Data Resource Profile: PROSPERED Longitudinal Social Policy Databases

There is consistent evidence of social inequalities in health across a variety of outcomes and settings.1–5 However, the paucity of evidence that can be used to inform social policies to improve population health led the World Health Organization (WHO) Commission on the Social Determinants of Health to focus one of its three overarching recommendations on the need to markedly improve our ability to ‘assess the impact of action’.6 In its final report, the WHO Commission urged nations to ‘invest in generating and sharing new evidence (…) on the effectiveness of measures to reduce health inequalities through acting on social determinants’.6 These calls have been echoed, reflecting growing interest in evaluating the population health effects of different social policy approaches.7–10 Social policies have the potential to influence population health through their impact on the social determinants of health, among other mechanisms. For example, a poverty policy, like the provision of a guaranteed minimum wage or an unemployment benefit, might be designed to reduce poverty by increasing levels of income; in turn, greater income is consistently associated with better health outcomes,5,11 with some evidence that this effect is causal.12,13 Similarly, an education policy that makes secondary education free might lower the cost of education and increase individuals’ educational attainment, which is associated with better health outcomes in observational and quasi-experimental studies.14–23 By examining not only if a social policy influences a particular health outcome, but also through what mechanisms and for which population segments, policy evaluation can help to identify interventions for addressing the social determinants of health.24 Micro-level social interventions have been the primary source of evidence for informing how public policy can be used to influence the social determinants of health.12,25,26 Evaluations of sub-national social programmes have become increasingly common; this includes the rapid increase in randomized field trials of social interventions.27–29 There is also interest in measuring the impact of national-level social policy reforms, including those that target income and poverty, employment and work conditions, gender equity and discrimination, and the family environment.30 This is the level of jurisdiction at which many social policies of interest are implemented, in both federal and unitary states. Furthermore, states’ compliance with international obligations and standards established by treaties and declarations often begins at the national level when states adopt new laws and policies. These national-level social policies have the potential to influence larger segments of the population than more local interventions, increasing the potential generalizability of an evaluation study. Evaluations of national-level social policy reforms are, however, relatively rare. Several factors likely contribute to this. Despite the rapidly changing policy environment, even on a national scale, few initiatives have systematically quantified the social policy approaches that different countries have adopted in a longitudinal manner conducive to policy evaluation. Additionally, experimental evaluation is infeasible for policies that affect the entire population in a particular country for a variety of practical reasons, leaving observational (i.e. quasi-experimental) methods for policy evaluation.31 Building evidence on the impact of public policies on population health and health equity was the impetus for two recent research initiatives. The Maternal and Child Health Equity (MACHEquity) Project, which extended from 2011 to 2016, focused on the relation between social policies and major causes of global morbidity and mortality among children and women. After its completion, this project expanded into the Policy-Relevant Observational Studies for Population Health Equity and Responsible Development (PROSPERED) Project, funded through 2021, which broadened its ambit to include different domains of public policy and other outcomes prioritized by the United Nations Sustainable Development Goals. A key component of both initiatives was building policy data infrastructure, specifically the collection of longitudinal social policy data across countries, primarily from original legislation. This work was carried out as a collaboration between the Institute for Health and Social Policy at McGill University and the WORLD Policy Analysis Center (WORLD), and built on policy data created by WORLD.32,33 We constructed databases on a series of national-level social policies enacted between 1995 and 2016 for up to 193 countries. The decision about which policy domains and indicators to prioritize was informed by various factors including: (i) theoretical considerations about their potential effects on social determinants of health; (ii) global consensus on their value, often underlined by international treaties and agreements; and (iii) considerations of data availability. As the initial project, MACHEQUITY, aimed at measuring the impact of social policies on maternal and child health, the first data collection effort focused on policy themes immediately relevant to maternal and child health. With PROSPERED, we are taking a broader approach that considers, in addition to social policies, the impact of policies and programmes defined more broadly, including those related to health care coverage and health services, public health programmes and the environment, among other areas. Under the theme of work-family policies, data on maternity, paternity and parental leaves, leave for health needs of children and adult family members and personal sick leave were collected for 193 countries, with a data point in every year between 1995 and 2016; data on breastfeeding breaks at work were collected for 193 countries between years 1995 and 2014; and data on minimum wage policies were collected for 121 low- and middle-income countries (LMICs) covered by at least one demographic and health survey (DHS),34 for the years 1999 to 2013, as we envisioned measuring policy effects on health using data from these surveys. Under the theme of child protection, we collected data on minimum age of marriage laws for 121 LMICs for the years 1995 to 2012; and child labour laws for 33 LMICs covered by at least one DHS child labour module. Indicators on work-family policies and child protection policies are presented in Tables 1 and 2. Work-family policies Y/N, yes/no; min, minimum; max, maximum; GDP, gross domestic product; PPP, purchasing power parity. 2013 only. Work-family policies Y/N, yes/no; min, minimum; max, maximum; GDP, gross domestic product; PPP, purchasing power parity. 2013 only. Child protection policies Y/N, yes/no. 1999-2012. Child protection policies Y/N, yes/no. 1999-2012. Original national legislation was our preferred source of information. Full-text copies of relevant legislation, in addition to information on amendments and repeals, were located mainly through the ILO's NATLEX and TRAVAIL databases. When full-text legislation was not available through these databases, we located laws through national government websites, Lexadin World Law Guide and the World Legal Information Institute. When online copies were not available, hard copies of legislation were obtained through university libraries. If full-text laws could not be obtained, secondary sources such as national reports on policies and laws to the UN or other similar global and regional bodies were used instead, after a review of their reliability and of the consistency and comparability of their methodology across countries. We particularly favoured secondary sources that either directly quoted source laws or at least clearly listed them. These were also always consulted to corroborate information available through primary sources. Full lists of such sources by policy areas can be found on the PROSPERED website [https://www.prosperedproject.com/database-descriptions/]. Coding is the process by which an individual researcher takes a piece of information from legislation, policy or any other source and translates it into a set of characteristics that can be quantitatively analysed. Our pre-defined coding rules range from computational guidelines on how to consistently convert the length of paid leave when given in different time units (days vs weeks vs months) or capture a wage replacement rate during paid leave if this rate varies for workers with different work tenure, to more conceptual rules on how to categorize light work with regards to child labour policies or what classifies as an exception to minimum age of marriage. They are formulated based on previous research, international consensus on definitions and considerations for user-friendliness of data for future research projects. For each country, two researchers from our team coded data sources independently according to these coding rules and compared their results to ensure accuracy. Although the degree of inter-rater concordance was not measured, if coding required a judgment call by the coder, the rules underlying such decisions were systematically described in a continuously updated codebook and applied consistently across countries. Summaries of these codebooks, in the form of data dictionaries, are delivered when data are downloaded, whereas more detailed methods are available upon request if researchers need further clarification. Coding was conducted in the original language of the document by team members fluent in the language; when this was not possible, we used a version translated into one of the official UN languages. To code our databases, we started with the most recently available cross-sectional policy databases developed by WORLD.35 In the first wave of coding, the most recent year we used as our basis was 2012. We reviewed the date of the sources used; when a national law used in the 2012 databases had been enacted before 1995 and had not been amended or repealed since, it was assumed that its provisions remained applicable from 1995 through 2012. The same text was therefore used to code all applicable variables for that particular country between 1995 and 2012. When a national law used to code the 2012 data was enacted sometime between 1995 and 2012, the same text was used to code variables in the years after the law was enacted, and we searched for the legislation that was in force in the preceding years. All variables for the remaining years were coded based on the former legislation. During the second wave of coding, some datasets were expanded with new variables and countries, and the majority of them were updated with additional years to a more recent date: 2013 for minimum wage policy covering 121 LMICs; 2014 for breastfeeding breaks at work policy covering 193 UN Member States; and 2016 for maternity leave, covering 193 UN Member States. During this phase, datasets on parental and paternity leave, sick leave and family health leave policies, covering 193 countries, were also constructed using the same process described above. Longitudinal social policy databases can be used by both academic and non-academic audiences. For academic research, this data resource allows for locating data points where national policies have changed in order to measure the impact of this change on chosen outcomes, using either methods that rely on cross-country comparisons or within country comparisons across time. To illustrate, our research group conducted a series of analyses that leveraged changes in national-level maternity leave policies to examine impacts on breastfeeding,36 immunization coverage,37 child growth38 and infant mortality.39 Data collected routinely as part of the DHS were used for measuring outcomes. Through the work-family policy database, we identified countries that had changed their maternity leave entitlements, specifically by increasing the duration or wage replacement rate. We then employed a difference-in-differences approach to evaluate the changes in outcomes in ‘treated’ countries that reformed their maternity leave policies compared with otherwise similar control countries that did not, which substituted for the counterfactual. In terms of infant mortality, we found that each additional month of paid maternity leave was associated with 7.9 fewer infant deaths per 1000 live births, a relative reduction in infant mortality of 13%.39 In terms of potential mechanisms, we found that each additional full-time equivalent week of paid maternity leave was associated with an increase in DTP1, 2 and 3 vaccination coverage by 1.38 [95% confidence interval (CI) = 1.18, 1.57), 1.62 (CI = 1.34, 1.91) and 2.17 (CI = 1.76, 2.58) percentage points, respectively, whereas we found no evidence for an effect of maternity leave on the probability of receiving vaccinations for BCG or polio.37 In another study based on 38 LMICs, a 1-month increase in the duration of maternity leave was associated with a 7.4 (95% CI = 3.2, 11.7) percentage point increase in the prevalence of early initiation of breastfeeding, a 5.9 (CI = 2.0, 9.8) percentage point increase in the prevalence of exclusive breastfeeding and a 2.2-month increase (95% CI = 1.1, 3.4) in breastfeeding duration.36 For policy makers and advocates, comparative social policy data collected across countries and over time can support a quantitative approach to monitoring compliance with global human rights treaties as well as progress towards goals and norms set by international declarations and agreements.40–42 To this end, our research group has contributed to No Ceilings, a joint initiative with the Bill and Melinda Gates Foundation, the Clinton Foundation and the WORLD Policy Analysis Center. The purpose of this initiative was to measure progress in women’s legal rights since the United Nations Fourth World Conference on Women, in Beijing in 1995, where delegates representing 189 nations agreed to a Platform for Action that called for the ‘full and equal participation of women in political, civil, economic, social and cultural life’. By using longitudinal data we have produced policy briefs and contributed to a policy report on the evolution of policies across countries in the areas of child marriage, minimum wage, breastfeeding breaks at work and maternity and parental leaves.43 Subsequently, two peer-reviewed journal articles, one comparing countries’ progress on the evolution of breastfeeding breaks policies44 and another on laws to prevent child marriage,45 stemmed from these briefs. Our longitudinal social policy data are a valuable resource, which offers data ready for download for almost all countries in the world for a multitude of policy areas spanning more than 20 years. The majority of the data are based on primary legislation interpreted by the same team of researchers, rather than questionnaires filled in by separate in-country specialists. This adds to the comparability and consistency of the data despite the great differences among nations’ law-making and policy design approaches. Until now, existing data resources on work-family policies have been limited either by region46–48 or by countries’ income-levels,49 or were only available as reports instead of downloadable datasets and/or did not include data points for multiple years.50–52 In addition, many of our databases include detailed variables allowing for an in-depth look into social policy differences across countries and time. Finally, our database on minimum age of marriage is a unique source, as there are no other longitudinal sources specifically on these policies covering so many countries. One major limitation is that the data resource records the existence of legal rights and regulations; therefore it cannot be used independently to measure the implementation level of these laws and policies or policy enforcement. When used for policy evaluation, it shows the impact of the adoption of a law (i.e. the intent-to-treat effect) rather than full implementation of this law or policy. However, it can also be used to estimate the level of implementation by matching policy information with available data on uptake or other proxies for enforcement. For example, if the prevalence of child marriage does not drop following an increase in the legal minimum age of marriage in a country, one possible explanation for this gap could be the lack of enforcement. Another constraint is that we focus on policies at the national level, hence not accounting for sub-national variations at the state or provincial level, or variation through industry-specific legislation or collective agreements. All the longitudinal social policy databases are available via the PROSPERED website [prosperedproject.com] as well as via the Harvard Dataverse [https://dataverse.harvard.edu/dataverse/3po]. They can be instantly downloaded with accompanying variable dictionaries after filling a short form with name, e-mail address, institution and country. Additionally, for the social policy areas covered by PROSPERED, the WORLD Policy Analysis Center website [worldpolicycenter.org] has expanded current data covering education, child labour, work-family legislation, adult labour working conditions and child marriage in 193 countries. In addition to areas of data shared between WORLD and PROSPERED, legal indicators are also available on a broad array of areas relevant to gender, discrimination and sexual harassment at work, poverty reduction, disability, sexual orientation and gender identity, migration, the environment and constitutional rights.53–61 Independent researchers could use these data and follow PROSPERED’s methodology to make these data longitudinal. The website allows users to visualize the data in maps, charts and tables, download datasets and explore findings in more detail in fact sheets, policy briefs and journal articles. In terms of future prospects, we remain strongly committed to expanding our databases by collecting policy data for ensuing years after 2016, as well as integrating other public policy areas that are relevant to health and other outcomes. We are currently exploring the feasibility of including policy indicators related to early childhood education and care, clean air and food labelling regulations. Initiated in 2012, longitudinal social policy databases were generated to meet research needs for comparable and reliable global data on national-level public policies that could have potential effects on the social determinants of health. PROSPERED databases are a unique resource that covers a series of policies under the themes of work-family and child protection for up to 193 countries and two decades. Databases have been created primarily by collecting and analysing information from national legislation. The main categories of data are labour rights protections: around birth, including the duration and wage replacement rates for maternity, paternity and parental leaves, as well as breastfeeding breaks at work policies; around sickness, including leave for personal sickness or family member’s sickness; and protections for children against early marriage and child labour. Longitudinal social policy data can be used for descriptive purposes, for example to monitor progress toward international benchmarks and targets, as well as to estimate the impact of policy reforms on health and health inequalities. Data (1995–2016) are available for download via the PROSPERED website [www.prosperedproject.com] and the Harvard Dataverse [https://dataverse.harvard.edu/dataverse/3po]. This work was supported by the: Canadian Institutes of Health Research (ROH-115209 and FRN 148467); Bill and Melinda Gates Foundation (OPP 1107826); and William and Flora Hewlett Foundation (2014–9620). Conflict of interest: None declared.

Récupéré en direct depuis OpenAlex et désinversé. Les résumés ne sont pas conservés dans cette base de données : les index inversés représentent 8,6 Go des 9,3 Go de texte de la base, et le serveur dispose de 13 Go libres.