How does public policy impact trafficking victimization?: An exact matching study in the EU

Abstract

Trafficking in human beings (THB) is a widespread, transnational issue in the European Union (EU). Member states act as source, transit, and destination countries for intra-EU trafficking, in addition to being a major destination region for external THB victims. This study presents a new dataset of THB victims observed in each EU member state per year and by type of exploitation going back as far as 2001 and employs exact matching methods to test the link between different prostitution policies and Roma secondary education attainment rates on observed THB victimization. The paper also builds off previous literature to compare how different legal prostitution models and THB supply factors are expected to influence various types of THB. The results indicate that legalized prostitution and lower educational attainment among the Roma community increase observed THB victimization, especially THB for the purpose of sexual exploitation. The paper does not find that the Swedish model significantly increases or decreases observed THB victimization. In demonstrating how matching methods can be utilized to uncover policy patterns in THB outcomes, this study provides a blueprint for how other hidden phenomena, such as corruption or migration, can be robustly and empirically tested.

1 Introduction

Trafficking in human beings (THB) represents both the persistence of modern slavery and an intensifying rise in the exploitation of vulnerable populations and individuals. Whether in its form as trafficking for the purpose of sexual exploitation (commonly referred to as “sex trafficking”), forced labor, labor exploitation, forced begging, forced criminality, forced marriage, organ trafficking, or its other types, THB poses a social welfare, security, crime, and economic policy problem. THB, as a political and legal problem, is particularly complex in the EU since THB is pervasive, exists in both transnational and internal trafficking forms, and requires member state and EU-level policy leadership and legal responses. Indeed, the European Commission (2020) noted that approximately half of all EU THB victims are EU citizens, one-third of EU victims are trafficked internally while the other two-thirds are trafficked in other member states, and THB criminal networks are increasingly sophisticated and operate across state borders (3). 7482 adult THB victims were recognized in 2018 and 11,062 adult THB victims were recognized in 2019 across twenty-three EU countries (this study). THB is also a highly lucrative illicit industry, with forced labor profits alone making $150.2 billion per year (ILO, 2014), 50% more than the value of the global arms trade (SIPRI, 2021).

THB is recognized as a primarily transnational and international issue as the phenomenon frequently involves the movement of victims across borders. However, most global THB is in fact internal and intra-regional (UNODC, 2009) with the distinction between domestic and external trafficking becoming blurred in the EU since the Schengen Area provides most EU citizens with freedom of movement throughout the bloc (European Commission, 2021n). The freedom of movement in the EU complicates not only how THB is identified, but also the effectiveness of domestic member state policy in combatting THB and protecting victims within their borders. Internal trafficking can spill over into surrounding states as potential victims are recruited for domestic exploitation but may be moved across borders for re-trafficking. In the sex market alone, upwards of 67% of external THB victims may have entered the sex market in their home country before being “re-trafficked” abroad (Kelly, 2002, 31). The transnational nature of THB and the re-trafficking phenomenon makes the movement of vulnerable individuals (even within their own state) the distinguishing feature of trafficking. The European Commission’s Directive 2011/36/EU defines THB as “the recruitment, transportation, transfer, harboring or receipt of persons, including exchange or transfer of control over that person, 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” (European Parliament and European Council 2011).

In response to the challenges of THB, prostitution models—such as legalized prostitution and the Swedish model—are designed to constrain the sex market in a way that regulates its legality and/or lowers THB for the purpose of sexual exploitation. The Swedish model (also referred to as the Nordic model) decriminalizes the act of prostitution while criminalizing the purchase of sex (Friesendorf, 2007). And legalized prostitution regulates the sex industry but legalizes the purchase and sale of sex (ibid). The Swedish model is preferred by “abolitionists” who believe prostitution is equivalent to rape and the sex industry profits from gender-based violence, while legalization is favored by advocates who believe prostitution can be liberating and that regulating the sex industry by requiring sex workers to formally register, pay taxes, and (sometimes) receive welfare benefits reduces the risk of exploitation within prostitution (ibid; Cho, 2016). Policymakers also try to address THB “push” and “pull” factors and labor market vulnerabilities to confront labor exploitation and other types of THB. THB is, more broadly, driven by numerous “push” and “pull” factors that incentivize individuals into vulnerable work and citizenship situations, and/or illicit industries. “Push” factors increase the vulnerability, and therefore subsequent dependency, of potential victims on traffickers and may hinder their ability to seek help, while “pull” factors in destination countries, such as demand for seasonal work and higher wages, encourage illegal or risky migration (Paraskevas & Brookes, 2018). Other “push” factors include social exclusion, lack of education, poverty, unemployment, conflict, political unrest, discrimination, marginalization, informal work, and migrant status (ibid; Andreas, 2008; Boyle & Shields, 2018).

Since “push” and “pull” factors are therefore regarded as the key determinants of the supply of THB victims, it is crucial to empirically test whether this is, in reality, reflected in the data on THB supply. Towards this end, this study is the first of its kind to test the causal relationship of different prostitution models and a lack of educational attainment among Roma communities via exact matching, and it is also the first to use a newly-collected dataset on the quantity of trafficking victims identified in each of the twenty-seven EU member states. This paper also attempts to surpass the normative debate over how the sex industry affects women’s autonomy and gender equality to empirically and quantitatively evaluate how each of these models affects THB victimization. In the following sections, I will review the literature on how public policy and legal models are theorized to influence, and hopefully minimize, THB before presenting hypotheses on how prostitution models and macroeconomic inequalities affect THB victimization, discussing this paper’s matching methods, and introducing my original dataset. Quantitative results are then explained before elaborating on how matching methods can further scientific inquiry into hidden economic activity, including illegal acts such as corruption and smuggling.

2 Literature review

This section is divided into two parts: one comparing prostitution models, and the other assessing how socioeconomic inequalities affect THB supply. This is helpful because the prostitution policy debate typically only addresses THB for the purpose of sexual exploitation, while socioeconomic “push” factors are related to all forms of THB and are especially important in the labor exploitation conversation. Since this paper considers causal mechanisms for both types of THB, I investigate each perspective and their connections to the “push” and “pull” factors for multiple types of THB victimization.

2.1 Prostitution policy comparison

The “Nordic” or “Swedish” model of prostitution bans prostitution while embracing gender equality values by criminalizing the purchase of sex but decriminalizing the sale of sex (Crowdhurst & Skilbrei, 2022, pp. 98–100). The policy, first implemented in Sweden with the passage of the 1999 Sex Purchase Act, was meant to be exported to other countries, influence international norms on prostitution and sex work, and abolish prostitution by limiting the sex market and changing societal attitudes towards prostitution (ibid). In one form or another, the policy has been adopted by Norway, Iceland, Canada, France, and Ireland (ibid). Regulated and legalized prostitution instead permits individuals to buy and sell sex within the regulated sex market, requires the licensure of brothels, and gives local police, inspectors, tax authorities, and the criminal justice system oversight on prostitution (Huisman & Kleemans, 2014). Even though legalization has practical elements of regulation, it is driven by a moral perspective that views prostitution as “regular labor, governed by market forces” and therefore no different from other types of legitimate, state-regulated work (Brants, 1998, p. 623). The Swedish model and legalized prostitution therefore overlap in decriminalizing the sale of sex, but diverge on perspectives of how buyers should be treated by the criminal justice system and how the state should engage with prostitution and sex work more broadly.

Scholars who support legalized prostitution argue legalization provides prostitutes legal avenues to report abuse to the police and to have dignity as official workers in a regulated profession, improves health outcomes amongst sex workers, gives brothel owners the freedom to report suspected trafficking cases, and strengthens criminal codes against forced prostitution (Zeegers & Althoff, 2015). They criticize the conflation of all prostitution with sex trafficking, which typically follows from abolitionist thinking on prostitution (ibid). These scholars accept, however, that the regulatory aspect of legalization can also have drawbacks to its own aims, such as driving smaller brothels out of the market and therefore part of the sex market into illicit and unregulated territory (Brants, 1998; Huisman & Nelen, 2014; Post et al., 2019). And they recognize that well-known cases, such as the Dürdan case in the Netherlands, demonstrate that severe forms of sexual exploitation can still occur under legalized, regulated sex markets and that a high percentage of sex workers may be coerced into the industry (Zeegers & Althoff, 2015, p. 367). Support for legalization is also indirectly advanced by criticizing weaknesses in the Swedish model.

Scholars have identified flaws in the Swedish/Nordic model, usually directly in the Swedish context. They claim that while street prostitution may have declined under the Swedish model, it may have increased in other hidden forms and indoor spaces, resulting in further vulnerabilities for prostitutes and shorter time spans negotiating with clients – although this evidence relies exclusively on interviews and media sources (Kingston & Thomas, 2019, p. 429; Levy & Jakobsson, 2014). There are also opportunities for prostitutes to be criminalized through immigration policies instead of prostitution ones, as many prostitutes in the Nordic countries are foreign and may face deportation instead of state help in leaving the sex market (Vuolajärvi, 2019). Furthermore, the criminalization of buyers may promote sex tourism into states with more relaxed regulations, such as the observed inflow of Swedish sex buyers to Latvia and Denmark (ICMPD 2009, p. 107), and the subsequent growth of prostitution demand in Norway and Denmark after Sweden adopted its namesake model in 1999 (CATWA 2017, p. 13). The efficacy of the Swedish model within a country may therefore be weakened within a state by lackadaisical implementation or countervailing immigration policies, and across entire regions by a lack of prostitution policy synchronization.

Just as there are arguments against the efficacy and moral positioning with the Swedish model, there are counterpoints to legalized sex work. Scholars who oppose legalized prostitution demonstrate that this policy can incentivize underground prostitution as sex workers wish to evade taxes (Friesendorf, 2007; Business Insider, 2019), foreign prostitutes may not have the required visa status to register with the state and receive protections (Friesendorf, 2007), organized crime related to trafficking may become more profitable and subsequently produce upticks in concealed prostitution activity (ibid), the sex market may enlarge (ibid), and only as little as 10% of sex workers may officially register (Business Insider, 2019). Street prostitution has also been shown to persist in countries with legalized prostitution, as is exemplified by the situation in Riga, Latvia (Baltic News Network, 2020). On the issue of THB victimization and forced prostitution, quantitative empirical studies have produced evidence that legalized prostitution is correlated with THB inflows (Cho et al., 2013; Jakobsson & Kotsadam, 2013) and has not led to improvements in victim protection policies (Cho, 2016). These studies therefore suggest that efforts to root out forced prostitution through legalized oversight of the sex market have not been successful in practice, and may have incentivized traffickers to move their activities to states with forms of legalization. Lastly, Huisman and Kleemans (2014), in a study on the Netherlands, find that countries with legalized prostitution continue to struggle with organized trafficking groups, the regulation of the sex market does not create the transparency necessary to report, identify, and separate sex trafficking from legal sex work; and prosecuting trafficking could potentially be more difficult under policies of legalized prostitution.

Finally, the Swedish model may be better at preventing forced prostitution. There is consistent evidence that the Swedish model has been effective at reducing street prostitution (ICMPD 2009; Socialstyrelsen, 2004), lowering the quantity of prostitutes (an indirect proxy for THB supply), and slowing or ending expansion of the sex market (Coalition Against Trafficking in Women Australia, 2017, pp. 12–16). Both France and Sweden have witnessed a decrease in prostitution demand and a less successful trafficking industry for sexual exploitation under the model (Human Thread Foundation, 2019).

2.2 Socioeconomic inequalities, education, and THB supply

While prostitution policies account for how policymakers can address the demand for prostitution, they tend to ignore the social and economic “push” factors that drive the supply of prostitution (Friesendorf, 2007). Policy interventions which address the macroeconomic inequalities driving socioeconomic “push” and “pull” factors are therefore needed to address the demand for labor exploitation and other forms of exploitation (such as forced begging and forced criminality) and the supply of THB victims. Poverty is one of the most well-cited “push” factors throughout the literature as it can be the difference between a migrant finding regular work or labor exploitation abroad. As is argued by Andreas (2008), impoverished migrants are more likely to end up as victims of forced labor than other migrants because they lack the social capital to navigate informal recruitment and seasonal employment safely. Since both legitimate and illegitimate migrant recruitment firms use informal recruitment strategies and subcontracting to hire migrants, poverty can make potential THB victims more dependent on exploitative work options and more vulnerable to recruitment by hidden THB networks; furthermore, countries with higher poverty rates are correlated with more THB victims (ibid; David et al., 2019).

Some scholars in the forced labor and THB for the purpose of labor exploitation literature, however, disagree on whether migrants’ social networks protect or endanger those at risk of labor exploitation. Indeed, not all migrants are exposed to THB for labor exploitation risks, and individuals who face labor exploitation may be citizens, regular migrants, or irregular migrants (van Meeteren & Wiering, 2019, p. 108). The picture that emerges of how “push” and “pull” factors influence THB for the purpose of labor exploitation, and THB more broadly, therefore requires a more nuanced assessment of how poverty, a lack of opportunity, social exclusion, and societal marginalization influences THB victimization. For instance, Ollus et al. (2013) found in Estonia and van Meeteren and Wiering (2019) in the Netherlands that THB for labor exploitation victims typically found their exploitative job abroad based off the recommendations of employers and bosses by their migrant social network. Yet, having a good social network is also a crucial factor in whether or not a migrant has options to leave the exploitative situation (Ollus et al., 2013). Complicating this understanding of the role of social networks in labor THB is that the goal of most victims is to earn enough money to pay off their debts to the traffickers and send remittances to their families at home (Villacampa, 2022).

Part of the reason that the scholarly image of THB for labor exploitation is murky is that it is an under-researched area of THB, it is conceptualized differently across states (Cockbain et al., 2018); and because victim service professionals tend to be trained to identify victims of sex trafficking—not labor exploitation—and to only be on the lookout for female victims when they are aware of the prevalence of labor exploitation (Villacampa, 2022). Nonetheless, forced labor can be identified as involving “physical or sexual violence or the threat of such violence, restriction of movement of the worker, debt bondage or bonded labour, withholding wages or refusing to pay the worker at all, retention of passports and identity documents, and the threat of denunciation to authorities” (Ollus et al., 2013, p. 14). And labor exploitation more generally encompasses immoral labor practices like overworking employees and underpayment or withheld wages, but also the more severe instances of forced labor and THB for labor exploitation; it can happen in legal and illicit business practices (Davies & Ollus, 2019). While knowledge of sex trafficking is widespread, THB for labor exploitation is rarely known or identified, even by police officers and judges (Ollus et al., 2013). Migrants must typically pay recruitment firms expensive sums to be moved and then employed abroad (ibid). But this implies that migrants from poor, but not extremely poor, backgrounds are those “pulled” into THB for labor exploitation; they must be wealthy enough to incur these upfront costs, but poor enough not to be able to leave the situation abroad once there (van Meeteren & Wiering, 2019).

On the THB for sexual exploitation side, women’s poverty and socioeconomic exclusion may also explain movements into sex industries outside of their home country. The prevalence of foreign prostitutes or sex workers in EU member states, particularly in western Europe, suggests that economic transition, poverty, and conflict dislocation can drive women into the foreign sex industry under conditions of “coerced” willingness (Kelly, 2002). Ethnic minorities like the Roma are particularly vulnerable to THB for the purpose of sexual exploitation or forced begging because they face an added layer of institutional discrimination that undermines the state’s willingness to prevent THB among their communities (Jovanovic, 2015). Furthermore, in the seven EU member states studied by Dimitrova et al. (2015), the Roma were found to be overrepresented in sexual exploitation victims’ data and constituted 90% of forced begging victims (14). Empowering the Roma community, especially Roma women, may therefore be an imperative THB prevention strategy, and this is therefore why this minority group is the subject of further policy exploration in this paper, along with the fact that the Roma community is the largest ethnic minority in Europe (Friesendorf, 2007, p. 389; European Parliament, 2022).

Education can be a particularly powerful means of confronting the “push” factor of modern slavery. Especially in Roma communities—which are significantly more likely to lack formal education and to rely on contributions from children’s wages and work in informal sectors (Pantea, 2006)—education can be a tool for tackling ethnic discrimination, socioeconomic exclusion from the labor market and other parts of society, and poverty (ERRC, 2014). Roma children face exclusion from formal education for a variety of reasons; structural reasons include consignment to remedial classes or institutions designated for mentally disabled students, economic reasons such as a lack of electricity and funds for school materials, and by choice to perform household duties (typically for girls) or work in informal jobs (like agriculture or construction for boys) and begging (Pantea, 2006, pp. 21–24; UNICEF, 2011, p. 19; ERRC, 2014). To put into context the gravity of the educational hurdles faced by Roma children, less than two-thirds of Roma children finish primary school (Ivanov et al., 2006, p. 29) and as little as 18% of Roma children ever enroll in secondary school (UNICEF, 2011, p. 2).

The literature therefore suggests that member states should address sources of poverty and socioeconomic inequality to reduce the supply of THB victims for all types of exploitation. While the demand for seasonal and migrant labor is unlikely to be addressed by working against poverty and social exclusion, empowering vulnerable communities that are mostly likely to engage in irregular and informal work via education lowers their potential dependency on trafficking networks and raises their social capital and regular work opportunities.

Hypotheses

Based on the literature’s prediction that the Swedish model may be better in limiting prostitution demand, and findings by Jakobsson and Kotsadam (2013) that states with legalization are correlated with greater THB inflows, I posit that legalized prostitution increases the quantity of observed sexual exploitation victims while the Swedish model decreases the quantity of observed sexual exploitation victims.

I do not expect legalized prostitution and the Swedish model to have an effect on labor or other types of exploitation (such as forced criminality and forced begging) as they are designed only to address THB driven by forced and hidden prostitution.

H₁

Legalized prostitution increases the quantity of observed sexual exploitation victims and has no effect on the quantity of labor or other types of THB exploitation victims.

H₂

The Swedish model decreases the quantity of observed sexual exploitation victims and has no effect on the quantity of labor or other types of THB exploitation victims.

Finally, the unique vulnerability of marginalized communities indicates that greater educational inclusion experienced by the Roma should lower the quantity of all types of THB victimization as this community is more likely to experience “push” and “pull” factors. The Roma community is the focus of this hypothesis due to data availability and the fact they are the largest ethnic minority in the EU (European Parliament, 2022).

H₃

Greater secondary educational attainment of the Roma community decreases the quantity of observed sexual, labor, and other exploitation THB victims.

3 Methodology and data

This study utilizes OLS regressions with exact matching to establish the independent effect of legalized prostitution, the Swedish model, and high educational attainment among the Roma community on the observed quantity of THB victims by type of exploitation. Pre-processing observational data via matching methods allows the researcher to establish the average treatment effect on the treated (ATT) of the independent variable as if a randomized experiment were performed since exact balance is achieved on observed covariates and on average balance is achieved on unobserved covariates (Ho et al., 2007; King, 2020). Matching on covariates therefore acts “as if” random assignment of the treatment occurred and eliminates parametric assumptions (Ho et al., 2007). Matching, as a blocked experiment, also removes the possibility for bias from researcher discretion, lowers standard errors, and is more robust when compared to completely randomized experiments (King, 2020). The exact matching method was chosen for this paper because it suffers less from the “curse of dimensionality” and model dependence than Mahalanobis distance matching and allows the pre-processed matches to be paired (ibid). The MatchIt software for R (Ho et al., 2011; Greifer 2021) was used to generate the OLS regression results with exact matching. The basic regression equations were as follows:

$$\begin{aligned} & Y \, exploitation \, victims = \beta_{0} + \beta_{1} X \, Intervention \\ & \quad + \beta_{2} National \, rapporteur + \beta_{3} Origin \, country \\ & \quad + \beta_{4} Transit \, country + \beta_{5} GDP \, per \, capita \\ & \quad + \beta_{6} Inflation + \beta_{7} Year \, dummy + \beta_{8} Country \, dummy + \varepsilon_{i} \\ \end{aligned}$$

Each equation was run separately with the quantity of sexual, labor, and other (which encompasses forced labor, forced begging, and miscellaneous exploitation types as outlined in Appendix A) exploitation victims as the dependent variable. Information on how member states were coded as having legalized prostitution or the Swedish model for each country-year is provided in Appendix B. Data for the a later additional socioeconomic control variable (the regional gender employment gap) were sourced from Eurostat (Eurostat, 2021b). And data for the educational attainment rates amongst the Roma were sourced from the European Union Agency for Fundamental Rights (2011).

The presence of a sufficient national rapporteur was included as a control variable because this monitoring mechanism may affect how well THB victims are identified by the state. Each member state’s type of THB (as an origin, transit, and/or destination country) is also controlled for since states may be predisposed to experience THB as a source, transit, or destination country based on their wealth and inequality levels (Frank, 2013a, 2013b). And country-specific characteristics, like GDP per capita and inflation, were included to account for how countries with more resources and a stronger economy may have greater state capacity in anti-THB efforts or may be better at detecting THB victims (Bowersox, 2018; Jakobsson & Kotsadam, 2013). Dummy variables for each year and country were included to control for time and place fixed effects. Also, age of consent was a control variable in regressions where the quantity of sexual exploitation victims was the dependent variable since varying standards of statutory rape may affect who is registered as a THB victim. And the percentage of the population that achieved tertiary education was controlled for in the macroeconomic inequality regressions as relative educational attainment may affect the magnitude of socioeconomic inequalities on THB victimization. GDP per capita and annual inflation rate data were sourced from the World Bank (2021a; 2021b), the age of consent data were sourced from the Age of Consent legal group, and some of the THB state type coding was based on Frank’s (2013a, 2013b) dataset.

Before each regression was performed, the country-year observations were exact matched on the presence of a sufficient national rapporteur, origin country, destination country, and year covariates. Balance on the covariates is depicted in Appendix D. The application of exact matching methods to a study on THB is exciting as it is the first blocked experiment within the THB literature. As such, it may be the sole study that can establish a causal link between the policy interventions of interest and observed THB victimization, particularly since Jakobsson and Kotsadam (2013) and Cho (2016) suffer from parametric assumptions, model dependence, and the limitations of the 3P, UNODC, and ILO datasets. This is also the first study to establish a causal link between various socioeconomic “push” factors and observed THB victimization. And given the widespread concerns anti-THB scholars have that current recommendations and conclusions are based in non-empirical and purely qualitative analysis, and that the illicit nature of THB and the vulnerable nature of its victims constrains available data collection, this paper takes on Cockbain and Kleemans’ (2019) and Davy’s (2016) call to provide robust scientific evidence with transparent data collection methods. Information on how each country-year observation was coded on these variables and data sources for all variables are detailed further in Appendix B.

Moreover, this study’s pairing of a new, cross-country comparative dataset with exact matching rectifies a particularly challenging problem the THB literature faces – how to apply quantitative analysis and inferential statistics to the field, and how to do so by type of trafficking victimization (Cockbain & Bowers, 2019, 13). For instance, in undertaking a systematic review of previous human trafficking literature, Cockbain et al. (2018) found that there were far more qualitative than quantitative studies on trafficking, those that were quantitative barely met basic scientific requirements, and the studies overall focused on sex trafficking to the detriment of inquiry into trafficking for the purpose of labor exploitation or domestic servitude. Villacampa et al. (2023) also note that THB studies struggle to gather consistent quantitative data, and that those who do so successfully typically focus on sex trafficking and female victims; data on men and boys who are THB victims is sometimes even discriminatorily excluded from empirical studies (161–163). In a field where only 12% of papers from Europe meet minimum scientific standards (Cockbain et al., 2018, p. 332), it is crucial to advance the scholarship forward with rigorous empirical approaches to all forms of THB.

4 New dataset: dependent variable

A new dataset that compiles the quantity of THB victims per year and by type of exploitation in each EU member state was completed for this paper using annual reports from national rapporteurs or equivalent mechanisms, and directly contacting relevant officials for the anonymized, aggregated data in instances where these reports were insufficient. The unit of observation is therefore the country-year quantity of THB victims by type of exploitation.

This data collection endeavor was important because victims’ data is only sporadically reported by the European Commission and is consistently, but not transparently and potentially inaccurately, reported by the UNODC. As the UNODC (2014) admit in their 2014 report, a high proportion of states only partially report THB data to the UNODC and their annual questionnaire only asks for statistics on the quantity of victims formally identified (18). Since the UNODC does not disclose whether countries are instructed to provide data from certain entities before others, it is unclear and impossible to independently determine how comparable victims’ data are between countries and over time in their data set. By reporting further details on data collection and sources for each member state in Appendix A, and prioritizing victim metrics reported by national rapporteurs and nationally recognized victim assistance shelters throughout the data collection process, I hope this dataset will open up opportunities for more accurate and transparent empirical studies to be conducted on THB outcomes beyond this current publication. Baseline descriptive statistics from the new dataset are listed in Appendix C.

The new dataset presented here is crucial to advancing empirical and inferential investigations of how countries’ legal approaches to prostitution affect the scale of sex and human trafficking. Prominent critics, such as Weitzer (2015), have critiqued such investigations for lacking empirical and quantitative analysis, using poor data sources (such as those provided by the European Commission and UNODC) when they are rooted in statistical analysis, and lumping estimates for different types of trafficking together (84–86). My new dataset, however, distinguishes between trafficking for the purposes of sexual, labor, and other types of exploitation, sources primary data directly from the ministries responsible for anti-trafficking measures, and allows me to undertake both an empirical and quantitatively rigorous analysis. Furthermore, Weitzer (2015) criticizes these data sources directly for relying on a variety of sources that are not directly comparable (such as data reported by the media, NGOs, and different governmental departments; 84–85). While at times this new dataset had to rely on data from NGOs (as in the case of data collection on Austria, France, and Italy), this was limited to cases where the government requested designated NGOs to collect and store the data on their behalf. In all other cases, the data was collected by contacting the ministry or department tasked with preventing and tracking human trafficking by the state and/or using and verifying these authorities’ online archives and reports.

The closest comparison to my own dataset is that created by Villacampa et al. (2023). In a very innovative approach, the authors sampled 757 professional bodies and authorities in Spain that work directly with THB victims and asked them to report in their questionnaire the number of victims they worked with in the past two years by type of THB exploitation. These authors’ data collection work is impressive; however, CITCO’s measures, instead of those recorded by Villacampa et al. (2023), were employed in my own dataset because CITCO acts as Spain’s national rapporteur, and is therefore the entity designated by the EU as responsible for official observed statistics on THB victimization.¹ Jiang and LaFree (2017) also tried to operationalize observed THB victimization using administrative data (crime and policing statistics) reported to the Eleventh United Nations Survey of Crime Trends and Operations of Criminal Justice System. While the present author has not checked the validity of that survey, it may be flawed like those conducted by the UNODC, and Jiang and LaFree (2017) do not suggest that they checked the validity of the UN survey’s reported data with the official reports produced by national police bodies. My dataset is also more robust since it required me to communicate directly with the ministries and departments responsible for THB victimization data collection, sparked a direct conversation between the policy officials and myself on which statistics are preferable in empirical studies, and uses data produced by the entities responsible for member state THB data collection under Directive 2011/36/EU. I note, however, that caution should still be applied to this new dataset. The dataset is an approximation of observed THB victimization and, no matter how legitimate the data source, relies on the reporting mechanisms of individual states. It is my hope that the new dataset presented here goes one step further in making the study of THB more empirical, scientific, quantitatively-driven, and transparent for analysis, criticism, and fruitful debate; albeit my claims can only be applied to observed THB.

In sum, by providing a new dataset that rectifies systemic errors made in data collection and sourcing by the UNODC and European Commission, scholars can more accurately analyze the effect of prostitution model type (and other policy interventions) on THB outcomes.

5 Results

Table 1² shows the OLS regression results of legalized prostitution and the Swedish prostitution model on the quantity of observed sexual, labor, and other exploitation THB victims (after exact matching was performed, as is true for all the regression results presented here). Legalized prostitution has a statistically significant, positive effect on observed sexual and labor exploitation victimization, with this prostitution policy correlated with hundreds of more THB victims in each member state per year, therefore rejecting the null first hypothesis. The Swedish prostitution model does not have a statistically significant effect on any category of observed THB victimization, and therefore I fail to reject the null hypothesis that the Swedish model has no effect on the quantity of THB victims observed.

Table 1 Effect of prostitution models on trafficking victims by type of exploitation

Table 2 displays the same regression results when a control variable for the size of the gender employment gap and the interaction effect between the prostitution model of interest in the regression and the gender employment gap. The gender employment gap measure deployed in these regression models estimates regional disparities in male and female employment at the NUTS 2 regional level within a country. Even with these additional controls, legalized prostitution has a positive, statistically significant, and even more practically significant effect on the quantity of observed sexual and labor exploitation victims. The effects from the Swedish model are not statistically significant, as well as the coefficients on the gender employment gap (GEG) and the interaction term of the GEG with legalized prostitution or the Swedish model.

Table 2 Effect of prostitution models on trafficking victims by type of exploitation, controlling for gender employment gap and interaction variable

Lastly, Table 3³ displays the regression results of high rates of Roma secondary education on the different categories of observed THB victimization. A high proportion of Roma individuals completing school has a negative and statistically significant effect on observed sexual exploitation victimization. This outcome is consistent with the predictions from the literature as the Roma community are overrepresented in sexual exploitation THB. The presence of a national rapporteur also had a negative, statistically significant effect on observed THB victimization and this effect was to the magnitude of nearly a thousand victims per year in the sexual exploitation THB categories.

Table 3 Effect of Roma educational attainment on trafficking victims by type of exploitation

5.1 Discussion

The baseline results in Tables 1 and 2 provide evidence in favor of the legalized prostitution arm of H₁ that legalized prostitution increases the quantity of observed sexual exploitation THB victims. This is consistent with previous findings by Jakobsson and Kotsadam (2013) that legalized prostitution is correlated with higher THB inflows and with Cho (2016) that legalization does not improve victim protection. The results in Table 2 lend even more credence to the statistically significant, positive correlation between legalized prostitution and a higher observed quantity of sexual exploitation victims because its models included controls for the gender employment gap (GEG) and interaction effect of the legal prostitution models and the GEG. These controls were important to include because GEG can act as a “push” factor for internal trafficking, or trafficking within a specific EU member state. Thus, by controlling for country characteristics – such as GDP per capita and inflation – which may drive trafficking to other member states, and an internal country characteristic which may foster internal displacement or desires to take risks in illicit or exploitative work conditions, the models in Table 2 should display a more unbiased effect of legalized prostitution (or the Swedish model) on the quantity of observed THB victims. And, surprisingly, Tables 1 and 2 suggest that legalized prostitution can affect observed THB for the purpose of labor exploitation in addition to THB for the purpose of sexual exploitation.

The observation that legalized prostitution can affect other types of THB besides THB for the purpose of sexual exploitation demonstrates that the transnational, professional nature of trafficking may mean that drivers of supply and demand for one type of THB may spill over into other sectors and forms of abuse. While not explicitly outlined in much of the labor exploitation literature, similar mechanisms which facilitate THB for the purpose of labor exploitation—such as predatory recruitment firms—may also be supplying potential sex trafficking victims to some member states, particularly if sex work is a recognized occupation. This warrants further study into overlaps between different forms of THB and how the scale of trafficking networks for sexual and labor exploitation within a country may produce “positive” externalities for the other form of trafficking, or how legal loopholes and institutional capacity failures within a state may help both illicit industries thrive under the radar. The evidence presented here may also reflect some noise in the data since the outcome is observed THB; i.e., some of the correlation between legalized prostitution and observed victimization for THB for the purpose of labor exploitation could be attributed to which states are better at monitoring THB victimization.

The results in Tables 1 and 2 are null for the Swedish model, and therefore I cannot make any conclusions on if the Swedish model in practice prevents, avoids, or even increases observed THB victimization. The fact that the Swedish model does not significantly increase or decrease observed THB victimization does not side with one side of the literature on this prostitution model or the other. It is, however, compatible with the literature at large, which is frequently overcome with conflicting empirical interpretations of realities on the ground in states with the Swedish model, and yet there is consensus among scholars that more nuance is needed in understanding the model’s implications (Jakobsson & Kotsadam, 2013; Kingston & Thomas, 2019; Vuolajärvi, 2019). Further empirical study on the long-term and wider-scale impact of the Swedish model on the scale of THB victimization is therefore also necessary.

In sum, the implications of the results in Tables 1, 2 and 3 suggest a potential need for amendments to legalized prostitution to provide more victim protection and reduce spillover effects into labor exploitation, a deeper analysis of the effects of the Swedish model on THB victimization, and an expansion of educational access and support to marginalized communities to reduce the supply of THB victims to traffickers. The output in Table 3 is particularly useful in both a) demonstrating the statistical and practical effect that THB supply factors can have on the scale of observed (sex) trafficking, and b) demonstrating that policymakers should place careful attention on marginalized communities at higher risk of THB. This is acknowledged by the European Parliament which, in a 2022 resolution, declared that educational opportunities and educational inclusion of Roma children is critical to improving the employability, health outcomes, and reduced segregation and discrimination of Roma residents and citizens (European Parliament, 2022). While this paper focused on how higher rates of educational attainment among the Roma community may lower THB for the purpose of sexual exploitation, future studies should evaluate if disparate educational attainment rates within EU member states and between member states is also having a practically and statistically significant effect on the size of THB victimization.

5.2 Ramifications for research on other “hidden” phenomena

The employment of matching methods in this paper can act as a guide for other comparative politics and legal studies which wish to establish causal links between relevant policy areas and hidden phenomena, such as illegal migration, corruption, or clientelism. Datasets that aggregate proxies for the hidden phenomena outcomes, or turn to non-traditional measurement sources like individual ministry reports and NGOs, to the country-year level and seek to evaluate how binary policy differences across countries influence such outcomes would benefit from considering the use of matching methods in their analysis. The combination of exact matching methods and the construction of a new dataset on a hidden topic (THB victimization) that required parsing through the transparency and accountability of data collection by government and non-state actors is therefore a blueprint for how other policy studies on hard-to-define subjects can be approached from a quantitative perspective. Particularly since natural experiments and randomized experiments are more challenging to come across or conduct with these “slippery” topics in comparative politics, larger discussions about how original datasets can be built to operationalize these features in their observed outcomes is called for.

6 Conclusion

Based on this study’s analysis, EU policymakers should consider the implications of legalizing the purchase of sex on THB when designing prostitution policies, and the consequences of educational disparities and the marginalization of certain ethnic communities when making decisions on anti-THB policies. The research also suggests EU policymakers should work across member states to prevent “underground” trafficking and establish robust national rapporteurs and referral mechanisms. Using exact matching data pre-processing with OLS regressions, a positive link is established between legalized prostitution and upticks in observed sexual and labor exploitation THB victimization. I also provide the first quantitative test of the link between “push” factors like educational attainment rates among the Roma community on observed THB victimization using a causal mechanism such as exact matching. The evidence in this last test implies that increasing Roma educational attainment can dramatically reduce THB for the purpose of sexual exploitation. Prevention efforts at the source, through education and prevention campaigns, should therefore also not be discounted in fighting THB.

In designing a novel dataset on THB victimization in all twenty-seven EU member states, this paper breaks major ground in establishing how administrative data sources from government offices and NGOs can be utilized to construct empirical databases which can estimate illicit activities and outcomes. The new dataset also offers opportunities for future studies to elucidate further inferences as to how specific victim protection, assistance, and integration projects affect latent THB victimization. The dataset goes beyond the Counter Trafficking Data Collaborative’s (CTDC) THB victims data and is more robust and transparent than the UNODC’s unreliable victims’ data. And in demonstrating that matching methods is applicable to THB research, this paper can become a model for blocked experiments on THB and other policy realms related to crime and migration. Finally, the importance of national rapporteurs in the EU are reiterated in the results section and offer hope that strengthening their offices through resource investment and EU and member state funding can tangibly change the situation of THB on the ground through evidence-based assessments and policy recommendations.

Appendices

Appendix A: THB victims data collection methodology

In an effort to rectify the lack of data on THB victims in EU member states in databases like those compiled by the UNODC and Counter Trafficking Data Collective (CTDC), I undertook the task of looking through each national rapporteur or equivalent mechanism’s online annual reports and reporting of their registered/identified victims’ metrics. In instances where the annual reports were not publicly accessible via ministry or rapporteur webpages, or there appeared to be older reports that were not online, I directly contacted the relevant officials for access to the anonymized, aggregated data on THB victims registered in the member state per year, and by type of exploitation. Following the layout of the CTDC’s global dataset, which was aggregated to the country-year level by Bowersox (Forthcoming), I collected data on the number of trafficking victims that experienced sexual exploitation, forced labour, forced labour and sexual exploitation, forced marriage, organ removal, or other types of exploitation. Detailed below are the steps I took in collecting and aggregating the THB victims’ data from each member state.

Austria

The annual reports and victims’ data were sourced from LEFÖ, a victims’ assistance NGO that is contracted out by Austria’s Ministry of the Interior to provide services to recognised trafficking victims. The total number of THB victims was provided for each year, as well as a percentage breakdown by type of exploitation. The researcher approximated the number of victims as close as possible using the percentages information, but for 2017 the rounding was slightly off. Data on forced labour victims only became available beginning with the 2014 reports. Domestic servitude was classified under the ‘other types of exploitation’ (other) category.

Belgium

The victims data were found in MYRIA’s annual reports. MYRIA is Belgium’s national rapporteur and is the state’s Federal Migration Centre. Forced criminality and forced begging were classified under ‘other’. The data on the quantity of victims registered by type of exploitation is collected by the Public Prosecutor’s office and then supplied to MYRIA. If the number of victims registered in a given year changed in later reports (since many reports listed statistics for multiple years), then the data provided in the most recent report were used and entered in the data collection process.

Bulgaria

The data on the quantity of THB victims identified in Bulgaria for the years 2013 and 2018–2020 were sourced from the National Commission for Combating Trafficking in Human Beings’ (NCCTHB) annual reports. The NCCTHB is Bulgaria’s national rapporteur. Data for the years 2014–2017 was sourced directly from Bulgaria’s webpage on the European Commission “Together Against Trafficking in Human Beings” website. The statistics listed on the European Commission webpage matched those published in later annual reports from the NCCTHB, and were most likely supplied directly to the European Commission by NCCTHB. There were two types of victims data provided in the NCCTHB’s annual reports—those supplied by the public prosecutor’s office and those supplied by ANCTHB on the number of trafficking ‘signals’ received. For consistency and to overlap with what was reported to the European Commission for the years 2014–2017, the data recorded by the public prosecutor’s office were included in this study.

Victims that were classified as having experienced domestic servitude and the selling of their newborn babies were categorised under ‘other’.

Croatia

The victims data were sent directly from the Ministry of the Interior to the researcher. If the Ministry of the Interior noted that one or more of the forced labour victims also experienced sexual exploitation, then those victims were classified under forced labour and sexual exploitation, and the same respective number of victims was subtracted from the total number of forced labour victims, thereby keeping the classifications the same across countries and the total quantity of THB victims equal to its true amount.

Victims that were listed as having experienced transit, forced criminality, THB attempts, the establishment of forced slavery, and were children of forced laborers were categorized under ‘other’.

Cyprus

The victims data were directly supplied to the researcher by the Coordinator of the Multidisciplinary Coordinating Group. The Coordinator is also an administrative officer within the Ministry of Interior. The only discrepancy is that in data supplied to the European Commission THB webpage for 2013, victims that were recorded by the Coordinator as ‘other’ in the table sent to the researcher were instead recorded in a further breakdown in categories such as ‘sexual and labour exploitation’, and ‘forced marriage’. For consistency across the years, the simple ‘other’ categorisation was used to record the 2013 victims, as was also done for all the other years covered in the data sent by the Coordinator to the author (2006–2020). The ‘other’ classification, as measured by the Coordinator/Ministry of Interior, includes forced criminality, forced marriage, illegal adoptions, cigarette smuggling, and other forms of exploitation that do not fall in the other categories.

Czech Republic

The annual reports and victims data were sourced from the Ministry of the Interior’s online archives. The data supplied from the Programme on Support and Protection of Victims of Trafficking in Human Beings (Mol) to the Ministry of the Interior were used instead of the data provided by the police because the police statistics did not disaggregate the THB victims data by type of exploitation. It should be noted that the total number of THB victims registered by the police are much higher than the quantity of victims helped by Mol each year. Thank you to Ján Sebo for assisting with translations of the Ministry of the Interior’s reports.

Denmark

The victims data were supplied by the Center Mod Menneskehandel (CMM). CMM reported their annual victims data from 2007 to 2016 to the European Commission, which were subsequently published on the Commission’s “Together Against Trafficking in Human Beings” website. Data for the years 2017–2019 were then taken directly from CMM’s annual reports.

Victims that were classified as having experienced forced criminality or unspecified/unknown types of exploitation were categorized under ‘other’.

Estonia

The Social Insurance Board keeps track of THB victims assisted each year by their services, and provides this data to the Ministry of Justice. Unfortunately, the Ministry of Justice’s annual reports only include statistics on the number of crimes and the number of victims starting in 2019. Before then, only data on the number of reported crimes (which could have had more than one victim, or no victim if it the crime was unsuccessful) are available. Additionally, the 2019 report did not disaggregate THB victims by their type of exploitation, but the Social Insurance Board confirmed in an email to the author that 99% of the victims they assist experience solely sexual exploitation, so all of the 2019 victims were recorded as sexual exploitation victims.

Finland

Pre-2014 data on the quantity of THB victims registered in Finland were taken from the 2014 report issued by the Ombudsman for Minorities (now called the Ombudsman for Non-Discrimination). The researcher recorded the number of victims approved by the National Assistance System for Victims (not just presented to it, but approved for services) each year. Police and border patrol statistics were also available, but these were not as comprehensive and only related to preliminary investigations. Furthermore, data for 2014 and later had to be sourced directly from the National Assistance System for Victims since other Ombudsman reports were not available online. Thus, it was preferable to use the pre-2014 data provided by the National Assistance System for Victims to have more comprehensive and consistent data on the whole. Thank you to Hans Nasman for helping with translations of the Ombudsman for Minorities’ and National Assistance System for Victims’ reports.

France

Victims data from France could only be sourced for the year 2019. CNCDH, France’s national rapporteur, encouraged the researcher to reach out to Contre La Traite, a Catholic NGO that is the only entity that keeps data on the quantity of victims assisted in France in any given year. The Director of Contre La Traite sent the author a link to the 2019 report, but other reports with data outside of 2019 could not be found on Contre La Traite’s webpage. Victims that experienced forced criminality and forced begging were classified under ‘other’.

Germany

Annual reports and victims data from Germany were sourced from BKA, the federal police. From the annual reports, the researcher learned that Germany only started investigating and reporting data on forced begging, forced criminality, and organ trafficking in the 2016/2017 time period. Victims that experienced forced begging and forced criminality were classified under ‘other’.

Greece

The Hellenic Police have published victims’ statistics from 2003 to 2020 on their website. The records from 2013 to 2020 include a breakdown of the quantity of victims by type of exploitation, and therefore the data from these years were included in the study. Victims that experienced forced begging were classified under ‘other’.

Hungary

Data on the quantity of victims reported from 2018 to 2021 to the victim identification service run by the Ministry of Justice (called the EKAT system) were directly provided to the author by the Ministry of the Interior’s Unit for Counter Trafficking and Horizontal Issues. Unfortunately, this data could not be used because several agencies that report to the EKAT system had not completed the victims’ application sections that asked for the type of exploitation they experienced. Therefore, only the total number of THB victims was reported and could not be used in this study’s final report. However, the total number of THB victims registered via EKAT was included in the basic descriptive statistics given in the policy brief.

Ireland

Annual reports and victims data were sourced from the Blueblindfold agency of the Department of Justice and Equality. Victims are registered and reported by An Garda Síochána, the national police, to Blueblindfold. Victims that experienced forced criminality and forced begging were classified under ‘other’, as well as those that experienced a mix of forced criminality, forced labour, and sexual exploitation. Data on victims registered with NGOs, instead of with An Garda, were available for the years 2014 and earlier, but the data on victims registered or suspected/presumed by An Garda were recorded for consistency across the years.

Italy

Victims data were given to the researcher directly from Numero Verde, the national anti-trafficking helpline that is both a part of Osservatorio Interventi Tratta and a government service under the Department of Equal Opportunities. Two separate statistics on human trafficking victims’ were provided to the researcher: one on the quantity of victims taken ‘in the charge’ of Numero Verde—implying that they went through the organisation’s programmes of emersion, assistance, and social integration—and the other on the quantity of victims that were reported and assessed under Italy’s various anti-trafficking programmes. The latter of these two measures on the number of victims recognized each year was used in this study because it aligns closer to the metrics used by other national rapporteurs. For example, the statistics from CoMensha are those reported to the organisation, not only those that were provided assistance by them (Nationaal Rapporteur Mensenhandel en Seksueel Geweld tegen Kinderen, 2020, p. 27). Additionally, it should be noted that Numero Verde reports within its statistics the victims that are assisted by other NGOs under the “Fuori Tratta” project, which is spearheaded by Cooperativa Sociale Dedalus (Campania).

Latvia

Victims data were given to the author directly from the National Anti-Trafficking Coordinator. Only the statistics on victims that were formally identified by the National Anti-Trafficking Coordinator were included in the study. Victims that experienced domestic servitude and forced criminality were classified under ‘other’.

Lithuania

Victims statistics from Lithuania were taken from the UNODC database. The Ministry of the Interior confirmed that the data reported on Lithuania by the UNODC had been directly supplied by them, and were therefore just as accurate as any secondary report they could send to the author. The UNODC data only has data up until 2018, so data on victims registered in 2019 was taken from the Ministry of the Interior’s annual report.

Luxembourg

Reports and victims data were sourced from the Commission consultative des Droits de l’Homme (CCDH). The CCDH’s reports that cover 2014–2016 and 2017–2018 were used to record victims data for the country-year observations from 2010 to 2018. Victims that experienced forced begging and forced criminality were classified under ‘other’.

Malta

The data related to THB victims by type of exploitation were directly supplied to the researcher by an official at the Ministry for Home Affairs, National Security, and Law Enforcement. Victims that experienced domestic servitude were classified under ‘other’.

Netherlands

Victims data were sourced from National Rapporteur reports. The data provided to the National Rapporteur from the Coordinator Center against Human Trafficking (CoMensha) were used instead of those supplied by the state police because only CoMensha disaggregated the victims data by type of exploitation. Victims that experienced exploitation within and outside the sex industry were classified under forced labour and sexual exploitation for the years 2012–2014. Victims that experienced unknown, forced criminality and forced organ removal, forced criminality alone, sexual exploitation and organ removal, sexual exploitation, forced labour, and forced criminality; and sexual exploitation and forced criminality were classified under ‘other’. Thank you to Frederic Ysewijn for assisting with translations of the National Rapporteur’s reports.

Poland

Annual reports and victims data were sourced from the government’s anti-trafficking portal (Handel Ludzmi). The data provided to Handel Ludzmi from KCIK, a help and consultation centre for victims, were used instead of those supplied by the general prosecutor’s office because the general prosecutor’s office data only became available beginning in 2015 and the author relied on statistics reported by victims assistance centres for many of the member states. The KCIK data for the years 2012 and 2013 could not be used because they were not disaggregated by type of exploitation. Victims that experienced forced begging, domestic slavery, treatment that degrades human dignity, and violence were classified under ‘other’.

Portugal

Annual reports and victims data were sourced from the Observatory of Human Trafficking. From these reports, data covering the years 2009–2019 were discovered. For the years 2009 and 2010, statistics on the number of “signals” and “confirmations” had to be aggregated by hand by the researcher, in order to keep the data consistent with the later reports. Victims that experienced forced labour and forced criminality were classified under forced labour. Victims that experienced forced begging, forced criminality, forced adoption, forced surrogacy, slavery, and other were classified under ‘other’.

Romania

Annual reports and victims data from the national referral mechanism were sourced from the Ministry of the Interior’s National Agency against Human Trafficking. Victims of pornography were classified under sexual exploitation. Victims that experienced forced begging, forced thievery, and other types of exploitation were classified under ‘other’. The 2012 annual report had data on the number of victims identified and the number actually trafficked in that calendar year; the statistics on the number of victims identified were recorded for consistency across the annual reports. It should also be noted that the National Agency against Human Trafficking automatically includes minors in their statistics.

Slovakia

The victims data were sourced from the Ministry of the Interior and Information Centre for Combating Trafficking in Human Beings and for Crime Prevention. Victims that experienced forced labour and forced marriage were classified under forced marriage. Victims that experienced forced begging, economic exploitation, forced criminality, and slavery were classified under ‘other’.

Slovenia

The annual reports and victims data were provided by the National Working Group for Combating Trafficking in Human Beings. The Slovenian police reported the statistics on victims to the working group. Victims that experienced servitude, forced criminality, and forced begging were classified under the ‘other’.

Spain

The online Ministry of the Interior reports on human trafficking were consulted for the victims data. Victims that experienced forced criminality and forced begging were classified under ‘other’.

Sweden

Sweden’s Gender Equality Agency (Jämställdhets Myndigheten) was consulted for victims data. The authority provided the author with infographics displaying statistics on the suspected cases of victims identified in 2018 and 2019, broken down by sexual exploitation, labor exploitation, forced begging, and forced crime. The total number of victims identified by type of exploitation did not equal the summation of the victims identified by type of exploitation; the difference between these two numbers was the quantity of victims whose type of exploitation was unknown. Victims that experienced forced begging, forced criminality, and unknown types of exploitation were categorized as ‘other.’

Appendix B: Coding methodology

2.1 Sufficient national rapporteur

In order to code whether or not a member state had a sufficient national rapporteur, public information on the European Commission’s “Together Against Human Trafficking” website and periodic monitoring reports by the Council of Europe’s GRETA taskforce were evaluated in addition to outside information on the listed “national rapporteur” or “equivalent mechanism’s” websites. Below are details on how each of the twenty-seven member states were given their sufficient national rapporteur classification. Those that were considered sufficient national rapporteurs (or borderline sufficient) were coded a “1” for the binary national rapporteur variable for all years where the national rapporteur existed, whereas those that were insufficient national rapporteurs (or borderline insufficient) were coded a “0” for the binary national rapporteur variable.

Austria

Insufficient national rapporteur. No national rapporteur is listed on the European Commission’s webpage and GRETA recognizes that Austria has a National Coordinator, but no separate national rapporteur (GRETA, 2020b, p. 11). The author had to gather annual victims’ data reports from LEFÖ, a victims’ assistance NGO, in order to have a source on THB monitoring for Austria.

Belgium

Sufficient national rapporteur. MYRIA, the Federal Migration Centre is fully recognized by the European Commission and GRETA, and the author found no reason that MYRIA should not be considered in full national rapporteur compliance. MYRIA is the national rapporteur along with the Interdepartmental Coordination Unit, both of which were established as national rapporteurs in 2014 but were defined by Royal Decree as the national rapporteur going all the way back to 2004 under the joint umbrella of the Centre for Equal Opportunities and Opposition to Racism, and fulfilled monitoring roles back to 1996 (Myria, 2021; GRETA, 2017e, p. 6; European Commission, 2021b). Belgium is therefore considered to have a sufficient national rapporteur since 1996.

Bulgaria

Sufficient national rapporteur. The National Commission for Combating Trafficking in Human Beings (NCCTHB) is fully recognized by the European Commission and GRETA. However, GRETA notes that it could be made more independent, although this is not a sufficient reason to dock its merits for the purpose of this study (GRETA, 2016a, p. 12). The NCCTHB was established as national rapporteur in 2004 under the Combating Trafficking in Human Beings Act (European Commission, 2021a; NCCTHB 2020).

Croatia

Borderline insufficient national rapporteur. The National Coordinator is listed as the equivalent mechanism on the European Commission webpage. However, GRETA notes in its third evaluation report that Croatia does not yet have a national rapporteur, and they encourage them to appoint one and the Croatian authorities indicated they plan to (GRETA, 2020e, p. 11). The author coded Croatia as a borderline insufficient case because the Office for Human Rights and Rights of National Minorities was able to supply data on THB victims.

Cyprus

Insufficient national rapporteur. There is conflicting information available on Cyprus. The Multidisciplinary Coordinator Group is listed as the equivalent mechanism on the European Commission webpage, but GRETA notes the Ombudsman acts as the external evaluator, which they consider to be the equivalent mechanism (GRETA, 2020d, p. 10). GRETA also noted in the most recent 2020 report that Cyprus still needs to amend Law 60(I)/2014 for the Ombudsman to be effective in this position (ibid). It was difficult to source publicly available information on either the Multidisciplinary Coordinator Group (MCG) or the Ombudsman’s work to counter human trafficking. In the end, an NGO that sits on the MCG had to be contacted for information on the Office of Combating Trafficking in Human Beings of the Cyprus Police and National Coordinator. Given how disjointed Cyprus’s approach to THB is, considering that there is a separate external evaluator, the MCG oversees national action plans, and there are separate roles for the National Coordinator and Cyprus Police; plus the fact that the external evaluator has not been granted all the powers necessary, Cyprus is classified as having an insufficient national rapporteur (GRETA, 2020d).

Czech Republic

Sufficient national rapporteur. GRETA designates the Head of the Crime Prevention Department of the Ministry of the Interior as the national rapporteur. The national rapporteur was established in 2003 (The Ministry of the Interior of the Czech Republic, 2021). GRETA notes that it should be made to be an independent monitor separate from the National Coordinator, but Czech authorities replied that the combined functions and dependent nature of the position allowed for better coordination and facilitated insights across the anti-trafficking agenda (GRETA, 2020a, p. 23). As with Bulgaria, the national rapporteur’s dependent nature is not enough of a warrant to dock its merits.

Denmark

Sufficient national rapporteur. The CMM is listed as an equivalent mechanism on the European Commission webpage. GRETA criticises that the CMM shares its responsibilities with the National Investigation Centre of the National Police (NEC) and that it is not a fully independent monitor, but these are insufficient reasons to dock its merits for the purpose of this study (GRETA, 2021, p. 9). The CMM was established in 2007 (CMM, 2021).

Estonia

Sufficient national rapporteur. The National Coordinator in the Ministry of Justice is the equivalent mechanism. GRETA notes that the National Coordinator is not a fully independent monitor, but again this is an insufficient reason to disqualify the equivalent mechanism as a sufficient national rapporteur (GRETA, 2018c, p. 21). The researcher’s own email interactions with the Ministry of Justice confirm this coding classification. The National Coordinator position was created in 2006 (European Commission, 2021c).

Finland

Sufficient national rapporteur. The Non-Discrimination Ombudsman acts as the national rapporteur. GRETA recognizes the Non-Discrimination Ombudsman as such, and the author found no reason that the Non-Discrimination Ombudsman should not be considered in full national rapporteur compliance (GRETA, 2019b, p. 11). The national rapporteur position was created in 2009, and so Finland is coded as having an insufficient national rapporteur in the years leading up to 2008 (European Commission, 2021d).

France

Borderline insufficient national rapporteur. The European Commission and GRETA recognise the National Consultative Commission on Human Rights (CNCDH) as the national rapporteur. However, the CNCDH relies on Contre La Traite to report victims data and does not have an established national referral mechanism for reporting victims. Because of these two shortcomings combined, CNCDH is classified as a borderline insufficient national rapporteur.

Germany

Insufficient national rapporteur. Both the European Commission and GRETA note that Germany has yet to establish a national rapporteur. In the meantime, BKA report data related to THB crimes, investigations, and victims.

Greece

Sufficient national rapporteur. The Office of the National Rapporteur within the Ministry of Foreign Affairs is the national rapporteur and is recognised by the European Commission and GRETA. GRETA notes that the Office includes three designated civil servants that work on THB issues and monitoring, but the Office needs further financial and human resources investments to be completely effective (GRETA, 2017d, p. 21). Despite the resource needs of the national rapporteur, it is seen as sufficient for the purpose of this study. The National Rapporteur position was established by a ministerial decision in 2013, and therefore Greece is coded as having an insufficient national rapporteur before that year (European Commission, 2021e).

Hungary

Borderline sufficient national rapporteur. The Anti-Trafficking Coordinator within the Ministry of the Interior is the national rapporteur and is recognized by the European Commission, and this designation was confirmed via email correspondence by the author. GRETA’s most recent report insinuates that Hungary does not have a national rapporteur, but the European Commission website and communication directly with the Ministry of the Interior refutes this (GRETA, 2019d, p. 54). But, the author still considers Hungary to be a ‘borderline’ sufficient national rapporteur because of GRETA’s lack of recognition. The Anti-Trafficking Coordinator position was created in 2008 (European Commission, 2021f).

Ireland

Borderline insufficient national rapporteur. The Anti-Human Trafficking Unit (AHTU) in the Department of Justice and Equality is listed as the equivalent mechanism on the European Commission webpage. In GRETA’s most recent evaluation report in 2017, they noted that Ireland did not have a national rapporteur or equivalent mechanism (GRETA, 2017c, p. 11). And while the European Commission notes that AHTU was established in 2008, it is unclear if it was/is immediately recognised as an equivalent mechanism. Since the ways in which AHTU fulfils its mandate outside of the annual reports on the Blueblindfold website are unclear, Ireland is seen as not having a sufficient national rapporteur. It is coded as ‘borderline’ insufficient because there is cooperation between the An Garda Síochána (national police force), Anti-Human Trafficking Team in the Health Service, and AHTU—at least according to the information on the European Commission webpage. This information could not be confirmed by the author (European Commission 2021m).

Italy

Insufficient national rapporteur. The European Commission webpage lists the Department of Equal Opportunities as the equivalent mechanism, but it is not recognised by GRETA, the U.S. State Department TIP reports, and the Department failed to respond to the author’s requests for data and communication (GRETA, 2019a, p. 12; U.S. Department of State, 2020). Due to all of these factors, the Department of Equal Opportunities is not a sufficient national rapporteur.

Latvia

Borderline insufficient national rapporteur. Latvia has a National Anti-Trafficking Coordinator, but not a national rapporteur. This was reiterated in GRETA’s most recent report (GRETA, 2017b, p. 9). Latvia is designated a ‘borderline’ insufficient national rapporteur because the National Coordinator was able to supply victims data when requested.

Lithuania

Sufficient national rapporteur. The Ministry of the Interior is the national rapporteur. It is recognized as such by the European Commission and GRETA, and this designation was reconfirmed in communication directly with the Ministry by the author. While a National Coordinator position within the Ministry of the Interior has existed since 2007 (European Commission, 2021g), the national rapporteur functions were not finalized until Resolution No. 1161 was passed in 2016 (and therefore Lithuania was coded as having an insufficient national rapporteur before 2016) (GRETA, 2019c).

Luxembourg

Sufficient national rapporteur. The National Human Rights Commission (CCDH) is the national rapporteur. It is recognized as such by the European Commission and GRETA, with GRETA even noting that CCDH is a fully independent monitor (GRETA, 2018f, p. 8). However, GRETA notes the institution could work to have fewer gaps in its data collection, but this is an insufficient reason to disqualify the merits of the national rapporteur for the purpose of this study (ibid, 11). The CCDH was established as the national rapporteur in 2014, and therefore Luxembourg is coded as having an insufficient national rapporteur in earlier years (ibid, 8).

Malta

Insufficient national rapporteur. The Human Trafficking Monitoring Committee is listed on the European Commission webpage as the equivalent mechanism. GRETA does not fully recognise this body as a national rapporteur, and the author had to reach out to the Ministry for Home Affairs, National Security and Law Enforcement instead for victims’ data. Furthermore, GRETA notes in its most recent report on Malta that the Committee and the National Anti-trafficking Coordinator positions have been vacant since 2013 (GRETA, 2017a, p. 8). There is no publicly available information on if this monitoring committee even exists.

Netherlands

Sufficient national rapporteur. The Office of the National Rapporteur is recognised by the European Commission and GRETA. GRETA notes that the national rapporteur needs to publish data on a quicker timetable and that the Office may need more financial and human resources investment in order to produce timely research, but these are insufficient reasons to disqualify the merits of the national rapporteur for the purpose of this study (GRETA, 2018e, pp. 14–15). The Office of the National Rapporteur was created in 2000 (European Commission, 2021h).

Poland

Sufficient national rapporteur. The Inter-Ministerial Committee for Combating and Preventing Trafficking in Human Beings is recognised by the European Commission and GRETA. GRETA notes that Polish authorities should review its effectiveness at critically evaluating state institutions because it is not a fully independent monitor, but this is an insufficient reason to disqualify its merits for the purpose of this study (GRETA 2017f, pp. 8–9). The resources separately provided by Handel Ludzmi, which appears to be a cooperation of multiple government agencies, is comprehensive.

Portugal

Sufficient national rapporteur. The Office of the National Rapporteur is recognised by GRETA and the European Commission. Between the Office of the National Rapporteur and the Observatory on Trafficking in Human Beings (OTSH) that is attached to the government, Portugal provides comprehensive annual reports and is deemed a sufficient national rapporteur. OTSH confirmed that Mr. Albano is the current national rapporteur under the Commission for Gender Equality (CIG). CIG was established as the national rapporteur beginning in 2008 (European Commission, 2021i).

Romania

Sufficient national rapporteur. The National Agency against Trafficking in Human Beings is the national rapporteur. It is recognised by the European Commission and GRETA as such. However, GRETA notes that Romanian authorities should review its effectiveness at critically evaluating state institutions because it is not a fully independent monitor, but this is an insufficient reason to disqualify its merits for the purpose of this study (GRETA, 2016b, p. 8). Reports are thorough and comprehensive. The National Agency against Trafficking in Human Beings was granted the duties of a national rapporteur under Government Decision no. 460/2011 (European Commission, 2021j).

Slovakia

Sufficient national rapporteur. The Information Centre for Combating Trafficking in Human Beings and for Crime Prevention is the national rapporteur. It is recognised as such by the European Commission and GRETA, although GRETA notes that it is not a fully independent monitor (GRETA, 2020c, p. 8). As with other member states evaluated, this issue of independent/dependent monitor is outside the scope of this study so Slovakia is still coded as having a sufficient national rapporteur. The Information Centre for Combating Trafficking in Human Beings and for Crime Prevention was created in 2009 (European Commission, 2021k).

Slovenia

Sufficient national rapporteur. The National Anti-Trafficking Coordinator and Anti-Trafficking Service in the Ministry of the Interior act as the national rapporteur. They are recognized as equivalent mechanisms by the European Commission and GRETA, although GRETA notes that it is not a fully independent monitor (GRETA, 2018a, p. 9). As with other member states evaluated, this issue of independent/dependent monitor is outside the scope of this study so Slovenia is still coded as having a sufficient national rapporteur. The National Anti-Trafficking Coordinator position was created in 2002 (European Commission, 2021l).

Spain

Borderline sufficient national rapporteur. There is shared competence between the Private Office of the Secretary of State for Security and the Centre for Intelligence against Terrorism and Organised Crime (CITCO) as the national rapporteur. Both are recognised by the European Commission and GRETA as the joint national rapporteur. GRETA also notes that further cooperation is needed between these entities and the national and regional authorities, and the national rapporteur should be an independent monitor (GRETA, 2018d, 11). Interestingly, the Ministry of the Interior compiles data on anti-THB activities and THB victims, but the reports were sent to the author by CITCO. Spain is therefore a borderline sufficient case because there seems to be a disjunction in which agencies fulfil which activities and criteria of the national rapporteur which clouds cohesive action against THB. Both the Private Office of the Secretary of State for Security and CITCO were established as national rapporteurs in 2014 (ibid, 9).

Sweden

Borderline sufficient national rapporteur. The Swedish Police Authority is the official national rapporteur in Sweden. The Swedish Police Authority is recognised as the national rapporteur by the European Commission and GRETA, but, as with many other member states, GRETA recommends that an independent institution oversee THB monitoring (GRETA, 2018b, p. 10). The Gender Equality Agency acts as the National Coordinator, although the authority could not provide data at the time of data collection because new reports in line with GDPR are set to be released at the end of summer 2021. Data instead had to be sourced through the National Support Program (NSP), which is run by the Swedish Civil Society Platform against Human Trafficking. While the author cannot accurately judge the data collection methods of the Gender Equality Agency since data were not provided by them, this plus the dependence between institutions are not enough to code Sweden as an insufficient national rapporteur. It is therefore regarded as borderline sufficient. The Swedish Police Authority was established as the national rapporteur in 1998 (ABbilgi Merkezl 2021).

2.2 Prostitution model

Recent periodicals and NGO publications were sourced in determining each member state’s prostitution models and laws. The Nordic model, as was described in above, criminalises the purchase of sex, but decriminalises the sale of sex. France, Ireland, and Sweden have adopted the Nordic model, which was substantiated by Nordic Model Now (2021). Complete criminalisation goes one step further than the Nordic model because it criminalises both the purchase and sale of sex, and therefore punishes prostitutes/sex workers in addition to sex buyers. Romania and Lithuania have adopted complete criminalisation, which was confirmed by Danna (2014, p. 12). Unregulated legalisation implies that member states have adopted prostitution policies that in practice have legalised prostitution to a certain extent. There may still be laws that outlaw running a brothel, soliciting or purchasing sex in public, and/or engaging in prostitution with a trafficking victim, but the state does not pursue criminal charges against those who buy sex. Most member states fall into this third category, including Belgium, Bulgaria, Croatia, Cyprus, the Czech Republic, Denmark, Estonia, Finland, Italy, Luxembourg, Malta, Poland, Portugal, Slovakia, Slovenia, and Spain. This classification for these countries was substantiated by Danna (2014), Carabott (2019), Clarke and Østerby (2016), Kathimerini Cyprus (2020), Radačić and Pajnik (2018), McCarthy (2018) and Reinboth and Woolley (2014). Finally, legalised prostitution is both legal (the purchase and sale of sex), and it is regulated by the state with sex workers entering an official registry. Austria, Germany, Greece, Hungary, Latvia, and the Netherlands have legalized prostitution and this was confirmed by Danna (2014) and McCarthy (2018).

Prostitution was legalised in Germany in 2002 (Business Insider, 2019), the Netherlands in 2000 (ibid), Greece in 1999 (NSWP, 2017), Hungary in 1999 (ICRSE, 2020), Latvia in 2008 (Republic of Latvia Cabinet, 2008), and since at least the 1970s in Austria, although prostitution policies are set and controlled at the Land level (Wagenaar et al., 2017). The Nordic model was adopted by Sweden in 1999 (Ekberg, 2004), France in 2014 (O’Sullivan, 2013), and Ireland in 2017 (Mulvihill, 2017).

2.3 Regional gender employment gap

Data on the regional gender employment gap in EU member states at the NUTS level two were sourced from Eurostat. The median regional gender employment gap across the dataset was 21.2% (author’s assessment). A member state with a regional gender employment gap level above 21.2% was therefore classified as having a high gap, whereas any member state below 21.2% was classified as having a low gap. (Eurostat, 2021b).

2.4 Regional employment disparity

Data on the regional employment disparity in EU member states at the NUTS level two were sourced from Eurostat. The median regional employment disparity across the dataset was 4.35% (author’s assessment). A member state with a regional employment disparity level above 4.35% was therefore classified as having a high disparity, whereas any member state below 4.35% was classified as having a low disparity. (Eurostat, 2021a).

2.5 Social exclusion risk

Data on the social exclusion risk in EU member states were sourced from Eurostat. Social exclusion risk is defined by Eurostat as “the sum of persons who are: at risk of poverty after social transfers, severely materially deprived or living in households with very low work intensity (Eurostat, 2021c).” The median social exclusion risk across the dataset was 20.8% (author’s assessment). A member state with a social exclusion risk rate above 20.8% was therefore classified as having a high risk rate, whereas any member state below 20.8% was classified as having a low risk rate. (Eurostat, 2021c).

2.6 Educational attainment by the Roma community

Data on the percentage of the Roma community in a given EU member state that finished school after sixteen years-of-age or are still in education were taken from the European Union Agency for Fundamental Rights. The median educational attainment rate in the Roma community across the dataset was 24% (author’s assessment). A member state with an educational attainment rate by the Roma higher than 24% was therefore classified as having high educational attainment amongst the Roma community, whereas any member state below 24% was classified as having a low educational attainment amongst the Roma community. Since this data were only provided for one year (2011), this same educational attainment rate was applied across the years for each member state as the unit of analysis was country-year. (European Union Agency for Fundamental Rights, 2011).

2.7 THB country type

Whether a member state was classified as a destination, transit, and/or origin (sometimes also called ‘source’ in the THB literature) country across its country-year observations for the matching and control variables depended on how officials within the member state had described it to the European Commission. The European Commission’s anti-THB webpage for each member state almost always described the member state with these labels. In the few cases where the European Commission webpages left it ambiguous or unclear if the member state was a transit country, the researcher looked to Frank’s (2013a, 2013b) dataset and whether or not he coded the state as a transit country. Based on Frank’s (2013a, 2013b) dataset, Denmark, Italy, Ireland, and Germany were coded as transit states, and Malta was not.

2.8 Percentage of the population with tertiary education

Eurostat’s data on the percentage of each member state that has attained tertiary education were utilised in this study. (Eurostat, 2021d).

2.9 Age of consent

The age of consent data were sourced from the Age of Consent legal group. (Age of Consent, 2021).

2.10 GDP per capita

Annual GDP per capita data in terms of constant 2010 USD were sourced from the World Bank’s Development Indicators database. (World Bank Development Indicators, 2021a).

2.11 Inflation

Annual inflation rates data (as GDP deflator) were sourced from the World Bank’s Development Indicators database. (World Bank Development Indicators, 2021b).

2.12 3P overall score

The 3P Index was developed by Seo-Young Cho, Axel Dreher, and Eric Neumayer and released in 2010 and 2015. The 3P Index ranks each state on a scale from 1 to 5 on their compliance level for trafficking prevention, victim protection, and trafficker prosecution. The aggregate 3P score is an addition of each member state’s individual score on prevention, protection, and prosecution metrics. Additional regression equations with aggregate 3P scores as the main independent variable of interest were run with the UNDOC data and included in Appendix I (Cho et al., 2014).

Appendix C: Descriptive statistics table

Variable	N	N = 249
Labor exploitation victims	227
Median (IQR)		16 (4, 58)
Mean (SD)		49 (78)
Unknown		22
Sexual exploitation victims	236
Median (IQR)		30 (10, 181)
Mean (SD)		164 (274)
Unknown		13
Other exploitation victims	203
Median (IQR)		7 (1, 32)
Mean (SD)		60 (260)
Unknown		46
Age of consent	244
Median (IQR)		15.00 (15.00, 16.00)
Mean (SD)		15.46 (1.17)
Unknown		5
Gender employment gap	120
Median (IQR)		21 (15, 29)
Mean (SD)		24 (12)
Unknown		129
Percent of the population with tertiary education	196
Median (IQR)		27 (21, 33)
Mean (SD)		27 (7)
Unknown		53
GDP per capita	239
Median (IQR)		28,874 (20,013, 47,020)
Mean (SD)		34,837 (21,363)
Unknown		10
Inflation	239
Median (IQR)		1.62 (0.77, 2.43)
Mean (SD)		1.67 (1.55)
Unknown		10
Proportion of Roma community that finished education	85
Median (IQR)		24 (10, 41)
Mean (SD)		24 (15)
Unknown		164
Total quantity of sexual exploitation, labor exploitation, other exploitation, organ removal, and forced marriage victims together	131
Median (IQR)		52 (22, 326)
Mean (SD)		284 (566)
Unknown		118

Appendix D: Matching and balancing on covariates

To determine whether or not the exact matching performed has balance and avoids statistical bias, I evaluated QQ plots and summaries of the absolute standardized mean difference on matching covariates. If good matching has been performed, then the observations should cluster around the diagonal line in the QQ plots and the covariates’ absolute standardized mean difference should go to zero in the summary plots. As can be seen in the figures below, the exact matching was successful.

See the Figs. 1, 2, 3, 4, 5 and 6.

Fig. 1

QQ plots—matching on national rapporteur, origin country, destination country, and year for legalized prostitution regressions

Fig. 2

Summary plot—matching on national rapporteur, origin country, destination country, and year for legalized prostitution regressions

Fig. 3

QQ plots—matching on national rapporteur, origin country, destination country, and year for Swedish prostitution model regressions

Fig. 4

Summary plot—matching on national rapporteur, origin country, destination country, and year for Swedish prostitution model regressions

Fig. 5

QQ plots—matching on national rapporteur, origin country, destination country, and year for Roma educational attainment regressions

Fig. 6

Summary plot—matching on national rapporteur, origin country, destination country, and year for Roma educational attainment regressions