Abortion Referendums: Is There a Recipe for Success?

The topic of abortion has been among the most salient and polarizing issues on the European continent since the 1960s. However, abortion referendums have not received much attention from political scientists— to date, there is no comparative analysis of such popular votes. This article goes beyond the usual models of single case/country analyses and examines nationwide referendums on this policy in four European countries from the last four decades. The analysis focuses on twelve referendums and aims to identify the factors behind the pro- choice or pro- life outcome of a referendum by exploring the impact of relative consensus within the parliament, support of the head of government and medical experts, and the level of secularization of the country. The findings suggest that these factors have a substantial influence on the outcome.

relative consensus in the parliament, support for the head of government, support for medical experts, and level of secularization.
The next two sections of this article review the literature on the discourse around abortion and referendum process and derive four testable hypotheses. The following section describes the research design with a particular focus on case selection, criteria for operationalization, and the QCA method. Next, the main findings of the analysis are presented. The final section discusses the key findings, potential implications, and avenues for further research.

Abortion Discourse
The issue of abortion has been discussed from a variety of perspectives: social, medical, ethical, religious, legal, and feminist, to name a few. It would be safe to assume that the only uncontested aspect of abortion is its multifaceted nature. The abortion debate differs across countries since this topic involves historical and cultural meanings, reflects different stages of demographic transition, and struggles over gender roles as well as relations between church and state (Kulczycki 1999). For instance, in Portugal, debate around the right to terminate a pregnancy has centered around a "question of modernity" and "fulfillment of the democratic promises" of the 1974 revolution by allowing the people to have the final say on such an important matter (Manuel and Tollefsen 2008). In Italy, the debate was seen as a manifestation of a "bottom-up" democracy, where women's movement groups forced political institutions to discuss social and gender issues (Calloni 2001). In Ireland, there has been a general debate on the legitimacy and appropriateness of the law dealing with social issues such as abortion (Cacciaguidi-Fahy 2005).
Until the 1930s, abortion was criminalized in all Western countries. Historically, Western Europe, together with the United States, played a vanguard role in abortion law reform. Western societies have progressed greatly in this issue: abortion is no longer viewed as a criminal act, while access to safe abortion is seen as a human right. However, abortion remains a sensitive and ethically contentious issue, even in countries where the procedure has been legal for decades (Jones and Chaloner 2007). The rights of prospective fathers have become an increasingly recognized aspect of the topic (Hill 2001), as well as sexselective abortions (Muižnieks 2014), or provisions for conscientious objection to performing abortion that are available, for instance, in most EU countries and many U.S. states (Bertelsen 2013;Savulescu and Schuklenk 2017). 2 The multisided issue of abortion is also heavily debated in the political sphere. In the United States, the two established political parties have volleyed over the abortion issue for decades: Democrats generally being in favor of abortion rights, while most Republicans remain against. However, a sizeable share of Republicans and Democrats alike do not agree with their party's dominant position on the issue (Diamant 2020). In Poland, where one of the strictest abortion legislations is in force, the topic remains a political football match played by the church and state, with serious implications for women and society. Throughout the world, political parties have significantly influenced abortion politics and debate along with doctors, philosophers, and theologians (Kulczycki 1999). Moreover, as one comparative study on abortion policies shows, it is not the (changing) public opinion on the topic that serves as a key element explaining variation in legislation across countries, but rather the political influence of the main actors involved (Blofield 2008). 3 In most countries, the issue of abortion has been resolved within the legislative body; however, in some cases, it has been via referendum. Despite the voluminous research on abortion devoted to the latter path, to date, there is no comparative study of abortion referendums investigating the reasons behind the adoption or rejection of this specific type of policy. This is rather surprising since referendums differ significantly from other decision-making processes-referendums are accompanied by longer campaigns, have a higher likelihood of unforeseen events, and political parties are often internally divided over the policy (de Vreese and Semetko 2004;LeDuc 2002). Moreover, previous research acknowledges that the very nature of the policy type, put to a popular vote, influences the whole referendum process and its outcome. This is because every topic voted upon brings a distinct package of attributes. For instance, a referendum on a new nuclear power plant and a referendum on same-sex marriage would set completely different argumentation lines and result in variant modes of campaigning, actors involved, perceived political pressure for parties, levels of societal saliency, and degrees of emotions (Silagadze and Gherghina 2020a). Therefore, earlier studies have analyzed referendums separately, including EU/ NATO membership referendums (Hobolt 2009;Qvortrup 2016), constitutional popular votes (Anckar 2014;Tierney 2012), ethnic/sovereignty referendums (Mendez and Germann 2018;Qvortrup 2014), and fiscal/budget referendums (Feld and Matsusaka 2003;Kriesi 2012). In spite of the consensus regarding the peculiarity of both the topic of abortion and the referendum process, the crosscountry dynamics of abortion referendums have been largely neglected. This study addresses this gap in the literature and explores all referendums that have been held on this topic in Europe by combining factors from both referendum research and abortion discourse. A comparative study on abortion referendums that allows for going beyond idiosyncratic national context is an important endeavor precisely because abortion is a multidimensional contemporary issue, and referendums are used more frequently than ever for resolving contentious issues.

Abortion and the Referendum Process
From cross-country studies on popular votes, we know that the role of political parties in referendums is pivotal, although their level of control varies depending on the type of referendum and context (LeDuc 2003;). Referendums may serve as a crisis-solving mechanism for the political parties enabling them, for instance, to decouple a controversial topic from the upcoming election. In addition, parliament's highly divisive decisions might not be regarded as fully legitimate until they are contested in a referendum (Björklund 1982;Morel 1993Morel , 2007. There is increasing evidence that referendums are often instrumentalized by political parties and used primarily as an electoral strategy to promote the political agenda of the initiator, increase their popularity, and weaken the position of political opponents (Hollander 2019;Silagadze and Gherghina 2020b). Parties play a central role not only in the campaigning and outcome of the referendum, but also in the subsequent stage of implementation of the people's will-there have been a series of adopted referendums that have never been implemented (Bassanini 2012;Muntean, Pop-Eleches, and Popescu 2010) or attempts to overturn an inconvenient result, as in the case of Brexit.
There are various mechanisms that allow political parties to influence the referendum process regardless of its initiator. One of the components of parties' clout over the vote are the cues that they provide to citizens. When parties take a clear stance over the referendum topic, they perform best in mobilizing their electorate for the issue at stake as opposed to being internally divided (de Vreese and Semetko 2004). Division may occur when a party sends mixed signals to their voters and, thus, diminishes their influence or even discourages individuals from voting due to political ambiguity increasing uncertainty among citizens. Not surprisingly, referendums that are accompanied by clear and coherent political cues are more likely to be adopted compared to those with inconsistent messages (Silagadze and Gherghina 2018). yet divisions within a party over an issue is one of the most frequent reasons for calling a referendum in the first place (LeDuc 2002).
Closely related to party cues, the size of the parliamentary majority in favor of the proposal is another influential element for the outcome of a referendum. Initiators with a comfortable majority in the parliament are five times more likely to succeed compared to the initiators from the parliamentary minority (Silagadze and Gherghina 2018). However, referendums often occur precisely because of the lack of parliamentary majority in favor of a proposition. A consensual referendum that is backed by all parties is an exception rather than a rule (Henderson 2004). Consequently, to leave parties completely out of the analytical framework in cases where there is no formal majority would be a drastic oversimplification of the studied phenomenon. This is especially true regarding divisive and contested moral policies, including abortion, which provoke most conflicts within parties and where reaching consensus is highly challenging (Engeli and Varone 2012;Mooney 2001).
The lack of parliamentary majority for a certain topic does not necessarily mean the absence of any kind of consensus among the parties. Political actors have strong incentives to attract as many allies as possible in referendum campaigns. Since, in referendums, the political elite do not control the outcome, their best chance to influence the polls is to form a large camp that effectively campaigns either for or against the ballot measure (Bernhard 2019). Hence, the factor that is being tested in this study is relative consensus within the parliament for the issue at stake. Relative consensus is defined as a similar position echoed among at least three out of five of the largest parties, thus, probably not qualifying for the formal majority necessary to push through the legislation, but, nevertheless, exhibiting powerful support for the issue by potentially mobilizing the electorate behind the chosen position. The very nature of the referendum, which represents a decision between two alternative policies, pushes parties toward building consensus; on many occasions, several parties align on the same side, even if they are fierce competitors in other settings. 4 Moreover, parties have a strategic impetus for reaching relative consensus if they perceive similar opportunities from the referendum and aim at maximizing their chances for success. The more parties in favor of a referendum proposition, the higher the chances are for it to be adopted. Cross-party coalitions enjoy enhanced influence over their voters as they naturally tend to activate more constituents in the support of their position. The mechanism is twofold. First, parties and politicians are elected representatives who enjoy a certain degree of authority and standing among their voters. Accordingly, voters tend to follow their recommendation or at least consider it as a source of information. Second, considerable resources and funds (aside from staff, volunteers, know-how, and reputation) that parties have at their disposal fuel the campaign in combination with an increased media presence.
The role of political consensus for the outcome of a referendum has been widely acknowledged in previous research (Henderson 2004;Hobolt 2009;Kissane 2009). Political actors ultimately aim to win direct-democratic votes; consequently, they have a strong incentive to build relative consensus within the parliamentary arena that later evolves into intra-party coalitions during the referendums campaign, targeting the maximum number of voters. In their analysis of all referendums and initiatives voted on in Switzerland since 1947-1.485 acts- Trechsel and Sciarini (1998) conclude that there is a straightforward dependency between the level of consensus that was reached in the parliamentary process and the chances for an adopted referendum. Similarly, Kriesi (2005Kriesi ( , 2006 showed that the chances of success at the polls increased the larger a given camp was. Strategic coalition forming during a referendum tends to continue beyond parliamentary walls. Often parties and movements organize themselves together and campaign on the same side. Establishing and expanding links with movements enriches a party's support base (de Vreese and Semetko 2004;Hobolt 2006). If minority governments are routinely assigned with the task of ruling over the entire country, would it be unreasonable to assume that the same constellation of powers has significant influence over the outcome of a single referendum? In line with these arguments, I hypothesize that: Hypothesis 1: Relative consensus within the parliament on a referendum question favors its adoption.
After studying popular votes in 39 countries from the period between 1975, LeDuc (2003 infers that "in referendums, as in elections, the messenger often matters as much as the message." Although there are no candidate names on the ballot, the arguments are delivered by established political leaders, most prominently by the head of government. Their messages have an impact on voters for two interconnected reasons. First, "the heuristics, shortcuts and cues a voter can use in an election are not different in principle from those that can be used in popular votes," implying that voters rely on shortcuts provided by the parties and their leaders during the referendum campaign (Lutz 2007, 631). Second, the premier often enjoys a distinguished degree of media coverage, thus, having a substantially higher influence on the voters' choices. In general, senior officials and party leaders receive a high level of visibility since mediated politics is vastly personalized and tends to merely focus on a few important figures (Mazzoleni and Schulz 1999;Rahat and Sheafer 2007).
Various studies suggest that the position of prime minister could have a significant influence on voters due to the high media presence and authority that this statesman enjoys. This happens due to, for example, political standing and seniority-head of the cabinet and its members, as well as party leaders receive additional coverage in television, newspapers, and radio during elections and routine periods across countries (van Aelst et al. 2008). Furthermore, the premier has the "incumbency bonus" that guarantees vast attention from the media on a variety of issues well beyond the referendum topic (De Swert and Walgrave 2002). Compared to other political actors, the head of government thus has a significant edge in articulating their position on a plethora of occasions. This leads to the second hypothesis: Hypothesis 2: Support from the head of government for a referendum proposition favors its adoption. 5 Earlier studies have emphasized the powerful interests within the medical community in the issue of abortion (Latham 2002;Sheldon 1997;Stetson 2001). Physicians articulated their claim on a monopoly on abortion as early as the nineteenth century as part of the strategy to safeguard the professionalization of medicine and secure their supremacy over midwives (Mohr 1978). Historically, abortion was seen primarily as a medical issue, rather than political or social, thus, the medical community 6 determined its regulation (Sheldon 1997). In some countries, doctors' associations played a vital role in the initial liberalization of abortion laws, in others they were united to oppose the first reforms (Blofield 2008). An analysis of the reproduction policies in France and Switzerland upholds the distinctive role that the medical community has played, successfully imposing numerous preferences in regulating abortion since the 1970s (Engeli 2009).
Doctors are experts in the field of reproduction, holding retaining social authority, prestige, and trust. They are the ones who perform the procedure, and are therefore, in most cases, the first and last points of visit for women seeking assistance. Furthermore, the topic of abortion is often framed within medical terminology and reasoning, so physicians tend to stand between feminists and religious actors in the debate and are present in the whole referendum campaign, keeping in mind their corporate interests and maintaining their medical authority on the issue (Engeli and Varone 2012). Medical experts remain at the forefront of abortion debates because their expertise is considered indispensable, even by the political elite (Griessler and Hadolt 2006). "The intervention of the medical community was legitimized by their scientific expertise and their social prestige. Being at the same time the main implementer of the regulation and the tenants of the technological development, the medical community benefited from both consequential institutional and prestige resources" (Engeli 2012, 340-1). This suggests a highly influential role of these experts in the abortion discourse, and I expect: Hypothesis 3: Support of medical experts for the referendum question favors its adoption.
According to Inglehart's research (1977,1990) on postmaterialism, a "silent revolution" took place in Western societies manifesting in gradual value change. Postmaterialism is a broad term, but, in essence, it can be boiled down to two main aspects: the shift away from materialist values and the empowerment of an individual. This article focuses on the latter concept. As Inglehart (2000, 224) explains, "the postmodern world view is linked with declining acceptance of rigid religious norms concerning sex and reproduction and a diminishing need for absolute rules," and modern Western societies emphasizes individual freedom and self-expression. This suggests an incremental increase in the percentage of nonchurchgoers in these societies, along with the decline of the church's influence over people's lives. However, the development is not so straightforward-while religion is declining mainly in economically developed countries, it is persisting and even increasing in poorer countries (Norris and Inglehart 2004).
Many Christian denominations, with their traditional values, consider various societal developments, such as same-sex marriage, artificial fertilization, and the legalization of abortion, to be morally unacceptable, and they openly oppose them. The Roman Catholic Church has been the most visible transnational actor in the abortion debate; its opposition to liberalization remains ferocious and constant (Kulczycki 1999). 7 Consequently, nonreligious societies are likely to go in the opposite direction and allow for more liberal legislation. Existing literature confirms a strong correlation between the proportion of the population that is religious and the restrictiveness of abortion policy in the Western world (Cook, Jelen, and Wilcox 1992). 8 Following this reasoning, I hypothesize that: Hypothesis 4a: A high level of secularization favors the adoption of the pro-choice referendum. Hypothesis 4b: A low level of secularization favors the adoption of the pro-life referendum.

Data and Method
I investigate under what conditions legality of abortion was upheld in a popular vote and when it was undermined. To test these hypotheses, this article focuses on twelve national referendums from the last four decades in four European countries: five referendums in Switzerland, four in Ireland, two in Portugal, and one in Italy. In total, only five countries in Europe have ever held referendums on abortion, all of them predominantly Catholic. Other states have resolved the issue of abortion through regular legislative processes. Liechtenstein is not included in this analysis, in any of the surveys used for this article, for two reasons. First, it is a monarchy in which the Prince of Liechtenstein is one of Europe's most powerful monarchs with extensive veto powers. Second, it is a microstate with less than 40,000 inhabitants and, within this distinct political system, direct democracy functions differently than in its neighboring states (Marxer 2018). Three further referendums were excluded from the analysis as they did not directly address the question of banning or legalizing abortion. Rather, they dealt with the right to information relating to abortion and contraception and the freedom to travel for abortions abroad (Ireland in 1992), and whether the performance of abortions should be restricted solely to public hospitals (Italy in 1981). Accordingly, the entire universe of European abortion referendums is comprised of 18 popular votes, of which twelve are included in this analysis due to the outlined reasons. Table 1 illustrates the twelve identified cases that exhibit diversity in their institutional design. In Ireland, all votes were mandatory and binding, in Switzerland and Italy they were bottom-up and binding, in Portugal they were top-down and nonbinding. Moreover, the types of government vary: Portugal is a semi-presidential republic, Italy and Ireland are parliamentary democracies, and Switzerland is a semi-direct democratic federal republic. Of the twelve referendums, five had pro-choice wording (see Appendix 2) and seven had prolife wording, aiming at either banning abortion or tightening access to it. 9

Country Context
Until 1978, abortion was illegal in Italy, and performing or receiving one was punishable with up to five years of imprisonment. In 1975, the small Radical Party, together with one influential feminist group (Movimento della Liberazione delle Donne Italiane), launched a successful signature collection for an abortion referendum. A year later, 50,000 women marched for pro-choice rights in Rome (Ginsborg 1990). The political parties were unwilling to put the issue on a referendum and, thus, adopted a law that allowed for voluntary termination of pregnancy (DiMarco 2009). In 1981, the Roman Catholic Church, in cooperation with right-wing parties, attempted to repeal the law in a popular vote, but the referendum was soundly rejected by citizens (Ginsborg 2003).
In Ireland, abortion had been illegal since 1861; however, the debate around this topic was politicized in a different manner than in other countries. In the United States and Britain, the topic became salient as a result of a pro-choice movement, whereas in Ireland, it occurred in the aftermath of a pro-life movement. The organization, Pro-Life Amendment Campaign (PLAC), was actively engaged in convincing the leaders of main political parties of the need for a pro-life amendment to the Constitution. As a result, in the course of the 1982 general elections, the leaders of the two major parties, Fianna Fáil and Fine Gael, pledged to initiate the amendment if elected (O'Leary and Hesketh Note: a It is not the exact formulation of the referendum question but rather its essence. For instance, in the 2008 Portuguese referendum, voters were asked: "Are you in agreement with the decriminalization of the voluntary interruption of pregnancy, if carried out, by the woman's choice, in the first ten weeks in a legally authorized health institution?" In the table, it is designated in a more concise way: "Abortion in the first ten weeks of pregnancy on request, pro-choice." 1988). 10 Thirty-five years after the first referendum on abortion in 1983, Ireland voted to remove the Eighth Amendment in 2018. In Portugal, abortion was completely illegal until 1984. Later, the law allowed for abortion in four cases: rape, risk to the mother's life or her physical or mental health, or fetal malformation. In fact, hospitals and doctors interpreted the law in the most restrictive way: only severe mental illness was considered to be sufficient grounds for mental health risk. Besides, many hospitals did not perform legal abortions because of conscientious objection (Vilar 2002). Portuguese citizens were asked to vote twice on the matter of abortion. The first referendum, held in 1998, was rejected. Almost a decade later, another referendum granted women the right to terminate a pregnancy on request.
In Switzerland, the law from 1942 was in force, which stipulated a penalty of up to five years' imprisonment and a substantial fine for a woman terminating pregnancy unless in cases where a woman's health was in danger. However, the official ban was ignored in most of the Swiss regions, and no woman was convicted of an abortion-related offence since 1988 (Swiss Info, September 30 2002). It took four referendums over a span of 30 years until abortion finally was legalized in 2002.
When it comes to the referendum process and practice itself, these countries differ considerably. Aside from the two abortion referendums, Portugal has held only one national referendum in its modern history-in November 1998 on regionalization. In sharp contrast, Italy and Ireland are two of the most frequent users of the referendum tool in Europe, whereas Switzerland is the world leader in referendum practice. In the case of Portugal, the reason behind its rare use of referendums lies in the country's constitution. The 1976 Constitution made no provision for national referendums (Uleri 1996). Indeed, Article 115 explicitly forbids the use of national referendums on laws or treaties already passed by the National Assembly (Freire and Baum 2003). Contrarily, in Italy, the referendum instrument plays a significant role in the political process. A substantial number of Italian referendums have been called on the basis of Article 75 of the constitution which allows for an "abrogative referendum." Here, a referendum requested by 500,000 citizens or five Regional Councils may repeal an existing law, in whole or in part. Additionally, Article 138 states that constitutional amendments-already passed by Parliament but not yet in force-may be put to a popular referendum when such request is made by onefifth of the members of the Chamber of Deputies or the Senate, 500,000 voters or five Regional Councils (Uleri 2002).
In Ireland, according to the 1937 Constitution, the referendum tool may be applied on two occasions. First, under Article 46.2, any amendment to the Constitution requires approval of the people voting in the referendum on that particular issue. Second, Article 27 contains provision for a rejective referendum on ordinary legislation in the event of a clash between the Dail and the Seanad.
All Irish referendums were held in accordance with the first provision, no rejective referendum has ever taken place (Gallagher 1996). Switzerland is known for a wide range of direct-democratic tools. The Swiss Constitution provides for mandatory constitutional referendum (Article 123), according to which any changes to the federal constitution must be approved by a double majority-a simple majority of the people and a majority of the cantons. In addition, 30,000 citizens or eight cantons may ask for a rejective referendum on the law or decree. With 100,000 signatures a citizens' initiative can be launched. On the cantonal and local levels, direct democratic institutions are, in general, more elaborate (Trechsel and Kriesi 1996).

Qualitative Comparative Analysis
Referendums represent a sophisticated process with numerous interfering variables and, thus, do not tend to be easily explained by a sole factor or individual case. To assess the causal complexity of referendum outcomes, QCA serves as an adequate tool since it allows for identifying, in a systematic manner, the configurational impact of various factors and assessing the multicausality of a given phenomenon. QCA, based on Boolean algebra and set theory, integrates qualitative and quantitative research methods. It preserves the richness of an indepth qualitative approach combined with formalized cross-case comparisons, hence, strengthening the capacity for generalization (Ragin 2008). QCA thus allows for abstraction from the idiosyncrasies of single cases and generates comprehensive explanations of social phenomena by identifying alternative combinations of conditions that can produce a given outcome (Legewie 2013). Crisp-set QCA is the most widely used technique within this method (Rihoux and De Meur 2009) and is applied to this study due to the nature of the datamost conditions are dichotomous.
Necessity and sufficiency are two central concepts within QCA terminology since the goal of QCA is to identify what conditions or combinations of conditions are necessary or sufficient for the outcome. Condition A is necessary for outcome y if y cannot occur in the absence of A, but A alone is not enough to produce y (in QCA terms, y is a subset of A). Condition A is sufficient if y always occurs when A is present, however, other conditions besides A may also produce y (Rihoux and Ragin 2009). Empirically, it is quite rare to identify a necessary condition since factors usually influence the occurrence of an event or phenomenon in conjunction; hence, most conditions or combinations of conditions are "quasi-necessary" or "quasi-sufficient" (Legewie 2013;Ragin 2006).
The truth table analysis is the core element of the QCA data analysis, which serves to identify "causal recipes" (combinations of conditions) that are sufficient for the outcome. The distinguished feature of the fs/QCA truth table is that it includes all the combinations with empirical correspondents as well as those combinations that do not appear in reality (logical remainders), thus providing a complete picture of the limited diversity, which researchers nearly always face. The software finds the complex, intermediate, and parsimonious solutions; the latter includes both empirical cases and logical remainders. The intermediate solution is recommended as the main point of reference for interpreting QCA results (Ragin 2008) and is the main focus of this study. 11 Classic crisp-set QCA analysis consists of three consecutive steps. The first step is generating a truth table to get an overview of the data and possible patterns. The second step is the necessity analysis, which shows which conditions are necessary or sufficient. There is a relatively strict rule, according to which a high consistency score of at least .9 is a prerequisite for a necessary condition; for a sufficient condition, a benchmark consistency value of .75 is required (Schneider and Wagemann 2007). Consistency measures the degree to which a relation of necessity or sufficiency is met within a given dataset, its values ranging from 0 to 1, where 0 indicates no consistency and 1 indicates perfect consistency (Ragin 2006). It resembles the notion of significance in statistical models (Thiem 2010). After the necessity or sufficiency relationship has been identified, the next step is a truth table analysis, which provides causal recipes for various outcomes. The software also displays the recipes' raw and unique coverage. Raw coverage shows to what extent each recipe can explain the outcome; the lower a coverage score, the less empirically relevant the causal recipe. Unique coverage illustrates the proportion of cases that can be explained exclusively by a particular recipe. In essence, coverage provides a measure of empirical relevance, analogous to R 2 in statistical models (Thiem 2010).

Operationalization
This study aims to explain the outcome of abortion referendums. In a referendum on the issue of abortion, the public could be asked two fundamentally different questions-whether or not to ban or liberalize abortion. Therefore, the outcome of such a referendum is operationalized as a dichotomy: 1 if it is pro-choice and 0 if pro-life. For instance, if a referendum directed at banning abortion is rejected, it is coded as a pro-choice outcome. This method of coding offers a more nuanced approach compared to simply viewing referendums as adopted or rejected regardless of the actual wording of the proposition.
Four conditions are also dichotomized according to relevant thresholds and coded based on pro-choice versus pro-life attitudes. A relative consensus within the parliament (A) is present when three of the five largest parties in the parliament are in favor or against the referendum proposition, and absent if less than three of those parties are in agreement. For example, in a pro-choice 11 It must be acknowledged that there is no consensus among QCA experts on which solution is the most appropriate (see e.g., Lucas and Szatrowski 2014). The complex and parsimonious solutions are presented in Appendices 7 and 8. referendum, if there were a relative consensus within the parliament for the pro-choice proposition, it would be coded as 1 (in the same direction as the outcome); if there were a relative consensus against it or no relative consensus, it would be coded as 0. If a party were divided over the issue, it would be counted as .5 points. If, for instance, out of five parties in the legislative chamber, two were united in favor of a proposition and two were divided (2*.5 points), it would be counted as three parties in favor. Admittedly, this approach might seem somewhat unusual. However, first, we are dealing with the five largest parties that in most cases account for between 80 and 95 percent of seats. To completely disregard the role of parties that internally were not fully in accord, regarding the referendum question, would be to simplify the process dramatically and miss important insights. Although the effect of mixed party cues on voters is lower than the effect of clear party cues, it is not null (Higley and McAllister 2002). Second, there is some degree of inter party conflict in the vast majority of all referendums, not only with polarizing issues (Sinnott 2002). With the topic of abortion, which transcends traditional partisan cleavages, politicians tend to act according to their own personal belief system, their "conscience." Consequently, in referendums on issues linked to morality, it is not rare that prominent party members campaign for the opposite blocks. It is reasonable to assume that every elected representative can influence their voters, especially when they are visible in the media giving their opinion on a highly intimate question that divides even families.
The second condition is support for the head of government (B), which is operationalized as 1 if they take a pro-choice stance and 0 if they choose a prolife position. 12 The third condition is support of medical experts (C). In a prochoice referendum, if there was a consensus among the medical community in favor of the pro-choice policy, a score of 1 would be assigned. Any other constellation (divided or consensus against) would be coded as 0. In a pro-life referendum, (C) is coded as 0 if there is a consensus for the pro-life measure, 1 in any other constellation (divided or consensus against). The fourth condition is a high level of secularization (D). This is measured using survey questions about church attendance, as is typically done in debates about secularization (Norris and Inglehart 2004). The condition is coded as 1 (favoring a pro-choice outcome) if more than 60 percent of respondents from respective surveys state that they never to go to church or go less frequently than several times a year, and as 0 if fewer than 60 percent respond in this way. 13 Detailed information on all sources used for the coding of conditions is provided in Appendix 1.

Analysis and Results
The first step in the QCA analysis was generating a truth table (Table 2), which offers an overview of the data and allows detection of possible patterns (raw data and coding are listed in Appendix 2). There are no contradictory configurations: instances where the same configuration of values leads to different outcomes. Moreover, I observe three "perfect pro-choice cases:" two referendums in Switzerland in 2002 and the 2007 referendum in Portugal, corroborating the theory that a value of 1 for all the conditions leads to an outcome of 1 (all pro-choice conditions leading to a pro-choice outcome). The dataset does not include a "perfect pro-life case," one in which all conditions have a value of 0, leading to an outcome of 0 (all pro-life conditions leading to a pro-life outcome).
Consider the 2008 Portuguese referendum. The debate over abortion heated up during the 2005 parliamentary elections as the new leader of the Socialist Party, Jose Socrates, promised to hold a new referendum if elected as premier. Consequently, the ruling Socialist Party was united with the pro-choice prime minister, while the Social Democratic Party faced opposition from its members 13 The following sources were used for this condition: European Values Study and International Social Survey Programmes (see Appendix 1). For older cases in Switzerland, a secondary source was applied (Lachat 2012) due to the fact that Switzerland was not part of the above-mentioned surveys. The threshold is set at 60 percent since this figure indicated the clear majority in the society. However, the analysis with thresholds of 51 percent or 55 percent exhibits the same results due to the nature of the data. For the selected cases, the share of nonchurchgoers was either 60-70 percent or below 35 percent.  (Manuel and Tollefsen 2008). Doctors were also organized in favor of abortion rights (Whitten 2009). The campaign was centered around compassion, communal values, and human suffering. Many citizens were terrified by the idea that women found guilty of abortion faced up to three years in jail or a 20,000-dollar fine. In addition, each year 10,000 (mostly poor) Portuguese women who could not afford to travel to Spain for abortions were hospitalized in the aftermath of back street abortions, many dying (Manuel and Tollefsen 2008). Moreover, a highly salient trial in Maia showed the brutal picture of illegal abortions in Portugal. A nurse who performed abortions in her home was condemned to eight years in prison-a longer sentence than for sexual abuse (Vilar 2002). The "yes" side, which united feminists, doctors, and most of the political parties, won the referendum with around 60 percent. The second step in QCA analysis is a necessity analysis, which shows that condition A (relative consensus within the parliament) is both a necessary and sufficient condition for both pro-choice and pro-life outcomes with consistency and coverage values of 1.0 (see Appendices 3 and 5). Conditions C (support of medical experts) and D (high level of secularization) exhibit a high level of consistency, .88 and .77, respectively.

Pro-choice Recipes
The truth table analysis provides the following causal recipes for nine prochoice outcomes 14 : The first recipe can be understood as relative support within the parliament in favor of a pro-choice policy (A) together with either support of medical experts for pro-choice policy (C) or a high level of secularization in the country (D), leading to the adoption of a pro-choice policy. Raw coverage, which shows to what extent each recipe can explain the outcome, is rather high for each recipe, .77 and .88, respectively. Both causal recipes have a consistency of 1, meaning that each of them is a sufficient, but not necessary, conjectural cause for the pro-choice outcome of a referendum, and therefore, the consistency of the entire proposition is also 1. The solution coverage, indicating what proportion of membership in the outcome can be explained by membership in the causal recipes, also equals 1 (see Appendices 4 and 7).

Pro-choice Recipes in Practice
Here, two referendums that resulted in the adoption of a pro-choice policy are considered in more detail to shed light on how these recipes work in practice: 14 A small letter indicates the absence of the condition and a capital letter signals its presence.
(1) AC + AD the 1981 referendum in Italy and the 2018 referendum in Ireland, which follow the first recipe (AC), and the 1978 and 1985 referendums in Switzerland, which adhere to the second recipe (AD).
In 1981, the Roman Catholic Church in Italy attempted to repeal a law from 1978 that allowed for abortion on request. The referendum campaign was promoted by the Vatican, but only the Christian Democrats with Prime Minister Arnaldo Forlani and the neo-Fascists, MSI, supported it (Ginsborg 2003). Although the majority of doctors in Italy were male and Catholic, they were not united behind the yes campaign (DiMarco 2009). The initiative was crushingly defeated with 67.5 percent of votes against. Despite the high level of religiosity in Italy, with one-third of population attending church services every week and only 15 percent replying never (International Social Survey Programme: Social Networks Support and Systems- ISSP 1986ISSP 1988, the country took a more secular path in defining family by approving the right to divorce in 1974 and defending the right to abortion in 1981. Thirty-five years after the first referendum on abortion, Ireland voted to remove the Eighth Amendment which granted an equal right to life to the mother and unborn. The country was not necessarily less religious; rather, preceding events and the general environment were different. In 2013, after an incident where a woman died in the aftermath of denied abortion during a miscarriage, the strict abortion law was amended allowing for abortion if doctors considered a woman's life to be at risk from pregnancy complications or suicide. In 2016, the UN called on Ireland to reverse its strict abortion provisions. A year later, a citizens' assembly established by the government recommended unrestricted access to abortion during early pregnancy (Henley 2018). The political landscape had changed too. Already during the 2016 parliamentary elections, various parties made explicit manifesto commitments to hold a referendum on repealing the Eighth Amendment, including Labour, the Greens, Social Democrats, Sinn Féin, and the Workers' Party. Fine Gael took the alternative path by committing to convene the Citizens' Assembly to deliberate on the issue (Field 2018). The majority of the political parties in Ireland supported the yes campaign, including Fine Gael, Sinn Féin, Labour, and the Green Party. Most politicians supporting the No campaign came from Fianna Fáil, though its leader Micheál Martin and other prominent members supported the yes campaign. The only registered political party to fully support the No side was Renua Ireland, which received around 2 percent of votes in the previous elections (Field 2018). Prime Minister Leo Varadkar from Fine Gael, a fierce advocate of liberalization, described abortion in Ireland as "unsafe, unregulated and unlawful," and he had the full support of the medical community of the country (Henley 2018). In the end, the amendment passed with 66.4 percent approval.
Swiss citizens voted four times in a referendum on abortion over a 30-year time span until it was legalized in 2002. The constellation around the topic has changed with time, though the level of religiosity has remained low. According to ISSP (2000), almost half of the Swiss population was found to never attend church activities, whereas only around 12 percent did on a weekly or monthly basis. Earlier, the percentage of people who never attended religious services or attended less than several times a year was around 60 percent (International Social Survey Programme 1989). No precise data is available for the 1970s. However, according to a survey among voters, around 70 percent exhibited low church attendance, albeit certain confessional differences were visible: among Protestants, the share of frequent churchgoers was considerably lower than among Catholics as early as the 1970s (Lachat 2012). 15 In Switzerland, the first referendum on abortion liberalization was rejected in September 1977. After the initiative was defeated, both opponents and proponents were confronted with the question of how to deal with the law that was adopted by the Parliament in June 1977, allowing abortion for sociomedical reason-in cases of life endangerment for a woman, or if pregnancy would result in a "not otherwise preventable social emergency situation for a pregnant woman," requiring a social report and consent from two doctors (BBl 1977 III 88, Art. 3 and 4). The majority of the parliamentary parties viewed the legislation as an insufficient compromise. The Christian Democrats and the Evangelical People's Party (EVP) supported the bill, while all other parties were against it (Nebel and Hurka 2015, 75;Schmitter 2014). The proposition had the full approval of the medical community and was also backed by the head of the Federal Council, Christian Democrat Willi Ritschard (Engeli and Varone 2012). However, the legislation was criticized by both conservatives and liberals: for one side it was too progressive, for the other it was not far-reaching enough (Gindulis 2003). In the end, the referendum was decisively rejected by Swiss citizens on May 28, 1978.
Another referendum on abortion took place in Switzerland in June 1985 as a result of the initiative "Right to Life" launched by various pro-life organizations advocating a ban on abortion. The only political party supporting the initiative was the Christian Democratic People's Party (CVP) with its president of the Federal Council, Kurt Furgler, a professed opponent of abortion (Schmitter 2014). The Swiss parliament and government addressed the voters with the recommendation to vote against the proposition. The citizens followed this advice and rejected the initiative by nearly 70 percent (Gindulis 2003).

Pro-life Recipes
There are three cases that resulted in the adoption of the pro-life policy. The analysis of the truth table reveals the following intermediate solution: 15 It is beyond the scale of this article to account for vote differences between Protestant and Catholic cantons. Moreover, it is noteworthy that two Swiss cantons did not allow women to vote until the early 1990s: Appenzell Ausserrhoden (1989) and Appenzell Innerrhoden (1991).
(2) abc + abd The absence of relative consensus in favor of a pro-life policy (a) together with a pro-choice standpoint of the head of government (b) combined with either the absence of support from medical experts (c) or a low level of secularization (d) leads to the adoption of a pro-life policy. Two cases follow the first recipe, the 1977 referendum in Switzerland and the 1998 referendum in Portugal. The 1983 referendum in Ireland follows the second recipe. The recipes' raw coverage ranges from .33 (abd) to .66 (abc), and both the solution coverage and consistency have a value of 1 (see Appendices 6 and 8). Admittedly, these recipes are somewhat puzzling since one would expect some of the mentioned ingredients to produce a pro-choice outcome. One should be highly cautious about drawing any generalizations from only three cases. However, a possible explanation might be the time variable. It is striking that all three referendums that produced a pro-life outcome were the first popular votes held on this topic in the respective countries. It is safe to assume that the societies were more traditional and the Church enjoyed a higher influence in the 1970s and 1980s than in the 2000s. Moreover, pro-choice groups learned from their experience and were able to organize themselves more efficiently in the subsequent votes.

Pro-life Recipes in Practice
On September 25, 1977 Swiss citizens voted in a referendum on the full liberalization of abortion, which was initiated by feminist groups. The federal Parliament was unable to agree on whether to recommend the proposal, thus, for the first time Swiss citizens voted on a private initiative without Parliament's advice (Gindulis 2003, 159;The New York Times, September 24, 1977). The government, the Federal Council, however, recommended to reject the initiative without offering a counter-proposal (Schmitter 2014, 72). The head of the Federal Council, Christian Democrat Kurt Furgler spoke out against the liberalization. The opponents of the initiative had powerful financial resources at their disposal and support from various political parties including the CVP, the Swiss People's Party (SVP), and the EVP (Schmitter 2014). The Liberals (FDP), Social Democrats opposed free abortion and favored a more moderate liberalization, and the medical community was divided (Engeli and Varone 2012). In the end, the initiative was narrowly defeated with 51.7 percent of votes against the liberalization.
In February 1998, the National Assembly of Portugal passed a bill that would provide for an abortion in the first ten weeks of pregnancy under certain conditions. However, Prime Minister Antonio Guterres, a pro-life Roman Catholic, was against the bill and called for a referendum. The tactic of calling a referendum sparked a lot of controversy, some arguing that Guterres violated the Constitution since it did not explicitly allow for measures adopted by the parliament to be subjected to a popular vote (Freire and Baum 2003). Thus, the ruling Socialist Party was divided with the pro-life prime minister and the majority of pro-choice MPs. The Social Democratic Party was mainly pro-life, Christian Democrats (People's Party) were united against liberalization, and the Democratic Unity Coalition (between Communists and Greens) was in favor (Manuel and Tollefsen 2008). The medical community was divided, though the mainstream discourse among medical professionals was against abortion (Whitten 2009). In the end, the "yes" side lost by a 2 percent margin with a low voter turnout of 32 percent, suggesting that the Church had been successful in mobilizing, whereas moderates opted for staying at home (Blofield 2008).
Even though abortion had been illegal in Ireland for over a century, a referendum on a pro-life amendment to the Constitution was held in 1983 as a result of the pro-choice movement. Fianna Fáil campaigned in favor of the amendment. Fine Gael and Labour were not so unanimous. In fact, their members opposed each other in the campaign. yet, shortly before the voting day, four party leaders addressed the nation with the following message: Haughey from Fianna Fáil was for "yes;" Prime Minister FitzGerald from Fine Gael, as well as Labour's and Workers' party leaders were for "No" (O' Leary and Hesketh 1988). The topic was highly divisive not only among politicians but doctors, lawyers, and many other professions were split and campaigned for opposite camps (O'Carroll 1991, 55). In the end, the amendment passed with 67 percent approval. The vote reflected the high level of religiosity in the country-more than 80 percent of the Irish population attended religious services on a weekly basis, suggesting a very low level of secularization, which could be translated into a low acceptance of abortion rights (EVS-European Values Study 1981-Integrated Dataset 2011. It is important to note that in the preceding parliamentary elections, Fianna Fáil gained 45 percent of votes, while Fine Gael and Labour gathered around 52 percent together, suggesting that even if the latter two parties were divided in their positions, the pro-life side would have had a clear majority.

Conclusion
The topic of abortion remains a salient issue since the idea that women have the right to terminate pregnancies has not yet reached the level of global institutionalization, and future referendums on this issue can be expected. This article aimed to identify multiple factors that contributed to the pro-choice or pro-life outcome of referendums in Europe since the 1970s. The analytical framework accounts for the idiosyncratic nature of the abortion policy, by including level of secularization and support of the medical community and bridges it with the insights from referendum research; that is, the role of consensus and position of the head of government. The findings indicate that relative consensus in the parliament, support of medical experts, and level of secularization play a major role in the adoption of a policy. A referendum had a pro-choice outcome in cases when relative consensus within the parliament was in favor of a pro-choice policy together with either support of medical experts or a high level of secularization in the country. No empirical evidence was found for support of the head of government.
These findings are relevant to the study of referendums since they provide first grounds to include the concept of relative consensus into future theoretical explanatory models. The factor of relative consensus explains the outcome in all referendums: it was the common denominator among all pro-choice outcomes, whereas its absence was the common denominator in all pro-life outcomes. So far, the concept of relative consensus has been applied in other fields (Cao et al. 2004;Miethe 1984) but has not found its way into political science due to its diffuse character. Relative consensus is different from parliamentary majority since in referendums there are usually only two sides, with all the votes reallocated to the "yes" or "No" side. This means that the classic representation formula in the parliament or exact percentage of a party is not as important as its mobilization effect, clear stance, and persuasive power.
Based on previous research about the relationship between consensus and adoption of a referendum (Kriesi 2006;Trechsel and Sciarini 1998), we know that in campaigns the number and variety of actors matter: the more, the better. This is because in popular votes it is about reaching the critical mass; the goal is to communicate with one voice through as many channels as possible. Consequently, a reshuffling of political forces occurs. It might seem somewhat simplistic, but, indeed, a referendum offers a simple choice with predominantly two single options. Thus, the whole electorate is divided not among all political parties competing with each other in elections, but rather between two groups with often heterogeneous team members. It might appear to be too early to discuss seriously the possibility of building a "theory" or even a "model" based on this factor. Nevertheless, it is plausible that relative consensus exhibits theoretical relevance by introducing another dimension to the study of referendums.
Admittedly, this article has some limitations. It considers a fairly restricted universe of cases both in content and quantity and, thus, some caution must be exercised when generalizing the results. Moreover, the operationalization of the relative consensus requires further elaboration, although the findings suggest the existence of certain threshold levels. This study is merely a first tentative step in introducing this factor in the explanatory models of referendums. Another limitation is the lack of control for interdependency and time. Interdependency of cases within a country refers to the possibility of one referendum affecting another. The time variable accounts for historical changes and the general process of modernization of social relations, including the status of women in the family and society. In spite of these points, the study serves as a good point of departure for future analyses expanding the unit of analysis to popular votes on other moral issues: for example, the death penalty, same-sex marriage, and divorce. Another avenue for further research would be to elaborate on the unexpectedly low influence of the head of government on the outcome of a referendum.