Effects of the Convention on the Rights of the Child on child mortality and vaccination rates: a synthetic control analysis

The Convention on the Rights of the Child (CRC) is one of nine core international human rights treaties currently in effect and the only one that focuses entirely on children. The United Nations General Assembly adopted the CRC in 1989 and the CRC took effect on September 2, 1990, after 20 countries had ratified it. The CRC is now the most widely adopted human rights treaty, as 192 of the 193 United Nations member states are parties to it, along with both non-member observer states (the Holy See and the State of Palestine), the Cook Islands, and Niue; the United States is the only United Nations member state that is not a party to the CRC [1].

The CRC addresses a wide range of children’s rights, including civil and political rights and economic, social, and cultural rights. The economic and social rights include rights relating to health, social security, an adequate standard of living, and education. With respect to health, in particular, Article 24 of the CRC obligates states parties to take appropriate measures to diminish infant and child mortality, to provide necessary health care, to combat disease and malnutrition, and to develop preventive health care.

Ironically, the near universal adoption of the CRC has made it challenging to assess its effectiveness. Certainly, most indicators of child well-being have improved since the CRC became effective in 1990. Infant mortality and under-5 mortality have decreased by 53% and 56%, respectively [2]. The prevalence of under-5 stunting (more than two standard deviations below median height-for-age) has decreased by 56% [3]. Vaccination rates for the third dose of diphtheria, tetanus, and pertussis vaccine (DTP3), for the first dose of measles-containing-vaccine (MCV1), for the third dose of polio vaccine (Pol3), and for Baccille Calmette Guerin vaccine (BCG, to prevent tuberculosis) have increased by 11%, 12%, 10%, and 7%, respectively [4]. And access to basic water and sanitation have increased by 8% and 9%, respectively, since 2000 [5]. Of course, one cannot necessarily attribute this progress to the CRC. Most measures of child well-being have generally improved over time, even prior to the CRC. And other global efforts to improve the lives of children in the past 30 years may have contributed to this progress, notably the United Nations Children’s Fund’s universal childhood immunization campaign in 1985, the World Summit for Children in 1990, and the Millennium Development Goals in 2000.

Many country-specific studies have described particular legislative and administrative actions to implement the CRC that seem likely to have improved children’s rights [6‐15]. Several of those studies have also noted corresponding improvements in some measures of child well-being in those countries, including vaccination rates [11]; female genital mutilation rates, adoptions, and juvenile detention and imprisonment rates [14]; child poverty and hunger rates, access to basic sanitation, school enrolment rates, and standardized test performance [9]; and child mortality and malnutrition rates [10]. But one cannot generalize the results of these studies to CRC states parties generally, nor can one even be certain that those changes would not have occurred in those specific countries absent their adoption of the CRC.

Over the past 20 years, many studies have estimated the effects on various outcomes of adoption of other core human rights treaties, including the International Covenant on Civil and Political Rights (ICCPR) [16‐26], the Convention against Torture and Other Cruel, Inhuman or Degrading Treatment or Punishment (CAT) [16, 18‐20, 22, 25‐29], the Convention on the Elimination of All Forms of Discrimination against Women (CEDAW) [16, 20, 22, 26, 30, 31], the International Convention on the Elimination of All Forms of Racial Discrimination (CERD) [16, 26], and groups of those treaties considered collectively [32‐34]. Those studies generally have not found significant effects of adoption of those treaties on the outcomes considered and, when they have, they have found significant harmful effects almost as often as significant beneficial effects. However, the results have varied by treaty, with analyses of the ICCPR generally finding no effects on civil and political rights, analyses of CEDAW generally finding beneficial effects on women’s rights, and analyses of CAT generally finding harmful effects on freedom from torture.

Only a few prior studies have considered the effects of adoption of core human rights treaties on health outcomes. Gray, Kittilson, and Sandholtz found significant beneficial effects of CEDAW ratification on female life expectancy [31]. Simmons found significant beneficial effects of CEDAW ratification on family planning access [26]. Palmer et al. did not find any significant associations between ratification of six core human rights treaties (CRC, the International Convention on Economic, Social and Cultural Rights (ICESCR), CEDAW, CAT, CERD, and ICCPR) and HIV prevalence, maternal mortality, infant mortality, under-5 mortality, or life expectancy [34]. And, in the only prior study to consider the individual effects of CRC adoption on any outcomes, Simmons also found that CRC ratification had weakly significant, beneficial effects on MCV1 vaccinations and no significant effects on DTP3 vaccinations [26]. The fact that the CRC has been almost universally adopted has likely dissuaded researchers from considering it as often as other core human rights treaties (“Most countries have. ratified the convention on the rights of the child, thus making any comparison useless.” ([34], p., 1991)). It is not clear why so few studies have analysed the effects on health outcomes of other core human rights treaties, such as CEDAW and the ICESCR.

A principal challenge in any analysis of the effects of adoption of human rights treaties is addressing the potential selection effects. Countries that adopt a human rights treaty may already be more committed to improving relevant outcomes under the treaty, making it difficult to determine whether any improvement in those outcomes is due to the country’s adoption of the treaty or to the country’s pre-existing commitment that led it to adopt the treaty. Some analyses have attempted to address these potential selection effects with either instrumental variables [16, 17, 26, 27] or propensity score or similar matching methods [20, 22‐24]. Most analyses have also included lagged dependent variables, which control for the country’s pre-existing commitment to some extent [17, 19, 21‐26, 30, 32, 33]. Of these approaches, matching methods seem most promising, because of the difficulty of identifying valid instruments. However, applying matching methods to the CRC is difficult, because most countries adopted it within a few years after its effective date, thus leaving few untreated units for matching.

In this study, we attempted to estimate the effects of CRC adoption on child mortality and vaccination rates by focusing on those first few years after its effective date, when there was maximal variability in its adoption status across countries. We did so using synthetic control methods, a more advanced matching technique developed by Abadie and Gardeazabal [35] and Abadie, Diamond, and Hainmueller [36] and extended to panels with multiple treated units by Cavallo, Galiani, Noy, and Pantano [37]. Specifically, we sought to determine whether countries that adopted the CRC immediately in 1990 achieved larger reductions in child mortality rates and increases in vaccination rates over the next several years than similar countries that could have adopted the CRC in 1990, but did not do so until at least 1993.

We needed to identify a time period with maximal variability in terms of countries’ adoption of the CRC. Throughout this article, we use the word “adopted” to indicate that the country became legally bound under the treaty, whether through ratification, acceptance, accession, or succession; adoption does not include merely signing the treaty without ratification. The first column of Table 1 shows that about one-third of countries (62) adopted the CRC in 1990, the year in which it became effective; another one-third of countries (62) adopted the CRC in 1991 or 1992; and another one-third of countries (70) adopted the CRC in 1993 or later. (In addition to the 192 United Nations member states that have adopted the CRC, Table 1 includes the Cook Islands and Niue; it does not include the Holy See or the State of Palestine.) Therefore, we compared the group that adopted in 1990 (the treatment group) with the group that adopted in 1993 or later (the control group); we would have preferred a longer time difference between the two groups’ adoption of the CRC, but that would have left too few countries in the control group. Because we wanted to compare similar countries, we further limited the control group to countries that existed in 1990; the second column of Table 1 shows the numbers of those countries by adoption year. And, because we expect any effects of CRC adoption on child mortality and vaccination rates to be larger in less developed countries, we further limited both groups to countries with under-5 mortality rates greater than 10 per 1000 live births in 1990; the third column of Table 1 shows the numbers of those countries by adoption year. Thus, 56 and 43 countries remained in the treatment and control groups, respectively.

Scholars of international law and international relations have often questioned the effectiveness of international human rights treaties in changing countries’ behaviours [17‐21, 33]. There are typically few penalties for noncompliance due to a lack of meaningful enforcement mechanisms and the unwillingness of the international community to use even those mechanisms that are available. Certainly, many countries adopt human rights treaties sincerely and make reasonable efforts to honour their commitments under the treaties. But other countries may ratify treaties strategically, with no intention of keeping their commitments, but merely to gain some benefit, such as admission to an international organization, or to avoid criticism for continuing human rights violations [26]. As summarized above, quantitative analyses of the effects of adoption of human rights treaties have often failed to support their effectiveness, although the results of those analyses have varied by treaty and analysis method.

To our knowledge, ours is only the second quantitative analysis of the effects of CRC adoption and the first analysis of the effects of adoption of any human rights treaty using synthetic control methods. Using instrumental variable methods, Simmons found weakly significant, beneficial effects of CRC ratification on MCV1 vaccination rates and no significant effects of CRC ratification on DTP3 vaccination rates [26]. Our year fixed effect analyses were generally consistent with those results, as we found positive relationships between CRC adoption and vaccination rates for those vaccines and the others that we considered, although those relationships never quite reached standard significance levels. However, in our synthetic control analyses, the relationships between CRC adoption and vaccination rates were often significant during the first seven years after adoption, despite the fact that our sample sizes were relatively small because of missing predictor variable data for many countries.

Even in our synthetic control analyses, we did not find any significant relationships between CRC adoption and child mortality rates. The distinction in our results between vaccination and mortality rates is not surprising. A country can improve its vaccination rates relatively quickly, by improving its efforts in areas such as professional knowledge exchanges, autonomy for its public health managers, and coordination with international agencies [48]. Child mortality is more complex, with a variety of proximate causes such as maternal age, parity, and birth interval; environmental contamination; nutrition; injury; and preventive health care, including vaccinations [49]. Therefore, one might not expect significant improvements even in infant mortality rates within the three- to five-year period after CRC adoption to which our analyses were effectively limited.

Djibouti’s experience illustrates the relative ease of increasing vaccination rates quickly, as compared with the challenge of making rapid improvements in mortality rates. In 1999, Djibouti’s vaccination rates for BCG, DTP3, MCV1, and Pol3 ranged from 23 to 26%, its infant mortality rate was 81 per 1000 live births, and its under-5 mortality rate was 103 per 1000 live births [2, 4]. In 2007, in its second periodic report submitted under the CRC, the Government of Djibouti described the measures it had taken to address those issues in the past decade, with assistance from UNICEF, the World Health Organization, the World Bank, and the United Nations Population Fund. Those measures included rehabilitating and refitting health-care facilities, establishing mobile surgery and primary health care clinics, making generic medicines available at affordable prices, training and placing gynaecologists and midwives in health-care centres, applying malnutrition protocols at health-care centres, adopting the Integrated Management of Childhood Illness programme, implementing a national immunization programme, launching a parental education programme, introducing a safe motherhood programme, and distributing mosquito nets to poor households [50]. Vaccination rates for each of those four vaccines (BCG, DTP3, MCV1, and Pol3) increased to at least 63% by 2003, representing an average increase of about 42% in just four years [4]. By comparison, infant and under-5 mortality rates decreased to 75 and 94, respectively, in 2003, representing decreases of only about 8% from 1999 [2]. In its 2012 annual report on Djibouti, UNICEF explained that efforts to reduce child mortality continued to face challenges such as high child malnutrition, high maternal mortality, very early pregnancies, and lack of access to improved sanitation facilities [51]. Djibouti has continued to make progress on reducing child mortality, with infant and under-5 mortality rates of 54 and 65, respectively, in 2015 [2], but UNICEF explains that child malnutrition, in particular, remains a major public health concern in Djibouti and contributes to about 35% of the causes of under-5 mortality [52].

The fact that both Simmons’ study and our study found some significant beneficial effects of the CRC is promising and suggests that the CRC has been important not merely as a normative statement of children’s rights, but also as a catalyst for real improvements in the lives of children in less developed countries. The 4 to 5% improvements in vaccination rates associated with CRC adoption that we found represented an additional five to seven million children vaccinated each year in less developed countries. Just for the vaccines considered in this study, those increases might have been associated with about 100,000 fewer deaths each year from diphtheria, tetanus, and pertussis, about 70,000 fewer deaths from measles, about 3000 fewer deaths from polio, and about 2000 fewer deaths from tuberculosis [53].

Our vaccination rate results are limited to the 59 less developed countries included in those analyses: countries that adopted the CRC either in 1990 or after 1992 and that had data available for the predictor variables prior to 1990. There is no reason to expect the time limitation to significantly affect the results; countries that adopted the CRC in 1991 or 1992 should not be very different than countries that adopted it earlier or later. The data limitation may be more significant, but the countries that were included and excluded from our vaccination rate analyses because of available predictor variable data were fairly well balanced by world region. The one exception to that balance is that only one of the seven applicable countries from Europe and Central Asia had sufficient predictor variable data available to be included in our vaccination rate analyses. Therefore, our results may not extend to countries in that region, but should generally apply to less developed countries from other regions.

Because some of the control group countries adopted the CRC as early as 1993, our results are also effectively limited to the period from 1991 until about 1995, when the difference between the two groups should be most meaningful. As noted above, this fact may have contributed to our inability to detect any significant relationships between CRC adoption and mortality rates, because any effects on mortality rates may take longer to materialize. Also, the fact that the treatment group countries had larger improvements in vaccination rates during the period from 1991 until 1997 does not necessarily mean that those differences would have persisted if the control group countries had never adopted the CRC.

A reviewer suggested that one limitation of our study is that we assumed that the data were normally distributed in the analysis for easier interpretation and that there may be a need to check the validity of this assumption; we do not believe that any such assumption was necessary for any of our analytical methods.

The limited variability in the timing of countries’ CRC adoptions and the lack of consistent, annual, country-level data prior to 1990 on most child indicators will make it challenging to further assess the effects of CRC adoption with quantitative methods. The CRC’s periodic reporting process should continue to enable qualitative analyses of the impact of CRC adoption, as all countries must report every five years on the actions they are taking to implement the CRC. The difficulty, of course, is that one can never be certain that the CRC is causing a country to take a particular action. However, beyond the CRC, our results suggest that it would be worthwhile to apply synthetic control methods to other human rights treaties, including studies of the effects of ICESCR and CEDAW adoption on health outcomes. Most other human rights treaties have much greater variability in countries’ adoption timing than the CRC does, which should allow even better synthetic control matching than in our analyses here. Perhaps other treaties will also fare better in synthetic control analyses than in instrumental variable or propensity score analyses, as the CRC did in this study.

Of course, researchers continue to develop new methods for these evaluations of binary policy interventions. Samartsidis et al. described the use of a latent factor model, or interactive effects model, which is a more flexible version of difference-in-difference modelling that allows the effect of unobserved confounder variables to vary over time [54]. They also explained Brodersen et al.’s causal impact model, which estimates a Bayesian model for the treated unit’s outcome that includes a time series component relating pre- and post-treatment outcomes for the treated unit and a regression component regressing post-treatment outcomes for the treated unit on pre-treatment outcomes for the control units to provide a counterfactual [55]. King, Lucas, and Nielsen introduced a matching frontier that allows researchers to choose solutions that provide the maximum possible predictor balance for any given sample size, without the need for the iterative process of parameter selection and balance checking that is typical of many matching methods [56]. And Doudchenko and Imbens proposed a new estimator that generalizes difference-in-difference, matching, and synthetic control methods [57]. Some of these approaches may also prove useful in estimating the effects of adoption of the CRC and other human rights treaties on health and other outcomes.

In our preferred analyses using synthetic control methods, we find that CRC adoption may be related to additional reductions in infant and under-5 mortality rates of about 1 to 2 deaths per 1000 live births, on average, during the first three years after adoption, although those relationships are not statistically significant. And we find that CRC adoption is related to additional increases in vaccination rates for the five vaccines that we considered of about 4 to 5%, on average, during the first three years after adoption and that those relationships remain significant for up to seven years after adoption. These findings are important for both policy and research purposes. From a policy perspective, our results further support the effectiveness of adoption of core human rights treaties in promoting the right to health in less developed countries and, in particular, the effectiveness of CRC adoption in promoting children’s right to health. And from a research perspective, our results show the advantages of using synthetic control methods in studies of the effects of adoption of human rights treaties on health and other outcomes. Our analyses using other methods that have most commonly been used in these studies, including year fixed effect regressions, instrumental variables, and propensity score matching, did not find any consistent significant relationships between CRC adoption and mortality or vaccination rates.