise carefully the relationships between various policy instruments, their objectives and the life chances of those most likely to be affected by them.2 17 These suggestions notwithstanding, it bears repeating that the task of linking population health and health inequalities to features of the broad institutional context is extremely challenging in the light of the complexity of causal pathways.

Limitations

This review was limited to empirical studies published in peer-review journals. Other potential sources of evidence such as the grey literature, books and book chapters82 were excluded. Moreover, although the peer-reviewed literature is thought to uphold certain standards of research quality, it may overrepresent studies that produce positive and significant findings. Finally, the method we employed to determine the between-regime overlap in CI of point estimates (ie, if the proportion overlap is >50%, then p>0.05) when significance tests were not undertaken in the published material may be conservative.30 31 Nevertheless, only six contrasts (4%) would have been of borderline significance if we had implemented a slightly more relaxed set of criteria. Even more important, however, is that this method is not intended to give precise p values or replace statistical calculations.30 31

Conclusion

Studies that find evidence of the salutary health effects of the social democratic welfare regime, compared with others, are more likely to examine population health, rather than socioeconomic inequalities in health; assess mortality rather than morbidity; and use specific policy instruments rather than classifying countries according to a specific welfare regime typology. When possible, future public health research on comparative health inequalities should consider both their absolute and relative dimensions, and adopt a more targeted method of assessing the health effects of policy ‘exposures’ by incorporating individual-level longitudinal data.

What is already known on this subject

  • Welfare regime theory is increasingly being used within the public health literature to examine cross-national differences in population health and health inequalities.

What this study adds

  • This comprehensive review identified 33 empirical studies that explicitly compare health outcomes (including health inequalities) on the basis of welfare state theory.

  • The analysis of health outcomes, and especially health inequalities, by regime does not necessarily produce results consistent with welfare regime theory.

  • Studies that analysed mortality using variables that described how welfare states are determined or what they actually do were more likely to support hypotheses aligning with welfare regime theory.

Policy implications

  • Assessing the health impact of specific policies or policy approaches is a promising direction for future research on the links between welfare regimes and health.

References

View Abstract

Supplementary materials

  • Supplementary Data

    This web only file has been produced by the BMJ Publishing Group from an electronic file supplied by the author(s) and has not been edited for content.

    Files in this Data Supplement:

Footnotes

  • Previous versions of this paper were presented at the Research Data Centres Conference on Health over the Life Course, University of Western Ontario, London, Ontario, 14–16 October 2009 and at the International Sociological Association (RC-15) Conference in Gothenberg, Sweden, July 2010.

  • Funding Funding for the project was provided by Population Change and Lifecourse Strategic Knowledge Cluster through the Social Sciences and Humanities Research Council of Canada. The funding source has no role in the study design; data collection or analysis, interpretation of data, writing of the report or decision to submit the paper for publication.

  • Competing interests None.

  • Provenance and peer review Not commissioned; internally peer reviewed.

Request Permissions

If you wish to reuse any or all of this article please use the link below which will take you to the Copyright Clearance Center’s RightsLink service. You will be able to get a quick price and instant permission to reuse the content in many different ways.

Linked Articles