International Human Trafficking: Measuring Clandestinity by the Structural Equation Approach

Worldwide human trafficking (HT) is the third most often registered international criminal activity, ranked only after drug and weapon trafficking. The aim of the paper is to measure the extent of HT inflows to destination countries. It proposes the application of the Multiple Indicators Multiple Causes (MIMIC) structural equation model in order to include potential causes and indicators in one model and generate an index of the intensity of HT in destination countries. Thus, we account for the unobservable nature of the crime as well as for visible aspects that both shape the extent of it.By including both dimensions of the trafficking process the model is applied over a period of ten years. The resulting measure orders 142 countries between 2000 and 2010 according to their potential of being a destination country based on characteristics of the trafficking process. The results are that OECD countries are the most likely destination countries while developing countries are less likely.


Introduction
Since human trafficking is the third largest kind of illicit international commerce, after illegal drug and weapon smuggling (U.S.Department of State, 2004), it creates an underground economy of illegal labor markets and businesses where immense profits and great suffering go hand in hand.Profit estimates range from 1 billion dollars (Belser, 2005) to 31.61 billion dollars at any given time (ILO, 2005; Interpol, 2012).This money is augmented by tax evasion and presumably used to finance the illegal businesses that traffic individuals, as well as other associated activities.The trafficked are abused through exploitation and coercion and deprived of the freedom to move or choose their place of living (Gallagher, 2009).Like other transnational criminal activities, it links with the corruption of civil society as it bypasses borders, undermines state sovereignty, and threatens state governance and human security (Shelley, 1999).
The international nature of this crime means that an international response is needed in order to address policy approaches and legal measures successfully, particularly due to the fact that the main problem in this field is the availability of comparable data.A comparable international measure of trafficking intensity is currently still missing.This article provides a first attempt at providing the literature with a new means of measuring human trafficking by using the structural equation approach. 1he Multiple Indicators Multiple Causes (MIMIC) model is a special case of the structural equation model which uses existing data to derive a measurement of the extent of human trafficking.Thereby an index of human trafficking "based on estimated parameters that relate directly to the causes and indicators" (Dreher, Kotsogiannis, &  McCorriston, 2007, p. 445) can be drawn up.This approach has been used in economics by several authors to explore latent phenomena such as the shadow economy (Schneider & Enste 2000, 2002) or corruption (Dreher  et al., 2007).
In this article we understand human trafficking in accordance with the international definition of trafficking in persons. 2 This source is not only the first successful international agreement on the common elements and implications of human trafficking, but it also provides an important working basis for the present research. 3he main elements are the inclusion of all forms of enslavement and the focus on the exploitation of victims through coercion or deception.They acknowledge that most of the victims of human trafficking are made vulnerable by migration, are not willingly enslaved, and that it is a clandestine business of internationally active criminal networks.To render the clandestine phenomenon of human trafficking visible, the objective of this article is to measure human trafficking by addressing the extent of victim exploitation in destination countries based on observed causes and indicators through the MIMIC structural equation approach.This allows us to explore the structural relationship between the causes and indicators of human trafficking, which to the best of our knowledge has not been investigated before.
In order to describe human trafficking as precisely as possible, aspects of both indicators and causes must be included, particularly since these have been neglected by earlier studies, which used multivariate approaches and focussed on factors that may cause human trafficking (e.g.Cho, 2015).The main idea behind the MIMIC model is to examine the relationship between an unobservable variable, e.g. the shadow economy, corruption, human trafficking, etc., and a set of observable variables (causes and indicators) using covariance information (Buehn &  Schneider, 2012). 4The flexibility in estimating the correlations of observable factors is one of the main advantages of the MIMIC approach.We disentangle these relationships and derive an index of the extent of trafficking in persons to destination countries by applying this single latent variable structural equation method.This method provides a detailed analysis of human trafficking, which sheds further light on the mechanism behind the trafficking process.
To summarize in advance the essential findings, the MIMIC estimates support the main assumptions as to the determinants and indicators of trafficking in persons.That is, specifically, that richer countries, investment relations, and opportunities for low-skilled labor positively correlate with human trafficking whereas language differences are negatively correlated.Indications of human trafficking inflows are the crime rate, legal measures against human trafficking, and the number of migrants registered in the countries.The human trafficking intensity index shows the prevalence of human trafficking to 142 destination countries for the period 2000-2010.
Section 2 explains the MIMIC model of human trafficking and presents the dynamics between indicators and causes of human trafficking.Thereafter, Section 3 discusses the results, the measurement of human trafficking, and country rankings.Section 4 concludes.
There are several reasons for the application of the MIMIC model in the context of human trafficking.Firstly, human trafficking is an economically significant criminal activity with huge profits linked to tax evasion.Secondly, international human trafficking receives increased attention in the global policy arena and the international community is willing to fight it.This has already spurred an increase in studies analyzing the underlying processes in law, political science, and economics.However, international trafficking in human beings is a multidimensional, unobserved phenomenon in which the whole process happens in the underground economy and neither traffickers nor victims are easy to identify.A latent variable approach such as the MIMIC is thus well suited to ad-Nations Office on Drugs and Crime (UNODC) initiated the program against Trafficking in Human Beings (GPAT); and the U.S. State Department publishes the yearly Trafficking in Persons reports comparing countries' legal responses towards human trafficking. 2The definition is presented in the Palermo Protocol Article 3 (UN, 2000). 3Article 3 of the protocol states that "human trafficking is the recruitment, transportation, transfer, harboring or receipt of persons, by means of the threat or use of force or other forms of coercion, of abduction, of fraud, of deception, of the abuse of power or of a position of vulnerability or of the giving or receiving of payments or benefits to achieve the consent of a person having control over another person, for the purpose of exploitation.Exploitation shall include, at a minimum, the exploitation or the prostitution of others or other forms of sexual exploitation, forced labor or services, slavery or practices similar to slavery, servitude or the removal of organs" (United Nations, 2000). 4See Appendix for a detailed description of the MIMIC model and the generation of the factors score for the final index. 5Additionally, Di Tommaso, Raiser and Weeks (2007) and Kuklys (2004) analyzed institutional change in Eastern Europe and welfare measurement.Buehn  and Eichler (2009) explored the connection between smuggling illegal and legal goods, and Buehn and Farzanegan (2013) developed an index of global air pollution.
Social Inclusion, 2017, Volume 5, Issue 2, Pages 39-58 dress its unobservable nature.Through the simultaneous consideration of key determinants and indicators, light is shed on the presence and magnitude of human trafficking to a country.Thirdly, the ability to estimate the parameters of a single structural equation has greater value than estimating numerous regressions.The MIMIC approach is based on the assumption that causal factors of latent phenomena are not to be considered independently.Human trafficking is a process with many facets in which several factors shape the incentive structure of all the actors involved, i.e. traffickers, victims and governments. 6he application of the MIMIC model to human trafficking focusses on the extent of human trafficking in destination countries. 7We identify what drives people to exploit vulnerable individuals, i.e. the demand structure (pull factors), and what puts people in this vulnerable position, i.e. the potential supply (push factors).In economic terms, human trafficking is located in a market setting where supply and demand are met at the expense of vulnerable individuals.The main reason for the abuse and exploitation of people is global income disparity.Emigration is propelled by economic factors that drive people to migrate and take risks in order to find more prosperous living conditions.In particular, traffickers use their victims' vulnerability and bring them to countries where both the demand for cheap labor and exploitation profits are high.These are the key factors that help identify the indicators and causes of human trafficking to destination countries.

Indicators
The extent of human trafficking is not directly measurable so indicators have to be identified that are a function of human trafficking in destination countries.To the best of our knowledge we are the first to determine multiple indicators to measure its extent in destination countries.There are many aspects correlated with human trafficking that could partially indicate its prevalence in a country.After extensive research of anecdotal and governmental evidence (e.g.U.S. Department of State, 2010,  2011, 2012, 2013), it becomes apparent that human trafficking has a two-sided nature.Some consequences are visible, but its illegal nature requires most of the action to be covert.We identify four indicators of human trafficking that reflect the intensity of human trafficking in destination countries: (1) the crime rate per 100,000 people, (2) the 3P-index of anti-trafficking policies8 , (3) the num-ber of identified human trafficking victims, and (4) the number of migrants registered in a country.In order to test the relationship between the indicators and human trafficking intensity (η) the following measurement model is implemented: (Ad1): Awareness within countries plays an important role in the identification of human trafficking.Given that human trafficking is a large-scale illegal business, its infrastructure must be highly developed.The criminology literature supports this claim and stresses the link between the transport of illegal migrants, human trafficking, and organized crime throughout the entire process of deceiving, transporting, and exploiting people (Salt,  2000; Salt & Stein, 1997; Schloenhardt, 2001).The extent of human trafficking in a country contributes to the overall prevalence of crime.Despite the hidden nature of the phenomenon, we are able to reveal it by looking at the occurrence of crime, measured as the level of crime in the country.We use the crime rate per 100,000 people taken from United Nations Surveys on Crime Trends and the Operations of Criminal Justice System (United Nations Office on Drugs and Crime [UNODC], 2008), which is the most complete set of cross-country crime data available.
(Ad2): The legal fight against crime is an indicator of the extent of human trafficking in the country.In order to explain this hypothesis, we look to the theory presented in the literature (e.g.Hathaway, 2007; Simmons,  2000) as to why states sign international treaties and enforce them domestically.This would suggest that countries which more rigorously prevent, suppress, and punish trafficking in persons are incentivized to do so because they have a higher intrinsic motivation for respecting human rights and expect the collateral consequences (such as international reputation and a strong civil society) to be more important than the costs of enforcement.The effect of enforcement on the incidence of violence, however, is not easily determined.In a meta-study which looked exclusively at the drug market, Werb et al.  (2011) found that interference in the market for drugs leads to an increase in violent incidents.We apply this argument to the trafficking market and argue that where human trafficking is combated by stronger law enforce-6 Assumptions made about the effects of the latent variable have to be considered carefully.Cliff (1983, p. 120) argues that there might be relevant divergence between the observed indicators and the latent phenomenon.This is especially important when interpreting correlations and model estimates established from the latent variable specifications and relating them directly to the unobserved phenomenon.However it should be noted that this is not a major problem here because the model is tested on several different specifications and model applications, which show that the underlying assumptions seem to be valid.Nevertheless, estimation models are no more than approximations of the unknown real social phenomenon and have to be interpreted cautiously, especially when the core component being estimated (HT) is not observable. 7Since October 2013, the Walk Free Foundation (2013) has measured the prevalence of people in slavery for 162 countries.This approach is based on risk characteristics of countries at one point in time.In contrast to the human trafficking intensity measure provided here, the measurement does not take into account development over time and is restricted to an analysis of the preceding year rather than of the last decade, and thus comparisons are difficult.
ment in terms of prosecution, protection, and prevention, the more incidences of trafficking will be registered in the country.A measure of these anti-trafficking instruments is the 3P-index provided by Cho, Dreher and Neumayer (2014). 9It is available for over 180 countries for the 2001 to 2013 period.The higher the score a country receives in the 3P-index (on a scale of 3 to 15), the more rigorously the anti-trafficking instruments are implemented.Our argument is supported by the following observations: the Convention Against Transnational Crime and the Trafficking Protocol are the results of international observations "that technological advances, combined with the ever-growing inter-dependence of economies, is offering criminal groups unprecedented lucrative opportunities" (Betti, 2001, p. 1).During the negotiations and the implementation of the Convention Against Transnational Crime and the Trafficking Protocol, public awareness of the topic increased substantially.Non-governmental organizations intensified public awareness campaigns and media coverage of human trafficking as an international criminal activity became ubiquitous. 10The increased salience of the topic forced policy makers to react and intensify the fight against it (Burstein,  2003, and sources cited there).In a simple correlation test between the 3P-index and the number of identified victims, we find a positive and significant correlation.We therefore argue that the 3P-index is a good indicator of human trafficking in the country.Ad(3): The number of identified victims in these countries gives an indication of the true number of victims.Although it is important to note that identification of victims, prosecution of traffickers, and prevention of the crime largely depend on the awareness of the existence of human trafficking in the wider public, as well as in legal institutions (Tyldum & Brunovskis, 2005), this number is an important sign of the existence of trafficking in persons.The indicator of the observed number of victims is a proxy for the total extent of the issue: it is only the tip of the iceberg.This observed number is affected by the quality of the law enforcement institutions in the destination country.Presumably the numbers are larger in countries with better institutions and therefore not necessarily where trafficking is more prevalent.However, the number of identified victims should be larger where the pool of all victims is larger, which suggests a positive correlation between the real extent and identified victims.In its global reports on trafficking in persons (UNODC,  2009, 2012), the UNODC provides the number of identified victims as a share of the total population for a large set of countries. 11Ad(4): Finally, as argued before, victim exploitation happens parallel to migration flows and we therefore use the number of migrants (in logs) as an indicator of human trafficking.The number of international migrants is available from the World Bank Development Indicators (World Bank, 2012).In order to obtain data for each year of the sample, we interpolate from the number of refugees as counted every five years by the Office of the United Nations High Commissioner for Human Rights.

Causes
The structural model includes not only all causes of human trafficking that influence the vulnerability of individuals and thereby pull them towards promising destination countries but also criminal aspects of the phenomenon.Since the application of the common definition of human trafficking in 2000, the number of studies on the causes of human trafficking has increased substantially (e.g.Akee, Basu, Bedi, & Chau, 2010; Cho, 2015;  Cho, Dreher, & Neumayer, 2013; Hernandez & Rudolph,  2015). 12The basic causes used in the modeling process of the MIMIC model are: (1) income per capita in logs, (2) foreign direct investment flows into destination countries (in logs), (3) employment in agriculture as a percentage of total employment in these countries, and (4) language fractionalization within the respective destination countries.The structural model is as follows: (Ad1): The main economic reason behind the existence of both human traffickers and an easily exploited population is the movement of people from regions of lower labor productivity to those of higher productivity.Thus, the ideal destination country for human trafficking is a high-income country, which can accordingly become a breeding ground for this type of activity.We use income measured by GDP per capita (in logs) taken from the World Bank's (2012) Development Indicators (WDI) as a proxy for the economic pull factor.
(Ad2): It is noted that international investment relations lead to increased cultural, social, and economic interrelation between countries.These are therefore are an additional pull factor for human trafficking, which can be interpreted as the negative externality of increased international connectedness and as being facilitated by different aspects of globalization processes.Interconnect- 9 We apply the overall index (3P-index) in order to avoid judging the importance of each of the single components.The fight against human trafficking is based on all three equally important aspects. 10See Ditmore and Wijers (2003) for details on the negotiations of the Trafficking Protocol.For an example of media coverage see Spiegel Online (2014). 11An overview of all variables used can be found in Table A1. 12All determinants used in the empirical literature so far are tested in the meta-study by Cho (2015) where she identifies robust causes of human trafficking flows.Contrary to Cho's approach, we focus on causal and indicating variables at the same time and apply the MIMIC model.In this way we add one layer of information to the one she uses.Nevertheless, the push and pull factors identified in Cho's study (2015) have been considered in various tests of our MIMIC model.In the final model we decided to focus on the determinants that have been robustly identified as causes of human trafficking in our setting and are in accordance with the model fit of our estimation technique.
edness facilitates transport via the establishment of international trade routes and investment connections.International crime groups are large-scale business operations that are active in both the official and the informal economies, corrupting officials and legal networks (UNODC, 2010).The variable used is the share of foreign direct investment (FDI) (in logs), which shows these connections (UNCTAD, 2012).13(Ad3): Most cases of human trafficking involve migrant workers in economic sectors such as agriculture and construction (Zhang, 2012).They account for 18 percent of identified cases of human trafficking according to the UNODC ( 2009).The increased chances of employment caused by the increased demand for cheap unskilled labor increases the attractiveness of countries as destinations for migrant workers (Hernandez & Rudolph,  2015). 14In addition, high demand in the commercial sex market or other informal markets increases the probability of people being pushed towards these locations (Cho  et al., 2013; Danailova-Trainor & Belser, 2006; Jakobson  & Kotsadam, 2013).Given the scarcity of data in this area, we are obliged to refrain from using sexual exploitation data and use agricultural employment given as a percentage of total employment, in data provided by the World Bank (2012). 15Ad4): On the other hand, trafficking may be limited by technological advances and personal contacts.These increase the availability of information on migration opportunities and job offers and thus presumably reduce the risk of being trafficked.Therefore information flows should have a restricting effect on trafficking.We em-ploy the language component of the distance-adjusted ethno-linguistic fractionalization index (DELF) developed by Kollo (2012)  16 using the variable in the WDI data set (World Bank, 2012).
Figure 1 shows the respective path diagram and expected relations for the MIMIC model of human trafficking, which combines the measurement and the structural model.

Results
Before estimating the MIMIC models, we use multiple imputation with fixed country and year effects to balance the sample and control for unobserved factors.Subsequently, the MIMIC models are estimated using a maximum likelihood estimator with missing values.After testing for the robustness of the specification and evaluating the model fit, the final indices are generated for the years 2000 to 2010.In this way country rankings for every single country-year combination are generated, which makes it possible to assess the development of the extent of human trafficking over time.gether with the overall fit of the model, both these aspects are enough to reject or confirm the assumed relationships (Bollen, 1989).The judgment of the quality of the model is based on whether the estimated covariance is equal or close to the true sample covariance.17

MIMIC Estimation Results
Table 1 shows the results of the MIMIC estimation.The fit indicators show a good model fit.In particular, the chisquare statistic with a p-value of 0.02 indicates a good model fit.The results are point estimates.The observed correlation of GDP per capita is positive and significant at the 1 percent level, indicating that wealthier countries are more often the destination of human trafficking.Investment flows into destination countries are also positively associated with human trafficking (at the 5 percent significance level).More international business and investments are correlated with illicit human movement in the form of human trafficking.The robust positive relation observed (significant at the 1 percent level) to the share of employment in agriculture endorses the positive relation between human trafficking and opportunities for cheap employment (low skilled labor), which provides more potential placements for use by traffickers to exploit people.Linguistic fractionalization within countries has a negative and significant relation to trafficking at the 10 percent level.This suggests that less diverse countries pull more human trafficking into the country.
Turning to the measurement model, we find that all indicators match our expectations.One of the indicators of the latent variable has to be normalized and used as an anchor variable for the scale and identification: the crime rate. 18We follow the literature by using the indicator with the largest standardized coefficient (0.891***) as the anchor variable (e.g.Buehn & Schneider, 2012;  Dreher et al., 2007; Schneider et al., 2010).All four indicators are positively related to the extent of human trafficking, which is in line with theoretical considerations and economic intuition.The 3P-index of anti-trafficking policies turns positive and significant at the 1 percent level.This shows the importance of the extent of antitrafficking policies, which protect victims, prosecute traffickers, and prevent human trafficking, as a reflection of the intensity of human trafficking in the country.The number of identified victims as a share of the population is not statistically significant.The low quality of the data would explain this finding (Laczko & Gozdziak, 2005).Finally and importantly, the share of migrants in the country positively indicates the prevalence of human trafficking (although this relation is not statistically significant at conventional levels).
We have carried out a number of robustness tests to check whether the results are valid under a variety of circumstances (Table 2).The results do not depend on the choice of estimation model and discount the three highest and lowest ranking countries (outliers) (Buehn & Farzanegan, 2012).In order to rule out endogeneity concerns, we have also estimated the results following Dreher et al. (2007) and lagged all quantitative causal variables by one period before estimating the model.We have also followed Dreher et al. (2007)  and tested whether it is a problem that some of the indicators could well be causal variables (such as the evaluates the fit of the model based on the deviance between the estimated and the real covariance.Brown and Cudeck (1993)  assume that RMSEA values smaller than 0.05 imply a good model fit, which corresponds to a probability close to 1.The two fit indices suggested by Bentler (1990) are the comparative fit index (CFI) and the Tucker-Lewis index (TLI).They indicate a good model fit with values close to 1 (Hu & Bentler, 1999).The coefficient of determination (CD) is similar to the R-squared with higher values showing better fit.crime rate or migration). 19The results are qualitatively unchanged. 20

Human Trafficking Index
The human trafficking index provides an assessment of human trafficking relative to other countries in all dimensions of the process.As described in Section 2, human trafficking includes all forms of exploitation and all groups of actors involved, traffickers as well as victims.We focus on the destination countries where exploitation takes place.Thus, we conceive of the extent of human trafficking in these countries as a bundle of actions and decisions.This includes the decision of traffickers to send individuals to these countries and exploit them.Additionally, it includes the extent of a market failure which provides opportunities for the exploitation of victims through the demand for cheap labor as well as the share of the population which is vulnerable and marginalized and therefore easily exploited.This shows that creating a clear picture of the extent and dimensions of human trafficking in countries is still very difficult.However, the derived index and subsequent ranking of countries is superior to existing measurements of specific aspects of human trafficking.For example, the tier score of the U.S. Department of State (2001, 2002, 2003, 2004, 2005, 2006, 2007, 2008,  2009, 2010, 2011, 2012), which only addresses compliance with U.S. policies, has been criticized as a one-sided approach that reflects political interest rather than the will to produce a transparent independent score of the variety of international policy approaches (Simmons &  Lloyd, 2010). 21By contrast, the underlying data implemented in the construction of our intensity index (human trafficking index) are based on publicly available data of causes and indicators that influence human trafficking by the means which have been described.The relevance of the measure is assessed in the following Table 3, where we show the pairwise correlations of the human trafficking index with the 3P-antitrafficking policy index, the tier rank of the U.S. Department of State, and the citation index compiled by the UNODC (2006, 2009, 2012). 22 19 These results are available from the authors upon request. 20An important issue in using the MIMIC approach is whether the assumption that the variables identify one latent factor holds.Using a principal component analysis (PCA), results show that one component is sufficiently explained by the variation in variables, suggesting that our assumption of one latent factor (human trafficking intensity) holds. 21 Simmons and Lloyd (2010) criticize the ranking of countries by the Department of State for mirroring the political interests of the U.S.A. and suggest that it eventually serves to get other countries to comply with the norms set out by the country depicted as the world's referee. 22We do not include the Global Slavery Index in our assessment here, since it mainly refers to countries where victims originate.Notes: * p < 0.01; The tier rank is reversed and shows higher values the better countries fulfil the Victims of Trafficking and Violence Protection Act (TVPA) standards.CI: destination/origin refers to the citation index compiled by the UNODC.The countries are ranked in a five-category scale with the highest value indicating that the probability of being a destination/origin country is "very high".
The correlations show that the human trafficking index has a high positive correlation with the 3P-index.Countries that rank high in fighting human trafficking have a higher intensity of trafficking inflows (as assumed in Section 2 and shown in Table 2).Similarly, we see a positive correlation between the tier rank and the citation index of destination countries, as well as a negative correlation to the citation index of countries of origin.One could argue that the correlation between the latter three indices is only modest (0.59 to 0.39, respectively).However, this is attributed to the different aspects of human trafficking that they measure.The tier rank addresses compliance with the TVPA and thus political decisions to fight human trafficking.This is only one aspect of the process of trafficking in human beings and a partial aspect of our intensity index, which also captures country characteristics, as well as the criminal dimension and vulnerability of victims.The same holds true for the citation indices: both indices capture aspects of human trafficking which are visible to society and therefore receive public attention.In the MIMIC model the crime rate, numbers of identified victims, and migration levels are used as indicators and combined with the causes-this ensures that this aspect of the trafficking process is included in the intensity index to provide a holistic picture.

Country Rankings
Looking at Table 4 we find mainly OECD countries leading the ranking of destination countries.The table shows the twenty top and bottom ranking countries in the years 2001, 2005, 2010.At the lower end of the ranking, we find Asian, Latin-American and Sub-Saharan African countries with a low prevalence of trafficking into their countries.
Unsurprisingly industrialized countries are at the top of the ranking.This is in line with the observations in the UNODC report (2006, p. 17).Countries in North America and Europe, as well as Australia are reported to be the top destination countries for human trafficking.The fact that Scandinavian countries rank so highly indicates that there are hidden activities taking place in these countries which enable traffickers to exploit individuals.This is less surprising than it seems when their geographical location is taken into account, particularly considering that they are often used as role models for institutional quality.They are close to Russia and Eastern European countries, which have lower economic opportunities and, in Russia's case at times, a higher level of persecution of individuals (traffickers and victims alike).Thus, they are very attractive destinations for the vulnerable and desperate in these countries as well as Middle Eastern, African, and Asian nations.It is also in line with the observation that the share of identified victims in these countries is quite high, especially at the beginning of the century. 23ooking at major OECD countries, the United Statesthe country which has pushed initiatives against human trafficking such as the implementation of anti-trafficking instruments-consistently ranks among the top 20 countries and even leads the ranking in 2010.This relatively constant position suggests that despite the United States' intense anti-trafficking efforts and awareness campaigns run locally and internationally, the magnitude of the problem within its borders seems to be stable and has even worsened recently with a higher intensity in 2010 (index value of 79).Finally, Germany ranks among the top ten countries throughout the years, which confirms reports of a high magnitude of human trafficking in the country.
At the lower end of the list are mainly low-income countries, which act as a source rather than a destination for victims of human trafficking.There are some (rather surprising) variations in the ranking of some countries.We attribute these to different aspects of the trafficking process.Firstly, from the perspective of the traffickers, it should make sense to use established trafficking routes.However, as a way of hiding from criminal justice systems, variations in destination countries may serve to reduce the chances of getting caught.Secondly, from the perspective of the victims, the same holds true for migration routes and observed travel patterns.Finally, from the perspective of a state fighting human trafficking, a series of legal successes in one year may serve to make the problem more visible but may not be repeated in the next.In countries such as Brunei Darussalam or the United Arab Emirates, an autocratic state may be able to hide the problem from the public eye more effectively than democratic states with checks and balances.

Conclusion
Human trafficking is a global phenomenon of large proportions.People are exploited, live in inhumane condi-tions, and lack food or access to health facilities, especially in high-income countries.The article disentangles the relationships between indicators and causes of human trafficking by employing a structural equation approach (MIMIC) and ranking 142 countries over a tenyear period according to the prevalence of human trafficking within their borders.
Our approach goes beyond existing studies by including both causal factors and indicators while acknowledging the illicit nature of the phenomenon.The causes mirror the incentive structure for traffickers, which is intended to maximize their chances of making high profits while maintaining a low probability of detection.Furthermore, the causes also capture the vulnerability of trafficking victims by addressing their incentives to move in the first place, a decision which leaves them vulnerable to false promises of better opportunities.By travelling from a poorer country to a wealthier one, individuals are inherently vulnerable: this is especially the case as linguistic fractionalization is negatively related to intensity.The indicators, on the other hand, show the results in countries where illegal trafficking of human beings takes place.Human trafficking is observed in countries with high crime levels, e.g. a larger underground economy running parallel to migration flows.The numbers of identified victims together with the number of migrants are indicators of the phenomenon and describe the visible extent of the problem.The dimensions of the fight against human trafficking (3P-index) quantify the application of anti-trafficking policies within countries.
The pattern of the development of human trafficking over the observed period is in line with expectations.Developed countries rank high and display a large amount of trafficking within their borders.These countries are the primary targets for traffickers, presumably because the potential for large profits is greatest in wealthier countries.The lowest ranking countries are mainly countries in Sub-Saharan Africa, Asia, and Latin America, where many people flee their miserable living conditions and look for other opportunities in more prosperous countries.

Table 1 .
MIMIC estimates.The crime rate is chosen as the anchor variable and normalized to 1.If the model fits the data perfectly and the parameter values are known, the sample covariance matrix equals the covariance matrix implied by the model with small Chi-square values.The root mean squared error of approximation (RMSEA) Notes: Absolute z-statistics in parentheses; * p < 0.1, ** p < 0.05, *** p < 0.01;

Table 3 .
Pairwise correlations of human trafficking indices.

Table 4 .
Highest and lowest ranking countries in specific years.