The impact of social assistance programs on population health: a systematic review of research in high-income countries

Decades of epidemiological research has demonstrated that socioeconomic resources such as wealth, income, and employment – often referred to as the social determinants of health – are “fundamental causes” of health inequalities [1, 2]. They are fundamental in the sense that they influence the everyday conditions, experiences, and exposures that influence health status. Put simply, those with fewer socioeconomic resources get sicker and die sooner than those higher up in the socioeconomic hierarchy. These findings have led to a broad consensus in the field of public health: social policies that shape the extent to which socioeconomic advantage and disadvantage occur in society offer the most effective, if politically contentious, strategy for reducing health inequalities [3]. Indeed, the final report of the World Health Organization Commission on the Social Determinants of Health concluded that the emphasis in public health must shift from individual-level interventions that aim to modify people’s behaviours to societal-level interventions that ameliorate their everyday socioeconomic conditions [4].

One of the most important ways that societies intervene to buffer the adverse consequences of socioeconomic disadvantage is through the provision of social assistance [5]. Social assistance refers to government programs that provide a minimum level of income support to individuals and households living in poverty. These programs lend support either in the form of direct cash transfers or through a variety of in-kind benefits (e.g. food stamps and rent subsidies). Social assistance has been shown to strengthen the purchasing power of the poor and raise their material standards of living [6, 7]. From a public health point of view, the supplemental provision of income can also enable people to avoid harmful exposures and adopt practices beneficial to their health [8]. Thus, theory predicts that social assistance programs offer an important means of protecting the health of socioeconomically disadvantaged groups and mitigating the extent of socioeconomic health inequalities [9, 10].

While there is widespread theoretical support for the role of social assistance as a policy lever with which to improve population health and promote health equity, it is unclear what the extant evidence demonstrates empirically. At the same time, there is growing concern that existing programs provide insufficient levels of protection and that such inadequacies in the social safety net produce extraordinary costs, both human and economic [11–13]. Such concern has, in some cases, prompted calls for a major overhaul of traditional social assistance schemes. In Canada, Finland, and the Netherlands, for example, governments have conducted small-scale experiments to explore the potential benefits of alternative systems of income provision, such as unconditional basic income [14]. At this critical juncture, there is a need to take stock of the extent to which existing social assistance programs are succeeding (or not) at promoting population health and health equity.

Given the clear implications of recent political developments for the health of socioeconomically vulnerable populations and the lack of clarity on the state of the extant evidence, our aim in this paper is to conduct a systematic review of peer-reviewed research that has examined the health impact of social assistance programs. We focus on programs that provide direct financial assistance rather than aid in the form of in-kind benefits. Previous reviews have evaluated the health impact of other sources of income maintenance, including food stamps [15], low-income tax credits [16], minimum wage laws [17], and unemployment insurance systems [18]. Furthermore, to avoid overlap with similar reviews in low- and middle-income countries [19], we restrict our analysis to high-income countries with well-established welfare state systems (i.e. Australia, Austria, Belgium, Canada, Denmark, Finland, France, Germany, Greece, Iceland, Ireland, Italy, Luxembourg, Netherlands, New Zealand, Norway, Portugal, Spain, Sweden, Switzerland, United Kingdom, United States). To our knowledge, this is the first systematic review to evaluate the health impact of social assistance transfers in high-income countries.

There are several important insights to be gained from our systematic review. Most notably, the results of our review suggest that social assistance recipients tend to exhibit worse health outcomes relative to their non-recipient counterparts. This appears to be the case even after controlling for key demographic and socioeconomic characteristics. This is somewhat puzzling, given that public health theory would predict that these programs are beneficial to health status [1, 9]. The observation that those receiving benefits are faring worse than seemingly comparable non-recipients may reflect that there are, in fact, systematic differences between these populations that are not readily observable using the data upon which these studies rely. There are least two major sources of confounding that could be biasing the results of the reviewed studies. Firstly, individuals who suffer from pre-existing health problems may be selecting into social assistance programs as a means of accessing ancillary benefits that are otherwise out of reach (e.g. health insurance coverage). Secondly, pre-existing health problems may contribute to adverse socioeconomic experiences such as job loss which in turn predict social assistance status. Indeed, there is evidence suggesting that those who suffer from psychological problems have a greater likelihood of experiencing socioeconomic disadvantage and selecting into social assistance [41–43]. In a similar vein, findings from the extant literature indicate that problematic risk behaviours such as binge drinking may predict later life socioeconomic hardship, thereby influencing social assistance status [44, 45]. In addition, individuals with unreported material resources such as savings and family wealth may be opting out of, or be ineligible for, these benefits. In all three cases, these residual sources of confounding are likely to bias results towards a negative association between income support and health status.

Alternatively, these findings may reflect the fact that social assistance is increasingly conditional on a range of punitive, work-related obligations that compel entry into precarious employment conditions [46–48]. While these measures have been shown to marginally improve employment outcomes among welfare recipients, the terms of their attachment to the labour market tend to be short-lived and produce their own set of adverse socioeconomic consequences, including higher rates of in-work poverty [49]. In fact, recent evidence suggests that these precarious working conditions may pose an equal if not greater risk to health status than the experience of unemployment [50–52]. Based on these findings, we might expect social assistance programs that compel marginal labour market attachment to produce negligible or even negative returns to health. Indeed, evidence from the broader literature demonstrates that alternative income maintenance programs which place fewer behavioural requirements on recipients and provide more generous benefit levels than social assistance programs (e.g. unemployment benefits, earned income tax credits, and unconditional cash transfers) have a positive effect on individual health [53–57]. The finding here that social assistance programs are not similarly associated with positive health outcomes may reflect that, unlike other forms of income maintenance, the scope and generosity of existing social assistance programs are insufficient to offset the negative health consequences of the severe socioeconomic disadvantage that renders one eligible for such programs.

In contrast to the puzzling findings reported in descriptive studies, evidence from experimental and quasi-experimental studies of welfare reform in the United States conform to our theoretical expectations. When benefits were reduced and work conditionalities were intensified, there were observable declines in the health status of the socioeconomically disadvantaged groups who tend to be the principal recipients of welfare; namely, poor and low-educated single mothers. Welfare reform is often assumed to promote work and earnings by encouraging reattachment to the labour market. However, the results of existing evaluations suggest that these returns are lukewarm at best [58, 59]. Furthermore, many households affected by welfare reform experienced heightened levels of material hardship [60]. Often, this was because women who were forced to leave welfare ended up in low-paying, insecure jobs [61]. The results of our systematic review lend support to this view by demonstrating that these reforms have had a negative impact on health status, an outcome that is sensitive to material conditions. Thus, while the main finding that social assistance programs do not appear to be succeeding at maintaining the health of the poor frustrates prevailing public health theory, our review provides some evidence suggesting that a reversal of these earlier welfare reforms and a resulting increase in the scope and generosity of social assistance benefits may have a positive effect on the health of socioeconomically disadvantaged populations.

There are several limitations to our analysis. First, we were not able to identify and include studies evaluating policy experiments involving the expansion of social assistance programs. Policy reforms in high-income countries have overwhelmingly involved the retrenchment of established levels of social protection [62–64]. Consequently, there are few examples of expansionary policymaking available for evaluation. Second, we restricted our search to peer-reviewed journal articles. Evidence collected in books, reports, and working papers were excluded from the review. We also restricted our search to English-language publications. This may explain why most of the studies included in the review were from English-speaking countries characterized by relatively weak welfare state infrastructures, with a majority being from the United States. Finally, due to heterogeneity across studies both in policy exposures and health outcomes, we were not able to conduct a meta-analysis of their results.

The overall results of our systematic review suggest that evidence on the health impacts of social assistance remains patchy. Rigorous evaluations of these programs are particularly lacking. Few of the studies accounted for systematic differences between social assistance recipients and their non-recipient counterparts. Fewer still adopted the strongest available methods and study designs to evaluate the health effects of policies. We believe there are several principal reasons for the lack of available evidence on the question examined in this review. It may be the case that existing sources of data provide insufficient information for the conduct of rigorous policy evaluations. For example, population-based health surveys tend to provide little if any information on the benefit characteristics of respondents. In addition, while those working in the field of public health may be increasingly familiar with appropriate statistical techniques to evaluate societal-level policy interventions [65–67], social assistance programs may not be particularly amenable to the application of such methods. For example, many of the best available methods (e.g. regression discontinuity, difference-in-differences, and interrupted time series designs) require researchers to identify moments of large-scale policy change. In contrast to other areas of public policymaking, such as tobacco or food labelling, social assistance programs are rarely affected by such abrupt punctuations. A notable exception in this regard is welfare reform in the United States, for which there is evidence that we have reviewed here [23, 31, 36, 37]. Finally, institutional barriers associated with the conduct of politically sensitive research may be standing in the way of the generation and dissemination of evidence on social assistance programs. Tackling the structural determinants of health requires large-scale government interventions (e.g. greater income redistribution and labour market regulation) [3]. Such efforts can attract opposition from political actors who oppose such a role for governments [68–74]. Many epidemiologists and other scientists contributing to the health inequalities literature may, in turn, feel that conducting and disseminating research of this nature is too political or, by virtue of the political opposition they believe it might face, too challenging to undertake [75, 76].

Notwithstanding these important challenges, there is a growing need for evidence on the health effects of social assistance and similar social policies [77]. While governments often identify health equity as an important priority, their choice of interventions have largely relied on behavioural health promotion strategies that fail to account for the role of social policies as necessary levers to reduce health inequalities [78, 79]. Because efforts to eliminate or even reduce health inequalities are unlikely to be successful if they fail to intervene upon their fundamental causes, it is imperative that public health researchers examine these policies and identify the structural interventions that hold the greatest (and the least) promise for reducing health inequalities [80]. The paucity of such evidence is particularly problematic in light of growing evidence that, despite more than a decade of efforts to promote health equity, inequalities in major indicators of population health appear to be widening [81, 82]. These troubling findings may reflect underlying changes in the social and economic architectures of high-income countries, such as the retrenchment of social protection policies – including social assistance programs [49, 62, 83] – and concomitant increases in adverse socioeconomic experiences, such as poverty and unemployment [84, 85]. Taken together, these broader trends highlight a continuing need for solid evidence to marshal in support of interventions that target the fundamental determinants of health.