What explains cross-national variation in the liberalization of abortion rights policies? Asal, Brown, and Figueroa (Reference Asal, Brown and Figueroa2008) make an important contribution to the study of this question by suggesting that what we often assume—that abortion laws are heavily influenced by cultural factors such as religion—may not be the most accurate explanation, suggesting the need to explore possibilities beyond the influence of cultural variables on women's rights policies. Yet in their test of the impact of various factors on abortion liberalization, Asal, Brown, and Figueroa also find that the United Nations Convention on the Elimination of All Forms of Discrimination Against Women (CEDAW) is not significant and is not associated with the liberalization of abortion laws. Instead, they argue that the most important influence on abortion liberalization is the empowerment of women politically and economically.Footnote 1
While there may be several variables at work in the process of abortion liberalization, this study tests the possibility that ratification of CEDAW does influence abortion policies in countries. Scholars have identified various paths to compliance and ways in which human rights treaties may be used to create domestic change, including through hard or soft power mechanisms (Abbott and Snidal Reference Abbott and Snidal2000; Hafner-Burton Reference Hafner-Burton2012; Helfer and Voeten Reference Helfer and Voeten2014; Simmons Reference Simmons2009, Reference Simmons2010). Asal, Brown, and Figueroa suggest that CEDAW may influence abortion policies and therefore include CEDAW as a control variable.
This article takes issue with the way in which Asal, Brown, and Figueroa measure and operationalize CEDAW. Specifically, they measure CEDAW's impact as whether a state had ratified the document at one point in time (Reference Asal, Brown and Figueroa2008). Considering that only six states had not, at the point in their data, ratified CEDAW, and two of these—the United States and Palau—had signed the document but not ratified, there is a weak statistical case for studying CEDAW in this way. This study offers a more sophisticated empirical examination of CEDAW's relationship with abortion liberalization in a cross-national context.
In this article, we provide a better measure of CEDAW that accounts for variability between states’ levels of ratification—ratification, ratification with reservations, and ratification of the Optional Protocol—and in doing so, we more accurately test the relationship between CEDAW and abortion liberalization. Additionally, we make use of structural equation models to more adequately operationalize varying levels of abortion liberalization on a yearly level (1960–2011), as well as to test for nuanced differences in the effects of treaty ratification. In doing so, our findings challenge previous findings that found no effect for CEDAW ratification and demonstrate the need for further research within this domain. The article contributes to the debate about whether human rights treaties influence domestic policies and highlights the importance of measuring treaty ratification in a way that captures the full complexity of ratification's effects.
ASSESSING THE IMPACT OF HUMAN RIGHTS TREATIES
There is great debate about the effectiveness of human rights treaties in creating government compliance with human rights norms and laws. Many scholars agree that there is a “problem with compliance” when it comes to human rights treaties (Adams and Kang Reference Adams and Kang2007; Hillebrecht Reference Hillebrecht2012b). It is one thing to sign and ratify a piece of paper but quite another to comply with the norms embodied in the treaty. Because of this compliance problem, some scholars argue that the usefulness of treaties in promoting the desired behavior in states is severely limited or only effective in certain cases (Hafner-Burton and Tsutsui Reference Hafner-Burton and Tsutsui2007; Hill Reference Hill2010; Neumayer Reference Neumayer2005; von Stein Reference von Stein2005), with some even claiming that human rights treaties may worsen state abuses that treaties aim to prevent (Hathaway Reference Hathaway2002). However, others note that by ratifying human rights treaties, states open themselves to international scrutiny, and thus what appears as an increase in human rights abuses is only an increase in awareness of them (Goodman and Jinks Reference Goodman and Jinks2003).
Still other scholars are optimistic about human rights treaties and their impact. Some of the variation in optimism may be related to differences in approaches to studying treaties’ impact. Hafner-Burton and Ron (Reference Hafner-Burton and Ron2009) point out that qualitative approaches tend to be more optimistic, while quantitative approaches often find no relationship between treaty ratification and changes in state behavior. Yet there are examples of both quantitative and qualitative (and mixed-methods) studies that provide more reasons for optimism (Simmons Reference Simmons2010). Englehart and Miller (Reference Englehart and Miller2014) find a robust relationship between CEDAW ratification and women's political rights over time. They test other potential explanations and find them lacking, including the “world polity” theory that international law is manifesting rather than advancing change.
While acknowledging the limitations of treaties, Simmons (Reference Simmons2009) argues that states ratify and comply with international treaties because they recognize the reputational costs of not doing so. She argues that domestic actors can increase compliance by using treaties as leverage in order to apply pressure to states. Transnational advocacy networks also contribute to the process by which states are compelled to honor their commitments (Keck and Sikkink Reference Keck and Sikkink1998), and Goodman and Jinks (Reference Goodman and Jinks2003) agree that treaties are useful, especially so for changing norms. They note that even if Hathaway (Reference Hathaway2002) is correct in her assertion that treaties do not have an immediate positive impact on human rights practices, the long-term effects make them worthwhile (see also Heyns and Viljoen Reference Heyns and Viljoen2001). States may use treaties as a type of external signaling, but this does not mean that these commitments are meaningless (Lupu Reference Lupu2013b; Milewicz and Elsig Reference Milewicz and Elsig2014). The type of commitment matters, and there are varying levels of commitment ranging from “soft” to “hard” commitments (Abbott and Snidal Reference Abbott and Snidal2000; Cole Reference Cole2009). However, most scholars agree that the level of democracy and other circumstances in a country are also important in determining the likelihood of state compliance with human rights treaties (Hafner-Burton and Tsutsui Reference Hafner-Burton and Tsutsui2007; Hathaway Reference Hathaway2002; Landman Reference Landman2005; Neumayer Reference Neumayer2005). A closer look at what scholars have said about the influences on state practices toward women's rights reveals that there are similar debates over what inspires states to better their practices. The question of treaty effectiveness is also debated in this literature.
EXPLAINING VARIATION IN WOMEN'S RIGHTS
Scholars focus on different dependent variables when they study human rights treaty effects, at times focusing on policy change and at others on state human rights practices as relevant outputs. Different measures of women's rights have been associated with a variety of factors, from the number of women in parliament and the labor force (Asal, Brown, and Figueroa Reference Asal, Brown and Figueroa2008) to cultural influences including religion (Berer Reference Berer2002; Dillon Reference Dillon1996; Htun Reference Htun2003). Other explanations for why some countries implement women's rights policies include the conditions of debate within democratizing countries and whether women have access to debate venues (Walsh Reference Walsh2010), the structures of economic and political power within countries that give more influence to certain actors over others (Blofield Reference Blofield2006), and even international power structures in which women's rights offer a way of ranking states (Towns Reference Towns2010). Among these explanations, however, women's rights treaties are sometimes tested and discounted (Asal, Brown, and Figueroa Reference Asal, Brown and Figueroa2008) or are not included as potential influences. Yet, just as there are scholars who find that other human rights treaties are often effective in promoting human rights, there is also literature on the effectiveness of treaties in promoting and protecting women's rights more specifically.
Gray, Kittilson, and Sandholtz (Reference Gray, Kittilson and Sandholtz2006) emphasize that CEDAW, the central United Nations treaty on women's rights, can function not only as a way of putting legal pressure on states to protect and promote women's human rights but also as a measure of globalization and the spread of human rights norms. They find that globalization significantly improves women's lives overall and that CEDAW is particularly important in their operationalization of globalization. Their cross-national study finds that the ratification of CEDAW is associated with longer life expectancy, higher literacy, and higher participation in the economy and parliament among women, and, significantly, they measure ratification by dividing states into categories of states that ratify, states that ratify with reservations, and states that sign but do not ratify, thus capturing varying levels of commitment to CEDAW. Hill (Reference Hill2010) also finds that although ratification of the Convention Against Torture and International Covenant on Civil and Political Rights human rights treaties is associated with reduced respect for human rights, ratification of CEDAW is positively associated with respect for women's economic, social, and political rights (as measured in the CIRI data set), while Englehart and Miller (Reference Englehart and Miller2014) find that there is a robust relationship between CEDAW ratification and women's political rights but no relationship between CEDAW ratification and women's economic rights over time. Cole (Reference Cole2013) uses the same data set and finds that CEDAW has a positive relationship with women's rights practices but not policies. Prior research suggests not only that human rights treaties vary and thus sweeping claims about whether treaties work are complicated, but also that measuring human rights treaties’ effectiveness in promoting different types of rights is important.
Scholarship in the area of compliance identifies many mechanisms through which human rights treaties may influence domestic policies and practices pertaining to human rights (Hafner-Burton Reference Hafner-Burton2012; Simmons Reference Simmons2010). Some scholars suggest that compliance with human rights treaties relies on other factors such as democracy (Landman Reference Landman2005), which may be due to the way in which domestic actors can use human rights treaties as leverage for change (Simmons Reference Simmons2009). Other scholars find that compliance with human rights treaties relies on domestic elite incentive structures (Cardenas Reference Cardenas2007; Hillebrecht Reference Hillebrecht2012a) or domestic institutions such as constitutions and courts (Lupu Reference Lupu2013a; Sandholtz Reference Sandholtz2012). The research on compliance varies in whether it focuses on policy outcomes or human rights practices. This study focuses on policy outcomes and whether CEDAW has an influence on policy change in women's abortion rights.
The literature explaining women's rights and the impact of human rights treaties is diverse and divided. This article focuses on CEDAW and abortion liberalization and challenges Asal, Brown, and Figueroa's (Reference Asal, Brown and Figueroa2008) finding that any positive effect CEDAW may have on liberalizing abortion laws is overshadowed by the more important impact of women's representation in parliament and the workforce. Beyond simply concluding that CEDAW is unimportant in the liberalization of women's rights policies, one possible explanation for this is that CEDAW may have a positive effect on the general status of women in a country but is not associated with specific policies such as abortion (this would fit with Gray, Kittilson, and Sandholtz Reference Gray, Kittilson and Sandholtz2006 and Cole Reference Cole2013). Another potential explanation is that international treaties, because of strong domestic cultural norms (particularly religious cultural norms), do not affect abortion or that local factors such as women's representation in parliament are more influential than global factors like CEDAW (Boyle, Kim, and Longhofer Reference Boyle, Kim and Longhofer2015).Footnote 2 Yet another possibility, and the one that is put forth here, is that the way in which CEDAW is treated in empirical tests has important implications for whether its impact on abortion laws is captured quantitatively.
When measured in a way that accounts for variation in state ratification commitments (Cole Reference Cole2009), CEDAW is positively associated with abortion liberalization. Before moving on to an empirical test of this hypothesis, we will briefly discuss the CEDAW document and the enforcement mechanisms it contains, as well as how this may be seen as creating variation in state ratification commitments and having an impact on abortion policies.
ENFORCEMENT MECHANISMS OF CEDAW
The way one measures a variable can have important implications for the results of empirical tests. In the case of CEDAW (and other treaties), understanding the consequences or pressures states face when they decide whether to comply is essential to accurately measuring the treaty's impact on women's rights policies such as abortion laws. CEDAW has three enforcement mechanisms: state reports,Footnote 3 wherein states must file reports on CEDAW's domestic implementation every four years; an interstate arbitration procedure outlined in Article 29.1, under which states may challenge other states’ interpretations of CEDAWFootnote 4; and the Optional Protocol to CEDAW, which went into force in 2000 and must be ratified separately by states. The Optional Protocol allows private individuals or groups to file complaints against states for violation of the law under CEDAW; currently, 105 states have ratified the Optional Protocol.Footnote 5
The enforcement mechanisms described here have the potential to be powerful tools of protest against human rights violations. However, a major issue in CEDAW's enforcement is its reservation process. Reservations are any statements made by states upon ratification that modify or exclude the legal effect of the treaty on the state reserving (Vienna Convention 1969). While other international treaties allow reservations on aspects of the treaties according to very specific rules (for example, the approval by a certain number of other states parties), CEDAW allows states to place reservations on any article as long as the reservation does not conflict with the overall goals of CEDAW. Unfortunately, there is no specific way to determine whether reservations conflict with the overall goals of CEDAW, and often reservations do seem to conflict (Brandt and Kaplan Reference Brandt and Kaplan1995; Neumayer Reference Neumayer2007). For example, Bangladesh has reserved on Article 2 of CEDAW, which states that states parties condemn discrimination against women and will pursue means to eliminate discriminatory behaviors and laws. Placing a reservation on this article would seem to directly conflict with CEDAW's ultimate goal of eliminating all forms of discrimination through necessary and appropriate means. However, the CEDAW document provides no way to prevent states from making such reservations (except Article 29.1).Footnote 6 Neumayer (Reference Neumayer2007) points out that some scholars do not see reservations as a problem; rather when states reserve on a treaty, they are signaling a higher level of commitment to the treaty because they signal that they take their obligations seriously enough to qualify some of their commitment. Alternatively, states that have no intention of complying would not bother with reservations. Neumayer finds potential support for this in that liberal democratic states are more likely to reserve on human rights treaties. Landman (Reference Landman2005) finds that young democratic states are more likely than older democracies to ratify higher-cost treaties with fewer reservations, suggesting that a state's international status influences its decisions about reserving on treaties. However, these findings do not shed light on the level of commitment to the treaties; therefore, how reservations affect state compliance is an open question.
Clearly, there is more to CEDAW than simply whether states have ratified the treaty. CEDAW ratification, ratification of the Optional Protocol, and the placement of reservations are included in this study and allow for a fuller understanding of CEDAW's effect on domestic policy by accounting for the opportunities and challenges to implementing CEDAW.
CEDAW AND ABORTION
There are many ways in which the ratification of CEDAW may influence women's lives. This article studies whether CEDAW can help explain variation in the liberalization of abortion policies across countries. Three CEDAW articles in particular may be cited as supporting women's human right to abortion: Article 12.1, Article 14.2(a)(b), and Article 16.1(e). Article 12.1 mandates equality in health care services, and the Committee on CEDAW characterizes “the refusal of medical procedures that only women require, such as abortion, as sex discrimination” (Cook and Dickens Reference Cook and Dickens2003, 6), suggesting that a lack of abortion rights constitutes unequal health care services. In 2006, the Constitutional Court of Columbia found the country's criminal ban on all abortion to be a violation of CEDAW—particularly Article 12.1—and Columbia subsequently liberalized its abortion laws (UNIFEM 2009). Article 14.2(a)(b) requires that rural women in particular have the right to health care services and to otherwise benefit from development. Finally, Article 16.1(e) explicitly requires women to have legal and accessible ways to control their reproduction. Because this article grants the right to control one's reproduction and have access to the things necessary to this end, Article 16.1(e) is the article most often interpreted as granting a right to abortion access, though it does not mention abortion explicitly. Indeed, some states have specifically placed reservations on this article in order to exempt them from having to liberalize abortion laws in their countries.
Based on these and other articles, the Committee on CEDAW has interpreted the CEDAW document as supporting a right to abortion in several instances. For example, in its 2013 General Recommendation No. 30 on women in conflict and postconflict situations, the committee recommended that all states parties guarantee safe abortion access and postabortion care. While this may only apply in conflict and postconflict circumstances, there are also other times the committee has addressed abortion. Under the Optional Protocol's mechanism through which an individual can take a complaint against a state party to the CEDAW Committee, the committee ruled that Peru had violated a young girl's human rights under CEDAW because she had been denied an abortion. As Ngwena (Reference Ngwena2013) points out, this was a groundbreaking decision that could lead to changes in the law in Peru and beyond.
In February 2018, the CEDAW Committee responded to requests from organizations in Northern Ireland that invoked Article 8 of the Optional Protocol in order to call into question the United Kingdom's abortion policies in Northern Ireland. The CEDAW Committee found that the United Kingdom breaches the rights of women in Northern Ireland because of restrictive abortion laws that only grant a right to abortion in cases of a threat to a woman's life or a serious risk to health. The committee stated that “denial of abortion and criminalization of abortion amounts to discrimination against women because it is a denial of a service that only women need” (Committee on CEDAW 2018). The committee referred to Article 12 and Article 16.1(e) in its findings, as well as Articles 1, 2, 5, and 14. While it is too soon to know whether these interpretations of the Committee on CEDAW will lead to the liberalization of Peru's or Northern Ireland's laws, these examples reveal that CEDAW is interpreted and used to challenge restrictive domestic abortion laws in a number of ways, including through general recommendations, domestic courts, and the individual complaints mechanism of the Optional Protocol.
Beyond interpretations of the CEDAW document, abortion is an important issue to study from a women's human rights perspective because, unlike other reproductive rights such as contraception, abortion is directly relevant only to women. The consequences of lacking access to safe and legal abortion include not only the economic, social, health, cultural, and private consequences of unwanted pregnancies but also unsafe and illegal abortions, from which an estimated 68,000 women die per year (Grimes et al. Reference Grimes, Benson, Singh, Romero, Ganatra, Okonofua and Shah2006). As mentioned earlier, the Committee on CEDAW has specifically stated that the way in which abortion disproportionally affects women makes it discriminatory to deny women access to the medical procedure (Cook and Dickens Reference Cook and Dickens2003, 6). Scholars from all different disciplines, the Committee on CEDAW, and the United Nations General Assembly Sixty-Sixth Session have used the issues outlined here to support claims that in all or many circumstances it is a woman's human right to have access to abortion and that states—especially those that have ratified CEDAW—must amend their laws accordingly (Berer Reference Berer2002; Blofield Reference Blofield2008; Cook Reference Cook1993; Cook and Dickens Reference Cook and Dickens2003; Cook, Dickens, and Bliss Reference Cook, Dickens and Bliss1999; Copelon et al. Reference Copelon, Zampas, Brusie and deVore2005; Fuentes et al. Reference Fuentes, Paine and Elliott-Buettner2008; Grimes et al. Reference Grimes, Benson, Singh, Romero, Ganatra, Okonofua and Shah2006; Htun Reference Htun2003; Shalev Reference Shalev2000; UN General Assembly 2011; Wolman Reference Wolman2010; Zampas and Gher Reference Zampas and Gher2008).
The possibility that CEDAW could increase pressures on states—from below or from above—to work toward abortion liberalization is theoretically sound. Asal, Brown, and Figueroa (Reference Asal, Brown and Figueroa2008) made this argument when they included CEDAW as a potential variable in their analysis of the influences on abortion liberalization. The next step is to test this relationship using methods that account for the CEDAW variable in a way that captures the levels of commitment among states.
METHODS AND DATA
We combined data from numerous sources to test our expectation that CEDAW is positively associated with abortion law liberalization. To measure abortion laws cross-nationally, we obtained data documenting abortion liberalization from Boyle, Kim, and Longhofer's work (Reference Boyle, Kim and Longhofer2015), which codes for whether countries allow abortion to save the woman's life, to preserve physical health, to preserve mental health, in cases of rape or incest, in cases of fetal impairment, for economic or social reasons, or on request. The data set includes dichotomous indicators of each legal form of abortion (1 = legal) for most countries from 1960 through 2011 (Boyle, Kim, and Longhofer Reference Boyle, Kim and Longhofer2015). It should be noted that while Boyle et al.’s data set is useful for comparing general liberalization across abortion types, countries, and years, it does not have information on barriers to abortion access that are often present in countries that have otherwise legalized the procedure (see Levels, Sluiter, and Need Reference Levels, Sluiter and Need2014, who code for barriers to access within Western Europe).
We measured whether countries had ratified CEDAW, as well as to which extent they had ratified the treaty (e.g., with reservations, fully, and whether the country ratified the 2000 Optional Protocol) by coding the ratification status and date from the UN international treaties website.Footnote 7 Ratification status for each of the three types of ratification was coded dichotomously beginning on or after the year of ratification. States that placed reservations, including declarations that have the legal effect of a reservation (Neumayer Reference Neumayer2007), on any part of CEDAW were coded as reserving on CEDAW. Data on religious prevalence cross-nationally was obtained from the Association of Religion Data Archives National Religion data set, which includes cross-sectional time-series data of country religiosity spanning back (in many cases) to 1945. As we are primarily interested in the effect of Protestantism, Catholicism, and Islam predominance on abortion liberalization, we include percentage adherence measures for those three branches of religion in our data (see Dillon Reference Dillon1996 for the influence of religion on similar areas).
Fertility rates, the percentage of each state's parliament made up of female legislators, and inflation-adjusted gross domestic product (GDP) data were obtained from the World Bank database. We make use of the Polity IV data set to control for the level of democratization on abortion liberalization. Finally, Henisz's (Reference Henisz2017) POLCON V measure of political constraint for each country-year is used to capture the extent to which political actors within each state are able to exercise veto power over actions in the international sphere, such as treaty ratification.
Unlike many previous excursions into the effects of treaty ratification (but see Frank Reference Frank1999), we make use of structural equation modeling (SEM) to predict variability in abortion liberalization. Our rationale for SEM in these analyses is twofold. First, it allows us to specify a measurement model predicting variability in individual legalized abortion types as a function of the latent construct of abortion liberalization. Previous research (see Asal, Brown, and Figueroa Reference Asal, Brown and Figueroa2008) has made use of summative indices to get at overall abortion liberalization; however, this method assumes that all types of legalized abortion contribute equally to liberalization, which we will shortly show is not the case. Additionally, making use of the factor-based measure of liberalization SEM affords avoids conflating variability due to the construct in question (abortion liberalization) with that due to measurement error or randomness.
Second, SEM allows us to fit and test complex models of the determinants of abortion liberalization, including varying treaty ratification statuses. For example, we specify a structural model that tests for whether treaty ratification status (full ratification versus ratification with reservations) is mediational in nature, something that the standard family of regression models make difficult. Our structural models allow us to break down the variance in treaty ratification in both a mediational and direct effect model. Moreover, SEM allows us to test these hypotheses using nested models in order to more conservatively assess the impact that added variables have in reconstructing the data's variance-covariance matrix, rather than simply testing against the traditional null hypothesis that each slope coefficient is zero.
We coded and estimated our SEM models in the R statistical software package using the Lavaan SEM library (Roseel Reference Roseel2012), which acts as both a supplement and set of convenience functions for the SEM library originally implemented by Fox, Nie, and Byrnes (Reference Fox, Nie and Byrnes2017).
RESULTS
We begin our measurement of abortion liberalization by first presenting the measurement model for cross-national abortion liberalization. In seeking to establish a model of liberalization, the obvious first step is to include each legal abortion “type” worldwide, which past authors (e.g., Asal, Brown, and Figueroa Reference Asal, Brown and Figueroa2008) have used to create summative indices of liberalization. This is complicated by the fact that not all legal abortion procedures contribute equally to liberalization.
We ran an item response matrix (not shown) for each of the abortion types coded for in the UN abortion policies data set, as well as their frequency of occurrence from 1960 to 2011. Including each measure of abortion legalization in our measurement model quickly led to estimation issues, especially for the particular modes of abortion. For example, there were zero cases in which abortion to save the life of the mother was banned but other reasons for the procedure were legal. In such cases, the presence of zeroes will lead to instability in model estimates at best, and models that fail to converge at worst. Similar issues were found with the legality of abortion for physical health of the mother, which was highly related to other reasons for abortion (e.g., states that allow abortion for physical health reasons usually allow abortion in other cases). Finally, we found similar results for abortion “on demand”—no states allowed elective abortion while banning other forms. Because of this, we chose to use the legality of abortion for mental health reasons, rape or incest, fetal impairment, and economic and social reasons in our measurement model of abortion liberalization.Footnote 8
Our measurement model, including abortion to preserve the mental health of the woman, in instances of rape or incest, fetal impairment, and economic and social reasons is presented in Table 1. Because our indicators for the latent construct of abortion liberalization are dichotomous in nature, we made use of a diagonally weighted least square (DWLS) estimation procedure with thresholds for each set to 1.
Table 1. Abortion liberalization measurement model
** p < .01; * p < .05.
For model fit we report chi-square fit statistics, a measure of “badness” of fit in SEM (with nonsignificant values indicating better fit; see Kline Reference Kline2010), the comparative fit index (CFI), which indicates better fit as it approaches 1.0, and root mean square error of approximation (RMSEA), which typically indicates better fit at values below .08. As can be seen, our measurement model for abortion liberalization fits the data well for metrics other than chi-square (χ2 = 54.492, p < .05, CFI = 1.000, RMSEA = .063), with each of our manifest indicators loading highly on the underlying factor. The high chi-square value most likely derives from the size of this sample, thus we rely more on fit statistics that are more robust to large sample sizes (CFI, RMSEA) in evaluating model fit (see Kline Reference Kline2010).
We turn next to our structural models, which predict variation in latent abortion liberalization with the exogenous variables mentioned previously. Importantly, in all models we include a lagged abortion liberalization variable to account for ongoing shifts in liberalization over time. We believe this makes our tests more conservative, because if other factors (such as treaty ratification) have an effect on the level of abortion liberalization after accounting for the previous country-year's level of liberalization, that effect is more likely due to the treaty itself rather than trends in liberalization that were already occurring at that time.
We begin our models by first fitting a structural model accounting for lagged abortion liberalization and country characteristics, including fertility rate, the percentage of women in legislative bodies, GDP, level of democracy, political constraint, and levels of religiosity across Protestant, Catholic, and Muslim faiths. Note that our treaty variables in these models—CEDAW ratification with reservations and full ratification—have their path coefficients set to 0 in order to test in later models whether freeing those parameters significantly improves model fit. This nesting allows us to assess changes in model fit as we increasingly relax parameters, which makes for more conservative hypothesis tests, as we are then testing against the null hypothesis that the variables add no significantly better fit to the previous iteration rather than the default null stating that the estimated model is a better fit than the fully constrained model (all paths set to 0).
We test not only for the overall effect of CEDAW ratification on abortion liberalization (Model 2) but also for the effect of varying levels of ratification (full ratification and ratification with reservations, Model 3), and finally we implement a test assessing whether and how fully ratifying CEDAW mediates the impact of the treaty to a greater extent than would be expected from ratification with reservations (Model 4). In doing so, we aim to decompose the effect of ratification into the individual effects of levels of ratifying actions. Note that in all models we report standardized regression coefficients so as to gauge the relative impact that each variable has on abortion liberalization.
Our country-level factors show consistent effects across all four regression models presented in Table 2. As would be expected given time-series data, our lagged latent variable for abortion liberalization was consistently positive and significantly related to “present” values of abortion liberalization. Fertility rates were negative and significantly associated with abortion liberalization, indicating that as fertility rates increase, abortion liberalization tends to decrease. Increases in the percentage of parliament made up by women were consistently associated with to increased levels of abortion liberalization. Interestingly, we found no effect for democratization across any of our models, nor did we find effects for GDP.
Table 2. Nested structural equation model comparisons
Note: Reported coefficients are standardized for comparability of effects. Models estimated using DWLS.
** p < .01; * p < .05; dashed lines indicate slopes fixed to 0 for nested model comparisons.
Levels of religiosity, measured by the percentage of adherents to Protestantism, Catholicism, and Islam within each country, were consistently negative and significantly associated with abortion liberalization across these models. Among the three major religions, Protestantism had the smallest effect and Catholicism had the largest consistent effect, and Islam fell in between the two. Finally, the political constraint measure showed a consistent negative effect on liberalization, indicating that those countries with more internal political veto points were less likely to liberalize abortion within the timeframe in study.
We relax the constraint on the effect of general CEDAW ratification in Model 2, which shows a positive and significant relationship on liberalization (b = .052, p < .01). Note that CEDAW ratification is inclusive of both states that ratified CEDAW with reservations and those that ratified it fully. Removing the constraint on CEDAW ratification led to significantly improved model fit (Δχ2 = –8.949, p < .01), suggesting that CEDAW ratification may play an important role in the liberalization of abortion laws cross-nationally, even after accounting for prior values of abortion liberalization. That said, the standardized coefficient is small, indicating that CEDAW ratification on its own does not have a large impact on liberalization.
Model 3 is the first in which we consider not only the overall impact of CEDAW, but also the impact of full ratification (rather than with reservations) of the treaty. As before, CEDAW ratification at any level has a significant and positive impact on abortion liberalization (b = .052, p < .01). Accounting for full ratification, rather than ratification with reservations, also shows a significant and positive influence on liberalization (b = .146, p < .01). This indicates that it is not only ratification of CEDAW that influences abortion liberalization but also the extent to which states ratify that plays an important role. Moreover, the coefficient for full ratification is nearly three times as large as for ratification with reservations, indicating that those states more committed to the treaty from the outset see larger increases in abortion liberalization, even after accounting for prior levels of liberalization.
We want to assess not only the independent effects of ratification and ratification level, however, so we specified a test of whether ratification level mediates the impact of treaty ratification in general. For those unfamiliar with structural equation models, mediation models perform similarly to multiplicative interaction effects in traditional regression models, in that they test whether two (or more) variables interact in such a way to produce varying effects based on variability occurring simultaneously between the interactive measures. In our mediational model, we assess both the direct (ratification with reservations → liberalization) and indirect (ratification with reservations → full ratification → liberalization) effects of ratification in order to parse apart whether differential effects exist for those states that fully, rather than partially, ratify CEDAW.
Retaining the direct effects of both general CEDAW ratification and full ratification, we estimate the effect of general ratification on full ratification, as shown in Figure 1 and Model 4 of Table 2.
Figure 1. CEDAW ratification level mediation model. Path diagram omits exogenous control variables for sake of brevity. All coefficients are standardized.
As is apparent in the coefficients for Model 4, all of the effects of the previous model are identical, save for the direct effect of general ratification. Mediation models allow us to decompose the effect of some variable between its direct effect and its effect, mediated by some other variable. As can be seen in the attendant figure and table, the direct effect of general ratification actually reverses sign when that variable's impact is mediated through full ratification (the direct effect shifts from .052 to –.588). This indicates that the impact of CEDAW on abortion liberalization is more nuanced than previous scholarly treatments indicate. States that have only ratified with reservations tend to have lower levels of abortion liberalization than those states that fully ratified.
Together with the shift in sign for the direct effect of ratification, the path of the effect for general CEDAW ratification, when passed through full ratification, demonstrates that ratification depends on whether states fully sign on to CEDAW when considering the treaty, as the indirect effect of ratification with reservations, as mediated by full ratification, is positive and significant (b = .640, p < .01). This indicates that ratification of CEDAW is only associated with increases in abortion liberalization when countries have fully ratified. Moreover, given that the coefficients for the indirect (reservations → full ratification → liberalization; b = .640) effect of general ratification and for the independent effect of full ratification (b = .657) are both the largest in the model, save for prior levels of liberalization, this suggests that treaty ratification in this case has a nuanced, but large, impact on abortion policies. Finally, fit statistics indicate that this model fits significantly better than Model 4 (Δχ2 = –18843, p < .01) and on the whole fits the data well (CFI = 1.000, RMSEA = .104).
As stated in our hypotheses, we also seek to test the effect of CEDAW Optional Protocol ratification on abortion liberalization cross-nationally. To that end, we specify two additional nested models in Table 3—each of these models include the full specification from our previous modeling excursions but adds in the effect of the CEDAW Optional Protocol, which came into effect for ratifying countries starting in 2000.
Table 3. Nested CEDAW Optional Protocol models
Note: Reported coefficients are standardized for comparability of effects. Models estimated using DWLS.
** p < .01; * p < .05; dashed lines indicate slopes fixed to 0 for nested model comparisons.
As before, our full model specification includes the mediating effect of full ratification of CEDAW; in addition, we add Optional Protocol ratification as a mediated variable. Model 5, shown in Table 3, sets the mediated relationship between full ratification and the Optional Protocol to 0, with only the direct effect of Optional Protocol ratification estimated. Ratification of the Optional Protocol had a positive and significant effect on abortion liberalization (b = .044, p < .01), further indicating the importance of accounting for different levels of ratification in assessing the effects of treaties. That being said, the standardized coefficient for Optional Protocol ratification is small, and thus does not appear to have a sizeable impact on liberalization when accounting for prior levels of abortion liberalization and the effect of earlier CEDAW ratification.
Model 5 frees up the mediated relationship between the Optional Protocol and abortion liberalization, with full CEDAW ratification as the mediator. What is most interesting about this mediated relationship is that unlike the relationship between full and partial ratification, there appears not to be a mediated relationship occurring with regard to the Optional Protocol, as there was no significant relationship between full ratification and Optional Protocol adoption (b = .010, p > .05).Footnote 9 Moreover, the total effect for the Optional Protocol both directly and indirectly (through full ratification) impacting abortion liberalization continued to be small (b = .044, p < .01), and the addition of this mediation test did not lead to significantly better model fit (Δχ2 = –3.001, p > .05). In sum, it appears that the largest relationship between ratification levels and abortion liberalization can be found in those countries that have fully ratified the original CEDAW treaty. Note especially that both the mediated and direct effects of full ratification continue to have nearly the same level of association with liberalization as lagged abortion liberalization.
CONCLUSIONS
In our analyses, we find that states that ratify CEDAW without reservations have more liberal abortion laws. These effects are stable and consistent, remaining the same regardless of the state's level of democracy, number of women in parliament, GDP per capita, and percentage of the population that is Catholic, Protestant, or Muslim. Moreover, we find there to be interesting relationships between the various types of CEDAW ratification and liberalization. As our study shows, a simple ratify/nonratify dichotomous dummy variable cannot account for CEDAW ratification on the range of state behaviors toward abortion liberalization. Moreover, in contrast to previous research (particularly Asal, Brown, and Figueroa Reference Asal, Brown and Figueroa2008), these results suggest that CEDAW is an important factor in explaining cross-national liberalization of abortion laws and therefore could be an important factor in explaining when states are likely to create policy change in order to respect other rights of women as outlined in that treaty.
In particular, states that ratify CEDAW without reservations tend to have more liberal abortion laws than those that ratify with reservations or that do not ratify at all. This suggests that more attention ought to be paid to the constructs that allow states to reserve on articles within CEDAW, and revisions to these constructs could be made to make CEDAW more influential. Interestingly, though in a theoretical sense encouraging and pressuring states to ratify the Optional Protocol could help increase compliance with the treaty as a whole because of its mechanism allowing for individual women and groups to file complaints against states, adoption of that protocol does not appear to have nearly the effect that full ratification does.
The implications of this study are great; however, they are not complete. The study cannot determine the causal direction of the relationships among the different variables. This is due to a problem of endogeneity in our data. While states that ratify CEDAW without reservations and states that ratify the Optional Protocol have more liberal abortion laws, the analysis presented here cannot reveal a causal relationship between these variables. For one, more could be done to parse apart the longitudinal dynamics of ratification and abortion liberalization. Though we attempt to account for this by including lagged levels of abortion liberalization in our models, more context is needed to parse the relationship between ratification and liberalization.
That said, what our study is able to say with confidence is that ratification of CEDAW without reservations and ratification of CEDAW's Optional Protocol is consistently associated with more liberal abortion laws, with the former appearing to have more of an impact than the latter. We think our analyses make a good case that CEDAW, and treaties like it, may work to succeed in promoting women's rights. As demonstrated in this manuscript, the study of treaty effects on state behavior necessitates the disaggregation of the types of acceptance and their attendant conditions—we cannot simply treat ratification as an on-off switch. We believe that our research makes a good case for handling the study of treaty ratification in a more nuanced manner than has been the case in the past.
