Change in Abortion Regulation: Theoretical Approaches and Causal Inference

From this article's perspective, which is a macroquantitative approach, relevant existing studies can be grouped along three major lines. One research strand locates the regulation of abortion within “classic” comparative policy analysis. Studies explain the (relative) restrictiveness of national regulation using “conventional” theoretical approaches and respective explanatory figures (e.g., Asal, Brown, and Gibson Figueroa Reference Asal, Brown and Figueroa2008; Brookes Reference Brooks1992; Gindulis Reference Gindulis2002, Reference Gindulis2003, 52–53). These are economic development or modernization in a broader understanding, democratic transition, left-right party competition, institutional constraints (such as judicial review) or internationalization (e.g., in terms of ratification of human rights treaties). A related factor that has so far only been qualitatively researched is the role of interest groups, especially physicians' organizations (e.g., Engeli and Varone Reference Engeli and Varone2011; Halfmann Reference Halfmann2011, 66–97).

A second major theoretical perspective is morality policy analysis. Quantitative inquiries from this perspective investigate to what extent abortion regulation could be explained via religion (in various aspects such as the religion-state relationship), a religious-secular party cleavage, most notably in terms of the existence and the strength of religious-conservative parties, public opinion, and also mobilization and special interests. For the latter, the focus is more on the role of religious and especially Christian activism (like churches), including special activism targeting abortion (see especially Minkenberg Reference Minkenberg2002, Reference Minkenberg2003; also Engeli, Green-Pedersen, and Larsen Reference Engeli, Green-Pedersen and Larsen2012, 5–26).

Third, studies employ a gender (agency) perspective when trying to explain abortion regulation (changes) and shifts toward lower restrictiveness across states and over time (Githens and McBride Stetson Reference Githens and Stetson1996; McBride and Mazur Reference McBride and Mazur2010, 13–16; McBride Stetson Reference McBride Stetson2001). One focus of this research strand is the role of either general women's rights activism or specific reproductive rights activism for explaining abortion regulation. At the same time, studies are interested in the impact of female representation in the legislative and executive (bureaucratic) spheres. Finally, this approach stresses the crucial role of the interaction of these factors.

Albeit single explanatory factors belonging to the women's empowerment concept (such as the number of female representatives in national parliaments or the female employment rate) are considered in several large-N studies of the first two theoretical perspectives, this is not embedded in a more encompassing theoretical women's empowerment approach. Studies from the third research perspective (if cross-national), although being theoretically closer, refrained from large-N inquiry.

Change in Abortion Regulation: Concept and Measurement

Considerable achievements have been made in terms of the conceptualization and operationalization of the dependent variable abortion regulation for quantitative comparative analyses (see especially Gindulis Reference Gindulis2002, Reference Gindulis2003, 43–51; also Brooks Reference Brooks1992, 347; Glendon Reference Glendon1987, 14; Minkenberg Reference Minkenberg2002; Outshoorn Reference Outshoorn, Githens and Stetson1996, 149; Rolston and Eggert Reference Rolston and Eggert1994, xxi; see also Kreitzer Reference Kreitzer2015 for a recent overview on subnational U.S. abortion regulations' measurements). However, there is still the need for further sophistication and preciseness. A particular problem in existing research is a correct classification of national regulatory models in terms of their degree of restrictiveness; there are still remarkable inconsistencies in the literature. Research needs to integrate the aspect of more relevant “procedural obstacles” in aggregated figures to appropriately determine their legal interference intensity. Recent descriptive inference analyses have convincingly demonstrated the tremendous importance of such seemingly minor regulatory tools, not only for detecting legal change, but also for establishing more nuanced national regulatory models (e.g., Levels, Sluiter, and Need Reference Levels, Sluiter and Need2014). This would include procedural aspects uncommon in most states at this time, but which have existed in the past. By tackling these problems, this study also improves on the concept and operationalization of the dependent variable abortion regulation more generally.

Overall Assessment of Previous Research

Albeit large-N studies have taken into account single gender empowerment variables for explaining abortion regulation in cross-national perspective, such analyses have different theoretical foci, are cross-sectional only, and employ an underspecified operationalization of the dependent variable (e.g., Asal, Brown, and Gibson Figueroa Reference Asal, Brown and Figueroa2008). More sophisticated analyses that combine cross-sectional and longitudinal data that take gender empowerment factors into account are even rarer and quite old in the meantime (e.g., Gindulis Reference Gindulis2002, Reference Gindulis2003, 52–90)—that is, they are unable to take into account the remarkable dynamic of abortion regulation changes that unfolded in the first decade of the 21st century, even in Western states. At the same time, no attempts have been made yet to quantitatively test for the role of distinct women's empowerment factors.

Studies that systematically approach the regulation of abortion from a gender (agency) perspective so far refrained from employing a classical macroquantitative research design with the “state-year” as the unit of analysis (Githens and McBride Stetson Reference Githens and Stetson1996; McBride Stetson Reference McBride Stetson2001; McBride and Mazur Reference McBride and Mazur2010; see also Mazur Reference Mazur2002, 137–153). Albeit political representation of women is a prominent explanatory factor within that scope, this research employs a meso- and microperspective. It is mostly interested in figuring out the more precise mechanisms (such as relevant gender framing) and scope conditions (such as group characteristics or the specific policy environment) that are relevant for women's reproductive rights activism to entail legal change.

Overall, it is not unfair to argue that for the most part we still lack theoretically guided, systematic (large-N) analyses that allow for rigorous testing of the role of gender-specific explanatory factors for abortion regulation while controlling for other theoretically competing independent variables. This contribution aims at closing this research gap.

Abortion as a Gender Issue

As indicated in the introduction, we conceive abortion regulation to be at least a strong gender issue (meaning that it does not need to be defined as an exclusive gender issue necessarily. See, e.g., Mazur Reference Mazur2002, 137–53; McBride and Mazur Reference McBride and Mazur2010, 71–99; Warren Reference Warren2000; Woliver Reference Woliver2002, 15–17, 82–114; see also Superson Reference Superson, Veltman and Piper2014 for a recent encompassing discussion) as well as an important issue for gender equality.Footnote ¹

Parliamentary debates on abortion take place against the backdrop of patriarchal societies and the discourse is shaped by and often reproduces gender stereotypes. Restrictive regulations are recurrently legitimized by the belief that women are not capable of foreseeing their own emotional response to the procedure (Nossif Reference Nossif2007, 61). At the same time, restricting access to abortion affirms women's role as procreators rather than individuals with the right to make decisions about their own bodies. Celis (Reference Celis and Stetson2001, 55) observes that proponents and opponents of restrictive abortion regulation differed considerably in the ways in which they constructed women as a social category in the parliamentary debate on abortion in Belgium. Female members of parliament (but not exclusively female ones) constructed women as competent to take into account their own situation as well as a fetus' right to life when deciding to have an abortion or not (ibid., 55). Others have highlighted the important role female politicians played in pushing abortion regulations to become less restrictive in various national contexts (e.g., Köpl Reference Köpl and Stetson2001, 17, 28 for the case of Austria; Calloni Reference Calloni and Stetson2001, 188, 196, 201 for the case of Italy). Overall, it is plausible to hypothesize that with increased empowerment of women in a society, a woman's right to make decisions about her own body becomes more accepted.

Women's Empowerment/Emancipation

Accordingly, our theoretical framework is a gender perspective or, more precisely, a female empowerment or emancipation approach.Footnote ² The core argument is, of course, that there is a relationship between women's positions in state and society and the regulation of abortion. Where the position of women is enhanced, abortion regulation should become less restrictive ceteris paribus. As women were for the longest time and in all states considered here deprived of achieving a level of empowerment similar to men, the liberalization of abortion regulation only started in the 20th century (Corrêa Reference Corrêa, Smelser and Baltes2001; Petchesky Reference Petchesky1990, 101–138; Staggenborg Reference Staggenborg, Smelser and Baltes2001). We also argue that for each of the women's empowerment elements introduced below, there is a plausible direct causal mechanism linking it to abortion regulation restrictiveness.

This women's empowerment theoretical approach—on the one hand —is different from stressing either general or specific women's rights activism (agency) as expressed in social mobilization, interest group formation, and political action on the issue (McBride and Mazur Reference McBride and Mazur2010, 71–99; McBride Stetson Reference McBride Stetson2001). This means that the theoretical construct rather emphasizes the underlying economic and societal forces that might—among other impacts—result in an increased women's activism toward the abortion issue. On the other hand, the theory also stresses the direct representation of women in the political sphere—hence in the relevant decision-making platforms—as crucial. As a result, this theoretical perspective somehow “bridges” intervening forces that might have resulted in an increased presence of women in politics. It is rather concerned with the direct effects of the underlying factors.

Albeit gender empowerment as a (social science) concept in a more general sense has, of course, a long tradition in gender and politics research (see, e.g., England Reference England, Smelser and Baltes2001; Johnson Reference Johnson, Smelser and Baltes2001; Tong Reference Tong, Smelser and Baltes2001 with further references); it was not until the mid-1990s that the concept was pinpointed and operationalized for quantitative assessment. The United Nations Development Program (UNDP) specifies it as increasing the share of women in the sphere of political decision making with respect to the legislative as well as the executive (administrative) branch, in the sphere of economic decision making and as increasing female economic independence—especially in terms of achieving the same level of earned income as men do (see Ismail, Rasdi, and Jamal Reference Ismail, Rasdi and Abd. Jamal2011 for a brief discussion). Greater educational attainment should be added here. Current research on women's empowerment in development emphasizes educational attainment, political life, and wage work as crucial (e.g., Jejeebhoy and Santhya Reference Jejeebhoy, Santhya, Jejeebhoy, Kulkarni, Santhya and Mehrotra2014). Albeit we will not employ UNEP's indicators (see section 5.3), we nonetheless principally adhere to that concept.

Educational Empowerment

A first cornerstone of female empowerment in a society is educational attainment—that is, a development that increases women's overall level of education with the proponents' goal to achieve (at least) equal footage with men (e.g., Jejeebhoy and Santhya Reference Jejeebhoy, Santhya, Jejeebhoy, Kulkarni, Santhya and Mehrotra2014). Greater educational attainment is a prerequisite for a greater representation of women in the spheres of political and economic decision making and is also necessary to increase economic independence, thereby establishing the basis for work and incomes.

But there are also more direct links as to why women's educational attainment would impact the regulation of abortion. A higher level of education is usually affiliated with an increased level of information gathering, cosmopolitanism, and a greater interest in and awareness of societal and political developments including (more liberal) abortion regulation abroad. Better informed citizens—that is, not only women—are usually better equipped to understand political processes and decision making (and can, hence, influence from “below,” e.g., Galston Reference Galston2001). This refers not only to lobbying activities in a narrower sense, but to interest formulation and aggregation as prerequisites possibly (also) leading to electoral choices. In short, better educated women should be better able to advance their interests in a society and political system as compared to less-educated women.

Furthermore, a higher level of educational attainment creates more economic and “life-course” opportunities for women and is therefore associated with higher opportunity costs of having children and hence lower fertility (Bergman Reference Bergman2005, 30–31; Engelhardt and Prskawetz Reference Engelhardt and Prskawetz2004, 41). This should also result in a stronger demand for less restrictive abortion regulations. We formulate our first two hypotheses accordingly:

Economic Empowerment

A second major pillar is economic empowerment. Though the concepts of educational empowerment and economic empowerment might seem closely related, they are not congruent.

As indicated above, one of the central life course opportunities of educational empowerment for women is pursuing a professional career. Under these circumstances, having children implies higher opportunity costs for women being employed or self-employed (Bergman Reference Bergman2005, 30–31; Engelhardt and Prskawetz Reference Engelhardt and Prskawetz2004, 41; England Reference England, Smelser and Baltes2001; Kalist Reference Kalist2004). This does not necessarily create a sole demand for the legal opportunity to opt for or against children as such during life. What is also (or perhaps even more crucial) is the demand for a right when to have children during life course—that is, at an optimal point considering economic constraints and ambitions.

Second, women's economic empowerment breaks with traditional gender roles that had been prevalent in states (Bergman Reference Bergman2005, 41–43; England Reference England, Smelser and Baltes2001; Rau and Wazienski Reference Rau and Wazienski1999). Abandoning the male breadwinner model might not necessarily imply that economically active women refrain from having children (albeit this is an empirical reality for an ever larger share of women in Western societies during the last decades). But, working for wages certainly diminishes the option to have many children for the vast majority of women who cannot afford costly child care services. At the same time, this should work toward women's demand for less restrictive abortion regulation.

First and foremost, economic empowerment is expressed as economic independence and closing the salary gap to men (see above). Labor market participation (wage work) of women is a prerequisite for this. We formulate the hypotheses accordingly:

Political Empowerment

The third crucial element of gender empowerment is political empowerment. The link to abortion regulation is quite straightforward. When more women are directly represented in the (decisive) decision-making bodies, the chances are higher that their interests will be advanced and finally pushed through (by taking them into account on all stages of the decision-making process). In that regard, a “critical mass” argument is often raised, meaning that reaching a certain numerical threshold is a prerequisite for effective female interest representation at all the stages (e.g., Celis Reference Celis2006; Schulze Reference Schulze2013; Swiss, Fallon, and Burgos Reference Swiss, Fallon and Burgos2012; also Johnson Reference Johnson, Smelser and Baltes2001, but see Lovendusky Reference Lovenduski2001 for a critical assessment).

Although the gender and politics literature discusses whether the simple descriptive representation of women is sufficient to advance substantive representation (see as a brief summary Saward Reference Saward2008; also Celis Reference Celis2008; Lovendusky Reference Lovenduski2001; Mackay Reference Mackay2008; McBride and Mazur Reference McBride and Mazur2010, 13–16), we test this association in our third hypothesis:

Dependent Variable

There are various options for conceptualizing and measuring national abortion regulations. This is due to the complex or multifaceted nature of the subject (as legally expressed) that could hardly be adequately captured with simplistic classifications like “legal” or “illegal” (see references in the second section).

Hence, the first challenge is finding an acceptable compromise between, on the one hand, a too complex and, on the other hand, a too simplistic, “light-weighted” conceptualization and measurement. Second, an adequate conceptualization and measurement should encapsulate all regulatory facets considered relevant from a conceptual perspective into a single dimension—that is, the restrictiveness-permissiveness axis. Finally, to employ the concept as a dependent variables' measurement in statistical models, one also needs to have one single figure (only).

Taking all these challenges into account, it becomes obvious that a “perfect” approach is inconceivable. However, existing approaches leave vast room for improvements. Tackling the subject from the perspective of an adult woman who pursues a legal termination of her pregnancy, there are three main levels of state legal intervention regarding a woman's wish to have her pregnancy terminated that, if combined, express the overall restrictiveness of a national regulation. In that, we define abortion restrictiveness as the intensity of state legal interference in a women's decision to abort.Footnote ⁴

First, the most crucial regulatory aspect and in that “first” level (Level 1) to consider is whether an abortion is indeed completely illegal, whether it is bound to reasons, which means indication-based, or whether it is available upon the pregnant woman's request (choice)—the latter at least during the first gestation weeks. This forms a “natural” order from less to more restrictive.

The second relevant regulatory aspect to consider, which forms Level 2, is the time component, or the time limit expressed as maximum gestation week up to which an abortion could legally be performed, given the most liberal basic form available at Level 1 is a certain indication or choice.Footnote ⁵ We opt for four different stages of time-stringency based on abortion trimesters. Definite time limits earlier in the pregnancy are more restrictive than later ones.

Level 3—that is, the one with the least intensive intervention—could be termed procedural requirements or more general further hurdles for a woman. Several aspects, such as medical expert approval, are equally (i.e., unweighted) taken into account here forming four major blocs at Level 3. The more aspects that are present within each bloc, the more restrictive the regulation.

All three levels make up a three-digit/decimals additive index in which the information for each of the three levels forms the basis for one of the ciphers reaching from left to right. The value for Level 1 (being the digit) counts more than the one for Level 2 while the figure for Level 2 is mathematically more important than the one for Level 3 (the latter being the decimals). Overall, we end up with a figure that captures the intensity of state interference in a woman's decision to abort a pregnancy. Table 4 in the supplementary materials summarizes the concept and measurement of the dependent variable and provides the respective figures that range from 1 (abortion completely unrestricted) to 7 (abortion totally prohibited). Figure 1 visualizes the development of the abortion index in our sample of 16 European countries over time.

Note: 0 represents no state intervention whatsoever (i.e., no legal restrictions on abortion). 7 represents the strongest state intervention (i.e., a total prohibition of abortion). Source: Own data collection as part of the Comparative Analysis of Moral Policy Change (MORAPOL) project, see Knill, Adam, and Hurka Reference Knill, Adam and Hurka2015.

Figure 1. Abortion regulation's restrictiveness: State intervention intensity in a women's decision to abort in 16 countries from 1960 to 2010.

Independent Variables

The most important independent variables in our analysis are the three aspects of women's empowerment we focus on: educational, political and economic empowerment.

As a measure of educational empowerment, we employ the share of the female population (over the age of 25 years) that has completed tertiary education (Barro and Lee Reference Barro and Lee2013).Footnote ⁶ If we think of female empowerment as the advancement of women in relation to men, an alternative operationalization for educational empowerment is the percentage difference of completed tertiary education between men and women. A straightforward and established way of operationalizing women's political empowerment is using the share of seats held by women in national parliaments (IPU 2015; McBride and Mazur Reference McBride and Mazur2011; The World Bank 2015; UN Data 2015). Economic empowerment is operationalized as female labor force participation rate (Brady, Huber, and Stephens Reference Brady, Huber and Stephens2014; OECD Statistics 2015; The World Bank 2015). Labor force participation is an adequate proxy for economic empowerment since it is essentially a measure of the degree of departure from the traditional male breadwinner model that kept women financially dependent on men.

We also include several variables in our models as controls. We control for economic development, various aspects of religion, partisan difference, and institutional hurdles (political constraints). In the two additional model specifications we also control for a time trend via the inclusion of a time count variable and for country differences via the inclusion of country dummies (fixed country effects).

GDP per capita (Heston, Summers, and Aten Reference Heston, Summers and Aten2012) is included in order to ensure that our results are not driven by simple economic development (modernization). Religion is often argued to be crucial for abortion regulation's restrictiveness. Catholic countries have been shown to regulate more restrictively than Protestant countries (Castles Reference Castles1998; Minkenberg Reference Minkenberg2002, Reference Minkenberg2003). Therefore, we include the yearly share of the population belonging to the Catholic denomination (Maoz and Henderson Reference Maoz and Henderson2013). This is a significant improvement to previous research, operationalizing Catholicism as a dummy variable (Castles Reference Castles1998, 56) or a time invariant measure (Minkenberg Reference Minkenberg2002, Reference Minkenberg2003). Since we use yearly data on the national share of Catholics (five-year intervals, linearly imputed), this figure is also an admittedly crude measure of secularization.Footnote ⁷ Additionally, religious influence is controlled for by including a measure of the state-religion relationship (Fox Reference Fox2008, 32–61). It indicates whether state and church are completely separated, cooperate, or if there is a state religion. Since the relationship between the state and religions has been argued to mediate the ability of the Catholic Church to influence regulation (e.g., Minkenberg Reference Minkenberg2003), we also include an interaction term of Catholicism and the state-religion relationship.

Partisan influence, being a “classical” variable in comparative policy scholarship (e.g., Schmidt Reference Schmidt and Keman2002), is controlled for via the inclusion of a variable representing left and liberal parties' strength in national governments, because parties belonging to left and liberal party families tend to have socially progressive stands on issues such as abortion (Gindulis Reference Gindulis2002, 318). Data come from Heichel, Rinscheid, and Knill (Reference Heichel, Rinscheid and Knill2015).

Furthermore, the measure of partisan influence is complemented by a measure of political constraints, a variable taken from Henisz (Reference Henisz2002, Reference Henisz2013). Institutional constraints can limit the room for maneuver of political actors (Schmidt Reference Schmidt and Keman2002). Therefore, an interaction term of left and liberal parties in government and political constraints is included in the analysis (see also Fink Reference Fink2008, 79–80 for this interaction).