DEPARTMENT OF POLITICAL SCIENCE CORRUPTION AND SEX TRAFFICKING OF WOMEN AND GIRLS UNDER THE CONTEXT OF WOMEN’S POLITICAL EMPOWERMENT A quantitative cross-sectional study Anastasia Makhota Master’s Thesis: 30 credits Programme: Master’s Programme in International Administration and Global Governance Date: 22.05.2023 Supervisor: Amy Alexander and Jana Schwenk Words: 19957 Abstract All over the world, millions of women and girls are sexually exploited in a lucrative human trafficking business. They represent so-called “hidden populations”, where the exact number of victims is unknown. Hence, investigating sex trafficking requires an in-depth understanding of what factors drive this criminal activity and, therefore, under what contexts the situation of women can improve. Despite arguments in the literature that corruption is a root cause fuelling human trafficking, limited research has explored how corruption affects sex trafficking exclusively. The existing scholarship also discusses the importance of women’s political empowerment (WPE) in curbing corruption and advancing gender-specific policies, including violence against women. In this thesis, building on previous research, I first argue that corruption reduces the presence and effectiveness of anti-sex trafficking legislation and, second, develop theoretical arguments on how WPE might moderate this relationship. By conducting a quantitative cross-sectional analysis among 131 countries, I find empirical confirmation that higher levels of corruption lead to worse anti-sex trafficking laws and enforcement. The expectation that WPE would weaken the negative effect of corruption on the adoption and enforcement of anti-sex trafficking laws finds only weak support, and the finding is sensitive to the corruption measurement employed. Thus, this thesis contributes to the understanding of corruption as a key determinant of sex trafficking and the role of WPE in advancing anti-sex trafficking legislation and enforcement. Future research should conduct more rigorous tests on the moderating effect of WPE, considering the limitations of this study, and also explore in detail what corruption types contribute the most to sex trafficking of women. Keywords: Human trafficking, sex trafficking, anti-sex trafficking laws and enforcement, corruption, women’s political empowerment. 2 Acknowledgements I would like to express my deepest gratitude to my supervisors, Amy Alexander and Jana Schwenk, for their excellent academic guidance, contribution to the research process and valuable feedback. Their extensive expertise allowed me to advance this work and acquire new knowledge and skills in the process of writing this thesis. I could not have undertaken this journey without support from the Swedish Institute. I thank the Scholarship Committee for believing in me and granting me a highly prestigious scholarship for global leaders. My appreciation also goes out to my family and friends for their continuous support throughout my studies in Sweden. You were always there for me, through good times and bad. In particular, words cannot express my gratitude to my mother, Oxana. Thank you for encouraging me to follow my dream of studying abroad and supporting me in every possible way. 3 List of Abbreviations BCI - Bayesian Corruption Index CATW - Coalition Against Trafficking Women CEDAW - Convention on the Elimination of All Forms of Discrimination Against Women CPI - Corruption Perceptions Index EIGE - European Institute for Gender Equality GDPPC - Gross Domestic Product Per Capita IBA - International Bar Association OLS - Ordinary Least Squares QoG - Quality of Government SDGs - Sustainable Development Goals TI - Transparency International TIP - Trafficking in Persons TVPA - Victims of Trafficking and Violence Protection Act UN - United Nations UNODC - United Nations Office on Drugs and Crime VAWG - Violence Against Women and Girls V-Dem -Varieties of Democracy VIF - Variance Inflation Factor WPE - Women’s Political Empowerment WPEI - Women’s Political Empowerment Index 4 Table of Contents 1. Introduction ...................................................................................................................... 7 2. Literature Review ............................................................................................................ 9 2.1. Understanding Human Trafficking ........................................................................... 10 2.2. Sex Trafficking and Sexual Exploitation .................................................................. 11 2.3. Determinants of Human and Sex Trafficking ........................................................... 13 2.4. Corruption as a core push factor of Human Trafficking ........................................... 15 2.4.1. Corruption and Sexual Exploitation ...................................................................... 17 2.5. Women’s Political Empowerment, Corruption, and Violence Against Women ...... 18 2.5.1. Women’s Political Empowerment and Corruption ................................................ 19 2.5.2. Women’s Political Empowerment and Violence Against Women and Girls ......... 21 3. Theoretical Framework, the Gap, and Hypotheses .................................................... 22 3.1. Theory and the Gap ................................................................................................... 22 3.2. The Effect of Corruption on Anti-Sex Trafficking Legislation and Enforcement .... 23 3.3. The Effect of Corruption on Anti-Sex Trafficking Legislation and Enforcement Moderated by Women’s Political Empowerment ............................................................ 24 4. Research Design ............................................................................................................. 26 4.1. Data and Dataset ....................................................................................................... 27 4.1.1. Dependent Variable: Sex Trafficking Data and Operationalisation ..................... 27 4.1.2. Main Explanatory Variable: Corruption Data and Operationalisation................ 31 4.1.3. Moderating Variable: Women’s Political Empowerment Data and Operationalisation ........................................................................................................... 33 4.1.4. Control Variables................................................................................................... 34 4.2. Method ...................................................................................................................... 39 4.2.1. Lagging of variables .............................................................................................. 39 4.2.2. Statistical Method .................................................................................................. 40 5. Analysis and Results ...................................................................................................... 41 5.1. Validity and characteristics of the data ..................................................................... 41 5.2. Statistical Analysis .................................................................................................... 42 5.2.1 Multivariate regressions: Results for the Main Effect ............................................ 42 5.2.2. Multivariate Regressions: Results for the Interaction Effect ................................. 45 5.3. Robustness tests ........................................................................................................ 50 5 5.3.1. Regression diagnostics for OLS ............................................................................. 50 5.3.2. Ordered Logistic Regression ................................................................................. 51 5.3.3. Alternative measures of corruption and time lags ................................................. 51 5.3.4. Bigger sample excluding Share of protestants in 1980 ......................................... 52 6. Discussion of Findings and Limitations ....................................................................... 53 7. Conclusion ...................................................................................................................... 58 References ........................................................................................................................... 61 Appendix ............................................................................................................................. 74 Appendix 1: Original Scales of Variables ....................................................................... 74 Appendix 2: Descriptive Statistics ................................................................................... 75 Appendix 3: Correlation table .......................................................................................... 76 Appendix 4: Diagnostics .................................................................................................. 77 Appendix 5: Multivariate Regressions (OLS): Results for the Main Effect including Regions ............................................................................................................................ 80 Appendix 6: Robustness tests .......................................................................................... 82 6 1. Introduction Human trafficking is a crucial social issue on the development agenda across all countries and communities today. It is a multifaceted concept encompassing human rights, gender, migration, corruption, and law enforcement (Laczko, 2005). According to the Protocol to Prevent, Suppress and Punish Trafficking in Persons, especially Women and Children (2000), human trafficking comprises labour trafficking, sexual exploitation, slavery, and organ trafficking. The International Labour Organisation reports 25 million victims trafficked globally with purely exploitative motivations, while the whole trafficking business produces an estimated $150 billion in revenue per year (Niethammer, 2020). Women and girls account for the majority of human trafficking victims (72%), as well as identified sex trafficking victims (94%) (UNODC, 2018). Therefore, sexual exploitation and sex trafficking are blatant manifestations of sex- and gender-based violence and discrimination (CATW, 2023). They not only result from but also sustain violence against women (Ibid). Sex trafficking networks thrive on targeting marginalised and vulnerable women and girls who often are subjected to physical and psychological abuse by the exploiters, which has serious consequences even after the victim is released. Numerous studies focused on the determinants of human trafficking. For women, factors such as poverty, young age, race, lack of opportunities, and previously experienced sexual abuse play an essential role in this crime (Cho, 2015a; TIP Report, 2022; CATW, 2023). Besides individual-level factors, an extensive scholarship exists on corruption and trafficking in persons. Researchers argue that public sector corruption facilitates human trafficking and undermines human security (Zhang and Pineda, 2008; Uddin, 2014; Cho, 2015a). However, despite the research community recognising trafficking in persons as a gendered phenomenon (Uddin, 2014; Alexander and Ravlik, 2015), only one study attempted to quantitatively analyse the effect of corruption on sex trafficking of women and girls (Jonsson, 2018). Furthermore, several scholars argue that higher women’s political empowerment (WPE) reduces corruption and violence against women and, thus, is essential for success in anti-trafficking interventions (Uddin, 2014; Alexander and Ravlik, 2015; Alexander and Bågenholm, 2018; Ravlik, 2019). However, no study analysed how corruption affects sexual exploitation of women and girls and whether WPE affects this relationship. Based on the literature, there are reasons to expect a moderating effect of WPE on the corruption - sex trafficking nexus such that greater WPE weakens the negative relationship. Thus, this thesis aims to contribute to the growing literature 7 on corruption and sex trafficking by investigating first, how corruption and sex trafficking are related and, second, how WPE might moderate this relationship. To fulfil the research aim stated above, this thesis will answer the following research question: “How does women’s political empowerment moderate the relationship between corruption and anti-sex trafficking legislation and enforcement across countries?” It is important to clarify that while sex trafficking is the main concept, this study employs the adoption and enforcement of anti-sex trafficking laws data in the analysis as the only suitable measurement of the concept. Therefore, I use the expressions “adoption and enforcement of anti-sex trafficking laws”, “anti- sex trafficking legislation and enforcement”, and “regulation and enforcement of anti-sex trafficking laws” as synonymous throughout the thesis to refer to sex trafficking in general. In addition, the concept of corruption in this thesis refers to corruption in the public sector, which is discussed in further detail in the literature review. To answer the research question, I first develop theoretical arguments on how corruption affects anti-sex trafficking legislation and enforcement by reinforcing women’s marginalisation and vulnerability, facilitating the sexual exploitation of women and girls and weakening the criminal justice system. I further develop the theoretical framework arguing that WPE moderates the relationship between corruption and anti-sex trafficking legislation and enforcement by improving accountability and transparency of the system, developing women- friendly policies, including policies on violence against women and girls (VAWG), and mobilising civil society to advocate for marginalised groups’ interests, including VAWG. I test the theoretical arguments briefly outlined above through a quantitative cross-sectional study employing an Ordinary Least Squares (OLS) multivariate regression with an interaction term on a sample of 131 countries using data from secondary data sources. This thesis’s principal finding is that higher corruption levels lead to worse anti-sex trafficking legislation and enforcement. However, sex trafficking measurement used in the thesis is sensitive to types of corruption and various corruption indicators. Furthermore, there is some evidence in favour that under higher levels of WPE, the negative relationship between corruption and the adoption and enforcement of anti-sex trafficking laws across countries becomes weaker. Given the low significance level of the latter finding, conclusions should not be given too much weight. Nevertheless, these results are consistent with theoretical 8 expectations and also indicate that the level of WPE seems to matter more in moderating the effect of corruption on anti-sex trafficking laws and enforcement among countries that have moderate or high levels of corruption. Overall, this thesis brings valuable theoretical implications and suggests new directions for research. Complementing the young scholarship on sexual exploitation of women and its determinants, the study also recognises that women could theoretically be most vulnerable to particular types of corruption, which opens the door for future research. In addition, the thesis acknowledges that further research is needed on the moderating effect of WPE on the relationship between corruption and sex trafficking. It should directly test these mechanisms with consideration of the limitations of this study. Furthermore, the analysis results lead to conclusions about the critical role of state capacity in combating sex trafficking, with policy implications such as adoption of strong anti-corruption measures and collection of statistical data on sex trafficking prevalence in a country. Also, governments must strengthen their efforts to reach gender equality both in political office and civil society, participating in international agreements and developing comprehensive action plans at the state level. The structure of the thesis is as follows: the subsequent chapter discusses the existing State of the Art on human and sex trafficking in relation to corruption and engages literature on women’s political empowerment. After that, the theoretical framework is presented, ending with the two developed hypotheses to address the research question. The following section presents the research design and discusses data, its operationalisation and the statistical method. Subsequently, the statistical regression analysis and the results are presented. The following section discusses the main findings and limitations of the study. The last chapter offers conclusions and a summary of the findings, as well as policy implications and avenues for future research. 2. Literature Review To address the research question stated above, this study begins with a comprehensive review of the existing State of the Art. Starting with an introduction of vital concepts, the review offers a discussion of the phenomenon of human trafficking, moving towards the focus of this thesis: sex trafficking and the sexual exploitation of women and girls. By defining and shaping a better understanding of central concepts, it shows both the relevance and problematic nature of sexual 9 exploitation. Consequently, the review gives a broad perspective of the key determinants and drivers of human and sex trafficking in the modern world, revealing the most powerful push factor of trafficking in persons - corruption. As such, the literature review shows arguments that corruption is often gendered and, thus, affects the most vulnerable and marginalised groups, namely women and girls. Therefore, in states with rampant corruption, females become victims of sexual extortion and sex trafficking. Moreover, corruption and violence against women and girls fuel and maintain the existence of lucrative trafficking businesses. However, the review further presents findings that higher gender equality and women’s political empowerment, particularly, have a positive effect on curbing corruption and addressing the issue of violence against women on state and international levels. 2.1. Understanding Human Trafficking As mentioned above, human trafficking is a complex multidimensional concept. While international organisations and scholars may use other umbrella terms such as “trafficking in persons” or “modern slavery,” it is referred to the same type of crime when adults or children are exploited and forced to do labour or participate in commercial sex by traffickers, who profit at their cost (TIP Report, 2022). While there are numerous explanations for the human trafficking concept, the internationally accepted definition exists. In 2000, the United Nations presented a new multilateral treaty, the Convention against Transnational Organised Crime, which became the primary international instrument in combating human trafficking. This Convention was supplemented with three additional Palermo protocols the same year. One of the protocols, the Protocol to Prevent, Suppress and Punish Trafficking in Persons, especially Women and Children, is of particular interest here as it includes standardised terminology and comprehensively defines human trafficking. Also known as the UN Trafficking in Persons Protocol (the UN TIP Protocol), it has been ratified by 180 parties as of January 2023 (the UN Treaty Collection, 2023). According to the Protocol, the crime of trafficking in humans consists of three integral elements: the act, the means, and the purpose. Thus, human trafficking is defined as “the recruitment, transportation, transfer, harbouring or receipt of persons, by means of the threat or use of force or other forms of coercion, of abduction, of fraud, of deception, of the abuse of power or of a position of vulnerability or of the giving or receiving of payments or benefits to achieve the consent of a person having control over another person, for the purpose of exploitation” (UN, Article 3, paragraph (a) of the UN TIP Protocol, 2000, p. 2). The 10 exploitation comprises four types of human trafficking, namely labour trafficking, sexual exploitation, slavery, and organ trafficking (Ibid). One of the primary purposes of the Protocol is the protection of aggrieved people, which is reflected through its strong stance on victims’ consent and innocence. The document states that when above mentioned means are applied, it does not matter whether a victim gives consent (UN, Article 3, paragraph (b) of the UN TIP Protocol, 2000, p. 2). If traffickers use force, fraud, or other means, the victim’s consent is irrelevant. Thus, the exploitation principle comes as the core of human trafficking. The exploitative pattern of traffickers is the priority, as meaningful consent in this situation is highly questionable. In particular, children cannot legally consent to commercial sex, making it a crime by default (TIP Report, 2022). Furthermore, it should be underlined that human trafficking does not imply the physical movement of victims across borders or even to another location inside a country (Ibid). Often victims are exploited in their local areas, thus making exploitation a crime where the movement of people is an additional but not a necessary component (Ibid). However, Friman and Reich (2007) criticise the vagueness of wording used when describing forms of exploitation and sexual exploitation in particular. The authors state that this vagueness is deliberate, to incentivise a wider range of countries to ratify the Protocol and, thus, modify the definition of exploitation in line with their internal policies (Ibid). Thus, problem of capturing trafficking properly arises when governments act as traffickers, purposefully entrapping their residents to participate in sexual slavery, forced labour, or military arrangements (TIP Report, 2022). 2.2. Sex Trafficking and Sexual Exploitation Although modern slavery exists in many forms, scholars and governments often focus on and recognise two main types: forced labour and sex trafficking (Friman and Reich, 2007; TIP Report, 2022). Moreover, research largely agrees that trafficking is a gender-based phenomenon (Uddin, 2014; Alexander and Ravlik, 2015). Traffickers mostly focus on people in vulnerable positions, as their powerlessness makes them easy targets. Women and girls happen to be the most marginalised group and, hence, are disproportionately affected by trafficking (Ibid). Alexander and Ravlik (2015) emphasise that gender inequality is a great 11 catalyst for human trafficking. In her book, Ravlik (2019) further argues that it also prevents the efficient enforcement of anti-trafficking laws. As this thesis focuses on the sexual exploitation of women and girls, forced labour and other types of trafficking in persons will not be discussed in detail. With regard to sex trafficking, the UN TIP Protocol does not include a clear definition. The U.S. Department of State, in its annual TIP Report (2022), applies an “acts, means, and purpose framework” to define it. Thus, sex trafficking refers to acts of “recruitment, harbouring, transportation, provision, obtaining, patronising, or soliciting” that occur when a trafficker employs force, deception, or coercion to engage another adult or child in a commercial sex act (TIP Report, 2022, p. 33). Sex trafficking may take place in various settings, including private residences, hotels, brothels, massage parlours, and over the Internet (Ibid). Importantly, means used by traffickers often lead to serious physical, psychological, and reputational harm to victims. In addition, females can be exploited for years, being in debt bondage and, thus, manipulated by a trafficker, which is considered illegal activity (Ibid). McAlpine, Hossain, and Zimmerman (2016), based on the UN Secretary-General’s Bulletin on protection from sexual exploitation and abuse (2003), give the following definition to sexual exploitation: “any actual or attempted abuse of a position of vulnerability, differential power, or trust, for sexual purposes, including, but not limited to, profiting monetarily, socially or politically from the sexual exploitation of another.” This definition of sexual exploitation, along with the concept of sex trafficking, gives a better understanding of the issue and shows its broad application and, thus, complexity when addressing it in studies. Both concepts, sex trafficking and sexual exploitation, represent a challenge for research, especially to its empirical part, due to the topic’s sensitivity and, therefore, difficulties connected to data collection. CATW (2023) states that sexual exploitation and sex trafficking are manifestations of gender- based violence and discrimination. It is happening as a result of this violence and discrimination and sustaining it. Violating fundamental human rights, sex traffickers and pimps1 target and exploit the most vulnerable women and girls. They expose females to violence such as physical 1 According to Roots (2013, p.31), a pimp is a person who subsists on the income of prostitutes, encouraging prostitution by various means, including violence. 12 abuse, rape, beatings, malnutrition, detention, physical threats, torture, forced drug and alcohol usage, and humiliation (Ibid). Consequently, a source of and a result of violence and discrimination against women and girls is the sex trade. This extreme form of crime arises from sex trafficking, poverty, and lack of choices in life of women. At the moment, there is growing attention to the issues of sex trafficking and sexual exploitation, including the exploitation of prostitution. International law highlights the criminal nature of these forms of violence against women and girls, which is documented in several major international treaties such as the Convention for the Suppression of the Traffic in Persons and of the Exploitation of the Prostitution of Others (1949), the UN Trafficking in Persons Protocol, the Convention on the Elimination of All Forms of Discrimination Against Women (CEDAW). Therefore, eliminating all forms of violence against women, including sexual exploitation and sex trafficking, is on the global development agenda and exclusively included in Goal 5 (Gender Equality) of 17 Sustainable Development Goals (SDGs) (UN General Assembly, A/RES/70/1, 2015). 2.3. Determinants of Human and Sex Trafficking Every country globally is engaged in human trafficking in one way or another: either as a source of victims (sending or origin countries), a means of transit (transit countries) or as a final destination (destination countries) (Zhang and Pineda, 2008; CATW, 2023). Zhang and Pineda (2008) argue that trafficking tends to be from less economically developed countries to more advanced states. Nevertheless, trafficking in persons takes place both across borders and locally, without victims’ transportation. Thus, it is vital to identify factors that lead to human trafficking and sex trafficking in particular. While research on sexual exploitation is still in its early stages, it is possible to refer to the more established bulk of the literature concerning human trafficking and its determinants to build an understanding of which factors might play an important role in sex trafficking. Cho (2015a) discusses a broad set of push and pull factors affecting people’s migration and, therefore, influencing human trafficking. So-called push factors ascertain the supply of victims from sending states, while pull factors reflect the demand in destination countries met by the trafficking of victims (Ibid). The author offers a framework based on four pillars - migration, vulnerability, crime, and policy and institutional efforts - through which it is possible to explain various factors of trafficking in persons in origin and destination states. Among numerous 13 factors, some can be relevant to more than one pillar (Ibid). Firstly, trafficking is closely interlinked with migration, as the majority of victims of human trafficking are initially migrants (Avdan, 2012; Rao and Presenti, 2012; Cho, 2015a; Jonsson, 2018). It means that migratory drivers coincide with push and pull elements of trafficking. Thus, in the literature, scholars highlight such push factors as poverty and the low gross domestic product (GDP) per capita, high unemployment rates, and gender inequality (Ibid; Zhang and Pineda, 2008). Migrants looking for better opportunities in wealthier countries with a demand for a cheap labour force often fall into the traps of traffickers. Therefore, many migrants are vulnerable to trafficking, especially women and girls, who often become victims of gender-based violence and are exploited for commercial sex (Ibid). Moreover, the crime prevalence expands the scope of the trafficking in persons’ business (Cho, 2015a). Finally, regarding policy and institutional efforts, such factors as inadequate anti-trafficking measures, low levels of law enforcement, and corruption also positively affect human trafficking activity in both origin countries and destination countries (Ibid). In addition, Zhang and Pineda (2008) and Cho (2015a) mention globalisation as one of the most significant predictors of trafficking in persons. Higher awareness about the gap in living standards around the world, accompanied by better transportation and trade openness, encourages both migrants and traffickers (Ibid). However, Masci (2004) argues that globalisation, on the other hand, can enhance international cooperation in anti-trafficking efforts and raise living standards in developing countries. Along with the authors mentioned above, Ravlik (2019) in her book presents a triangulated nexus of anti-trafficking enforcement, where three major factors - organised crime, corruption, and human rights violations - facilitate trafficking and hinder the implementation of effective anti-trafficking legislation across countries. Indeed, this framework aligns with previous findings in the literature on human trafficking and its determinants. Scholars like Cho (2015a) and Zhang and Pineda (2008) emphasise, based on their quantitative analysis, that factors of crime and corruption are the most influential when it comes to trafficking in persons. Therefore, the difficulties in executing anti-trafficking laws are significantly greater in states with widespread corruption and criminal human rights violations (Ravlik, 2019). 14 2.4. Corruption as a core push factor of Human Trafficking Even though there are dozens of determinants affecting trafficking in persons, some play a more significant role. Authors like Bales (2007), Zhang and Pineda (2008), Studnicka (2010), Cho (2015a), Jonsson (2018), and Ravlik (2019) underline the significance of corruption as a factor influencing human trafficking, providing empirical support via quantitative analyses. Thus, there is a strong link between corruption and trafficking, which are both considered criminal activities. Hughes and Denisova (2001) especially bring readers’ attention to the sexual exploitation of women from Ukraine, revealing its nexus with political criminal networks and corruption. Nevertheless, despite that numerous studies have tried to address corruption, this phenomenon represents a challenge for research as corruption is hard to define and capture empirically. Scholars tend to identify such elements of corruption as bribery, patronage, cronyism, nepotism, and public funds embezzling, among others (Morgan, 1998 in Zhang and Pineda, 2008; Esarey and Chirillo, 2013). Commonly, corruption is defined as the abuse of public authority for private gain (IBA, 2016; Esarey and Schwindt-Bayer, 2018; Transparency International, 2023). Nevertheless, in the context of trafficking in persons, Zhang and Pineda (2008, p. 45) offer to understand corruption as “public officials seeking financial reward by either looking the other way or facilitating human trafficking activities.” Corruption opportunities exist in various public official positions and sectors of government. It has been found that corrupt officials dealing with human trafficking usually work at the border control, diplomatic corps, immigration services, local law enforcement agencies, justice, security, and armed forces (Council of Europe, 2002; Zhang and Pineda, 2008). However, according to the UNODC survey from 2009, the law enforcement sector is the most engaged in trafficking-related corruption (UNODC, 2011 in IBA, 2016). Thus, police, customs, border control, and immigration officers should be subjected to scrutiny (Ibid). Hence, trafficking cannot operate without the participation of corrupt officials (PACO, 2002; Zhang and Pineda, 2008). Such results are based on both quantitative analysis and qualitative research, suggesting that public sector corruption maintains and encourages the crime of human trafficking and sexual exploitation. Human trafficking-related corruption can be divided into three types depending on public officials’ role in crimes (IBA, 2016). Firstly, public servants may act as traffickers, directly 15 participating in the organisation of forced prostitution, commercial sexual exploitation of children, or forced labour. IBA (2016) reports major cases in Burundi, the USA, India, and other countries in Eastern and Central Europe and the Middle East, where public officials were alone or in collaboration with trafficking rings responsible for the sexual exploitation of women and children. As a terrifying example, in 2015, a human rights organisation discovered the sex trafficking of children from Ukrainian state-run orphanages (Disability Rights International, 2015 in IBA, 2016). Secondly, public officials can facilitate trafficking by recruiting and transporting victims, as well as by arranging illicit departures and entries and falsification of documents (IBA, 2016). Lastly, corrupt public servants can contribute to the impunity of trafficking criminals by hindering traffickers’ detection, investigation, prosecution, and sentencing (Ibid, p. 29). Law enforcement officers worldwide are heavily bribed to protect lucrative trafficking businesses and turn a blind eye toward criminal activity. While most of the abovementioned scholars conducted quantitative analyses to generate support for the negative relationship between corruption and trafficking, several qualitative works focus on country case studies. Thus, Ariadne et al. (2021) have taken a closer look at widespread corruption among government officials in Indonesian provinces. In this origin and destination state for trafficking, women and children account for most of the victims in forced labour and sexual exploitation, which results from and is fuelled by corruption and poverty (Ibid). Moreover, the authors state that government corruption hinders existing anti-trafficking policies from being effectively executed. Accordingly, Guth (2010, p. 148) states that about three million of the Philippine population are “at high risk of being trafficked at any given time” due to corruption at the municipal level, which limits citizens’ access to economic opportunity and legal equality, leading to migration and, thus, fostering trafficking in persons. A similar situation is developing in Nigeria (Agbu, 2003). The crucial role of corruption in facilitating human trafficking was also recognised internationally, and numerous developed countries include recommendations for combatting corruption leading to labour and sexual exploitation. Thus, The U.S. Department of State, in its annual TIP Report (2022), draws attention to this issue and emphasises the impunity of corrupt officials and traffickers reign supreme, who fearlessly encourage and maintain trafficking. In line with this, the International Bar Association (IBA) (2016) dedicated an immense report to human trafficking and public corruption, stating that corruption contributes 16 to developing deep connections between traffickers and those in charge of prosecuting offenders. In addition, Ravlik (2019), in the context of the triangulated nexus of anti-trafficking enforcement, argues that the difficulties in executing anti-trafficking laws are significantly greater in countries where corruption is widespread. 2.4.1. Corruption and Sexual Exploitation Scholars started to investigate the relationship between corruption and sexual exploitation quite recently. Empirically, this research faces challenges in capturing both the extent of corruption and sexual exploitation, as both activities are illegal in nature. Nonetheless, Jonsson (2018) brings a serious contribution to the field by confirming the relationship between law enforcement corruption and sex trafficking activity globally, arguing that countries with higher levels of police corruption have greater trafficking outflow. Thus, corrupted police fuel and maintain the sex trafficking of women and girls, receiving profits in the form of bribes and thus, facilitating impunity of criminals. While Jonsson (2018) provides important insights into the relationship between corruption and sex trafficking, the study suffers from the empirical limitation imposed by using data on reporting of all human trafficking cases from secondary sources as a proxy for sex trafficking. In line with this, Díaz Rivillas and Solano López (2020) discuss a dire situation in Latin America where corruption disproportionately affects women and girls. The authors conducted interviews with victims of sexual exploitation, heads of Trafficking and Anti-Corruption prosecutors’ offices, and coordinators of the National Programme for the Rescue and Support of Victims of Trafficking in Argentina. They found that public sector corruption is strongly related to the sexual exploitation and trafficking of women and girls. Corruption thrives, fuelling trafficking throughout the criminal chain up to justice and victim protection systems (Ibid). Another finding of the study in Latin America confirms Jonsson’s (2018) theory and shows that law enforcement corruption has the strongest effect on sex exploitation, and trafficking rings are mainly supported by local police and municipal inspectors. The interviews’ results reflect three main types of trafficking-related corruption according to IBA (2016) described above, where the most intimidating is the impunity of criminals and insecurity of victims unable to rely on the police and judicial system. 17 Recent research also shows that corruption is often gendered as sexual exploitation maintains and fuels corrupt networks (Eldén et al., 2020; Sundström and Wängnerud, 2021; Bicker Caarten et al., 2022). As part of the ring, public officials often use sexual extortion2 as a currency. In such situations, they use power disparity and victims’ vulnerability to require sex acts as a bribe. A study by Bicker Caarten et al. (2022) shows that young female migrants are more vulnerable to sextortion and, thus, might become further involved in sex trafficking (Jonsson, 2018). In line with this, Díaz Rivillas and Solano López (2020) state that traffickers often use sexual extortion to bribe the authorities. Thus, women and girls, usually poorly educated and coming from rural areas with no documents and sufficient finances, are the most vulnerable to sextortion and, as a consequence, potential sex trafficking (Feigenblatt, 2020; Hendry, 2021; Bicker Caarten et al., 2022). Therefore, it is important to consider that the prevalence of monetary corruption leads to higher sextortion, another form of corrupt activities (Aja-Eke et al., 2023). The combination of both can potentially have a stronger effect on sex trafficking of women and girls worldwide. Hence, it is crucial that the whole concept of corruption should be reshaped in future studies considering its gendered nature. The findings reviewed above raise the question of whether the sexual exploitation of women and girls worldwide fuelled by corruption can be changed by achieving higher levels of gender equality in civil society and politics. Thus, the following section explores the existing literature regarding the effects of women’s political empowerment on corruption and violence against women. 2.5. Women’s Political Empowerment, Corruption, and Violence Against Women In the early stages of trafficking research, Balos (2004) and Ekberg (2004) mentioned that gender inequality facilitates sex trafficking, as traffickers often use female vulnerability. However, since then, many authors have highlighted the research gap in the relationship between gender discrimination and human trafficking (Uddin, 2014; Cho, 2015a). In her book, Ravlik (2019) recently identified global drivers influencing anti-trafficking enforcement, where some attention is dedicated to gender equality as a factor related to the success of anti- 2 Sextortion lies at the intersection of corruption and gender-based violence (Eldén et al., 2020). According to Eldén et al. (2020, p. 9), this form of corruption “occurs when a person with entrusted authority abuses this authority to obtain a sexual favour in exchange for a service or benefit which is within their power to grant or withhold.” 18 trafficking legislation. Indeed, widespread gender disparity results in the creation of discriminatory policies, which disempower women (Ibid). For a long time, no clear definition and measurement of WPE existed, and scholars focused on a broader concept of women’s empowerment, under which women take ownership of their own lives and gain the ability to make strategic choices (EIGE, 2016). Recently, greater attention has been paid to the concept of women’s political empowerment in the context of women’s access to political power and influence over it (Malhotra et al., 2002; Alexander et al., 2016). Alexander et al. (2016, p. 433) highlight the importance of separating the notion of women’s political empowerment and define it as “the enhancement of assets, capabilities, and achievements of women to gain equality to men in influencing and exercising political authority worldwide.” Women can be politically empowered in different ways: formally and informally. Thus, holding positions as politicians in legislative bodies defines formal WPE, while political empowerment of civil society actors and citizens is connected to informal WPE (Ibid). Women’s political empowerment statistics are typically confined to their representation in parliament and overlook other significant empowerment areas. However, the Varieties of Democracy (V-Dem) project offers a comprehensive measure of WPE, the Women’s Political Empowerment Index (WPEI), which is based on women’s civil liberties, their participation in civil society organisations as well as the descriptive representation of women in the political office (Sundström et al., 2017; Coppedge et al., 2022, p. 302). Despite the development of WPEI as an advancement of measurements of WPE, few studies in human trafficking so far have employed it as a measurement (Ravlik, 2019), and none of these studies specifically investigated a potential moderating effect of WPE on the previously introduced relationship between corruption and sexual exploitation of women and girls. 2.5.1. Women’s Political Empowerment and Corruption A large body of research exists on the relationship between women’s political empowerment and corruption. Many scholars agree that female disempowerment and gender inequality foster corruption (Alexander and Bågenholm, 2018; Esarey and Schwindt-Bayer, 2018; Ottervik and Su, 2022). This corpus of literature is based on the premise that empowering women in government will mitigate corruption since women as agents are less prone to act corruptly (Ibid). However, it should be emphasised that some researchers suggest reverse causality when 19 corrupt political systems limit women’s political participation (Esarey and Schwindt-Bayer, 2018). Looking closely at formal WPE, there is a strong essentialist argument by Dollar et al. (2001) and Swamy et al. (2001) that higher female participation in the political office leads to lower corruption, in particular, due to women’s honesty and trustworthiness. Alexander (2021) expanded and enhanced the explanation of why women as agents confront corruption head-on. The author’s first assumption stems from the social role theory of gender (Eagly and Crowley, 1986): women are less tolerant of corrupt conduct due to their socialisation to act more ethically (Alexander, 2021). Women, in addition to being the “fairer” sex, tend to abstain from the risks connected with engaging in corruption in a primarily non-corrupt public environment. Second, when it comes to positions of authority, women are a marginalised group, making it difficult for them to enter patronage networks and therefore engage in power abuse (Ibid). Third, Alexander (2021) employs the “differential treatment theory of gender and corruption” by Schwindt-Bayer, Esarey, and Schumacher (2018, p. 60). Women’s involvement in corruption may result in harsher punishment from society because it is widely assumed that they are more trustworthy than men. Therefore, having “higher standards” to comply with prevents women from acting unethically (Alexander, 2021). At last, women rely on a functional state to supply essential public goods, strengthening their autonomy and, as a result, their presence in government (Ibid). Hence, it is not in the best interests of women to undermine good governance by aggravating corruption. Moreover, Alexander and Ravlik (2015) emphasise the vital role of “the women’s interest mechanism.”3 It suggests that women as agents have the ability to combat corruption indirectly. They encourage and advance policies for women and girls, which improves public service delivery and system transparency, resulting in a reduction of corruption (Alexander and Bågenholm, 2018). Confirming conclusions of numerous qualitative studies from the past, a recent quantitative analysis by Goel and Nelson (2023) shows that through their participation in parliaments, women in the legislative bodies indeed put pressure on corruption to decline. 3 According to Alexander and Ravlik (2015, p. 1), the women’s interest mechanism is “representation of more specific female interests to broader quality of government outcomes.” 20 2.5.2. Women’s Political Empowerment and Violence Against Women and Girls As stated in the section above, female leaders advocate for women-friendly policies that improve public service delivery and system accountability (Bågenholm and Alexander, 2018). In line with this, Meyer (2003), Gold and Haynie (2014), and Alexander and Ravlik (2015) state that the women’s interest mechanism includes but is not limited to health, violence against women and girls, employment discrimination, and sexual harassment because such policies affect female autonomy. Alexander and Ravlik (2015) argue that anti-trafficking enforcement is crucial for women’s autonomy protection when it comes to violence against women and girls, and thus, this policy should be considered among other women’s interests. Consequently, the question is whether women in political office successfully address the VAWG issue. In the framework of theory on women’s substantive representation4 (Pitkin, 1967), numerous scholars argue that female descriptive representation positively affects the substantive representation of women’s issues (Lovenduski and Norris, 2003; Childs and Krook, 2009; Wängnerud, 2009 in Alexander and Ravlik, 2015). Issues that more genuinely represent the concerns of women are more likely to be brought up, promoted, and supported by female politicians, in particular, due to shared experiences with gender discrimination (Wängnerud, 2009). Thus, based on the arguments presented above, Alexander and Ravlik (2015) state and confirm the hypothesis that a higher percentage of women in parliament leads to better anti-trafficking enforcement. However, in the case of violence against women and girls, not only a descriptive political representation of women but also other components of WPE play a crucial role in combatting gender inequality. Women’s civil movements have dedicated decades to campaigning, programming, and raising awareness on different levels to battle violence against women and girls (Michau et al., 2015). Numerous civil society initiatives have been introduced to help victims and survivors of violence with shelter, support, and legal counselling. Htun and Weldon (2012) presented breakthrough research at the time, establishing that women’s mobilisation in civil society exclusively affects the development of policies on violence against women via the incorporation of feminist concepts into international norms. Feminist autonomous movements are the main drivers of change because they formulate marginalised social group viewpoints, spread new ideas to the larger public, influence public opinion, and demand meaningful 4 Substantive representation as “acting in the interests of the represented in a manner responsive to them” (Pitkin, 1967, p. 209). 21 institutional reforms, progressive social policies5 and overall transformation of social practice (Ibid). Besides this cross-country research which covered 70 states, other scholars address the issue on a national level. For instance, Sinha et al. (2017) describe how Indian women’s movements initiated anti-rape, domestic abuse, and sexism initiatives aiming to change the robust patriarchal structure and cultural norms, fuelling action against VAWG in the country. A study by García-Moreno et al. (2015) states that women’s civil movements cause progressive government action on VAWG issues worldwide, which provides prospects for its abolition. The authors highlight that prevention policies, as well as education and empowerment of women, should be at the forefront of the battle. In line with this, Bertone (2004, cited in Alexander and Ravlik, 2015) argued that advocacy by transnational feminists over violence against women is strongly interconnected with raising international activity to fight the sex trafficking of women and girls. Thus, the conclusion is that the civil society movement component of WPE plays a crucial role, and the expectation is that it will affect sex trafficking prevalence and anti-trafficking laws and enforcement, which fall under the definition of women’s interests. 3. Theoretical Framework, the Gap, and Hypotheses 3.1. Theory and the Gap In summary, the literature provides compelling evidence to suggest that corruption should have a crucial effect on sex trafficking as well as on the adoption and enforcement of anti-sex trafficking laws across countries. Corruption fuels and maintains sexual exploitation. Scholars observe that women and girls are being exploited for the purposes of organised crime, which is fundamentally connected to corruption and how corruption is gendered in an exploitative way. Thus, based on previous research on human trafficking and corruption, as well as literature on women’s political empowerment and how it affects corruption and VAWG policies, I build a new theoretical argument. This thesis argues that corruption facilitates widespread sexual exploitation and lack of proper anti-sex trafficking legislation and law 5 Progressive social policies are defined by their explicit goal of improving the position of historically marginalised and excluded groups (Htun and Weldon, 2012, p 552). Such policies seek to reform and enhance society to promote peace, justice, and equality (Ibid). 22 enforcement. It further theorises that this relationship is moderated by whether or not women are politically empowered. It is important to note that by theorising and further testing the effects mentioned above, this thesis aims to address the existing gap in the literature. Despite the fact that several studies tested the relationship between corruption and human trafficking quantitatively, a clear effect of corruption on the sex trafficking of women and girls has not been analysed so far. Jonsson (2018) made a successful analysis of the effect of police corruption on sex trafficking. However, the author relied on data concerned reporting of all human trafficking cases as a proxy of sex trafficking in a relatively small sample of countries. This thesis instead focuses specifically on sex trafficking, testing the connection between corruption and sex trafficking more directly. In addition, a possible moderation effect of women’s political empowerment on the relationship between corruption and sex trafficking has never been assessed. Given the interdependence between WPE and corruption and sex trafficking separately, it is important to investigate the relationship between corruption and WPE on sex trafficking. Thus, this thesis has the ambition to be the first to test this configuration of relationships by conducting a quantitative analysis. Therefore, this thesis aims to contribute to existent literature by first examining the direct effect of corruption on sex trafficking and then developing and testing arguments on the moderating effect of WPE on this relationship. Thus, this thesis seeks to address the following research question: “How does women’s political empowerment moderate the relationship between corruption and anti-sex trafficking legislation and enforcement across countries?” 3.2. The Effect of Corruption on Anti-Sex Trafficking Legislation and Enforcement Based on the literature review, scholars strongly agree that corruption fuels and maintains human trafficking (Hughes and Denisova, 2001; Bales, 2007; Zhang and Pineda, 2008; Studnicka, 2010; Cho, 2015a; Ravlik, 2019). Díaz Rivillas and Solano López (2020) and Jonsson (2018) further argue that it also affects sexual exploitation and sex trafficking of women and girls. Both scholars and international organisations state that public servants use and reinforce females’ marginalisation and vulnerability by profiting from trafficking and receiving monetary bribes and sexual favours. Corrupted officials act as traffickers themselves, facilitate trafficking by recruiting and transporting victims, and contribute to the impunity of 23 trafficking criminals by impeding their detection and prosecution (IBA, 2016). Moreover, Ravlik (2019) presents evidence that corruption not only results in higher levels of trafficking but also hinders anti-trafficking enforcement across countries. Thus, taking into account that 72% of detected human trafficking victims are women and girls (UNODC, 2018 in CATW, 2023), it seems reasonable to expect that corruption will lead to higher sex trafficking and worse anti-sex trafficking legislation and enforcement across countries6. Therefore, the discussion based on previous literature results in the following hypothesis: H1: Higher levels of corruption are associated with lower levels of adoption and enforcement of anti-sex trafficking laws across countries. 3.3. The Effect of Corruption on Anti-Sex Trafficking Legislation and Enforcement Moderated by Women’s Political Empowerment Based on the arguments outlined above, there are strong reasons to assume a negative association between high levels of corruption and low levels of anti-sex trafficking legislation and enforcement. However, limited research has been done to determine under which contexts this relationship weakens or strengthens. First, a large number of studies have been dedicated to gender equality and corruption. Scholars state that higher female participation in political office leads to lower corruption due to women’s honesty, trustworthiness, and overall socialisation to act more ethically (Dollar et al., 2001; Swamy et al., 2001; Alexander, 2021). Empowered women, through their agency, also indirectly challenge corruption. They promote female-friendly policies strengthening their autonomy, which positively affect the political system’s transparency and accountability, resulting in lower levels of corruption (Alexander and Bågenholm, 2018). Second, the women’s interest mechanism plays a crucial role in combatting violence against women and girls. Female leaders, both in politics and civil society organisations, focus on protecting the rights and autonomy of women and girls as one of the most marginalised groups. Higher descriptive representation positively affects the representation of women’s issues 6 In this thesis, I focus on anti-sex trafficking laws and enforcement, as data on the number of sex trafficking victims is scarce and often assumed to be underreporting the actual number of victims of sex trafficking. A detailed review of data available and measurements is given in the Research Design section. 24 (Lovenduski and Norris, 2003; Childs and Krook, 2009; Wängnerud, 2009). In line with this, Alexander and Ravlik (2015) state that a higher percentage of women in parliament leads to better anti-trafficking enforcement. Furthermore, a cross-country analysis by Htun and Weldon (2012), along with other case studies, confirms that higher women’s mobilisation in civil society is positively associated with the development of policies on VAWG. Supporting the research presented above, this thesis theorises that when WPE is high, women increase their influence in both formal and informal politics, improve the system’s transparency and address the VAWG issue, thereby weakening the negative association between public sector corruption and the adoption and enforcement of anti-sex trafficking laws. Therefore, based on the advantages WPE has, there is evidence suggesting that the negative effect of corruption on the adoption and enforcement of anti-sex trafficking laws will be weaker under contexts of higher WPE. This leads to the following hypothesis: H2: Under higher levels of women’s political empowerment, the negative relationship between corruption and the adoption and enforcement of anti-sex trafficking laws across countries becomes weaker. 25 Figure 1. The moderating effect of Women’s Political Empowerment on Corruption’s effect on Adoption and Enforcement of Anti-Sex trafficking laws. Reinforces women’s marginalisation and vulnerability Adoption and Facilitates and maintains sex Enforcement of Corruption trafficking, sexual exploitation and extortion of women and girls Anti-Sex trafficking laws Weakens state criminal justice system Women’s Interest Mechanism Improve accountability and transparency of the system Develop women-friendly policies, including VAWG Mobilise civil society to advocate for marginalised groups’ interests, including VAWG Women’s Political Empowerment The model above reflects the presented theory, showing macro and micro relationships between the key concepts. However, it should be highlighted that it has been created for better visualisation of the theoretical background. Only the main variables in the bold, in shaded rectangles, will be formalised and tested in the research design. Information in other rectangles is provided to explain the expectations of how corruption affects the adoption and enforcement of anti-sex trafficking laws and why women’s political empowerment moderates this relationship. 4. Research Design In this section, I present the research design this thesis employs by discussing the underlying data and methods in detail. To test the above-developed hypotheses, I employ data on anti-sex trafficking laws and enforcement, corruption, women’s political empowerment, and additional 26 control factors for a sample of 131 countries7. Data was provided by the Quality of Government (QoG) cross-sectional and time-series dataset and the WomanStats Project cross-sectional dataset. In terms of the scope of the data, this study will use this set of countries based solely on data availability after merging the above-mentioned datasets. As no panel data is available for sex trafficking, the analysis is cross-sectional, using 2019 as the most recent year of data available for the measurement of sex trafficking. An explanation of the operationalisation of the variables will be provided below. In addition, it should be highlighted in advance that all independent variables, except for two controls, will be measured as lags. This choice will be elaborated on in the Statistical Method’s section. I test the hypotheses through a large-N cross-sectional study employing Ordinary Least Squares (OLS) method with interactive specifications. The method is chosen to test the effect of corruption on the adoption and enforcement of anti-sex trafficking laws moderated by levels of women’s political empowerment through an interaction, where corruption and WPE are multiplied to create a new independent variable (Gerring and Christenson, 2017). This is done to observe whether the strength of the relationship between corruption and the adoption and enforcement of anti-sex trafficking laws changes across levels of WPE. The choice and specifics of the statistical method will be discussed in detail. 4.1. Data and Dataset This study will use data from two primary data sources merged into a single dataset to test the hypotheses. Both sources, the Quality of Government cross-sectional and time-series dataset and the WomanStats Project dataset, are highly credible, well-known and widely used in the academic field (Teorell et al., 2022; WomanStats Project, 2023). This section offers a discussion on data availability and introduces the reasoning for choosing particular measurements as an operationalisation of the dependent and key independent variables as well as explanatory factors. It starts with a discussion of the main concepts: sex trafficking, corruption and women’s political empowerment. The section continues with a discussion on the decision to include each of the control variables and their measurement. 4.1.1. Dependent Variable: Sex Trafficking Data and Operationalisation Data 7 Given data availability, 131 countries is the resulting sample for which data on all relevant variables was available. 27 The absence of high-quality data remains a problem for comparative research on human trafficking, notwithstanding recent advancements. There is scarce information available regarding the number of people that work for criminal organisations engaged in such trafficking operations (Di Nicola, 2007 in Jonsson, 2018). Moreover, victims of human trafficking are considered to be “hidden populations” in regards to which researchers do not have data on the size and borders (Jonsson, 2018). Thus, estimating the number of trafficked victims represents a difficult task with a risk of either understatement or exaggeration. Previous studies mostly used the country classification data from the UNODC’s report on TIP (2006), which provides information on the trafficking reporting, avoiding tackling actual numbers of trafficked victims (Cho et al., 2013; Jakobsson and Kotsadam, 2013; Cho, 2015a; Masakure, 2017; Jonsson, 2018). However, this dataset on human trafficking has serious limitations and does not suit the purposes of this thesis. Given that the measurement is based on data from 2006, it can be considered outdated for conducting analysis in 2023, and, most importantly, it is focused on the prevalence of human trafficking, a measurement for which reliability can be questioned for the reasons stated above. In addition, this dataset incorporates all forms of human trafficking and does not focus solely on sexual exploitation of women and girls. Thus, based on existing limitations of data availability, this thesis turns to data on anti- trafficking legislation and enforcement. Currently, there are four different global measurements of anti-trafficking enforcement (Ravlik, 2019). The most frequently used and considered reliable is Tier Rankings reported by the US Department of State (Ibid; Lee, 2005; Wooditch, 2011; Van Dijk and Klerx-Van Mierlo, 2014; Hernandez and Rudolph, 2015). Each country is ranked among one of three tiers according to its compliance with the international standards of the Victims of Trafficking and Violence Protection Act (TVPA), such as laws recognising human trafficking as a crime, law enforcement, prosecution, victim protection, prevention policies (H.R.3244 - 106th Congress, 2000). The second measure, 3P Anti-trafficking Policy Index, developed by Cho (2015b), examines government anti-trafficking efforts in three major policy dimensions: prosecution of human trafficking criminals, prevention of human trafficking crime, and protection of human trafficking victims (Ravlik, 2019). The third measurement, the Government response to modern slavery index by the Walk Free Foundation, based on almost a hundred various indicators, is often criticised due to its non-transparent nature (Ibid). Unfortunately, all the measurements presented above incorporate both forced labour and forced sexual exploitation, making them unsuitable for this thesis’s analysis. 28 Operationalisation of the Concept While all of the anti-trafficking enforcement measurements above have benefits and drawbacks, they cover human trafficking as a whole phenomenon. Thus, after considering all available options in the field, it has been decided to use the trafficking scale developed by Hudson et al. (2012) for this thesis, which is based on anti-sex trafficking practices, laws and enforcement and incorporated into the WomanStats Database. Even though, like some previous measures, it focuses on policies, the fundamental benefit of this trafficking scale is that it is based on a unique and original dataset meant to represent the global situation of women (Ravlik, 2019). It is devoted exclusively to the sex trafficking of women and girls and provides a comprehensive and multifaceted assessment, incorporating data from various factors linked to female sexual exploitation (WomanStats, 2023). Although it was coded in four-year intervals since 2007, only a few studies have used the trafficking scale for quantitative research (Stearmer et al., 2007; Hudson et al., 2011), and none of them focused on the effect of corruption on anti-sex trafficking enforcement. This metric is not as widely used as the other global indices, but it has significant potential owing to its geographic scope (Ravlik, 2019) and specific focus on sex trafficking. Overall, the WomanStats Project is well-known globally as its unique data collection allows scholars to address various research questions concerning women. The data is coded into over 350 variables available for 176 countries, including laws, statistics, and practices (WomanStats, 2023). The project highlights the reliability of information gathered only from reputable sources such as governments, non-governmental and intergovernmental organisations, UN reports, CEDAW, country experts, and others, with a total of over 500 sources concerning the status of women (Ibid; Caprioli et al., 2009). However, Ravlik (2019) underlines one potential flaw of the trafficking scale - the lack of transparency in expert evaluations, the same issue as in the Tier Rankings. Considering this possible limitation, Hudson’s Trafficking Scale is still the most fitting to answer the thesis research question as it has an undeniable advantage: it focuses exclusively on the sex trafficking of women. In the WomanStats Database, indicators concerning women’s physical security and violence against women, in particular, are combined into a Trafficking Cluster. There are seven variables, which serve as sub-scales for Hudson’s Scale of Trafficking variable and measure different aspects of trafficking in women and girls, focusing on law, practice, enforcement, and Tier ranking (WomanStats, 2023). However, one sub-scale represents the prevalence of sex trafficking (Ibid). The project states that the sex trafficking scale is an “examination of the legal 29 framework the country has to combat sex trafficking, the enforcement of those laws, and the success of that enforcement in curbing trafficking” (Ibid). The variable has an ordinal scale with five scale points ranging from 0 to 4, where 0 reflects the best situation in a country and 4 - the worst. However, I recoded the variable to the following scale, such that higher values would represent more extensive measures against sex trafficking and more successful implementation (see the original scale in Appendix 1a): 1 = No anti-sex trafficking laws, non-compliance with the Trafficking Victims Protection Act of 2000 (TVPA) (Tier 3 ranking), no victim support, the government can facilitate trafficking; 2 = Limited anti-sex trafficking laws, non-compliance with the TVPA (Tier 2 ranking), limited initiative to comply; 3 = Anti-sex trafficking laws but not consistently enforced, non-compliance with the TVPA, but efforts are made to comply (Tier 2 ranking); 4 = Anti-sex trafficking laws, laws are enforced, but enforcement becomes weaker or considerable sex trafficking goes unnoticed by the authorities, full compliance with the TVPA (Tier 1 ranking); 5 = Anti-sex trafficking laws, laws are enforced, full compliance with the TVPA (Tier 1 ranking), sex trafficking is rare (WomanStats Project, 2023). It should be highlighted that the years for which data is available are limited (2007, 2011, 2015, 2019), which makes it challenging to conduct a time-series analysis. Thus, in this thesis, I use the variable coded for 2019 to conduct a cross-country study, given that this data offers a snapshot of the most recent time point. However, I acknowledge that it can be a possible limitation of the study, not controlling how previous rankings affect the rankings in 2019. Nevertheless, striving for a parsimonious model, I do not use data from other years. In Figure 2 below, the global distribution of the variable is presented for the year 2019. 30 Figure 2. Map based on Hudson’s Scale of Trafficking variable in 2019 (WomanStats Project, 2023). 4.1.2. Main Explanatory Variable: Corruption Data and Operationalisation As it has been discussed in the literature review section, corruption as a concept is hard to define and quantify due to its informal and illegal nature. However, there are several measures used by scholars (Kaufmann et al., 2007). Existing indicators often reflect corruption perception and can include various forms of misconduct as well as focus on a particular corruption type. The most commonly used measures are Transparency International’s (TI) Corruption Perceptions Index (CPI), the Bayesian Corruption Index (BCI), the World Bank’s Control of Corruption governance indicator and the V-Dem political corruption index. In this thesis, I chose to rely on the CPI, which is suitable for the analysis for several reasons. First and foremost, CPI reflects the theoretical expectations from the literature review well and matches the theory presented above, where public sector corruption is expected to negatively affect the adoption and enforcement of anti-sex trafficking laws across countries. Secondly, CPI is the world’s most extensively used global corruption assessment (TI, 2023), which is meticulously constructed and based on 13 highly reputable data sources, including the World Bank and the V-Dem Institute. It offers the most comprehensive overview of public sector 31 corruption and is regularly checked for robustness by the European Commission’s Joint Research Centre (Ibid). Finally, CPI has no distinction between administrative and political corruption (Coppedge et al., 2022), which is in line with other studies on human trafficking and corruption, showing that corruption in various aspects of the public sector can affect trafficking in persons8. The index is also suitable as it was used previously in the field of human trafficking by, e.g., Cho (2015a) and Zhang and Pineda (2008), which allows for easier comparison with previous results. Furthermore, CPI is expert-perception-based, meaning that experts and businesspeople answer survey questions about how high they believe corruption to be. It might be considered a more reliable measurement compared to other corruption indicators based on individual-level corruption experiences and perceptions that rely on surveys of individual in a country. However, it also might have potential limitations compared to more “objective” measurements, such as procurement data or conviction rates, data on which is still comparably rare and not available on a global scale. The CPI scale ranges between 0, highly corrupt, and 100, highly clean (Theorell et al., 2022). However, I standardise CPI to the scale of 0 to 1 to match the scale of the moderating variable, and name the variable TI Corruption in the statistical models. Some scholars also suggest standardising variables in terms of directionality. In this thesis, I reverse CPI to allow for easier testing of the formulated hypotheses and to make the interpretation more straightforward. In the resulting coding, 0 represents highly clean countries, and 1 represents highly corrupt ones. In addition, the CPI will be lagged back by five years, along with other independent variables. This choice will be explained in the method’s section. In order to test the robustness of results, CPI will be replaced by the BCI and the V-Dem political corruption index. It is important to note that the thesis does not aim to identify which type of misconduct among public sector corruption is most strongly associated with sex trafficking but rather chooses a measurement that might capture all possible types of corruption across the governmental realm. It is one of the possible limitations of this study and an 8 Indeed, of the various measures of corruption, CPI is the strongest correlate of the sex trafficking scale, as expected given the better alignment of its operationalisation of corruption with the theoretical descriptions of the actual types of corruption fuelling trafficking (see Correlation Table in Appendix 3). 32 opportunity for future research to investigate what type of corruption is especially conducive to sex trafficking. CPI, BCI, and the V-Dem political corruption index will be retrieved from the QoG Standard Dataset (Theorell et al., 2022). 4.1.3. Moderating Variable: Women’s Political Empowerment Data and Operationalisation Back in 2002, Malhotra et al. (2002) brought up the issue of measuring women’s empowerment as a factor of international development. Over time, other authors discussed the need for a new indicator, capturing WPE as an advancement of the concept. Thus, Alexander et al. (2016) were the first to define WPE as a more sophisticated concept than women’s representation in parliament and discuss the need for appropriate measurement. Recently, Sundström et al. (2017) from the V-Dem Institute presented a new measurement, the Women’s Political Empowerment Index (WPEI), corresponding to the requirements of scholars. In contrast to previous measurements that relied on singular indicators, it is aggregated as an average of three indices: women’s civil liberties, women’s civil society participation, and women’s political participation (Ibid). Indeed, WPE is seen as “a process of increasing capacity for women, leading to greater choice, agency, and participation in societal decision-making” (Coppedge et al., 2022, p. 302). Conveniently, WPEI is also available in the QoG Standard Dataset (Teorell et al., 2022) and will be retrieved from there. It is coded on an interval scale from 0 to 1, where 0 represents low WPE and 1 - high WPE in a country (Ibid). In addition, as it was briefly mentioned above, the WPEI will be lagged back by five years. This choice will be elaborated on in the method’s section. WPEI will be used as moderating variable based on the developed theoretical model. As the literature review has shown, there are strong arguments demonstrating the substantial influence WPE has on public sector corruption. Moreover, this operationalisation of WPE allows to expect its effect on sex trafficking and anti-sex trafficking legislation and enforcement, taking into account both formal and informal empowerment discussed before, as well as the civil liberties women have in societies. The WPEI measurement represents the most accurate situation on a country level showing how strongly females are empowered politically and in civil society across countries. Each of the indices has its weight, crucial for conducting the analysis. Women’s political participation index reflects female descriptive representation in the legislature and equality of total power 33 allocation (Coppedge et al., 2022, p. 303). In accordance with the literature, it should contribute to curbing corruption and improving women-friendly policies. Likewise, the Women’s civil society participation index should have an effect on the advancement of VAWG policies by letting women have a strong and free say in political debate, participate in civil society organisations and bring up VAWG issues in media (Ibid). Lastly, it has been decided to keep the Women’s civil liberties index, measuring domestic movement freedom, private property rights, freedom from forced labour, and access to justice (Ibid). Even though it has not been directly covered in the theory part, having civil liberties allow women to act proactive in both formal and informal politics and advocate for better policies, having confidence that their rights are respected and protected on the governmental level. At the same time, having civil liberties potentially can lead not only to better policies but also to lower sex trafficking flows (Cho, 2015a). 4.1.4. Control Variables Other potential interfering factors, called confounding or control variables, affecting the dependent and key independent variables must be considered. Consequently, it will be feasible to offer an interpretation of the results that is not influenced by those control variables. Chosen controls will help to avoid omitted variable bias and, being held at a constant, will isolate the effect of public sector corruption on anti-sex trafficking legislation and enforcement. As a result, the accuracy of the final model will be improved. All confounding variables, except ratification of CEDAW, are retrieved from the time-series QoG Standard Dataset (Teorell et al., 2022), where these variables will be lagged back by five years. Ratification of the CEDAW variable is retrieved from the WomanStats Dataset, where the only years the data is available are 2015 and 2022. Thus, for the analysis, I use 2015, the only year available before 2019, when extensiveness and implementation of sex-trafficking laws are measured. Two controls, Share of protestants in 1980 and Region, will not be lagged. The methods section below will explain the choice of lagging the variables. Economic Controls GDP Per Capita Human trafficking research generally asserts that socioeconomic status substantially impacts the outflow of trafficked people. Being an individual-level push and pull factor, low-income levels and poverty force potential victims out of their home countries, rendering them more susceptible to trafficking (Salt, 2000; Rao and Presenti, 2012; Cho, 2015a; Jonsson, 2018). 34 Victims of human trafficking usually are migrants who come from less advanced and lower- income nations and look for better economic opportunities in more affluent and developed parts of the world (Kangaspunta, 2003; Rao and Presenti, 2012; Cho, 2015a). Thus, given arguments in the literature that economic development is a major factor in trafficking outflows, I control for GDP per capita. Based on the literature, states with lower GDP per capita are expected to have larger trafficking outflows (Ibid). However, since my dependent variable is a complex trafficking scale measure more focused on law and enforcement, I expect that higher GDP per capita will lead to better anti-sex trafficking laws and their enforcement across countries. This assumption is based on the fact that wealthier countries are able to direct more resources to enhance law enforcement practices (Ravlik, 2019). Moreover, there is a strong agreement in the literature that countries with higher GDP tend to have lower corruption levels (Esarey and Chirillo, 2013; Esarey and Schwindt-Bayer, 2018; Esarey and Valdes, 2021). In addition, economic progress is often associated with political liberties, which promotes more gender parity in government and society (Ibid), affecting women’s political empowerment, the moderating variable. Following convention in the field, I further log-transform it to GDPPC (log) as recommended by the above-mentioned scholars (Jonsson, 2018; Ravlik, 2019), as log transformation helps to improve the normal distribution of the data. Globalisation Globalisation, positively affecting counties’ openness, encourages cross-border mobility and migration of people (Cho, 2013). However, transnational criminals have benefited greatly from globalisation in their business operations (Shelley, 2010). In the origin countries, globalisation with greater information exposure facilitated migration and, thus, human trafficking (Cho, 2013). However, in destination countries, information flows discourages human trafficking by raising public knowledge of the problem (Ibid). In this thesis, I operationalise globalisation through the Index of Globalisation (Globalisation). Using a scale of 1 to 100, it comprises three segments: economic, social and political globalisation, where the economic part is attributed a stronger weight (Teorell et al., 2022, p. 211). The political component focuses on membership in international organisations and international treaties, the economic one reflects both trade and financial flows, and social globalisation covers many aspects, from travelling and migration to global internet connections and freedom of citizens (Ibid). Thus, looking at the whole set of components, I expect better anti-sex trafficking laws and enforcement with higher levels of globalisation. While migration might affect the numbers of sex trafficking in the origin countries, my dependent variable is still strongly focused on policy outcomes. Thus, with better 35 performance in all other components, such as participation in international treaties, civil freedom, and trade openness essential for economic growth, the literature suggests that anti- sex trafficking legislation will improve (Ravlik, 2019). Moreover, globalisation is expected to affect WPE, improving gender equality situation positively (Ibid). In addition, many scholars argue that globalisation affects corruption across countries. While some suggested a positive relationship, most scholars agree that globalisation reduces corruption as intrastate institutions become more transparent under the pressure of the international community (Sung and Chu, 2003; Akhter, 2004 in Das and DiRienzo, 2009, p. 34). Political Controls Democratic Institutionalisation Strong democratic institutions implicitly mean a larger variety of laws protecting human rights and more rigorous enforcement of those laws in a country. Thus, one might assume that more democratised countries have better regulation and enforcement of anti-sex trafficking laws. Therefore, following examples of other scholars (Alexander and Ravlik, 2015; Jonsson, 2018), I add control for levels of democracy to capture heterogeneity across groups of countries and minimise bias (Ibid). In addition, due to increased electoral accountability, corruption is anticipated to be lower in a more democratic environment (Kolstad and Wiig, 2016; Esarey and Schwindt-Bayer, 2018). Hence, I control for the V-Dem electoral democracy index (Democracy), reflecting the extent to which electoral democracy as an ideal is achieved (Teorell et al., 2022). It is coded on an interval scale, from low to high (0-1), and incorporates indices measuring freedom of association, clean elections, freedom of expression, elected officials, and suffrage (Ibid, pp. 621-622). In addition, it is important to highlight that democracies are more inclusive of women in politics and civil society organisations, which affects WPE (Bush and Zetterberg, 2021). Political Stability and Absence of Violence/Terrorism McAlpine, Hossain, and Zimmerman (2016) provide evidence that human trafficking and sexual exploitation are common in conflict-affected areas worldwide. In addition, women and girls are more vulnerable to sexual violations in crisis settings (Ibid). According to the literature, population displacement brought on by military conflict, political disturbance, and civil unrest makes individuals more susceptible to sexual and forced labour exploitation (Plümper and Neumayer, 2006; Akee et al., 2010; Shelley, 2010). Thus, the expectation is that countries with less political stability have higher human trafficking outflows and worse 36 enforcement of anti-sex trafficking laws. I operationalise the variable as Political Stability and Absence of Violence/Terrorism (Political Stability). It measures “perceptions of the likelihood of political instability and/or politically motivated violence, including terrorism” (Teorell et al., 2022, p. 626). This variable from the Worldwide Governance Indicators ranges from about -2.5 to 2.5, with higher values corresponding to better governance performance (Kaufmann et al., 2010). In addition, conflict and corruption can lead to and aggravate each other (Andvig, 2007; TI, 2019). High corruption can cause instability in peaceful countries, while conflict can foster the development of new types of corruption due to instability and resource shortage (Ibid). Ratification of CEDAW Being part of international agreements and treaties, as well as compliance with international laws, is crucial for enhanced anti-trafficking enforcement as it leads to commitment to adopting relevant legislation on a state level (Ravlik, 2019). CEDAW, known as the International Bill of Rights for Women, is one of the most important international treaties protecting women’s rights and fighting against their discrimination. However, even though all States Parties of CEDAW are legally bound to “fulfil, protect, and respect women’s human rights” (UN Women, 2010), many of them ratified the treaty with reservations or did not ratify Optional Protocol, thereby avoiding punishment for non-compliance. Thus, I operationalise ratification of CEDAW as a scale created by Professor Senem Ertan (CEDAW), ranking the degree to which countries have committed themselves to CEDAW. I retrieved the variable from WomanStats Database and recoded it more intuitively (see the original scale in Appendix 1b): 1 = Did not ratify CEDAW; 2 = Ratified CEDAW, but with reservations; 3 = Ratified CEDAW, no reservations, did not ratify Optional Protocol; 4 = Ratified CEDAW, no reservations, ratified Optional Protocol. Thus, the expectation is that the better country’s compliance with CEDAW, the better anti-sex trafficking laws and enforcement it will have. In addition, ratification of the Convention offers females instruments to advance their political empowerment, relying on international legislation (Vahdati, 2021). The variable will be operationalised for the year 2015, the only year available prior to 2019. 37 Societal controls Population Size Several scholars have stated that countries with higher population density are anticipated to have more traffickers and trafficking victims (Cho et al., 2013; Jakobsson and Kotsadam, 2013). Rao and Presenti (2012) further argue that such states are usually origin countries, from where the outflow of migrants and trafficking happens. In general, trafficking studies include population size variables in their set of controls to consider the size of a specific country and to remove any biases that may exist due to the fact that some countries have more people than others (Jonsson, 2018; Ravlik, 2019). Population density is an important control when working with country-level statistics, according to the literature (Babbie, 2013 in Ravlik, 2019). Thus, I expect more populous countries to have higher sex trafficking levels and worse legal regulations and enforcement of anti-sex trafficking laws. In addition, some scholars highlighted that corruption is higher in countries with larger population (Chong Soh and Amin, 2020). I operationalise this control as a total population variable (Population (log)), which includes residents of all nationalities and legal statuses (Teorell et al., 2022). Following convention in the field, I log-transform the variable to improve the normal distribution of the data. Religion Numerous studies include religion as a proxy for culture to account for the effect of the population’s beliefs and attitudes among different countries. Some scholars use the percentage of Muslims, arguing that conservative attitude leads to restrictions of women’s economic involvement and mobility, thus lowering sex trafficking (La Porta et al., 1999; Cho, 2015a; Jonsson, 2018). Nevertheless, I operationalise religion as Share of protestants in 1980, in line with another group of scholars, who state that Protestant-dominated nations favour gender equality and have higher numbers of women politicians (Welzel et al., 2002). Moreover, numerous studies have concluded that Protestantism restrains corruption (Arrunada, 2010; Chase, 2010; Portilla, 2022). Thus, I expect countries with a higher percentage of protestants to have better anti-sex trafficking laws and enforcement. The variable is retrieved from the QoG Dataset and shows protestants as a percentage of population in 1980. Thus, it will be used as a historical variable in 1980 because it is heritage originating from a protestant culture that is assumed to affect gender equality and WPE in particular, not the current number of protestants. Extra controls 38 Geographic Region Cho (2015a) and Hernandez and Rudeolph (2015) state that geographic location affects human trafficking flows in general. Thus, to account for the regional effect and reduce biases in the dependent variable’s data, I include a variable Region of the Country (Region) with ten categories. Politico-geographic regions, based on geographical proximity and intrastate democratisation, are the following: Eastern Europe and post-Soviet Union, Latin America, North Africa and the Middle East, Sub-Saharan Africa, Western Europe and North America, East Asia, South-East Asia, South Asia, The Pacific, The Caribbean (Teorell et al., 2022). 4.2. Method 4.2.1. Lagging of variables Since sex trafficking of women and girls is a multifaceted phenomenon, I have to consider the possibility that corruption might not have an immediate effect on the sex trafficking situation. Importantly, I operationalise the dependent variable as the trafficking scale developed by Hudson et al. (2012), based on anti-trafficking practices, laws and law enforcement. Thus, considering that the trafficking scale focuses on the policy aspect, not the number of victims, I expect corruption to influence anti-sex trafficking legislation and enforcement over time. In this case, it is hard to provide straightforward support for lagging corruption back when dealing with sex trafficking policies since such research is in its infancy. However, looking at corruption studies more broadly, numerous scholars researching relationships between corruption and economic development, public spending and foreign aid (Delavallade, 2006; Charron, 2015; Aïssaoui and Fabian, 2022) include lagged corruption as a key independent variable to account for the time lag effect, persistence of corruption over time as well as to avoid potential omitted variable bias. In addition, looking at the research question of this study, which highlights the possible moderation effect WPE can have on the relationship between corruption and the adoption and enforcement of anti-sex trafficking laws across countries, I have to consider the slow effect of WPE on both anti-sex trafficking policy outcomes and corruption. In the case of VAWG policies, Htun and Weldon (2012) discussed that women’s mobilisation throughout society has an effect on such policies after some time. Once women are politically and informally empowered, it pushes the gradual policy development (Ibid). In the case of corruption, Wängnerud (2009) argued that increased WPE slowly instigates progress in levels of 39 corruption among countries, showing a standout example of Rwanda, where a high percentage of women in parliament resulted in a serious decrease in corruption over ten years period. In line with this, Dahlum et al. (2021) also lagged WPE by 5 and 10 years in their analysis, arguing that changes in the policy realm require time. Thus, the arguments presented above justify the lagging of variables within five and ten years. However, considering the most recent data on anti-sex trafficking laws and enforcement is from 2019, CPI cannot be lagged back by ten years due to a change of index methodology in 2012. Thus, I will use a 5-year lag in my analysis, which should be enough to see changes, considering, for instance, that an average parliamentary term for politicians is four years or that civic engagement brings results faster due to the widespread of social media and overall globalisation processes. To make the analysis more parsimonious, I lag both corruption and WPE by five years, assuming that WPE levels are expected to affect corruption the same year. As theoretical arguments show, women counteract corruption and push for more effective anti- sex trafficking laws. Therefore, they need to be empowered at the same time as corruption is happening for it to be effective. Following similar logic, all control variables are lagged by five years across all models, except the Percentage of protestants from 1980, which is introduced as a historical variable, and Regions, added to control for regional fixed effects. The CEDAW ratification is lagged by four years back, for which the most recent data is available. In addition, I will check the robustness of results with other corruption indices with lags for 5 and 10 years. Descriptive statistics of all variables in their final form are presented in Appendix 2. 4.2.2. Statistical Method In this thesis, the relationship between anti-sex trafficking laws and enforcement, corruption, and WPE will be quantitatively analysed by applying Ordinary Least Squares (OLS) regression with interactive specifications. Despite the fact that the dependent variable is ordinal, with five categorical outcomes, I apply OLS striving for straightforwardness and parsimony in my analysis and research. This decision finds theoretical support among scholars, who state that even though, as a rule of thumb, a model can be run with OLS if a dependent variable has six or more categories, some studies used a variable even with four categories and reached sufficient results (Stolle et al., 2008 in Jonsson, 2018). Nevertheless, I acknowledge that several types of analysis can be done with an ordinal dependent variable (Menard, 2002 in Jonsson, 2018) and that conventional interpretation of OLS is not applicable here. Hence, ordered logistic regression will be run to test the robustness of results that are based on OLS regressions. 40 I will conduct a cross-sectional analysis where the main year of analysis is 2019, while all independent variables are lagged back by five years, except for two controls, for the reasons stated in the previous section. This results in a total sample size of 131 observations (countries), enough to generalise from this sample to the population from which the sample was gathered (Field, 2018). The following estimation equations explain the main effect expected in the First Hypothesis, with and without control variables, where 𝑌𝑖 represents the level of anti-sex trafficking regulation and enforcement in a country i, and 𝜀𝑖 is the error term: 1) 𝑌𝑖 = 𝛼 + 𝛽1𝐶𝑜𝑟𝑟𝑢𝑝𝑡𝑖𝑜𝑛𝑖 + 𝜀𝑖 2) 𝑌𝑖 = 𝛼 + 𝛽1𝐶𝑜𝑟𝑟𝑢𝑝𝑡𝑖𝑜𝑛𝑖 + 𝛽2𝐺𝐷𝑃𝑖 + 𝛽3𝐺𝑙𝑜𝑏𝑎𝑙𝑖𝑠𝑎𝑡𝑖𝑜𝑛𝑖 + 𝛽4𝐷𝑒𝑚𝑜𝑐𝑟𝑎𝑐𝑦𝑖 + 𝛽5𝑃𝑜𝑙𝑖𝑡𝑖𝑐𝑎𝑙 𝑆𝑡𝑎𝑏𝑖𝑙𝑖𝑡𝑦𝑖 + 𝛽6𝐶𝐸𝐷𝐴𝑊𝑖 + 𝛽7𝑃𝑜𝑝𝑢𝑙𝑎𝑡𝑖𝑜𝑛𝑖 + 𝛽8𝑅𝑒𝑙𝑖𝑔𝑖𝑜𝑛𝑖 + 𝛽9𝑅𝑒𝑔𝑖𝑜𝑛𝑖+ 𝜀𝑖 The model specification below is employed to test the Second Hypothesis and determine whether WPE moderates the main effect between corruption and anti-sex trafficking legislation and enforcement. I introduce an interaction term between corruption and WPE by multiplying these variables. Thus, estimation equations with and without explanatory variables are the following: 1) 𝑌𝑖 = 𝛼 + 𝛽1𝐶𝑜𝑟𝑟𝑢𝑝𝑡𝑖𝑜𝑛𝑖 + 𝛽2𝑊𝑃𝐸𝑖 + 𝛽3𝐶𝑜𝑟𝑟𝑢𝑝𝑡𝑖𝑜𝑛𝑖 ∗ 𝑊𝑃𝐸𝑖 + 𝜀𝑖 2) 𝑌𝑖 = 𝛼 + 𝛽1𝐶𝑜𝑟𝑟𝑢𝑝𝑡𝑖𝑜𝑛𝑖 + 𝛽2𝑊𝑃𝐸𝑖 + 𝛽3𝐶𝑜𝑟𝑟𝑢𝑝𝑡𝑖𝑜𝑛𝑖 ∗ 𝑊𝑃𝐸𝑖 + 𝛽4𝐺𝐷𝑃𝑖 + 𝛽5𝐺𝑙𝑜𝑏𝑎𝑙𝑖𝑠𝑎𝑡𝑖𝑜𝑛𝑖 + 𝛽6𝐷𝑒𝑚𝑜𝑐𝑟𝑎𝑐𝑦𝑖 + 𝛽7𝑃𝑜𝑙𝑖𝑡𝑖𝑐𝑎𝑙 𝑆𝑡𝑎𝑏𝑖𝑙𝑖𝑡𝑦𝑖 + 𝛽8𝐶𝐸𝐷𝐴𝑊𝑖 + 𝛽9𝑃𝑜𝑝𝑢𝑙𝑎𝑡𝑖𝑜𝑛𝑖 + 𝛽10𝑅𝑒𝑙𝑖𝑔𝑖𝑜𝑛𝑖 + 𝛽11𝑅𝑒𝑔𝑖𝑜𝑛𝑖+ 𝜀𝑖 5. Analysis and Results In this section, the results of statistical analysis will be presented. The section begins with a discussion of the validity and characteristics of the data, followed by statistical analysis and results. The results of robustness tests are subsequently provided. 5.1. Validity and characteristics of the data When conducting an OLS regression, particular assumptions regarding the data must be checked. As a rule, scholars test assumptions such as linearity, normal distribution of residuals, 41 multicollinearity, and homoscedasticity. In addition, extreme outliers should be controlled for. However, due to the ordinal nature of the dependent variable, assumption of linearity, that for all values of the independent variable, the effect of the independent variable on the outcome will always stay the same, cannot be met (Kellstedt and Whitten, 2018, p. 211). Given that several of the control variables employed are theoretically close to each other and have been employed to explain variation in each other, a test for multicollinearity is in order. When checking for multicollinearity, a strong correlation between two independent variables should be avoided (Kellstedt and Whitten, 2018, p. 232). A high correlation was detected between Democracy and WPE, and Globalisation Index and GDP per capita (log), with Spearman’s coefficients of .854 and .858 accordingly (see Appendix 3), which is slightly above the acceptable threshold of .8 (Field, 2018). However, as a result of the more substantial multicollinearity check with variance inflation factor (VIF), none of the variables has a VIF score above 10 or tolerance below 0.1, meaning there is no perfect multicollinearity between chosen independent variables (Appendix 4a). Other assumptions will be tested in the robustness section after the main analysis is completed. 5.2. Statistical Analysis In this section, I conduct the analysis to test the stated hypotheses. Firstly, I run multivariate regressions to test whether corruption has a negative effect on anti-sex trafficking laws and enforcement (H1). Secondly, I conduct multivariate regressions with the interaction effect between WPE and corruption, expecting to see that under higher levels of women’s political empowerment, the negative relationship between corruption and the adoption and enforcement of anti-sex trafficking laws across countries becomes weaker (H2). Prior to the analysis, it should be noted that at Spearman’s correlation coefficient of -0.567 (see Appendix 3), there is a strong negative relationship between CPI in 2014 and the adoption and enforcement of anti-sex trafficking laws in 2019 across countries, which is in line with the first hypothesis and can be seen as indicative evidence in favour of it. 5.2.1 Multivariate regressions: Results for the Main Effect The outputs for the main effect’s multiple regressions (OLS) are presented in Table 1 below, where the first hypothesis is tested. Based on the theory, the expectation is that higher levels of 42 corruption lead to lower levels of adoption and enforcement of anti-sex trafficking laws across countries. Since my dependent variable is ordinal and categorical, with a scale from 1 to 5, positive values indicate a higher likelihood of a country being in a high category, meaning it is more likely to have better anti-sex trafficking laws and enforcement. In Table 1, the explanatory variables are added stepwise in groups that combine political and societal variables. Economic controls are added one by one because the Globalisation Index includes not only economic indicators but also political and societal ones. In Table 1 below, six models are presented, where Model 1 demonstrates a bivariate relationship, and Models 2-6 are multivariate regressions in which controls are added stepwise. Model 1 presents the bivariate regression, where the effect of corruption on anti-sex trafficking laws and enforcement is negative and significant at the 99% level (p <0.01) with a regression coefficient of -2.504. Thus, the null hypothesis, stating there is no relationship, can be rejected. This result supports the expectation of the first hypothesis: higher levels of corruption indeed are associated with lower levels of adoption and enforcement of anti-sex trafficking laws across countries. For instance, in Afghanistan, where Corruption Perceptions Index is high, 0.88, Hudson’s sex trafficking score is 2. At the same time, Sweden, where corruption is low, 0.13, scored 4 for anti-sex trafficking legislation and enforcement. In all the models, the main relationship demonstrates a negative effect of corruption on anti-sex trafficking legislation and enforcement. Once controlling for other explanatory variables and accounting for regional fixed effects, the main coefficient keeps its statistical significance at 95% level (Models 2-6). While the results of including most of the explanatory variables align with the theoretical expectation, some controls affect the dependent variable unexpectedly. In Model 2, GDP per capita (logged) is introduced, but the positive coefficient is significant only at 90% level, indicating that wealthier counties have better anti-sex trafficking laws and law enforcement. However, it falls from statistical significance in Models 3-6. In Model 3, I introduce the Index of Globalisation, which shows a positive and significant relationship. The result is consistent with the theoretical expectations that freedoms and openness to the world lead to better anti- sex trafficking legislation and enforcement (Ravlik, 2019). However, in the final Model 6, the coefficient is significant only at the 90% level. 43 Democracy, introduced in Model 4, also provides support for the theoretical argument that more democratic counties have better anti-sex trafficking legislation and enforcement. The coefficient is positive and significant at 95% level in Models 4 and 6, and at 99% in Model 5. Political stability and absence of violence/terrorism variable is added in Model 4 and significant at 95% level in Models 4 and 5, but only at 90% in Model 6. Compliance with CEDAW, also introduced in Model 4 along with other political controls, works in line with the theoretical background. The third category has a positive and significant effect on the dependent variable at 95% level across all models. This means that countries that are in the third category, i.e., ratified CEDAW without reservations (but not ratified Optional Protocol), compared to the first category of states that did not ratify CEDAW, are more likely to have more effective anti-sex trafficking laws and enforcement in all models that include the variable. Other categories, however, fail to reach statistical significance, except category four in Model 6, which is significant at the 90% level. Thus, being part of CEDAW, and actual compliance with it, leads to a better sex trafficking situation on a country level. Population size (logged) does not significantly affect the dependent variable. The share of protestants in 1980 shows 90% level significance with a negative coefficient in Model 5 but falls from statistical significance in the final Model 6. While a negative sign suggests that countries with stronger protestant values in 1980 have worse anti-sex trafficking policies and enforcement in 2019, given the lack of significance in the full model, no conclusions of its effects can be drawn here. Lastly, in Model 6, I account for regional fixed effects, controlling for which does not affect the significance of the main relationship. This result is in line with the first hypothesis and shows that higher levels of corruption five years prior lead to lower levels of adoption and enforcement of anti-sex trafficking laws across countries. Thus, I accept the first hypothesis. The adjusted R2 considerably improves with the introduction of explanatory variables and regional effects. In the bivariate Model 1, adjusted R2 has a value of 0.347, while in the final Model 6, it rises to 0.541. It means that the final model has higher explanatory power. 44 Table 1. Multivariate regressions (OLS): Results for the Main Effect DV: Anti-Sex Model 1 Model 2 Model 3 Model 4 Model 5 Model 6 Trafficking TI Corruption -2.504*** -1.762*** -1.199** -1.469** -1.840*** -1.552** (0.302) (0.489) (0.506) (0.588) (0.625) (0.720) GDPPC (log) 0.123* -0.033 0.026 0.037 -0.023 (0.064) (0.079) (0.081) (0.083) (0.097) Globalisation 0.026*** 0.022** 0.020** 0.022* (0.008) (0.009) (0.010) (0.011) Democracy 0.781** 0.834*** 0.776** (0.303) (0.304) (0.381) Political Stability -0.228** -0.262** -0.215* (0.099) (0.106) (0.120) CEDAW=2 0.526 0.452 0.522 (0.387) (0.392) (0.403) CEDAW=3 0.860** 0.865** 1.048** (0.403) (0.411) (0.422) CEDAW=4 0.551 0.544 0.692* (0.392) (0.397) (0.410) Population (log) -0.013 0.000 (0.043) (0.050) Share of protestants in -0.006* -0.006 1980 (0.003) (0.004) Regional Fixed No No No No No Yes Effects Constant 4.218*** 2.744*** 2.108*** 0.982 1.490 0.753 (0.180) (0.788) (0.789) (0.901) (1.150) (1.473) R2 0.347 0.365 0.411 0.477 0.490 0.541 N 131 131 131 131 131 131 Notes: Standard errors in parentheses. ***p < 0.01, **p < 0.05, *p < 0.10. All independent variables are lagged by five years but CEDAW (lagged by four years) and Share of protestants (historical variable from 1980). 5.2.2. Multivariate Regressions: Results for the Interaction Effect Table 2 presents results of the interaction between corruption and WPE in relation to the dependent variable, adoption and enforcement of anti-sex trafficking laws. The bivariate regression between corruption and the dependent variable is kept for visualisation and comparison of results (Model 1). Again, it is negative and significant at the 99% level. In Model 45 2, when Women’s Political Empowerment Index is introduced, the main relationship between corruption and anti-sex trafficking legislation and enforcement stays negative and significant at 99% level. In line with theoretical expectations, WPE Index has a positive sign, suggesting that higher levels of WPE are associated with better anti-sex trafficking legislation and enforcement, holding corruption constant. This relationship is significant at the 95% level. Based on the theory, I expect that WPE moderates the relationship between corruption and the adoption and enforcement of anti-sex trafficking laws, such as the higher WPE levels negate the negative effect of corruption. Thus, under higher levels of WPE, I expect a weaker relationship between corruption and the adoption and enforcement of anti-sex trafficking laws across countries. In order to test the second hypothesis, the interaction variable was created and introduced to the regression: Corruption*WPE, which is a combination of CPI and WPEI. The interaction is introduced in Model 3 and further tested in combination with other explanatory variables and regional fixed effects in Models 4-5. In Models 3-5, the main relationship between corruption and anti-sex trafficking laws and enforcement, holding WPE equal to zero, remains negative at the 95% level. In the absence of WPE, the corruption coefficient seems to show a stronger negative effect when adding controls as the strength of the relationship increases from -3.317 in Model 3 to -5.671 in Model 5. However, WPEI’s coefficient, when corruption is equal to zero, falls from statistical significance in Models 3 and 4, then becoming significant and negative at 90% level when accounting for regional fixed effects in the last model. Contrary to what might expect, WPEI has a negative effect on anti- sex trafficking laws and enforcement in the absence of corruption, indicating that in countries with very low levels of corruption, WPE actually decreases the likelihood of effective anti-sex trafficking policies and enforcement. However, since WPEI is only significant at p < 0.1, this result should not be given too much weight. Nevertheless, the most important results Table 2 illustrates is the interaction effect in relation to the relationship between corruption and anti-sex trafficking laws and enforcement. In all three Models 3-5, it remains positive but comes up as significant at the 90% level only in the final Model 5, when controlled for explanatory variables and regional fixed effects. According to the sign, it aligns with the theoretical expectations: WPE moderates the effect of corruption on anti-sex trafficking laws and enforcement, such that the negative effect of it becomes weaker if WPE is high. The visualisation of the interaction effect is presented in Figure 2, following Table 2. However, while I do not reject the second hypothesis, this finding should be interpreted 46 with caution. As the coefficient is only significant at p < 0.1, results should not be attributed too much weight as it is a high threshold for significance. This result will be further elaborated upon in the discussion section below. Table 2. Multivariate regressions (OLS): Results for Interaction Effect DV: Anti-Sex Trafficking Model 1 Model 2 Model 3 Model 4 Model 5 TI Corruption -2.504*** -2.100*** -3.317** -3.911** -5.671** (0.302) (0.347) (1.514) (1.827) (2.212) WPE 0.929** 0.0496 -1.460 -3.402* (0.412) (1.142) (1.487) (1.808) TI Corruption # WPE 1.513 2.537 5.047* (1.831) (2.114) (2.565) GDPPC (log) 0.039 -0.046 (0.086) (0.099) Globalisation 0.019** 0.018 (0.009) (0.011) Democracy 0.927** 1.129** (0.455) (0.489) Political Stability -0.300** -0.297** (0.115) (0.128) CEDAW=2 0.281 0.302 (0.418) (0.422) CEDAW=3 0.722* 0.840* (0.430) (0.437) CEDAW=4 0.384 0.480 (0.421) (0.427) Population (log) -0.014 -0.015 (0.044) (0.051) Share of protestants in 1980 -0.005 -0.004 (0.003) (0.004) Regional Fixed Effects No No No No Yes Constant 4.218*** 3.300*** 4.032*** 2.824* 4.318* (0.180) (0.444) (0.991) (1.640) (2.328) R2 0.347 0.372 0.375 0.497 0.557 N 131 131 131 131 131 Notes: Standard errors in parentheses. ***p < 0.01, **p < 0.05, *p < 0.10. All independent variables are lagged by five years but CEDAW (lagged by four years) and Share of protestants (historical variable from 1980). 47 In order to visualise the interaction effect, I plot the data and build a margins plot presented below (Figure 2). Figure 2. Margins plot: graph for the interaction effect. Notes: The graph displays the interaction between corruption and WPEI in 2014. WPEI (vdem_gender_l5) is shown at its minimum, mean and maximum values. The main finding from the previous section and Table 1 is that higher corruption, in general, means a lower likelihood of being in a high category of anti-sex trafficking laws and enforcement, which is in line with H1. The theoretical expectations of H2 that under higher levels of women’s political empowerment, the negative relationship between corruption and the adoption and enforcement of anti-sex trafficking laws across countries becomes weaker, find numerical confirmation in Table 2 and visual in the graph (Figure 2). With the low levels of WPE, the blue line is very steep, as countries become less likely to be in a higher category of anti-sex trafficking legislation very quickly when corruption is high. With relatively higher levels of WPE, red and green lines become less steep, which aligns with H2. High levels of WPE moderate the negative relationship between corruption and the adoption and enforcement of anti-sex trafficking laws such that the negative effect of corruption on anti-sex trafficking laws and enforcement becomes weaker. 48 To make the results more intuitive, I turn to country examples. States like Austria and Belgium have low corruption and high WPE and hence, score high on the anti-sex trafficking laws and enforcement scale. However, states like Colombia, Uganda and Zimbabwe, having high corruption levels, still get high scores in anti-sex trafficking legislation and enforcement since they have high levels of WPE. Thus, higher levels of WPE weaken the negative relationship between corruption and the adoption and enforcement of anti-sex trafficking laws among these highly corrupted states. It corresponds to the less steep red and green lines on the graph presented above. Nevertheless, when a country has low WPE and low corruption levels, it is more likely to have a higher score in anti-sex trafficking policies and enforcement efforts. Qatar is a good example here. At the same time, if a country has high corruption and low WPE, it is more likely to have a lower score in anti-sex trafficking policies and enforcement. Such situation can be observed in countries like Congo, Somalia, or Iran. Overall, level of WPE seems to matter more in moderating the effect of corruption on anti-sex trafficking laws and enforcement among states with moderate or high levels of corruption. In regards to the explanatory variables, they show the same directionality as in Table 1. The coefficient for the Globalisation Index is positive and significant at 95% level in Model 4, suggesting that higher globalisation levels are associated with better anti-sex trafficking laws and enforcement. However, it falls from statistical significance in the last model. The coefficients for Democracy and Political Stability and Absence of Violence/Terrorism remain significant at 95% level across all models. While Democracy has a positive sign aligning with the theoretical expectations, Political Stability is negative, suggesting results opposite to theoretical arguments. Regarding compliance with CEDAW, the coefficient of the third category has a positive and significant effect on the dependent variable at 90% level in Models 4-5. The adjusted R2 considerably improves from 0.347 in Model 1 to 0.557 in Model 5, where explanatory variables and regional fixed effects are introduced. Hence, the final model shows greater explanatory ability. To summarise, H1 is supported based on the analysis. Indeed, higher corruption leads to lower levels of the adoption and enforcement of anti-sex trafficking laws across countries. However, there is weak empirical support for H2. Thus, even though there is an indication that under higher levels of WPE, the negative relationship between corruption and the adoption and 49 enforcement of anti-sex trafficking laws across countries becomes weaker, this result should not be attributed too much weight. 5.3. Robustness tests This section provides the results of robustness tests conducted in order to further test the findings and results of OLS regressions. 5.3.1. Regression diagnostics for OLS As was discussed before the analysis, when running an OLS regression, particular assumptions regarding the data must be checked. Thus, in this section, I test assumptions of normal distribution of residuals and homoscedasticity and control for extreme outliers. Firstly, there should be a uniform error variance under the assumption of homoscedasticity; otherwise, the data will have unequal error variance and be heteroscedastic (Kellstedt and Whitten, 2018, p. 208). As the RVF plot (Figure 1 in Appendix 4b) showed some indications of unequal error variance, I conducted a model with robust standard errors to see potential changes in p-values. However, there are no changes compared to the full model. Secondly, I conduct an analysis of the residuals, which allows to see whether errors are normally distributed with a mean of zero. The histogram (Figure 2 in Appendix 4b) shows that the error distribution largely follows this assumption, although, from QQ-plot (Figure 3 in Appendix 4b), we see single cases on both left and right tails, which slightly skew the distribution. Finally, I check for outliers and influential observations with extreme values that deviate from the best-fit line, thus potentially affecting the results. By using measurement of Cook’s Distance, there are 10 outliers with Cook’s D value above 4/n (Appendix 4c) for the main model (Table 1). However, removing these from the regression does not change the significance of the results, even though the R-squared increases and standard errors slightly decrease (Ibid). From a theoretical perspective, looking at the reasons why these observations are outliers, countries such as Bahrein, Barbados, Guyana, and the Philippines have some contradictions, for example, being very low in democracy, with high corruption, experiencing political instability, but at the same time scoring high in sex trafficking legislation and enforcement. The USA, having the lowest score on CEDAW (the country did not ratify it), 50 scoring high in sex trafficking legislation and enforcement. The other five outliers, however, are hard to interpret. Thus, considering that the main effect does not change, I decide to keep outliers to avoid the sample reduction. For the model with interaction (Table 2), there are 11 outliers (see Appendix 4c), and removing them from the Model does change the outcome: while the R-squared increases, the interaction coefficient falls from statistical significance (Table 1 in Appendix 4c). Thus, the results are sensitive to the exclusion of some cases, suggesting that I should interpret it with caution. 5.3.2. Ordered Logistic Regression To ensure that the outputs were not influenced by the use of OLS, I run the regressions again employing ordered logistic regression. Due to the fact that the dependent variable is ordered and categorical, conducting ordered logistic regression may provide different results. From Table 1 in Appendix 6, it is clear that the ordered logistic regression shows similar results to OLS: directionality, the significance of the main effect at 95% level and the significance of the interaction term at 90% level when controlling for other factors are confirmed by the results of the ordered logistic regression. Thus, conducting the ordered logistic regression suggests the robustness of the results in the main text. 5.3.3. Alternative measures of corruption and time lags To check whether results depend on a measure of corruption and lags back in time, I employ OLS, where I replace CPI with Bayesian Corruption Index (BCI) corruption measure, which is available for both 2014 and 2009, allowing for an additional ten-year lag. BCI is an alternative measure developed by Standaert in 2015. While highly correlated with CPI, it was constructed using a different methodology (Standaert, 2015). The results are presented in Tables 2 and 3 of Appendix 6, with BCI lagged by 5 and 10 years back, respectively. Results of Table 2 (Appendix 6) show that while the bivariate relationship between BCI in 2014 and anti-sex trafficking laws and enforcement in 2019 is negative and significant at 99% level, after introducing all controls and regional fixed effects, it remains significant only at 90% level. The interaction term is also significant at 90% level in the full model. In Table 3 (Appendix 6), BCI and all controls except CEDAW, religion and region are lagged by 10 years. The main bivariate relationship is negative and significant at 99% level but falls from statistical significance after the introduction of all controls and accounting for regional fixed effects. The interaction term is insignificant. It seems corruption and other explanatory variables measured in 2009 do not 51 affect the adoption and enforcement of anti-sex trafficking laws in 2019 as strongly as measures from 2014. The reason might be that during those ten years, countries may undergo significant internal changes, and hence, measurements from 2009 cease to be relevant to the state of affairs in 2019. Overall, the results lead to the conclusion that the main relationship between corruption and anti-sex trafficking laws and enforcement is negative in all scenarios. While it is significant at 95% level with CPI, the results based on BCI in 2014 show significance only at the 90% level. Thus, the results seem sensitive to the measurement of corruption employed. This issue will be elaborated on in the discussion section below. As 10 years lag of corruption seems not to have an effect on sex-trafficking laws, I further conduct the last robustness test replacing CPI with the V-Dem corruption measure in 2014. After introducing the controls and interaction, neither the corruption coefficient nor the interaction terms reach significance. This suggests that anti-sex trafficking legislation and enforcement might be sensitive to particular types of corruption (Jonsson, 2018) or corruption measurement. In addition, Models without time lags can be found in Table 5 of Appendix 6. When not including a time lag, only the bivariate relationship between corruption and anti-sex trafficking laws and enforcement is significant at 99% level and negative. As expected, current corruption levels do not affect anti-sex trafficking laws and enforcement, providing insignificant outcome when controlling for other factors measured in 2019. Indeed, corruption does not have an immediate effect on the sex trafficking situation but rather confirms the suspicion that corruption and WPE have a delayed effect. 5.3.4. Bigger sample excluding Share of protestants in 1980 Once excluding the control variable Share of protestants in 1980, the sample size rises up to 160 observations. In models with such sample (Table 6 in Appendix 6), the main relationship between corruption and anti-sex trafficking legislation and enforcement remains significant at 95% level, while interaction becomes significant at 95% level. On the one hand, it might suggest that results are dependent on sample size. On the other hand, it is also possible that Share of protestants in 1980 is a relevant confounder. 52 6. Discussion of Findings and Limitations Hypothesis 1 The result of running regression models for the main effect (Table 1) provides support for the theory that higher levels of corruption are associated with lower levels of adoption and enforcement of anti-sex trafficking laws across countries. Indeed, in line with theoretical expectations, public sector corruption in 2014 has a significant effect on the state of affairs in sex trafficking of women and girls in 2019. When controlling for other explanatory variables and accounting for regional effects, the main effect remains significant at the 95% level. Thus, the result is consistent with existing research in the field of human trafficking. This thesis finds a confirmation that corruption is a strong determinant and core push factor of trafficking in persons, reaffirming the conclusions made by Bales (2007), Zhang and Pineda (2008), Studnicka (2010), Cho (2015), and Ravlik (2019). More importantly, it brings valuable implications to the young and more narrow research on sexual exploitation of women and girls. The study by Jonsson (2018) initiated the discussion on whether corruption affects sex trafficking. While Ravlik (2019) stated that corruption hinders anti-trafficking legislation and enforcement as a whole phenomenon, Jonsson (2018) used a proxy variable of trafficking in persons for sex trafficking. This thesis attempted to address these limitations by conducting a more straightforward analysis utilising a holistic measurement of public sector corruption and a more reliable measurement of sex trafficking. As such, this thesis succeeded in confirming the results by Jonsson (2018) and Ravlik (2019). Indeed, corruption reinforces females’ marginalisation and vulnerability in society, and weakens the effectiveness and transparency of government apparatus, therefore, facilitating and maintaining sex trafficking and undermining the adoption and enforcement of anti-sex trafficking laws. The model also revealed two significant cofounders: Democracy and Ratification of CEDAW. Both seem to matter a lot in explaining the variance made in the adoption and enforcement of anti-sex trafficking laws. In line with theoretical argumentation, states with higher democracy levels have advanced legal frameworks and law enforcement regarding sex trafficking (Alexander and Ravlik, 2015; Ravlik, 2019). At the same time, being part of CEDAW, the international agreement fighting women’s discrimination and actual compliance with it, also leads to a better sex trafficking situation on a country level. The latter is an important finding not addressed in previous quantitative studies on sex trafficking. 53 However, taking into account the results of the robustness tests, limitations of the study and the dataset, the findings must be interpreted with caution. The conclusion is that my first hypothesis receives strong support using Corruption Perceptions Index but much weaker or no support when employing other measurements of corruption. Thus, it might suggest that my result is dependent on CPI and overall sensitive to types of corruption (Jonsson, 2018) or measurement techniques of various corruption indicators. When testing the robustness with BCI from 2014, the main effect comes as significant only at 90% level when controlling for other factors. Employment of the V-Dem Political Corruption Index does not lead to significant results at all. Such results imply that differences in the indicators affect the analysis’ outcome. In the main model, CPI was chosen based on the theoretical expectations as it has no distinction between administrative and political corruption (Coppedge et al., 2022) and, thus, captures corruption in various aspects of the public sector that can potentially affect sex trafficking. However, CPI is a composite indicator with a unique methodology, combining perceived corruption evaluations from various sources. Both BCI and the V-Dem Political Corruption Index have their own models of indices construction. BCI is also a composite index of the perceived overall level of corruption, measuring the same concept as CPI. However, it represents the data unbiased by any modelling choices of the composer, and has an absolute scale9, and substantively relies on different model (Teorell et al., 2022). Apparently, corruption measurement plays an important role in my analysis, as the effect of the independent on the dependent variable varies. Moreover, employing the V-Dem Political Corruption Index can point not only to the model’s sensitivity towards different methodologies and data sources but also to various types of misconduct. As the index reflects the pervasiveness of political corruption, it slightly differs in what types of corruption it includes. While it is still a broad range of cases, including public sector corruption, some other aspects, such as executive corruption in higher echelons of power, might affect sex trafficking less, or affect more the legislation side rather than enforcement, which also has a strong weight in my dependent variable. In line with this, Jonsson (2018) argues that law enforcement corruption, especially among police officers, influences sex trafficking the most. While the author’s argument is concerned with sex trafficking prevalence and not legislation, it is still a relevant factor that deserves to be addressed in future research. 9 On a scale, it has an actual zero when according to all surveys, there is no corruption, and one when according to all surveys, there is the worst corruption (Teorell et al., 2022, p. 121). 54 This study could not address this issue since there is no disaggregated measurement that only measures corruption in law enforcement on a global scale. Hypothesis 2 The interaction effect between corruption and WPE is significant at 90% level in the full model, controlling for other factors. The analysis results are consistent with the theoretical expectations that under higher levels of women’s political empowerment, the negative relationship between corruption and the adoption and enforcement of anti-sex trafficking laws across countries becomes weaker. Similar to the effects of the main hypothesis, once replacing CPI by BCI in 2014, the interaction effect stays significant at 90% level. Consequently, there is some evidence in favour of the second hypothesis, however, given the low level of significance, not too much weight should be put on these conclusions. Therefore, it should be interpreted and discussed with caution. Rather, future research should aim to address the data and method limitations. Reflecting on the results, employing formal and informal WPE in the model indeed aligns with theories in the field of gender equality. As a moderation factor, WPE may have an effect such that through their agency, female activists in civil society advocate for VAWG policies, while women in political office counteract political corruption and advance women-friendly legislation (Htun and Weldon, 2012; Alexander and Ravlik, 2015; Alexander, 2021). However, future research needs to directly test these mechanisms. In addition, the results suggest that the level of WPE seems to matter more in moderating the effect of corruption on anti-sex trafficking laws and enforcement among countries that have moderate or high levels of corruption. When looking at the sample, states with relatively high corruption levels and high levels of WPE (e.g., Colombia, Uganda and Zimbabwe) still get high scores on Hudson’s sex trafficking scale. At the same time, when corruption is high and WPE levels are low (e.g., Congo, Somalia, and Iran), it is more likely that the country will have a lower score in anti-sex trafficking policies and enforcement. From the theoretical perspective, it seems reasonable that having low levels of corruption would lead to better policy outcomes in a country, notwithstanding moderating factors. As a rule, such countries have good governance and transparent systems, especially if they are democratic regimes. However, it is an interesting finding that even in corrupt settings, under high levels of WPE, it is possible to observe successful adoption and enforcement of anti-sex trafficking laws. The women’s 55 interest mechanism comes to the fore, providing indication that female political empowerment may potentially lead to positive policy outcomes despite a corrupt environment. WPE as Moderator: Possible reasons for weak empirical support The general theoretical idea of this thesis is that WPE, as a complex concept and measurement including several components, should have a moderating effect on the negative relationship between corruption and sex trafficking. While the results indicate some empirical evidence in favour of this, the evidence is rather weak. Considering possible reasons for such outcome, I revisit the operationalisation of the WPE concept. WPEI, constructed as an average of indices of women’s civil liberties, civil society participation, and political participation, could have been disaggregated. Potentially it could lead to different results as each component, whether it is civil liberties of women or political rights, works differently in affecting anti-sex trafficking laws and enforcement. This is one of the limitations of this study not addressed due to the limited time and scope of work and an opportunity for further research. The weak effects found could also be interpreted as evidence that there are other possible moderating effects. In both full models (Table 1 and 2), democracy was a strong and significant confounder, confirming theoretical expectations that with higher levels of democracy, countries tend to have better anti-sex trafficking legislation and enforcement. Thus, it might be an important contextual setting where higher levels of democracy will negate the negative effect of public sector corruption on the adoption and enforcement of anti-sex trafficking laws. Indeed, some studies mention that more democratised countries have better legal human rights protection and advanced legislation regarding women-friendly policies (Alexander and Ravlik, 2015). Moreover, the proportion of women in the cabinet in parliamentary democracies explains the amount of female-friendly social policies (Atchison and Down, 2009). Overall, democracies engage more women in politics and civil society organisations, which has an impact on WPE (Bush and Zetterberg, 2021). In addition, it has been widely acknowledged that democracy is important in combating corruption (Kolstad and Wiig, 2016; Esarey and Schwindt-Bayer, 2018). Another possible reason for the lack of stronger results in favour of the hypothesis might be a limited sample size. Introducing to the model one of the explanatory variables, Share of protestants in 1980, considerably lowers the number of observations from 160 to 131. As the robustness test shows, when excluding this variable, the main relationship between corruption 56 and anti-sex trafficking legislation and enforcement remains significant at 95% level, while interaction becomes significant at 95% level (Appendix 6, Table 6). On the one hand, it is a relevant and theoretically strong confounder. Protestant-dominated countries support gender equality and tend to have lower corruption (Welzel et al., 2002; Portilla, 2022). On the other hand, however, the results of analysis might be dependent on a sample size. As Share of protestants in 1980 has a low number of observations, excluding almost thirty cases from the model leads to results that are significant only at a higher significance threshold, while the effect direction remains unchanged. Thus, it is one of the limitations of this thesis, and it would be interesting to address it in future research if there is an opportunity to gather more data on the Share of protestants in 1980 across countries. Additional Limitations In addition to the limitations connected to methodological choices in this study, there are other aspects that possibly affected the final results. First, the operationalisation of sex trafficking of women and girls presents a challenge for scholars. As it has been discussed in the data section, victims of human trafficking are considered to be “hidden populations” with no veracious information on their size and borders (Jonsson, 2018). Currently, the only option for analysis is data on the country’s anti-sex trafficking legislation and enforcement, which is aggregated in Hudson’s scale of trafficking. However, this measurement has drawbacks, such as a lack of methodological transparency and the complexity of composition, including the prevalence of the sex trafficking component. From the WomanStats Codebook (2023), it is clear that when constructing the scale, coders strongly relied on Tier Rankings reported by the US Department of State, and this scale is also known for its issues with the lack of transparency in expert evaluations (Ravlik, 2019). Another restriction of this thesis that must be acknowledged is the cross-sectionality of the analysis. With the sex trafficking data available, conducting a time-series analysis is challenging. Hudson’s trafficking scale was not measured over successive years but every four years starting from 2007. Thus, controlling how previous levels of anti-sex trafficking enforcement affects the results in 2019 would represent additional difficulty when adding to the model with time lags, which is beyond the scope of this thesis. In addition, it is important to highlight that by using time-lags, I tried to take causal direction into consideration empirically. However, the issue of reverse causality should be acknowledged, as theoretically, 57 it becomes a vicious cycle between corruption and sex trafficking, with both factors mutually reinforcing each other (Ravlik, 2019). Furthermore, it is necessary to emphasise that the thesis did not aim to identify which type of misconduct in the public sector is most closely related to sex trafficking but rather to cover all conceivable examples in the public realm. This is one of the study’s shortcomings as well as an opportunity for future research. 7. Conclusion By developing a plausible theory and research method, this study aimed to contribute to the growing literature on corruption and sex trafficking by investigating first, how corruption and sex trafficking are related and, second, how WPE might moderate this relationship. The research field on this nexus is relatively young, and no study has analysed the moderating effect of WPE on corruption’s effect on sex trafficking regulation and prevalence. Thus, via a quantitative approach, this thesis applied Ordinary Least Squares regression with interactive specifications to answer the stated at the beginning research question: “How does women’s political empowerment moderate the relationship between corruption and anti-sex trafficking legislation and enforcement across countries?” To address the research question, two hypotheses have been tested. First, this thesis tested the negative relationship between public sector corruption and the adoption and enforcement of anti-sex trafficking laws assumed in previous literature. The analysis results align with theoretical expectations and show that higher levels of corruption five years prior lead to lower levels of adoption and enforcement of anti-sex trafficking laws across countries. Thus, it makes an important contribution to the sex trafficking scholarship. While previous literature argued that corruption is a strong determinant of human trafficking (Bales, 2007; Zhang and Pineda, 2008; Studnicka, 2010; Cho, 2015a; Ravlik, 2020), very limited research has been done on the actual relationship between corruption and sex trafficking of women and girls. Despite the fact that a large number of trafficking incidents are related to sexual exploitation (UNODC, 2011 in Jonsson, 2018), there was a need for a more substantial and straightforward analysis concerning sex trafficking itself, separated from general trafficking in persons. Thus, this thesis provides confirmation of the negative impact of high corruption levels on the adoption and enforcement of anti-sex trafficking laws. However, an important reservation has to be made. 58 Through the application of robustness tests, this thesis found out that this result depends on the Transparency International CPI measure and, hence, Hudson’s scale of trafficking is sensitive to the corruption measurement and types of corruption. Thus, this result should be considered with caution. This theoretical implication indicates that, theoretically, women could be most vulnerable to particular corruption types, thus, affecting the research outcome. This is a potential avenue for future research, and, as has been emphasised by Jonsson (2018), some types of public misconduct might have a stronger effect on sex trafficking, such as law enforcement corruption. Moreover, future studies in sex trafficking should consider reshaping the concept of corruption, taking into account its gendered nature and the prevalence of sextortion, which has been a silent form of corrupt activity for a long time. Creation of sexual corruption measurement and its incorporation into global corruption indicators is a potential agenda for discussion. However, it is still constrained by the lack of data on a global scale. Having found support for the first hypothesis, this thesis further tested whether, under higher levels of WPE, the negative relationship between corruption and the adoption and enforcement of anti-sex trafficking laws across countries becomes weaker. The analysis results are consistent with the expectations of the second hypothesis but only approach significance at the 90% level, controlling for other factors, and, therefore, must be interpreted with caution and not given too much weight. However, the results are also robust to the ordered logistic regression test and to conducting OLS with BCI measurement instead of CPI in 2014. This suggests that the developed theory has potential and should be further tested, taking into account the limitations of this study. Overall, given the results of the analysis, the main theoretical expectations find empirical support to a certain extent. An additional finding is that WPE seems to matter more in moderating the effect of corruption on anti-sex trafficking laws and enforcement among countries with moderate or high corruption levels. With high levels of WPE, even corrupted states tend to have successful anti-sex trafficking legislation and enforcement. At the same time, WPE does not seem to be a crucial moderator for counties where corruption is low. It should be noted, however, that states with low corruption and low WPE are relatively rare cases with specific governments (e.g., Qatar). Thus, reflecting on the implications of the results for the existing literature, this thesis suggests that as a combination of female political descriptive 59 representation, civil liberties and civil society participation, WPE seems to have moderating effect on the negative relationship between corruption and anti-sex trafficking laws and enforcement. This finding is new to the sex trafficking scholarship. Nevertheless, it should not be treated as a strong result due to its low significance but rather used as a start to explore this nexus by employing more advanced analysis and tackling existing limitations. Based on the conclusions from the analysis and discussion, this thesis suggests further exploring the theory of the moderating effect of WPE on the negative relationship between corruption and sex trafficking. More nuanced theoretical arguments and more direct tests of these mechanisms need to be conducted to determine whether particular types of corruption have a stronger impact on sex trafficking legislation and enforcement. Furthermore, disaggregating WPEI might lead to new results and conclusions, given that each of the indices inside can work differently in affecting anti-sex trafficking laws and enforcement. Finally, apart from valuable theoretical implications to the existing literature, this study has implications for politics and society more broadly. A key takeaway relates to the critical role of state capacity in strengthening anti-sex trafficking laws and enforcement. Sound anti- corruption measures should be applied, striving for transparency in the public sector, at all levels, from the government to the judiciary and law enforcement. This might allow to reduce opportunities for public servants to engage in illicit activities connected to sex trafficking as well as make it riskier for them due to stiffer penalties. In addition, ratification and a strong commitment to international treaties to combat discrimination against women and protect their rights are highly recommended for states. This suggestion is based on the finding that stronger compliance with CEDAW leads to better anti-sex trafficking legislation and enforcement across countries. Furthermore, states should strengthen their focus and efforts on reaching gender equality both in political office and civil society. Through their agency, empowered women will advocate for more gender-specific policies on violence against women and girls across all levels of society. Moreover, governments should pay close attention to gathering statistical evidence of sex trafficking prevalence as well as monitor and evaluate the work of anti-trafficking enforcement mechanisms. As opposed to the legal framework, this data is scarce and unreliable at the moment, hindering research in the field and effective government response. Hence, in the future, comprehensive action plans to combat sex trafficking, supported by statistical data, must be developed and applied to national policy agendas. 60 References Agbu, O. (2003) “Corruption and human trafficking: The Nigerian case”. West Africa Review, 4(1). Aïssaoui, R. and Fabian, F. (2022) “Globalization, economic development, and corruption: A cross-lagged contingency perspective,” Journal of International Business Policy, 5(1), pp. 1– 28. Aja-Eke, D., Gillanders, R., Ouedraogo, I., and Maiga, E. (2023) “Sextortion, Women’s Educational Attainment, and Corruption”. Available at SSRN 4318031. Akee, R.K.Q. et al. (2010) “Ethnic Fragmentation, Conflict, Displaced Persons and Human Trafficking: An Empirical Analysis,” in Frontiers of Economics and Globalization. Emerald Group Publishing Limited, pp. 691–716. Akhter, S.H. (2004) “Is globalisation what it’s cracked up to be? Economic freedom, corruption, and human development,” Journal of world business: JWB, 39(3), pp. 283–295. Alexander, A.C. (2021) “Gender, Gender Equality, and Corruption,” in The Oxford Handbook of the Quality of Government. Oxford University Press (Oxford Handbooks), pp. The Oxford Handbook of the Quality of Government, 2021. Alexander, A.C. and Bågenholm, A. (2018) “Does Gender Matter? Female Politicians’ Engagement in Anti-corruption Efforts,” in Gender and Corruption. Cham: Springer International Publishing (Political Corruption and Governance), pp. 171–189. Alexander, A.C., Bolzendahl, C. and Jalalzai, F. (2016) “Defining Women's Global Political Empowerment: Theories and Evidence,” Sociology Compass, 10(6), pp. 432–441. Alexander, A. C., and Ravlik, M. (2015) “Responsiveness to women’s interests as a quality of government mechanism: A global analysis of women’s presence in national legislatures and 61 anti-trafficking enforcement”. Paper presented at the American political science association meeting, San Francisco. Ariadne, E., Pratamawaty, B. B., and Limilia, P. (2021) “Human Trafficking in Indonesia, The Dialectic Of Poverty and Corruption”. Sosiohumaniora, 23(3), pp. 356–363. Arrunada, B. (2010) “Protestants and Catholics: Similar Work Ethic, Different Social Ethic,” The Economic journal (London), 120(547), pp. 890–918. Atchison, A. and Down, I. (2009) “Women Cabinet Ministers and Female-Friendly Social Policy,” Poverty & Public Policy, 1(2), pp. 1–23. Avdan, N. (2012) “Human trafficking and migration control policy: vicious or virtuous cycle?,” Journal of public policy, 32(3), pp. 171–205. Andvig, J.C. (2007) “Corruption and armed conflicts: some stirring around in the governance soup”. NUPI Working Papers, Norwegian Institute of International Affairs (NUPI), Oslo Babbie, E. R. (2013). The basics of social research. Boston: Cengage Learning. Bales, K. (2007) “What Predicts Human Trafficking?,” International journal of comparative and applied criminal justice, 31(2), pp. 269–279. Balos, B. (2004) “The wrong way to equality: Privileging consent in the trafficking of women for sexual exploitation”. Harvard Women’s Law Journal, 27, 137. Bertone, A. (2004) “Transnational Activism to Combat Trafficking in Persons.” Brown Journal of World Affairs Winter/Spring: pp. 1-14. Bicker Caarten, A., Van Heugten, L., and Merkle, O. (2022) “The intersection of corruption and gender-based violence: Examining the gendered experiences of sextortion during migration to South Africa”. African Journal of Reproductive Health, 26(6), pp. 45-54. 62 Bush, S.S. and Zetterberg, P. (2021) “Gender Quotas and International Reputation,” American journal of political science, 65(2), pp. 326–341. Caprioli, M., Hudson, V. M., McDermott, R., Ballif-Spanvill, B., Emmett, C. F., and Stearmer, S. M. (2009) “The Womanstats project database: Advancing an empirical research agenda”. Journal of Peace Research, 46(6), pp. 839–851. Charron, N. (2015) “Exploring the Impact of Foreign Aid on Corruption: Has the 'Anti- Corruption' Aid been Effective?” Chase, G. (2010) “Religion’s impact on corruption”. Journal of Business and Economics Perspectives, 5(1), pp. 102–112 Childs, S., and Krook, M. L. (2009) “Analysing women’s substantive representation: From critical mass to critical actors”. Government and Opposition, 44(2), pp. 125-145. Cho, S.-Y. (2015a) “Modeling for determinants of human trafficking: An empirical analysis,” Social inclusion, 3(1), pp. 2–21. Cho, S.-Y. (2015b) “Evaluating Policies Against Human Trafficking Worldwide: An Overview and Review of the 3P Index,” Journal of human trafficking, 1(1), pp. 86–99. Cho, S.-Y. (2013) “Integrating equality: Globalization, women’s rights, and human Trafficking”. International Studies Quarterly, 57, pp. 683-697. Cho, S.-Y., Dreher, A. and Neumayer, E. (2013) “Does Legalized Prostitution Increase Human Trafficking?,” World Development, 41(1), pp. 67–82. Chong Soh, Y. and Amin, M. (2020) “Does Being A Large Country Lead to Higher Corruption?” The World Bank. Retrieved March 15th, 2023 from: https://blogs.worldbank.org/developmenttalk/does-being-large-country-lead-higher- corruption 63 Coalition Against Trafficking Women (CATW) (2023) “Violence against women and girls”. Retrieved February 5th, 2023 from: https://catwinternational.org/the-challenge/violence- against-women-and-girls/ Coppedge, Michael, John Gerring, Carl Henrik Knutsen, Staffan I. Lindberg, Jan Teorell, David Altman, Michael Bernhard, Agnes Cornell, M. Steven Fish, Lisa Gastaldi, Haakon Gjerløw, Adam Glynn, Sandra Grahn, Allen Hicken, Katrin Kinzelbach, Kyle L. Marquardt, Kelly McMann, Valeriya Mechkova, Pamela Paxton, Daniel Pemstein, Johannes von Römer, Brigitte Seim, Rachel Sigman, Svend-Erik Skaaning, Jeffrey Staton, Eitan Tzelgov, Luca Uberti, Yi-ting Wang, Tore Wig, and Daniel Ziblatt. (2022). “V-Dem Codebook v12” Varieties of Democracy (V-Dem) Project. Council of Europe (2002) “Trafficking in Human Beings and Corruption”. Programme against Corruption and Organised Crime in South-eastern Europe (PACO). Retrieved February 10th, 2023 from: https://documentation.lastradainternational.org/lsidocs/297%20Trafficking%20and%20Corru ption%20(PACO).pdf Dahlum, S., Knutsen, C.H. and Mechkova, V. (2022) “Women’s political empowerment and economic growth,” World Development, 156, p. 105822. Das, Jayoti and DiRienzo, Cassandra (2009) “The Nonlinear Impact of Globalization on Corruption”. The International Journal of Business and Finance Research, Vol. 3, No. 2, pp. 33-46. Delavallade, C. (2006) “Corruption and distribution of public spending in developing countries”. J Econ Finan 30, pp. 222–239 Di Nicola, A. (2007) “Researching into human trafficking: Issues and problems”. In M. Lee (Ed.), Human Trafficking (pp. 49–72). Portland, USA: Taylor and Francis Díaz Rivillas B. and Solano López A. L. (2020) “Corrupt officials, an essential link in human trafficking”. Planeta Futuro (El País). Retrieved February 10th, 2023 from: https://europa.eu/capacity4dev/mujer-y-corrupci-n-los-impactos-diferenciados-de-la- 64 corrupci-n-en-am-rica-latina-y-el-caribe/wiki/corrupt-officials-essential-link-human- trafficking Disability Rights International (2015) “No Way Home: The Exploitation and Abuse of Children in Ukraine’s Orphanages”. Retrieved February 11th, 2023 from: www.driadvocacy. org/wp-content/uploads/No-Way-Home-final2.pdf. Dollar, D., Fisman, R. and Gatti, R. (2001) “Are women really the ‘fairer’ sex? Corruption and women in government,” Journal of economic behaviour & organisation, 46(4), pp. 423–429. Eagly, A. H., and Crowley, M. (1986) “Gender and helping behaviour: A meta-analytical review of the social psychological literature”. Psychological Bulletin, 100, pp. 283–308. Eldén, Åsa et al. (2020) Sextortion: corruption and gender-based violence. (Rapport till Expertgruppen för biståndsanalys; 2020:06). Ekberg, G. (2004) “The Swedish law that prohibits the purchase of sexual services: Best practices for prevention of prostitution and trafficking in human beings.” Violence Against Women, 10(10), pp. 1187–1218. Esarey, J. and Chirillo, G. (2013) “‘Fairer Sex’ or Purity Myth? Corruption, Gender, and Institutional Context,” Politics & Gender, 9(4), pp. 361–389. Esarey, J., and Valdes, N. (2021). “Do Legislative Gender Quotas Reduce Corruption?” Working paper. Retrieved February 26th, 2023 from: www.justinesarey.com/research European Institute for Gender Equality (EIGE) (2016). “Glossary: Empowerment of Women”. Retrieved February 26th, 2023 from: https://eige.europa.eu/thesaurus/terms/1102 García-Moreno, Claudia, Dr et al. (2015) “Addressing violence against women: a call to action,” The Lancet (British edition), 385(9978), pp. 1685–1695. Gerring, J. and Christenson, D. (2017) Applied social science methodology: an introductory guide. 65 Goel, R.K. and Nelson, M.A. (2023) “Women’s political empowerment: Influence of women in legislative versus executive branches in the fight against corruption,” Journal of policy modeling, 45(1), pp. 139–159. Gold, B.R. and Haynie, K.L. (2014) “Representing Women’s Interests and Intersections of Gender, Race, and Ethnicity in US State Legislatures,” in Representation. New York: Oxford University Press, pp. Representation, 2014. Guth, A.P. (2010) “Human trafficking in the Philippines: the need for an effective anti- corruption program,” Trends in organised crime, 13(2-3), pp. 147–166. Feigenblatt, H. (2020) “Breaking the silence around sextortion. The links between power, sex and corruption”. Transparency International Field, A.P. (2018) Discovering statistics using IBM SPSS statistics. 5. ed. Friman, H. R., and Reich, S. (Eds.). (2008) Human trafficking, human security, and the Balkans. University of Pittsburgh Press. Hendry, N. H. (2021) “Sextortion”. The Fourth Industrial Revolution and Its Impact on Ethics: Solving the Challenges of the Agenda 2030, pp. 315-320. Hernandez, D., and Rudolph, A. (2015) “Modern day slavery: What drives human trafficking in Europe?” European Journal of Political Economy, 38, pp. 118–139. H.R.3244 - 106th Congress (1999-2000) “Victims of Trafficking and Violence Protection Act of 2000.” Congress.gov, Library of Congress, 28 October 2000. Retrieved February 26th, 2023 from: https://www.congress.gov/bill/106th-congress/house-bill/3244. Htun, M. and Weldon, S.L. (2012) “The Civic Origins of Progressive Policy Change: Combating Violence against Women in Global Perspective, 1975–2005,” The American political science review, 106(3), pp. 548–569. 66 Hudson, V. M., Ballif-Spanvill, B., Caprioli, M., and Emmett, C. F. (2012) “Sex and world peace”. New York: Columbia University Press. Hudson, V. M., Bowen, D. L., and Nielsen, P. L. (2011) “What is the relationship between inequity in family law and violence against women? Approaching the issue of legal enclaves”. Politics and Gender, 7(4), pp. 453–492. Hughes, D and Denisova, T. (2001) “The Transnational Political Criminal Nexus of Trafficking in Women from Ukraine”. Trends in Organized Crime, vol. 6, no. 3-4. International Bar Association (2016) “Human Trafficking and Public Corruption” – A report by the IBA’s Presidential Task Force Against Human Trafficking. Retrieved February 15th, 2023 from: https://www.ibanet.org/MediaHandler?id=E34FFA1D-8038-4AEC-A631- E0E2A7E0AD86 Jakobsson, N., and Kotsadam, A. (2013) “The law and economics of international sex slavery: Prostitution laws and trafficking for sexual exploitation”. European Journal of Law and Economics, 35(1), 87–107 Jonsson, S. (2018) “The Complex Relationship between Police Corruption and Sex Trafficking in Origin Countries,” Journal Of Human Trafficking, 2018, Vol. 5, Iss. 2, pp. 109-.129, 5(2), pp. 109–129. Kangaspunta, K. (2003) “Mapping the inhuman trade: Preliminary findings of the database on trafficking in human beings”. Paper presented at the Forum on crime and society, Vienna, Austria. Kaufmann, D., Kraay A. and Mastruzzi M. (2010) “The Worldwide Governance Indicators: Methodology and Analytical Issues”. World Bank Policy Research Working Paper No. 5430. Retrieved March 1st, 2023 from: http://papers.ssrn.com/sol3/papers.cfm?abstract_id=1682130. 67 Kaufmann, D., Kraay, A., Mastruzzi, M. (2007) “Measuring Corruption: Myths and Realities.” Africa Region Findings & Good Practice Infobriefs; no. 273. World Bank, Washington, DC. © World Bank. Kellstedt, P.M. and Whitten, G.D. (2018) The Fundamentals of Political Science Research. Cambridge University Press. Kolstad, I. and Wiig, A. (2016) “Does democracy reduce corruption?,” Democratization, 23(7), pp. 1198–1215. La Porta, R. et al. (1999) “The quality of government,” Journal of law, economics, & organization, 15(1), pp. 222–279. Laczko, F. (2005) “Introduction: Data and Research on Human Trafficking,” In Data and research on human trafficking: A global survey. International migration, 43(1-2), pp. 5–16. Lee, J. J. (2005) “Human trafficking in East Asia: current trends, data collection, and knowledge gaps”. International Migration, 43(1‐2), 165-201. TIP report 2013: 29-38 Lovenduski, J., and Norris, P. (2003) “Westminster women: The politics of presence”. Political Studies, 51(1), pp. 84-102. Malhotra, A., S. Schuler and C. Boender (2002) “Measuring Women’s Empowerment as a Variable in International Development”. World Bank Workshop on Poverty and Gender, Washington, DC Masakure, O. (2017) “Human Trafficking and Education in Africa”, in “Security, Education and Development in Contemporary Africa”, International Political Economy of New Regionalism series, London; New York: Routledge, Taylor and Francis Masci, D. (2004) “Human trafficking and slavery: Are the world’s nations doing enough to stamp it out?” CQ Researcher, 14(12), 273–296. 68 McAlpine, A., Hossain, M. and Zimmerman, C. (2016) “Sex trafficking and sexual exploitation in settings affected by armed conflicts in Africa, Asia and the Middle East: systematic review,” BMC international health and human rights, 16(1), pp. 1–16. Michau, Lori, MA et al. (2015) “Prevention of violence against women and girls: lessons from practice,” The Lancet (British edition), 385(9978), pp. 1672–1684. Menard, S. (2001) Applied Logistic Regression Analysis. 2nd edn. Los Angeles: SAGE Publications Inc (Quantitative Applications in the Social Sciences). Meyer, B. (2003) “Much ado about nothing? Political representation policies and the influence of women parliamentarians in Germany”. Review of Policy Research, 20(3), 401-422. Morgan, A. L. (1998) “Corruption: Causes, consequences, and policy implications. A literature review”. The Asia Foundation Working Paper Series. Niethammer C. (2020) “Cracking The $150 Billion Business Of Human Trafficking”, Forbes. Retrieved February 1st, 2023 from: https://www.forbes.com/sites/carmenniethammer/2020/02/02/cracking-the-150-billion- business-of-human-trafficking/?sh=16218c464142 Ottervik, M. and Su, Z. (2022) “Gender system and corruption: Patriarchy as a predictor of ‘fairness,’” Governance (Oxford), 36(2), pp. 629–649. PACO (2002) “Trafficking in human beings and corruption”. Report on the regional seminar. Program against corruption and organised crime in South Eastern Europe. Economic Crime Division. Portoroz, Slovenia. Pitkin, H. F. (1967) “The Concept of Representation”. Berkeley: University of California Press. Plümper, T. and Neumayer, E. (2006) “The Unequal Burden of War: The effect of armed Conflict on the Gender Gap in Life Expectancy”, International Organization, Vol. 60, No. 3, pp. 723-754. 69 Portilla, J.G. (2022). Corruption and Religion (A), (B), (1). In: “Ye Shall Know Them by Their Fruits”. Contributions to Economics. Springer, Cham. Rao, S., and Presenti, C. (2012) “Understanding human trafficking origin: a cross-country empirical analysis”. Feminist Economics, 18(2), pp. 231–263. Ravlik, M. (2019). The Fight Against Human Trafficking. Cham: Springer International Publishing AG. Roots, K. (2013) “Trafficking or Pimping?: An Analysis of Canada’s Human Trafficking Legislation and its Implications,” Canadian journal of law and society, 28(1), pp. 21–41. Salt, J. (2000) “Trafficking and human smuggling: A European perspective”. International Migration, 38(3), pp. 31–56. Schwindt-Bayer, L.A., Esarey, J. and Schumacher, E. (2018) “Gender and Citizen Responses to Corruption among Politicians: The U.S. and Brazil,” in Gender and Corruption. Cham: Springer International Publishing (Political Corruption and Governance), pp. 59–82. Shelley, L. (2010) Human Trafficking. New York: Cambridge University Press. Sinha, P. et al. (2017) “Structural violence on women: An impediment to women empowerment,” Indian journal of community medicine, 42(3), pp. 134–137. Standaert, S. (2015) “Divining the level of corruption: A Bayesian state-space approach”. Elsevier. Journal of Comparative Economics. 43. 3. pp. 782-803. Stearmer, M., Emmett, C. F., and Stearmer, M. (2007) “The Great divide: Revealing differences in the Islamic world regarding the status of women and its impact on international peace”, pp. 1–22. Paper prepared for presentation at the annual meeting of the American Political Science Association. 70 Stolle, D., Soroka, S., and Johnston, R. (2008) “When does diversity erode trust? Neighborhood diversity, interpersonal trust and the mediating effect of social interactions”. Political Studies, 56(1), pp. 57–75. Studnicka, A. C. S. (2010) “Corruption and human trafficking in Brazil: Findings from a multi- modal approach”. European Journal of Criminology, 7(1), pp. 29–43. Sundström, A. et al. (2017) “Women’s political empowerment: A new global index, 1900- 2012,” World Development, 94 (June), p. 321. Sundström, A. and Wängnerud, L. (2021) “Sexual forms of corruption and sextortion: How to expand research in a sensitive area.” Sung, H-E, and Chu, D. (2003) “Does participation in the global economy reduce political corruption?” International Journal of Comparative Criminology, 3(2), pp. 94 – 118. Swamy, A. et al. (2001) “Gender and corruption,” Journal of development economics, 64(1), pp. 25–55. Teorell, Jan, Aksel Sundström, Sören Holmberg, Bo Rothstein, Natalia Alvarado Pachon & Cem Mert Dalli. 2022. The Quality of Government Standard Dataset, version jan22. University of Gothenburg: The Quality of Government Institute. Retrieved February 15th, 2023 from: https://www.gu.se/en/quality-government doi:10.18157/qogstdjan22 The United Nations Treaty Collection (2023). Status of treaties: Protocol to Prevent, Suppress and Punish Trafficking in Persons, Especially Women and Children, supplementing the United Nations Convention against Transnational Organized Crime. New York, 15 November 2000. Retrieved February 21st, 2023 from: https://treaties.un.org/pages/viewdetails.aspx?src=ind&mtdsg_no=xviii-12-a&chapter=18 The United Nations Secretary-General’s Bulletin on protection from sexual exploitation and abuse (2003). Retrieved February 21st, 2023 from: https://www.un.org/preventing-sexual- exploitation-and-abuse/content/documents 71 Trafficking in Persons Report (2022). The U.S. Department of State. Retrieved February 1st, 2023 from: https://www.state.gov/reports/2022-trafficking-in-persons-report/ Transparency International (2019) “Conflict at the bottom”. Retrieved March 15th, 2023 from: https://www.transparency.org/en/news/conflict-at-the-bottom Transparency International (2023) “What is corruption?” Retrieved March 15th, 2023 from: https://www.transparency.org/en/what-is-corruption Uddin, M.B. (2014) “Human Trafficking in South Asia: Issues of Corruption and Human Security,” International Journal of Social Work and Human Services Practice, 2(1), pp. 18– 27. United Nations (2000). Protocol to Prevent, Suppress and Punish Trafficking in Persons Especially Women and Children, supplementing the United Nations Convention against Transnational Organized Crime. ADOPTED 15 November 2000 BY General Assembly resolution 55/25. Retrieved February 3rd, 2023 from: https://www.ohchr.org/en/instruments-mechanisms/instruments/protocol-prevent-suppress- and-punish-trafficking-persons UN General Assembly (2015). Transforming our world: the 2030 Agenda for Sustainable Development, A/RES/70/1. Retrieved February 3rd, 2023 from: https://www.refworld.org/docid/57b6e3e44.html United Nations Office on Drugs and Crime (UNODC) (2011). “The Role of Corruption in Trafficking in Persons”. Issie Paper. Retrieved February 25th, 2023 from: https://www.unodc.org/documents/human-trafficking/2011/Issue_Paper_- _The_Role_of_Corruption_in_Trafficking_in_Persons.pdf UNODC (2006). “Trafficking in Persons: Global Patterns”. United Nations publication UNODC (2018). “Global Report on Trafficking in Persons 2018”. United Nations publication 72 UN Women (2010). “CEDAW & Women's Human Rights”. Retrieved February 23rd, 2023 from: https://asiapacific.unwomen.org/en/focus-areas/cedaw-human- rights/faq#:~:text=CEDAW%20legally%20binds%20all%20States,by%20private%20individ uals%20and%20organizations. Vahdati, M. (2021) “Women’s Political Empowerment and CEDAW: The Case of Iran and Turkey”. Wilson Center. Retrieved March 7th, 2023 from: https://www.wilsoncenter.org/article/womens-political-empowerment-and-cedaw-case-iran- and-turkey Van Dijk, J., and Klerx-Van Mierlo, F. (2014) “Quantitative indices for anti-human trafficking policies: Based on reports of the US State Department and the Council of Europe”. Crime, Law and Social Change, 61(2), pp. 229–250. Wängnerud, L. (2009) “Women in Parliaments: Descriptive and Substantive Representation.” Annual Review of Political Science 12: pp. 51–69. Welzel, C., Norris, P. and Inglehart, R. (2002) “Gender Equality and Democracy,” Comparative sociology, 1(3-4), pp. 321–345. Wooditch, A. (2011) “The efficacy of the trafficking in persons report: A review of the evidence”. Criminal Justice Policy Review, 22(4), pp. 471–493. WomanStats Project Database (2023). Retrieved February 5th, 2023 from: http://www.womanstats.org WomanStats Project (2023), “WomanStats Codebook.” Retrieved February 5th, 2023 from: http://www.womanstats.org/new/codebook/ Zhang, S.X. and Pineda, S.L. (2008) “Corruption as a Causal Factor in Human Trafficking,” in Organized Crime: Culture, Markets and Policies. New York, NY: Springer New York (Studies in Organized Crime), pp. 41–55. 73 Appendix Appendix 1: Original Scales of Variables Appendix 1a: Original Scale of the Dependent Variable Scale points for Hudson’s Sex Trafficking Scale (WomanStats Project, 2023): Scale Point 0: Anti-sex trafficking laws, laws are enforced, full compliance with the Trafficking Victims Protection Act of 2000 (TVPA) (Tier 1 ranking), sex trafficking is rare; Scale Point 1: Anti-sex trafficking laws, laws are enforced, law enforcement becomes weaker or considerable sex trafficking goes unnoticed by the authorities, full compliance with the TVPA (Tier 1 ranking); Scale Point 2: Anti-sex trafficking laws but not always enforced, non-compliance with the TVPA, but efforts are made to comply (Tier 2 ranking); Scale Point 3: Limited anti-sex trafficking laws, non-compliance with the TVPA (Tier 2 ranking), limited initiative to comply; Scale Point 4: No anti-sex trafficking laws, non-compliance with the TVPA (Tier 3 ranking), no victim support, the government can facilitate trafficking. Appendix 1b: Original Scale of Ratification of CEDAW Variable This scale by Professor Senem Ertan assesses the extent to which a country has committed to CEDAW (WomanStats Project, 2023): Scale Point 0: Ratified CEDAW, no reservation, ratified Optional Protocol; Scale Point 1: Ratified CEDAW, no reservations, did not ratify Optional Protocol; Scale Point 2: Ratified CEDAW, but with reservations; Scale Point 3: Did not ratify CEDAW. 74 Appendix 2: Descriptive Statistics Table 1. Descriptive statistics (2014 sample, general) Variable Obs Mean Std. Dev. Min Max Anti-Sex Trafficking 2019 172 2.727 .905 1 5 TI Corruption 166 .573 .199 .08 .92 Bayesian Corruption 172 .483 .164 .069 .744 V-Dem Corruption 169 .512 .306 .002 .967 WPE 169 .734 .186 .176 .966 GDPPC (log) 168 8.672 1.456 5.616 11.685 Globalisation 169 63.038 14.69 28.952 90.661 Democracy 169 .534 .259 .019 .916 Political Stability 172 -.184 .978 -2.749 1.468 CEDAW 172 3.035 .967 1 4 Population (log) 170 16.114 1.65 12.56 21.034 Share of protestants in 1980 139 12.429 20.658 0 97.8 The Region of the Country 172 4.058 2.349 1 10 Table 2. Descriptive statistics (2014 sample, N=131) Variable Obs Mean Std. Dev. Min Max Anti-Sex Trafficking 2019 131 2.817 .858 1 5 TI Corruption 131 .56 .202 .08 .92 Bayesian Corruption 131 .476 .173 .069 .744 V-Dem Corruption 131 .493 .307 .002 .967 WPE 131 .745 .17 .275 .966 GDPPC (log) 131 8.647 1.545 5.616 11.685 Globalisation 131 63.415 14.616 28.952 90.661 Democracy 131 .556 .252 .019 .916 Political Stability 131 -.184 .948 -2.699 1.468 CEDAW 131 2.985 .953 1 4 Population (log) 131 16.334 1.597 12.56 21.034 Share of protestants in 1980 131 12.164 20.683 0 97.8 The Region of the Country 131 4.359 2.079 1 10 75 Appendix 3: Correlation table 76 Appendix 3: Correlation table Spearman's rank correlation coefficients Variables (1) (2) (3) (4) (5) (6) (7) (8) (9) (10) (11) (12) (13) (1) Anti-Sex Trafficking 2019 1.000 (2) TI Corruption 2014 -0.567 1.000 (3) Bayesian Corruption 2014 -0.478 0.899 1.000 (4) V-Dem Corruption 2014 -0.533 0.905 0.842 1.000 (5) WPE 2014 0.522 -0.609 -0.428 -0.652 1.000 (6) GDPPC (log) 2014 0.537 -0.760 -0.700 -0.725 0.520 1.000 (7) Globalisation 2014 0.610 -0.782 -0.700 -0.727 0.613 0.858 1.000 (8) Democracy 2014 0.546 -0.611 -0.412 -0.660 0.854 0.521 0.592 1.000 (9) Political Stability 2014 0.382 -0.761 -0.691 -0.754 0.616 0.673 0.622 0.556 1.000 (10) CEDAW 2015 -0.032 0.034 0.129 0.045 0.208 -0.214 -0.145 0.201 0.112 1.000 (11) Population (log) 2014 -0.019 0.196 0.210 0.181 -0.091 -0.137 0.012 -0.085 -0.408 -0.164 1.000 (12) Share of protestants in 1980 0.148 -0.184 -0.084 -0.206 0.353 0.107 0.062 0.328 0.276 0.337 -0.148 1.000 (13) The Region of the Country 2019 0.036 -0.129 -0.190 -0.232 0.133 0.018 -0.008 0.127 0.181 -0.011 0.033 0.164 1.000 Spearman rho = 0.164 Note: Sample year: 2014. CEDAW is only available for 2015, Share of protestants is a historical variable from 1980. Appendix 4: Diagnostics Appendix 4a: VIF-Statistics Variance inflation factor VIF 1/VIF TI Corruption 6.952 .144 WPE 4.446 .225 GDPPC (log) 7.737 .129 Globalisation 8.958 .112 Democracy 4.56 .219 Political Stability 4.392 .228 2.cedaw l4 13.178 .076 3.cedaw l4 8.412 .119 4.cedaw l4 13.916 .072 Population (log) 2.157 .464 Religion: Protestant 1.874 .534 2.ht region 5.241 .191 3.ht region 6.438 .155 4.ht region 10.268 .097 5.ht region 5.906 .169 6.ht region 2.069 .483 7.ht region 3.452 .29 8.ht region 3.66 .273 9.ht region 1.486 .673 10.ht region 2.699 .371 Mean VIF 5.89 . Appendix 4b: Figure 1. Error distribution: RVF plot with a line. 77 Appendix 4b: Figure 2. Histogram. Distribution of the error term. Appendix 4b: Figure 3. QQ-plot. Distribution of the error term. Appendix 4c: List of Outliers for the main model Country name Bahrain Barbados Guyana Iran (Islamic Republic of) Japan Korea (the Republic of) Libya Philippines (the) United States of America (the) Venezuela (Bolivarian Republic of) Appendix 4c: List of Outliers for the model with interaction Table: List of Variables Country name Bahrain Barbados Bhutan Greece Guyana Iran (Islamic Republic of) 78 Japan Korea (the Republic of) Libya United States of America (the) Venezuela (Bolivarian Republic of) Appendix 4c: Table 1. The model with interaction excluding outliers (2014 sample) DV: Anti-Sex Trafficking 2019 Model 6 TI Corruption -3.454 (2.173) WPE -1.392 (1.745) TI Corruption # WPE 2.250 (2.568) GDPPC (log) 0.0185 (0.091) Globalisation 0.0073 (0.011) Democracy 1.079** (0.454) Political Stability -0.215* (0.117) CEDAW=2 0.415 (0.671) CEDAW=3 1.027 (0.644) CEDAW=4 0.669 (0.659) Population (log) 0.0201 (0.048) Share of protestants in 1980 -0.007** (0.003) Regional Fixed Effects Yes Constant 2.201 (2.249) R2 0.622 N 120 Notes: Standard errors in parentheses. ***p < 0.01, **p < 0.05, *p < 0.10. 79 Appendix 5: Multivariate Regressions (OLS): Results for the Main Effect including Regions Anti-Sex M1 M2 M3 M4 M5 M6 M7 M8 M9 Traff 2019 - - - TI - - - 2.504** 1.762** ** -0.826 -1.340 ** ** ** 1.840 ** -1.552** Corruption * * 1.199 1.469 1.460 * (0.302) (0.489) (0.506) (0.521) (0.593) (0.588) (0.591) (0.625) (0.720) GDPPC 0.123* -0.033 -0.004 0.011 0.026 0.022 0.037 -0.023 (log) (0.064) (0.079) (0.079) (0.079) (0.081) (0.084) (0.083) (0.097) Globalisatio 0.026** ** ** *** 0.021 0.020 0.022 0.023 ** 0.019** 0.022* n (0.008) (0.009) (0.009) (0.009) (0.009) (0.009) (0.011) 0.834** Democracy 0.699** 0.755** 0.781** 0.779** *** 0.776 (0.296) (0.295) (0.303) (0.305) (0.304) (0.381) Political - - - -0.170* ** ** ** -0.215 * Stability 0.228 0.236 0.262 (0.096) (0.099) (0.106) (0.106) (0.120) CEDAW=2 0.526 0.512 0.452 0.522 (0.387) (0.394) (0.392) (0.403) CEDAW=3 0.860** 0.839** 0.865** 1.048** (0.403) (0.414) (0.411) (0.422) CEDAW=4 0.551 0.533 0.544 0.692* (0.392) (0.401) (0.397) (0.410) Population -0.0101 -0.013 0.000 (log) (0.044) (0.043) (0.050) Share of protestants -0.006* -0.006 in 1980 (0.003) (0.004) 2. Latin 0.743** America (0.349) 3. North Africa & 0.809** the Middle East (0.396) 4. Sub- Saharan 0.593 Africa 80 (0.381) 5. Western Europe and 0.941*** North America (0.357) 6. East Asia 0.628 (0.459) 7. South- 0.673* East Asia (0.402) 8. South 0.383 Asia (0.463) 9. The -0.613 Pacific (0.751) 10. The 0.756 Caribbean (0.470) 4.218** 2.744** 2.108** Constant 1.619*** * * 1.747 ** 0.982 1.143 1.490 0.753 (0.180) (0.788) (0.789) (0.802) (0.799) (0.901) (1.143) (1.150) (1.473) r2 0.347 0.365 0.411 0.436 0.449 0.477 0.477 0.490 0.541 N 131 131 131 131 131 131 131 131 131 Notes: Standard errors in parentheses. ***p < 0.01, **p < 0.05, *p < 0.10. Sample year: 2014. 81 Appendix 6: Robustness tests Appendix 6: Table 1. Ordered Logistic Regression Results (2014 sample) DV: Anti-Sex Trafficking 2019 Model 1 Model 2 Model 3 Model 4 TI Corruption -7.145*** -5.091** -8.237* -18.18** (1.071) (2.395) (4.452) (7.890) GDPPC (log) -0.065 -0.125 (0.317) (0.333) Globalisation 0.073** 0.0633* (0.037) (0.038) Democracy 3.153** 4.255** (1.290) (1.697) Political Stability -0.684* -0.919** (0.400) (0.435) CEDAW=2 1.983 1.532 (1.502) (1.532) CEDAW=3 3.795** 3.481** (1.601) (1.645) CEDAW=4 2.519 2.134 (1.543) (1.575) Population (log) 0.033 0.003 (0.170) (0.174) Share of protestants in 1980 -0.018 -0.013 (0.012) (0.013) Regional Fixed Effects Yes Yes WPE 1.427 -10.60* (3.307) (6.328) TI Corruption # WPE 2.762 16.03* (5.407) (9.161) N 131 131 131 131 Notes: Standard errors in parentheses. ***p < 0.01, **p < 0.05, *p < 0.10. Models 1-2 test H1, Models 3-4 include interaction term and test H2. 82 Appendix 6: Table 2. OLS Results with BCI Corruption measure (2014 sample) DV: Anti-Sex Model 1 Model 2 Model 3 Model 4 Model 5 Model 6 Model 7 Trafficking 2019 Bayesian -2.609*** -0.859 -1.374** -1.632*** -1.237* -3.022** -4.635** Corruption (0.371) (0.528) (0.553) (0.580) (0.676) (1.518) (2.146) GDPPC (log) -0.014 0.033 0.0474 0.005 -0.011 (0.080) (0.080) (0.083) (0.095) (0.098) Globalisation 0.030*** 0.022** 0.020** 0.023** 0.021* (0.008) (0.009) (0.009) (0.011) (0.011) Democracy 1.028*** 1.128*** 1.003*** 1.290*** (0.301) (0.308) (0.375) (0.490) Political Stability -0.207** -0.226** -0.181 -0.256** (0.096) (0.102) (0.116) (0.127) CEDAW=2 0.508 0.444 0.507 0.367 (0.387) (0.393) (0.405) (0.418) CEDAW=3 0.835** 0.840** 1.007** 0.873** (0.402) (0.412) (0.424) (0.434) CEDAW=4 0.559 0.559 0.676 0.532 (0.392) (0.399) (0.412) (0.425) Population (log) -0.007 0.003 -0.014 (0.043) (0.051) (0.052) Share of protestants in -0.005 -0.005 -0.003 1980 (0.003) (0.004) (0.004) Regional Fixed Yes Yes Effects WPE 1.013 -2.226 (0.922) (1.489) Bayesian Corruption # 1.160 4.302* WPE (1.921) (2.578) Constant 4.058*** 1.427** 0.613 0.868 -0.001 3.100*** 2.292 (0.188) (0.708) (0.805) (1.062) (1.345) (0.765) (1.934) R2 0.278 0.397 0.477 0.487 0.536 0.361 0.548 N 131 131 131 131 131 131 131 Notes: Standard errors in parentheses. ***p < 0.01, **p < 0.05, *p < 0.10. Models 1-5 test H1, Models 6-7 include interaction term and test H2. 83 Appendix 6: Table 3. Robustness test with BCI Corruption measure (2009 sample) DV: Anti-Sex Model 1 Model 2 Model 3 Model 4 Model 5 Model 6 Model 7 Trafficking 2019 Bayesian -2.484*** -0.794 -1.210** -1.440** -1.116 -2.250 -2.710 Corruption (0.371) (0.524) (0.564) (0.597) (0.687) (1.462) (2.082) GDPPC (log) -0.062 -0.022 -0.027 -0.0841 -0.081 (0.086) (0.089) (0.093) (0.104) (0.106) Globalisation 0.033*** 0.027*** 0.027*** 0.0294*** 0.028** (0.008) (0.009) (0.009) (0.0106) (0.011) Democracy 0.721** 0.788** 0.763* 0.713 (0.301) (0.306) (0.387) (0.510) Political Stability -0.150* -0.173* -0.171* -0.196* (0.088) (0.094) (0.103) (0.106) CEDAW=2 0.491 0.377 0.479 0.488 (0.480) (0.495) (0.494) (0.497) CEDAW=3 0.745 0.675 0.876 0.901* (0.500) (0.524) (0.533) (0.537) CEDAW=4 0.530 0.462 0.609 0.615 (0.491) (0.510) (0.513) (0.517) Population (log) -0.024 -0.0203 -0.025 (0.047) (0.052) (0.053) Share of protestants in -0.005 -0.005 -0.004 1980 (0.003) (0.004) (0.004) Regional Fixed Yes Yes Effects WPE 1.321 -0.758 (0.922) (1.525) Bayesian Corruption # 0.420 2.109 WPE (1.820) (2.532) Constant 3.997*** 1.684** 0.991 1.606 0.895 2.783*** 1.627 (0.188) (0.692) (0.883) (1.232) (1.488) (0.774) (1.909) R2 0.258 0.386 0.437 0.447 0.513 0.352 0.518 N 131 130 130 130 130 131 130 Notes: Standard errors in parentheses. ***p < 0.01, **p < 0.05, *p < 0.10. Models 1-5 test H1, Models 6-7 include interaction term and test H2. 84 Appendix 6: Table 4. Robustness test with the V-Dem Corruption (2014 sample) DV: Anti-Sex Model 1 Model 2 Model 3 Model 4 Model 5 Model 6 Model 7 Trafficking 2019 V-Dem Corruption -1.511*** -0.530* -0.471 -0.526 -0.387 -0.108 -1.578 (0.207) (0.297) (0.366) (0.373) (0.406) (1.170) (1.559) GDPPC (log) -0.009 0.0600 0.0666 0.019 0.013 (0.079) (0.081) (0.085) (0.096) (0.099) Globalisation 0.029*** 0.025*** 0.026*** 0.027** 0.028** (0.008) (0.009) (0.009) (0.011) (0.011) Democracy 0.749** 0.775** 0.807* 0.989* (0.332) (0.335) (0.408) (0.526) Political Stability -0.171* -0.188* -0.145 -0.181 (0.102) (0.108) (0.121) (0.133) CEDAW=2 0.497 0.438 0.494 0.415 (0.394) (0.402) (0.409) (0.433) CEDAW=3 0.842** 0.822* 1.020** 0.949** (0.411) (0.423) (0.429) (0.448) CEDAW=4 0.563 0.543 0.662 0.588 (0.400) (0.410) (0.417) (0.438) Population (log) -0.016 -0.009 -0.018 (0.044) (0.051) (0.053) Share of -0.003 -0.003 -0.003 protestants in 1980 (0.003) (0.004) (0.004) Regional Fixed Yes Yes Effects WPE 1.737* -1.126 (0.892) (1.471) V-Dem Corruption -1.372 1.543 # WPE (1.435) (1.951) Constant 3.561*** 1.319** -0.102 0.171 -0.586 2.039*** 0.451 (0.120) (0.609) (0.760) (1.058) (1.328) (0.776) (1.858) R2 0.293 0.399 0.457 0.462 0.526 0.325 0.529 N 131 131 131 131 131 131 131 Notes: Standard errors in parentheses. ***p < 0.01, **p < 0.05, *p < 0.10. Models 1-5 test H1, Models 6-7 include interaction term and test H2. 85 Appendix 6: Table 5. OLS Results without time lags DV: Anti-Sex Trafficking 2019 Model 1 Model 2 Model 3 Model 4 TI Corruption -2.615*** -1.103 -2.625 -4.038 (0.313) (0.794) (1.785) (2.609) GDPPC (log) -0.025 -0.014 (0.113) (0.119) Globalisation 0.026** 0.024* (0.0121) (0.013) Democracy 0.530 0.449 (0.394) (0.513) Political Stability -0.132 -0.218 (0.132) (0.145) CEDAW=2 0.588 0.430 (0.403) (0.419) CEDAW=3 0.976** 0.831* (0.423) (0.436) CEDAW=4 0.692* 0.538 (0.412) (0.427) Population (log) 0.006 -0.016 (0.054) (0.057) Share of protestants in 1980 -0.004 -0.003 (0.004) (0.004) Regional Fixed Effects Yes Yes WPE 0.762 -1.796 (1.360) (2.182) TI Corruption # WPE 0.648 3.682 (2.155) (3.078) Constant 4.290*** 0.340 3.458*** 2.396 (0.186) (1.658) (1.180) (2.706) R2 0.351 0.526 0.381 0.534 N 131 130 131 130 Notes: Standard errors in parentheses. ***p < 0.01, **p < 0.05, *p < 0.10. 86 Appendix 6: Table 6. OLS Results excluding Share of protestants in 1980 (2014 sample) DV: Anti-Sex Model 1 Model 2 Model 3 Model 4 Trafficking 2019 TI Corruption -2.702*** -1.492** -3.555*** -6.134*** (0.279) (0.650) (1.346) (2.003) GDPPC (log) -0.067 -0.067 (0.089) (0.089) Globalisation 0.0232** 0.016 (0.009) (0.010) Democracy 0.842** 1.240*** (0.340) (0.431) Political Stability -0.160 -0.246** (0.108) (0.113) CEDAW=2 0.379 0.126 (0.355) (0.367) CEDAW=3 0.665* 0.442 (0.372) (0.379) CEDAW=4 0.305 0.070 (0.358) (0.369) Population (log) -0.024 -0.031 (0.046) (0.046) Regional Fixed Effects Yes Yes WPE -0.017 -3.727** (1.060) (1.686) TI Corruption # WPE 1.653 5.607** (1.612) (2.292) Constant 4.319*** 2.094 4.144*** 5.932*** (0.168) (1.285) (0.926) (2.044) R2 0.373 0.509 0.406 0.529 N 160 160 160 160 Notes: Standard errors in parentheses. ***p < 0.01, **p < 0.05, *p < 0.10. 87