Article Text

Download PDFPDF

Welfare generosity and population health among Canadian provinces: a time-series cross-sectional analysis, 1989–2009
  1. Edwin Ng1,
  2. Carles Muntaner2,3,4
  1. 1Centre for Research on Inner City Health, Li Ka Shing Knowledge Institute, St. Michael's Hospital, Toronto, Ontario, Canada
  2. 2Bloomberg School of Nursing, University of Toronto, Toronto, Ontario, Canada
  3. 3Dalla Lana School of Public Health, University of Toronto, Toronto, Ontario, Canada
  4. 4Department of Public Health Sciences, Korea University, Seoul, South Korea
  1. Correspondence to Dr Edwin Ng, Centre for Research on Inner City Health, Li Ka Shing Knowledge Institute, St. Michael's Hospital, 209 Victoria Street, 3rd Floor, Toronto, Ontario, Canada M5B 1C6; nged{at}smh.ca

Abstract

Background Recent work in comparative social epidemiology uses an expenditures approach to examine the link between welfare states and population health. More work is needed that examines the impact of disaggregated expenditures within nations. This study takes advantage of provincial differences within Canada to examine the effects of subnational expenditures and a provincial welfare generosity index on population health.

Methods Time-series cross-sectional data are retrieved from the Canadian Socio-Economic Information Management System II Tables for 1989–2009 (10 provinces and 21 years=210 cases). Expenditures are measured using 20 disaggregated indicators, total expenditures and a provincial welfare generosity index, a ombined measure of significant predictors. Health is measured as total, male and female age-standardised mortality rates per 1000 deaths. Estimation techniques include the Prais-Winsten regressions with panel-corrected SEs, a first-order autocorrelation correction model, and fixed-unit effects, adjusted for alternative factors.

Results Analyses reveal that four expenditures effectively reduce mortality rates: medical care, preventive care, other social services and postsecondary education. The provincial welfare generosity index has even larger effects. For an SD increase in the provincial welfare generosity index, total mortality rates are expected to decline by 0.44 SDs. Standardised effects are larger for women (β=−0.57, z(19)=−5.70, p<0.01) than for men (β=−0.38, z(19)=−5.59, p<0.01).

Conclusions Findings show that the expenditures approach can be effectively applied within the context of Canadian provinces, and that targeted spending on health, social services and education has salutary effects.

  • SOCIAL EPIDEMIOLOGY
  • PUBLIC HEALTH
  • PUBLIC HEALTH POLICY
View Full Text

Statistics from Altmetric.com

Introduction

An emerging body of research in comparative social epidemiology uses an expenditures approach to investigate the relationship between welfare states and population health.1 Rather than conceptualising welfare states as regimes (eg, national clusters based on historical legacies of social policy development)2 ,3 or institutions (eg, policy principles and characteristics such as qualifying criteria, replacement rates, duration and coverage),4–6 the expenditures approach uses information on public spending to quantitatively gauge the extensiveness of welfare generosity and its impact on population health.7 The argument is that government expenditures operate as macrosocial determinants of health (SDOH), and increases in welfare generosity are expected to improve population health.2 The mechanisms through which welfare generosity affects health are complex and varied, including the provision of benefits and services (eg, providing unemployment benefits and healthcare),8 development of basic capabilities (eg, creating opportunities for citizens to lead healthy lives)9 and increased command over resources (eg, empowering citizens with control over SDOH).10

Using expenditure data to conceptualise welfare states offers several advantages. Theoretically, government expenditures can be hypothesised as essential public goods that operate as health-promoting welfare resources.11–13 Collectively provided welfare resources are akin to SDOH and consist of laws, policies and programmes designed to prevent and manage common life events that adversely affect population health. Methodologically, government expenditures are considered a valid measure of welfare generosity.14 Allocation of governmental budgets is illustrative of spending priorities and welfare policy dynamics. Shifts in welfare effort might reflect corresponding shifts in social needs, labour market conditions or political power. On pragmatic grounds, aggregate expenditure data are widely available and allow for comparisons of social policy development across political jurisdictions and over time. Future expenditure studies will become more theoretically informed and methodologically sophisticated as disaggregated data become more available.14 Advantages aside, expenditure studies have also been long criticised for being unable to differentiate between effort and need (eg, bigger spending may simply reflect bigger problems)15 and for failing to account for qualitative dimensions of welfare generosity (eg, eligibility requirements, coverage and duration of benefits and services).4

Existing expenditure studies find that welfare generosity is generally predictive of favourable absolute health outcomes (eg, lower mortality rates, increased life expectancy).2 Among countries that invest a larger share of their economies to welfare resources, population health outcomes tend to improve. Puzzlingly, generous expenditures do not necessarily generate smaller health inequalities within nations.16 One of the most important contributions of recent studies has been the identification of which contexts and expenditures matter to health. Most studies have examined the cross-national impact of government expenditures among Organisation for Economic Co-operation and Development (OECD)3 ,17–24 and European countries.7 ,12 ,25 Comparatively less work has examined the expenditure–health link within subnational contexts (exceptions include Canada26 and the USA10 ,27). Similarly, studies have relied heavily on aggregate expenditures such as healthcare17–20 ,23 ,25 and social welfare,3 ,19 ,21 ,23 ,24 and less on disaggregated measures (exceptions include spending on primary, secondary and higher education;10 active labour market programmes;22 and police and roads27).

This study takes advantage of provincial differences within Canada to test the effects of a comprehensive set of disaggregated expenditures on population health. Widely considered as a liberal welfare state,28 Canada serves as an attractive case study to examine the expenditure–health link within a single country. First, provinces are subnational units with sufficient political autonomy to achieve different levels of welfare generosity and population health. Under Canada's Constitution, provinces are responsible for the provision and legislation of several health-promoting welfare resources (eg, health, social services and education). The exclusive and concurrent legislative powers of the federal and provincial governments are outlined in online supplementary table S1. Second, there is considerable variation in welfare generosity across Canadian provinces.29 Between 1989 and 2009, New Brunswick, Quebec and Newfoundland spent on average the most per capita on health ($2433), social services ($2077) and education ($2184), respectively. Over the same time period, in contrast, Prince Edward Island invested on average the least per capita on health ($2044) and social services ($786), and Ontario the least on education ($1498; authors’ calculations). Third, significant health differences exist across provinces.30 Trend data show that mortality rates have historically been lowest in Western provinces (eg, British Columbia and Alberta) and highest in Atlantic provinces (eg, Newfoundland and Nova Scotia). The availability of high-quality data among Canadian provinces allows us to use time-series cross-sectional (TSCS) methods to investigate trends shown in figure 1. Among Canadian provinces, between 1989 and 2009, total provincial expenditures in real per capita terms have steadily increased ($5910–$8651) while total age-standardised mortality rates have steadily decreased from 7.27 to 5.53 per 1000 deaths. In this study, we examine the effect of provincial expenditures on mortality rates, net of province-specific factors; construct and test the effect of a provincial welfare generosity index; and explore the differential effect of provincial welfare generosity on male and female mortality rates. Our hypotheses are twofold: (A) as provinces invest a larger share of their governmental budgets on health-promoting welfare resources, mortality rates decline; and (B) provincial welfare generosity on health, social services and education produces the largest effects. Given the dearth of previous scholarship on the possible connections between expenditures and health by gender, we refrain from developing a hypothesis.

Figure 1

Trends in total provincial expenditures in real per capita terms (arithmetic mean of Canadian provinces) and total age-standardised mortality rate per 1000 deaths in Canadian provinces, 1989–2009. Source: Authors’ tabulations based on Statistics Canada and its Canadian Socio-Economic Information Management System (CANSIM) tables 102-0504 and 385-0001.

Methods

This study uses a pooled TSCS analysis of Canadian provinces from 1989 to 2009, including Newfoundland, Prince Edward Island, Nova Scotia, New Brunswick, Quebec, Ontario, Manitoba, Saskatchewan, Alberta and British Columbia. Data are retrieved from Statistics Canada and its Canadian Socio-Economic Information Management System (CANSIM) II Tables. Our final sample includes 210 province-years as cases.

Dependent variables

We focus on total, male and female age-standardised mortality rates (all causes of death). These outcomes are the weighted average of age-specific mortality rates per 1000 persons, where the weights are the proportions of persons in corresponding age groups of a standard population (1991 Canadian Census population).

Independent variables

Provincial expenditure data are converted into real per capita terms using population data and the consumer price index (2002=100) for each province.31 We test the mortality effects of 20 disaggregated expenditures classified under three government functions: general services, community and social services, and economic services and other functions. Under government services, we test two measures: general government services and protection of persons and property. Fifteen indicators of community and social services are tested: hospital care, medical care, preventive care, other health services, social assistance, workers’ compensation benefits, other social services, elementary and secondary education, postsecondary education, special retraining, housing, environment, recreation and culture, and regional planning and development. Within economic services and other functions, three subfunctions are considered: transportation and communications; resource conservation and industrial development; and labour, employment and immigration. These government functions are tested as possible macro-SDOH that influence the availability and quality of welfare resources.11 ,13 For example, government services, community and social services, and economic functions, respectively, ensure the security of citizens (eg, policing and firefighting), provide essential services (eg, delivering healthcare) and protect against economic risks (eg, promoting fair employment conditions), which subsequently enhance population health.

In addition to testing these disaggregated expenditures, we regress mortality rates on two general measures of provincial spending. First, we test total government expenditures, which include the preceding disaggregated expenditures but exclude measures with missing data and non-service functions. Second, consistent with Brady's32 expenditure approach to poverty, we construct a provincial welfare generosity index by averaging the standard scores of medical care, preventive care, other social services and postsecondary education. As revealed below, these four disaggregated expenditures are predictive of lower total, male and female mortality rates.

In all models, we control for several extraneous variables. Transfers are federal equalisation payments made to less prosperous provinces to ensure that government expenditures are comparable across provinces.33 Debt charges refer to money owed by the province, and controls for the inverse association between public debt and government spending.31 Transfers and debt charges are both measured in real per capita terms. Dependency ratio is measured as the number of dependants (% of the provincial population under 18 and over 65 years of age) for every 100 workers. Dependency is controlled because a higher proportion of non-working citizens such as children and seniors may affect the demand and need for welfare benefits and services.34 Urban population (%) is added because residing in urban areas is associated with lower mortality due to greater access to welfare services.35 Immigrant population (%) accounts for the existence of the ‘healthy immigrant effect’ (eg, recent immigrants tend to be healthier than native-born Canadians) and for the uneven distribution of immigrants across provinces.36 Female labour force participation (%) is held constant to remove its positive and negative effects on expenditures and health, respectively.17 Unemployment rate (%) reflects the current state of provincial economies, and accounts for the inverse association between unemployment and poor health.37 Low income (%) is included as a relative measure of low socioeconomic status (ie, set at 50% of adjusted median household income), given the importance of income as an SDOH. Real gross domestic product per capita measures the average real income per person for each province and factors out the overall level of economic development for the province for a given year. A full description of variable definitions and data sources is available in the online supplementary table S2.

Statistical analyses consist of four steps. First, we present mean values and total and within-province SDs for the variables used in the analysis. Second, we examine the effect of each disaggregated expenditure on total, male and female mortality rates in separate regression models. Given the temporal dominant nature of our data set (21 years >10 provinces), we use the Prais-Winsten estimates with panel-corrected SE (PW-PCSE) with a first-order autocorrelation correction (AR1).38 ,39 PW-PCSE with AR1 is the recommended strategy to correct for panel heteroscedasticity and contemporaneous and serial correlation of errors. Fixed-unit effects (provincial dummy variables) are added to control for unobserved heterogeneity and omitted variable bias.39 Third, we regress mortality outcomes on a provincial welfare generosity index, which combines the significant predictors of lower mortality rates (medical care, preventive care, other social services and postsecondary education) into a single measure. Fourth, to demonstrate the robustness and convergence of our results, we perform several postestimations of provincial welfare generosity index models. In different model specifications, we add a linear trend, lag predictors by 1 year, test for parameter heterogeneity,39 run jackknife replications and generate feasible generalised least squares (FGLS) estimations. To compensate for the number of inferences being made (eg, multiple comparisons problem), we use a p value of 0.01 to determine statistical significance. All analyses are conducted using Stata/SE V.12.1.

Results

Table 1 presents the mean values and total and within-province SDs for the variables used in the analysis. Between 1989 and 2009, the average total, male and female age-standardised mortality rates were 6.47, 8.26 and 5.10 per 1000 deaths, respectively. As a percentage of provincial expenditures, 81% was allocated to community and social services, 13% to economic services and other functions, and 6% to general government services. In terms of disaggregated spending, provinces spent the most on average on elementary and secondary education ($991.07), medical care ($915.82) and hospital care ($885.18), and the least on other education services ($25.85), labour, employment and immigration ($29.56), and regional planning and development ($36.88). A correlation matrix is provided in online supplementary table S3.

Table 1

Means, total SDs and within-group SDs, 1989–2009*

Table 2 displays the regression results for total, male and female age-standardised mortality rates on 20 disaggregated expenditures and total expenditures. For simplicity, we display only unstandardised coefficients (b), z-scores (z) and standardised coefficients (β). In separate models, we find that four specific expenditures are predictive of lower total, male and female mortality rates: medical care, preventive care, other social services and postsecondary education, net of control variables. Provincial spending on medical care has the largest effect in lowering mortality rates compared with the other expenditures. If medical care expenditures are increased one SD (ie, $207 or 14% of total expenditures), total, male and female mortality rates should decline by about 0.20, 0.26 and 0.15 per 1000 deaths, holding other control variables constant. Table 2 also shows that total expenditures had no statistically significant effect on mortality outcomes.

Table 2

Separate PW-PCSE models of disaggregated provincial expenditures and total expenditures on total, male and female age-standardised mortality rates in Canadian provinces, 1989–2009†,‡,§

Table 3 displays the Prais-Winsten regressions of total, male and female mortality rates regressed on the provincial welfare generosity index while controlling for extraneous variables. Model 1 shows that the provincial welfare generosity index has a large negative effect on total mortality rates (β=−0.44, z(19)=−5.76, p<0.01). The effect of provincial welfare generosity index, as measured by its standardised coefficient, is actually larger than any of the four disaggregated expenditures (medical care, preventive care, other social services, and postsecondary education) used to construct the index. For a hypothetical SD increase in the provincial welfare generosity index, total mortality rates would decline by 0.25 per 1000 deaths. Models 2 and 3 show that provincial spending is beneficial to both genders, and the standardised is actually larger among women (β=−0.57, z(19)=−5.70, p<0.01) than among men (β=−0.38, z(19)=−5.59, p<0.01). Apart from the provincial welfare generosity index, none of the control variables reached statistical significance across all three models (results not shown). We did find, however, that female labour force participation is negatively associated with male mortality (β=−0.17, z(19)=−4.72, p<0.01) but had no such effect on female mortality (β=−0.07, z(19)=−1.34, p=0.18).

Table 3

PW-PCSE models of provincial welfare generosity index on total, male and female age-standardised mortality rates in Canadian provinces, 1989–2009†,‡

Postestimation results confirm the robustness and convergence of our provincial welfare generosity index models. Since results are consistent across all three outcomes, we present only total mortality results in the online supplementary table S4. Results were substantively identical when we detrended the data (model 1), lagged independent variables by 1 year (model 2), allowed parameter values to differ from 1989–1999 to 2000–2009 (model 3), excluded other predictors and provinces (model 4) and used the FGLS estimator (model 5).

Discussion

This study used TSCS data to examine the effect of provincial welfare generosity on total, male and female mortality rates among Canadian provinces. We find support for our two hypotheses: (A) provincial welfare generosity is a strong predictor of lower mortality rates; and (B) provincial expenditures on health (medical care and preventive care), social services (other social services) and education (postsecondary education) are the significant features of welfare generosity. We also find that provincial welfare generosity has a larger effect in reducing female mortality than male mortality rates. Our findings confirm previous expenditure–health findings1 and make two contributions to the extant literature. First, we show that expenditure–health relationships can be effectively investigated within subnational contexts. Provinces can be viewed as subnational welfare states that play key roles in the promotion and protection of population health. Second, our research addresses the need for more disaggregated expenditure studies. In doing so, we add a much-needed level of precision about what kinds of government expenditures actually matter to population health. In all, our findings are consistent with prevailing explanations about how collectively provided welfare resources operate as macrosocial determinants of population health.

First, health expenditures ensure that necessary health services are available to all citizens. In particular, we find that expenditures on medical care and preventive care are the central features of health spending that affect population health.26 Medical care spending includes outlays on drug programmes; dental and visiting-nurse services; and outpatient hospital care services, public residential care facilities, workers’ compensation boards and other public health institutions.33 Expenditures on preventive care prevent and mitigate the occurrence and impact of diseases including public health clinics; communicable disease control services; food and drug inspection services; and government agencies providing nursing, hygiene and nutrition advisory services.33 These results echo the work of Bunker,40 who found that advances in curative and medical care accounts for half of the 7.5 years of increased life expectancy in the USA since 1950. Provincial governments play key roles in treating and preventing disease through the provision of high-quality medical care and the initiation of primary, secondary and tertiary prevention interventions.8

Second, expenditures on social services offset and forestall situations where the well-being of citizens is threatened by circumstances beyond their control. Findings indicate that a disaggregated expenditure classified as other social services is a significant predictor of lower mortality rates. Other social services are intended to meet the needs of deprived populations (eg, elderly, physically and mentally impaired, households with dependent children and survivors of a deceased family member). Services and supports include lodging and board, legal aid, home care, transportation, counselling, nursery and daycare, alcohol and drug rehabilitation and the provision of essential goods (eg, food, clothing and fuel).33 Common and unexpected life events such as growing old and facing disability frequently expose citizens to greater risks like economic insecurity. To protect against such risks, provincial welfare generosity in the form of other social services has the effect of supporting citizens during vulnerable times and generating the additional effect of protecting population health.

Third, education expenditures cover the costs of developing, improving and operating education services and systems. Our analyses show that increased spending on postsecondary education is associated with reductions in mortality rates. Such expenditures support the attainment of credentials in universities, community colleges and specialised educational institutions.33 Also included are bursaries, scholarships and other types of financial assistance (eg, loan forgiveness, interest relief, refundable learning tax credits) that ease the financial burden associated with higher education.33 Plausible pathways through which higher education might promote population health include the building and upgrading of human capital, increasing social mobility, shaping future job opportunities, raising earning potential and protecting young adults from premature mortality risks (eg, unintentional injuries).10

Our expenditure study is the first to construct a comprehensive welfare generosity index. Interestingly, this index has an even larger effect in lowering mortality rates than any of its underlying components. There are two possible explanations for this result. First, the effect of welfare generosity on population health may be synergistic. Since our index combines the significant features of provincial spending, benefits and services are likely to promote health by meeting the needs of citizens in comprehensive and interconnected ways. For example, the combined effects of medical care and preventive care, which treat and prevent disease simultaneously, might be more effective in reducing mortality than either acting alone or separately. Welfare generosity as a whole appears to be more beneficial to population health than the sum of its disaggregated parts. A second possible explanation is that the effect of welfare generosity on population health is generalised to all citizens. For example, those in need benefit from receiving essential transfers and services. Those not in need might benefit as well from simply knowing that they are well secured and insured against expected and unexpected life events.11 If citizens view welfare generosity as effective social safety nets, welfare generosity might afford citizens with a greater sense of security, protection and assurance, and citizens then are better positioned to live healthier lives. If citizens view welfare generosity as a set of social protection policies and programs, welfare generosity might afford citizens with a greater sense of security and protection against situations that adversely affect their well-being as well as enabling them to live healthier lives.

We also find new evidence on welfare states, gender and population health. Our result that welfare generosity has a larger impact in lowering the mortality rates of women than men requires more theoretical and empirical attention. Our initial interpretation is that welfare resources may narrow unequal gender relations (eg, socioeconomic inequalities), which in turn benefits the health of women more than men.41 Two other findings are worth noting. First, total expenditures are not predictive of improved population health (see table 2). Other studies also find no association between total spending and population health.23 It seems that targeted spending on health-promoting welfare resources, not general increases in total expenditures, offers the greatest potential to yield the largest returns to the public's health. Second, female labour force participation only had a negative effect on male mortality and had no such impact on females. Among men, this might reflect the growing number of dual-income households which have increased from 47.1% to 64.4% from 1976 to 2008.42 Male mortality benefits from dual-income households by increasing material living conditions, easing work demands and buffering against economic stressors. Among women, in contrast, the non-significant finding is consistent with the double burden hypothesis.43 Women working full time in the labour force and pulling a second shift at home are often exposed to role conflict and overload, which subsequently fails to benefit female health.43

Conclusions

Our ultimate conclusion is that disaggregated levels of provincial welfare generosity are macrosocial determinants of population health among Canadian provinces. Provincial welfare generosity appears to operate as health-promoting welfare resources, which provide welfare benefits and services, ensuring that citizens possess basic capabilities and an adequate command over resources. Having said that, several limitations need to be acknowledged. As noted above, expenditure studies are confronted with two common limitations. First, expenditure data might conflate welfare effort with population need. We partially differentiate between effort and need by controlling for the number of dependants in each province. Second, our expenditure findings might be seen as overlooking institutional or qualitative dimensions of provincial welfare generosity. On the one hand, the institutional characteristics of certain benefits and services are uniform across provinces (eg, medical care and preventive care are delivered on the basis of universality and need). In contrast, other forms of welfare generosity are means-tested, selective or based on the ability to pay (eg, social services and postsecondary education). Further work is needed to determine whether provincial policy differences with respect to eligibility requirements, duration and coverage are also predictive of population health. Another possible limitation is that our study uses an index to summarise provincial welfare generosity. Composite measures have drawn criticisms for selecting inappropriate subindicators and using incorrect weights.44 We believe, however, that our index overcomes these issues by combining only significant expenditures and applying equal weights to each expenditure. Lastly, our study may be limited by the shorter than desired time frame to estimate the effects of provincial welfare generosity on mortality rates. Unfortunately, 1989 is the earliest year for which valid and reliable data on disaggregated expenditures are available from Statistics Canada.31

Having established the connection between provincial welfare generosity and population health, our forthcoming work will address two pressing topics. First, we will develop a fuller understanding on what kinds of expenditures matter to population health and augment our provincial welfare generosity by incorporating the significant features of federal government expenditures (eg, unemployment benefits, paid parental leave, sickness benefits, pensions and old age security). Second, we will apply a political epidemiological perspective to determine the roles played by collective political actors such as political parties and women in government in triggering welfare generosity and improving population health.

What is already known on this subject?

  • Recent work in comparative social epidemiology uses an expenditures approach to examine the link between welfare states and population health.

  • Connections between aggregate expenditures and population health across countries have been established. The connection between disaggregated spending and population health within nations has not as yet been established.

What this study adds?

  • This is the first study to conceptualise Canadian provinces as subnational welfare states and to test the effect of disaggregated expenditures and a provincial welfare generosity index on total, male and female mortality rates.

  • As provinces devote a larger share of their budgets to medical care, preventive care, other social services and postsecondary education, mortality rates decline.

Acknowledgments

EN gratefully acknowledges the support of The Canadian Institute for Health Research, Grant #96566, and the Ontario Ministry of Health and Long-Term Care. The authors gratefully acknowledge the comments and suggestions of the three referees.

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.

Footnotes

  • Contributors EN and CM both conceived the idea; while the former conducted the analyses and drafted the paper, the latter provided revisions.

  • Competing interests None declared.

  • Provenance and peer review Not commissioned; externally 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.