The Prosecution of State-Level Human Trafficking Cases in the United States

In an effort to combat human trafficking, the United States federal government and all fifty states passed new laws that criminalise human trafficking and support the identification and prosecution of human trafficking perpetrators. Despite the passage of these laws, only a small number of human trafficking cases have been prosecuted in the last fifteen years. Guided by the notion that prosecutors seek to avoid uncertainty when making decisions to pursue criminal prosecution, we explore how human trafficking crimes are indicted under these newly defined state laws. Using a sample of cases from twelve US counties and interviews with police, prosecutors and court personnel, we examine the factors that influence the decision to prosecute crimes investigated as human trafficking in state court. This research informs our understanding of why so few human trafficking cases are prosecuted and why human trafficking suspects are rarely convicted of trafficking offenses.


Introduction
In response to growing concerns about human trafficking, the United States [US] federal government passed the Victims of Trafficking and Violence Prevention Act in 2000 (TVPA). 1 Following the model of the TVPA, as of 1 The TVPA defines severe forms of trafficking in persons as 'sex trafficking in which a commercial sex act is induced by force, fraud, or coercion, or in which the person induced to perform such an act has not attained 18 years of age; or the recruitment, harboring, 2015, all fifty states had also passed laws criminalising acts of human trafficking. Federal and state legislation included enhanced penalties for perpetrators and increased protections for victims. 2 Further, many states have created task forces to help with prosecuting human trafficking cases, as it became apparent that states are most likely to locate and identify instances of human trafficking at the community level, and not all human trafficking cases meet federal jurisdiction. Since passage of the TVPA in 2000, 1,876 suspects have been prosecuted for federal human trafficking offences 3 and roughly 450 suspects have been prosecuted for state-level human trafficking crimes. 4 Understanding the context of human trafficking prosecutions is important since public officials and opponents of current anti-trafficking policies have argued that the relatively small number of prosecutions in the US provides evidence that the seriousness and prevalence of human trafficking has been exaggerated. 5 Additionally, human trafficking prosecutions and convictions are a common metric upon which a government's anti-trafficking response is judged. 6 The TVPA itself lists prosecution and punishment of human traffickers as one of its primary purposes, along with protection of identified human trafficking victims and prevention of human trafficking.
To date, there have been few empirical examinations of the factors that influence local prosecution of human trafficking crimes, and thus little is known about the effectiveness of state human trafficking laws. The few studies that have examined human trafficking prosecution have focussed on federal prosecution. 7 Studies examining state use of human trafficking laws generally focussed on ambiguity in these laws and the challenges of developing new legal standards. 8 For example, prosecutors are often unaware that their state has a trafficking law and are unfamiliar with the legal elements necessary to prove a trafficking charge. 9 Parallels are often drawn between human trafficking and other sensitive crimes like sexual assault and domestic violence, but research is needed to understand whether the same factors that inhibit the prosecution of other sensitive crimes explain prosecutorial decisions in new state human trafficking crimes.
Using data from a sample of state-level human trafficking cases, supplemented with information from interviews with police, prosecutors and other court personnel in twelve US counties, this exploratory study examines the type of charge used to prosecute state-level human trafficking cases and the factors that influence the decision to prosecute. Our analysis is guided by uncertainty-avoidance theory, which posits that prosecutors are less likely to prosecute cases if they are uncertain about the outcome of obtaining a conviction. 10 Although this study provides one of the first quantitative assessments of trafficking prosecutions to-date, there are a number of important limitations that must be taken into account when considering its findings and the conclusions drawn from them. The data collected for this study are from trafficking cases prosecuted between 2000 and 2010 and most likely representative of the 'first generation' of state-level human trafficking prosecutions. Since that time there have been numerous efforts to train state prosecutors about human trafficking and some states have developed specialisation among a small number of prosecutors who have experience developing human trafficking cases. 11 In the 'second generation' of human trafficking prosecutions we might expect more and a broader array of different types of trafficking cases to be pursued. In the data analysed here, state-level human trafficking prosecutions included only sex trafficking offenders. It is also possible that different factors may predict prosecution decisions about pursuing human trafficking charges in the second generation of cases. Additionally, the sample of jurisdictions studied here is not nationally representative; conclusions about the prosecution of human trafficking transportation, provision, or obtaining of a person for labor or services, through the use of force, fraud, or coercion for the purpose of subjection to involuntary servitude, peonage, debt bondage, or slavery' (TVPA of 2000, section 103, 8a and b). 2 B Stolz, 'Interpreting the US Human Trafficking Debate through the Lens of Symbolic Politics ', Law & Policy, vol. 29, 2007, 311-333. 3 See: Attorney General's Annual Report to Congress and Assessment of US Government Activities to Combat Trafficking in Persons, 2010Persons, , 2011Persons, , 2012Persons, , 2013Persons, , 2014Persons, and 2015 ', International Review of Victimology, vol. 11, 2004, 143-176. 6 For examples, see: US Department of State Trafficking in Persons Report series (2010)(2011)(2012)(2013)(2014)(2015); and the UN Office on Drugs and Crime Global Report on Trafficking in Persons, 2009, 2014. 7 H Clawson et al., Prosecuting Human Trafficking Cases: Lessons learned and promising practices, ICF International, Washington DC, 2008 A Reiger, 'Missing the Mark: Why the Trafficking Victims Protection Act fails to protect sex trafficking victims in the United States', Harvard Journal of Law and Gender, vol. 30, 2007, 231-256;M Wade, 'Prosecution of Trafficking in Human Beings Cases' in J Winterdyk, B Perrin and P Reichel (eds.), Human Trafficking: Exploring the international nature, concerns and complexities, CRC Press, Boca Raton, FL, 2012; A Farrell, C Owens and J McDevitt, 'New Laws But Few Cases: Understanding the challenges to the investigation and prosecution of human trafficking cases ', Crime, Law and Social Change, vol. 61, 2014, 139-168. 9 Newton, Mulcahy and Martin, 2010. 10 C Albonetti, 'Prosecutorial Discretion: The effects of uncertainty ', Law and Society Review, vol. 23, 1987, 291-313. 11 For example, the National Association of Attorneys General constituted a committee on human trafficking prosecutions in 2011. Their website houses numerous publications aimed at improving state prosecutor knowledge of human trafficking laws and the challenges of developing human trafficking cases. See: http://www.naag.org/naag/committees/naag-special-committees/human-traffickingcommittee.php cases are only generalisable to the twelve counties studied. Future research should attempt to replicate these findings using more recent prosecution data from a larger sample of jurisdictions.

Background
In the US prosecutors have considerable discretion in making decisions about criminal cases and are generally immune from review when rejecting charges. 12 Prosecutors may exercise discretion by declining to prosecute cases that are brought to their attention or by charging offenders with more or less serious crimes. 13 The power of prosecutorial discretion is kept in check by a unique set of ethical obligations. Prosecutors are responsible for the vigorous prosecution of offenders and to the service of justice, which requires consideration of the interests of those who they prosecute.
Research suggests that prosecutors commonly make charging decisions based on the likelihood of conviction. 14 Successful convictions are markers of ethical charging practices, by only bringing criminal charges where there is evidence of guilt beyond a reasonable doubt, and they are also a common measure of occupational success. 15 Research on prosecutorial decision making suggests prosecutors employ a 'downstream orientation', where they anticipate or predict how juries will interpret and respond to a case, when making decisions about whether or not to pursue prosecution. 16 The goal for reducing uncertainty is the likelihood of achieving a conviction at a jury trial, despite the fact that a majority of criminal cases are disposed of through a plea agreement. Research suggests that prosecutors' assessments of whether cases are likely to result in a conviction are primarily based on legally relevant factors, such as the strength of the evidence against the accused. 17 Other studies find that extra-legal factors such as suspect and victim characteristics and the victim's relationship to the suspect also influence the decision to prosecute a case. 18 Because prosecutors rarely have all of the information they need to make informed, rational decisions, they establish formal and informal case processing norms intended to absorb uncertainty by imposing a rationality on the decision making process. 19 For example, court workgroup members such as prosecutors, defence attorneys and judges establish recognised 'going rates', or informal norms concerning routine charges, plea agreements and punishment for criminal offences that are familiar to the court. Nardulli et al. (1988) uses the term 'consensus mode' to describe those cases where workgroup members apply going rates easily to facilitate decision making because the case is routine and well understood by all participants. 20 These cases are likely to end in pleas. When faced with criminal charges with less well established going rates, consensus commonly breaks down, pleas fail and trials are more likely. In these less established cases, even when there may be sufficient evidence to pursue prosecution, prosecutors may dismiss charges if they lack established going rates that would normally facilitate pleas, have concerns about how jury members will perceive victims or witnesses, or in the case of new crimes like human trafficking, concerns about the legitimacy of the offence itself. There are also practical conditions necessary to secure a conviction, particularly the cooperation of victims. Research confirms the importance of victim willingness 12 W Lafave, 'The Prosecutor's Discretion in the United States', The American Journal of Comparative Law, vol. 18, 1970, 532-548; C Spohn, D Beichner and E Davis-Frenzel, 'Prosecutorial Justifications for Sexual Assault Case Rejection: Guarding the "gateway to justice" ', Social Problems, vol. 48, 2001, 206-235. 13 J Spears and C Spohn, 'The Effect of Evidence Factors and Victim Characteristics on Prosecutors' Charging Decisions in Sexual Assault Cases', Justice Quarterly, vol. 14, 1997, 501-524. 14 D Beichner and C Spohn, 'Modeling the Effects of Victim Behavior and Moral Character on Prosecutors' Charging Decisions in Sexual Assault Cases ', Violence and Victims, vol. 27, 2012, 3-24;Spohn, Beichner and Davis-Frenzel, 2001. 15 A J Davis, 'The American Prosecutor: Independence, power, and the threat of tyranny ', Iowa Law Review, vol. 86, 2001, 393-466; D S Medwed, 'The Zeal Deal: Prosecutorial resistance to post-conviction claims of innocence', Boston University Law Review, vol. 34, 2004, 125-183. 16 L Frohmann, 'Convictability and Discordant Locales: Reproducing race, class and gender ideologies in prosecutorial decisionmaking ', Law and Society Review, vol. 31, 1997, 531-556. 17 Albonetti, 1987, finds that cases resulting in arrests and criminal charges were more likely to involve the presence of a weapon, the use of force, and severe victim injuries. Other legal factors unrelated to incident severity have still been shown to impact case outcomes. For example, both Albonetti, 1987, and L Mather, 'Comments on the History of Plea Bargaining', Law and Society Review, vol. 13, 1979, 281-285, found that defendants with a criminal history are more likely to be prosecuted. J Schmidt and E Steury, 'Prosecutorial Discretion in Filing Charges in Domestic Violence Cases ', Criminology, vol. 27, 1989, 487-510, find that offenders with a failure to appear in court are also more likely to be prosecuted.  Women, vol. 7, 2001, 46-59;T Schlesinger, 'Racial and Ethnic Disparity in Pretrial Criminal Processing', Justice Quarterly, vol. 22, 2005, 70-92;Spohn, Beichner, and Davis-Frenzel, 2001. to cooperate in the prosecution and victim credibility in prosecutor decisions to pursue charges in sensitive crime cases. 21 Human trafficking provides a unique opportunity to examine the process of uncertainty avoidance in an area of law where evidentiary burdens are less established due to the relatively new nature of the legal reform. Because human trafficking at the state level is new, prosecutors are less likely to have had experience prosecuting these types of cases and are therefore less likely to be able to predict how a judge or jury would interpret the evidence or perceive the victims. 22

Sources of Data
The data used in this study was collected as part of a larger project examining human trafficking prosecutions in a targeted sample of US counties. 23 Many human trafficking cases cannot be prosecuted at the federal level due to jurisdictional limitations and time and resource constraints. In the US, state prosecutions are becoming more common and are projected to make up the majority of human trafficking prosecutions in the coming years. 24 We reviewed the closed case records of human trafficking investigations conducted by law enforcement in 12 sampled counties and conducted 166 in-depth interviews with police, prosecutors, victim service providers and court officials involved in the investigation and prosecution of these cases. The interviews were intended to illuminate the process of investigating and preparing human trafficking cases for state-level prosecution.
Because relatively few local law enforcement agencies have investigated human trafficking cases, randomly selecting counties would have yielded sites that did not have the caseload and experience necessary to inform our research. Instead, we used a multi-stage cluster sampling approach where we first identified US counties where there was evidence that the police had investigated cases of human trafficking since the passage of the TVPA in 2000. We then grouped counties by state legislative characteristics and selected a targeted sample of twelve counties that varied by legislation (no state human trafficking law, basic state human trafficking law and comprehensive state human trafficking law). 25 Two hundred and fifty-four (254) human trafficking cases were identified across the study sites. 26 We drew a sample of approximately 15 cases per site that were stratified by year (2003)(2004)(2005)(2006)(2007)(2008)(2009)(2010) and type of trafficking (sex trafficking adult, sex trafficking minor, labour trafficking, and sex and labour trafficking combined). The present analysis focusses on the 150 suspects that were arrested by state and municipal law enforcement across our sampled cases to understand the factors that predict state prosecution. We coded detailed information about the characteristics of each studied case from police incident reports, investigative records, indictments and charging documents, court testimony records, and sentencing opinions.

Variables and Measurement
We examine three charge outcomes for each studied suspect: charges declined, charged with a human trafficking offence, or charged with another criminal offence. Outcomes were coded as 1 when charges were filed for a measured violation. See Table 1 for a complete list of all dependent and independent variables in this study. A majority of the state-level cases with identified human trafficking perpetrators were charged with offences other than human trafficking. Only 22% of the state-level cases with identified human trafficking perpetrators were charged with a state-level human trafficking offence (see Table 2). All of the suspects in our sample who were 21 C Albonetti, 'Criminality, Prosecutorial Screening and Uncertainty: Toward a theory of discretionary decision making in felony case processing ', Criminology, vol. 24, 1986, 623- Social Problems, vol. 38, 1991, 213-226. 22 Similarly, with hate crimes, prosecutors decided to prosecute crimes under hate crime legislation when they felt more confident 'decreasing the complexity of a case' and 'minimizing risk' of a not guilty verdict. It is possible for human trafficking cases to be investigated and prosecuted as other types of crime (e.g. promotion of prostitution). Cases involving human trafficking acts were identified in states without state laws defining human trafficking as a stand-alone crime. In those cases, perpetrators of human trafficking were prosecuted for other types of offences. 26 This included all of the investigations of human trafficking that had been opened and closed between the passage of the first state human trafficking laws (2003) and the beginning of data collection in 2010. charged with a state human trafficking offence were sex trafficking offenders. 27 Thirty-seven per cent (37%) of suspects were charged with promotion/compelling prostitution offences and an additional 16% of suspects were charged with prostitution offences. 28 Eleven per cent (11%) were charged with sexual offences such as sexual assault or sexual exploitation of a child. The remaining suspects were charged with other types of offences including conspiracy, kidnapping, and drug offences.
Empirical research suggests that the likelihood of prosecution increases with the existence of evidence supporting the prosecution, including physical, demonstrative, (e.g., photos, '911' emergency call tapes, medical reports) and digital evidence (e.g., emails, ATM transactions) that corroborates victim testimony or independently furthers the prosecution's case. 29 Additionally, prosecutors take into account the amount of evidence and the number of witnesses as well as the existence and willingness of a victim to cooperate with the prosecution. To capture information about the strength of the case we include a measure of whether the police collected physical, demonstrative, or digital evidence (coded 0 for no evidence collected and 1 for any physical, demonstrative or digital evidence). We also measured victim cooperation by whether or not a victim was interviewed and provided information to the police and prosecutors (coded 0 for no and 1 for yes). The data coding mechanisms employed to capture the quality of evidence were simplistic. Although necessary for an exploratory study where coding structures for the quality and quantity of these complex events have not been developed, dichotomous variables are not able to capture variation that likely exists in the quality of victim cooperation or the content of victim interviews.
Additionally, we coded for indicators of the means of human trafficking as specified in the TVPA and the reauthorisations of 2003, 2005, and 2008 to further operationalise the strength of a human trafficking case. Human trafficking indicators included: threatened or actual physical or non-physical harm, use or threatened use of law to exert pressure, demeaning or demoralising the victim, disorienting victims (e.g. isolation, restrict communication), diminishing resistance and debilitating (e.g. denying food, water, medical care, weakening with drugs or alcohol), deceiving (e.g. overstate risks of leaving and/or rewards of staying), dominating, intimidating and controlling (e.g. displaying weapons, rules and punishments), knowingly recruited, enticed, harboured, transported, provided, obtained, or maintained a person for purposes of commercial sex, knowingly benefited from participating in human trafficking, knew [or recklessly disregarded] that force, fraud, or coercion would be used to cause the person to engage in commercial sex. We used the force, fraud and coercion means framework from the TVPA because all states of the sites we studied used the TVPA as the basis for their own legislation. This coding allows us to measure the strength of a case across different states' unique legislation and elements. Evidence of human trafficking is coded as a count of the number of these human trafficking indicators identified in the case review, ranging from 0 to 10. Multiple elements, resulting in a higher value, suggest a stronger case. Modelling was conducted to control for the effect of a case being adjudicated in a particular state since state statutory definitions of human trafficking varied across study sites.
We examine characteristics of both the victim(s) and suspect(s) to determine the degree to which extra-legal factors influence prosecutor charging decisions in state human trafficking cases. Qualitative research suggests that prosecutors are more likely to bring charges in human trafficking cases involving minor victims because these victims are perceived as needing more protection. 30 Additionally, under federal law and many state laws, the evidentiary burden of proving human trafficking is reduced when the victim is a minor (for example, under federal law there is an exception to the force, fraud or coercion requirement for minor victims). Thus, we measure whether any of the victims in the state-level cases are minors (coded 0 for only victims who are 18 or over and 1 for at least one victim under 18). We further measure whether the victim(s) are female, male, or whether the case involves both male and female victims. 31 We also measure whether the case involved no victims, a single victim, or multiple victims (each coded as 0 for no and 1 for yes). 32 In addition to information about victims, we measure the influence of suspect race, coded as White, Black, Hispanic, Asian or other, and suspect age (coded as 0 if under 30 and 1 if 30 27 Despite the fact that all suspects charged with human trafficking offences were engaged in sex trafficking, we continue to use the umbrella term of human trafficking because state criminal codes varied in their classification of human trafficking crimes with some specifying separate sex and labour trafficking offences and others including sex and labour trafficking acts under a generic human trafficking offence. 28 Prostitution offences included prostitution or solicitation of prostitution, common night walking enticing a person for prostitution, keeping a house of prostitution, and living off the earnings of prostitution. The parties to such crimes can include a sex worker, customer and/or facilitator. We coded for transgender victims, but none were identified in the study cases. 32 Victim cooperation in many cases is more complicated than a dichotomous indication of no cooperation versus some cooperation. Additionally, cooperation is not static. Victims sometimes cooperated and provided full information at one point in the investigation only to disappear or recant their statements at later points. Unfortunately, we could not code for these complexities in a reliable way. Instead, we coded for indications in the case record that a victim was at any point in the process willing to cooperate and provide information to law enforcement and prosecutors compared to those cases where no victim was identified or willing to provide information. Future research is needed to explore how various forms or degrees of cooperation impact the decision making processes of prosecutors in human trafficking cases. and above). 33 We include a variable for missing data for suspect race and age to capture the effect of cases where this information was not available.
There is significant debate about the value or harm of arresting human trafficking victims. Since human trafficking victims often engage in illegal activity such as prostitution or illicit immigration during the course of their victimisation, law enforcement can use the power of arrest or detention to secure victims. Victim detention may occur because law enforcement does not recognise individuals as crime victims or because they do recognise their victimisation but utilise arrest as a mechanism to protect victims from retribution, and coerce their cooperation with ongoing investigations. Potential victims were arrested in nearly one-third of the cases analysed for this study. Many of these victims were initially identified as an offender rather than a victim. Though concerning, this finding is not surprising considering that units that commonly uncover sex trafficking, such as vice units, are generally tasked with making prostitution arrests. To determine whether arresting human trafficking victims promotes or hinders prosecution of human trafficking suspects, we measure victim arrest as whether or not any identified victim was arrested during the course of the investigation into a human trafficking offence (coded as 0 for no and 1 for yes).

Analytic Strategy
Multinomial logistic regression models are estimated to test the impact of various legal and extralegal factors on a polytomous outcome measure of whether a state human trafficking offender had all charges dismissed (coded 0), was charged with another type of crime (coded 1) or was charged with a human trafficking offence (coded 2). There were 150 cases with state charges (N=150). Multinomial regression is utilised here to explain the odds of a defendant having charges dismissed, being charged with a non-trafficking offence or being charged with a trafficking offense when specific conditions are present (e.g., victim who is a minor victim is arrested). We address the issue of non-independence related to multiple subjects being charged in the same case by using the 'cluster-robust standard error' option in Stata software, specifying the criminal case as a 'cluster' (a robust treatment of errors). This option adjusts the standard errors for all predictors to where they should be (without applying this adjustment, standard errors will be understated, thus leading, in some instances, to seemingly statistically significant findings that are actually not significant). In addition to correcting for case-level effects, we account for county-level effects, a second source of non-independence in our models, by including county-level dummy variables for each county. 34 Data from qualitative interviews was transcribed and uploaded to the software programme NVivo10 for coding and analysis. We developed a series of thematic codes to better understand the challenges state prosecutors faced in pursuing criminal charges against human trafficking offenders. Common themes about prosecution from the qualitative analysis are explored here to better understand and contextualise findings from the case analyses.

Results
To understand how legal and extra-legal factors influence the prosecution of state human trafficking cases a series of multinomial regression models are estimated. Table 3 presents findings from the multinomial regression models estimating whether suspects were not charged (reference category), charged with another type of crime, or charged with a human trafficking offence for state-level cases. The average age of suspects is 33. However, suspect age is not normally distributed. To address skewed distributions, we created a dummy variable separating young (under 30) and older (30 and above) suspects. 34 A key prerequisite of conducting multivariate modelling and associated statistical tests is that observations be independent of one another (i.e., standard errors of predictors must be independently distributed). When that assumption is violated, errors will be correlated at the group level. We encounter two such sources of non-independence in the current research. In criminal cases that contain multiple defendants, outcomes for individuals within those cases will be related (and therefore, not independent). Cases in the same county are also not independent. Counties have characteristics, attributes, and cultural norms inherent in their criminal justice process that could influence outcomes for individual defendants.

Human Trafficking Indicators
In the state-level human trafficking cases we studied, evidence of indicators of legal elements of human trafficking was not significantly associated with filing human trafficking charges. Further, somewhat surprisingly, when cases included more indicators of human trafficking elements, they were significantly less likely to result in the filing of other types of criminal charges as well. Because trafficking laws are difficult to navigate and require evidence that prosecutors and the police are often not used to obtaining, prosecutors may shy away from prosecuting cases that present with human trafficking indicators. A detective in one study site explained the reluctance of the state prosecutor in his district to charge human trafficking crimes: 'The trafficking law hasn't been used that much, so, as a prosecutor, you don't want to be the only one using it, and all of a sudden your case doesn't go forward.' Interviews with state prosecutors confirmed they were generally unfamiliar with human trafficking laws and struggled to define the concept of human trafficking beyond the prostitution of minors (just one of the legal elements of trafficking). Labour trafficking cases were particularly challenging for state prosecutors because some states defined labour trafficking offences separately from sex trafficking offences of adults, and required different legal elements for these crimes, particularly with minor victims.
Despite these challenges, state prosecutors who took the time to look into the details of a particular human trafficking case describe being moved by the incredible level of violence and coercion involved. In some cases, these facts prompted prosecutors to pursue human trafficking charges despite known impediments. The prosecutors we spoke with were often the first in their state to prosecute a case using state anti-trafficking laws. One prosecutor describes how she happened upon the state human trafficking offence and decided to pursue the charge.
I started bouncing the case around with a few colleagues and I'll be completely honest and I'm embarrassed to say it, but none of us were really aware of the state human trafficking law. It was pretty new at the time…once I read through it, I was like, 'this human trafficking offense is perfect.' I mean, it's exactly what it is.
Prosecutors who charged human trafficking suspects using untested state laws were often met with challenges when trying to explain and prove the facts of the new crime to judges and juries. Many state and local prosecutors were operating on their own with little to no source of legal guidance they could refer to for topics such as prosecutorial techniques, how to handle common defence tactics, or sample jury instructions. In every site, when prosecutors who took human trafficking cases to trial using state anti-trafficking laws were asked where they went for guidance on jury instructions, they said that they created them themselves and had wished they had a resource or fellow state prosecutors to consult.

The Presence of Evidence
Suspects were more likely to be charged with either a human trafficking offence or other types of charges when the case had physical, demonstrative, or digital evidence. This finding is consistent with previous research on other crimes suggesting that prosecutors are more likely to file charges when the evidence strongly supports conviction. 35 This is to be expected. However, despite the importance of evidence, roughly a third of the cases we reviewed had no evidence supporting victim testimony. A state prosecutor describes the challenges of pursuing trafficking charges when cases lack corroborating evidence.
The evidence in the end just wasn't strong. Probably because there were a lot of inconsistencies with the victim's statements and us not being able to prove the case beyond a reasonable doubt to a jury. Typically it would be the case that a victim gave an initial statement and as the investigation progresses that statement becomes something that we can't corroborate with other evidence and you need more than just one person saying this is what happened. Other witnesses that will corroborate the event are gone or not credible, so you have to weigh the credibility of all the witnesses that will be testifying to see if a jury is going to believe them or not, and that consists of their background, their criminal histories, their age, their relationship to the parties, you know whether they have a stake in the outcome of the case, things like that.
This quote illustrates the 'downstream orientation', where the prosecutor evaluates evidence based on how he or she believes it will be received by judges and juries. The need for substantial evidence of trafficking such as hotel receipts, photographs of injuries, Backpage/Craigslist advertisements, phone and text message records and financial records was a recurring theme in many of the interviews with police and prosecutors, even if there was a victim or witness willing to cooperate and testify in court. Prosecutors complained that they were not being referred cases with strong evidence. In most of the cases we reviewed, prosecutors did not get involved with an investigation until a suspect was arrested. When evidence from the original investigation was lacking, prosecutors sometimes sent the police back out into the field to collect additional evidence, but physical records and corroborating witnesses were often difficult to locate after some time had passed.
In addition to improving the likelihood that cases wou ld result in a conviction based on their own weight, prosecutors suggested that physical or corroborating evidence also improved the likelihood that the victim/witness would actually testify. One prosecutor expressed concern about a victim failing to appea r to testify at trial when her testimony was the main source of evidence. 'We're always looking for corroborative evidence, so that we're taking the burden off of the victim.' When physical or corroborating evidence is hard to come by, the case ends up res ting on the believability of the victim.

Victim Cooperation and Credibility
Previous research on prosecution of sexual assault and domestic violence cases suggests that victim cooperation is strongly associated with the decision to pursue prosecution. 36 When we included a measure of victim arrest, victim cooperation did not statistically predict the decision to pursue state charges in human trafficking cases. As described in more detail below, arresting victims appears to be a mechanism that law enforcement uses to secure victim cooperation. Despite the failure of victim cooperation to independently predict prosecution, in nearly every study site, prosecutors and police who were interviewed cited lack of victim cooperation as the biggest barrier to prosecution. For example, a police officer in a human trafficking unit described the challenge of moving a case forward to prosecution when a victim could not be located or was not willing to cooperate with law enforcement.
A domestic victim without a cell phone or way of contacting them, they're like gypsies. They move all over the place. Sometimes they go back home. Sometimes they run away again, they end up hooking up with some other trafficker. We end up with a whole different case, with a different exploiter because they hooked up with 35 See: Albonetti, 1986;Frohmann, 1997;and Spohn, Beichner and Davis-Frenzel, 2001. 36 Dawson and Dinovitzer, 2001;Spohn, Beichner and Davis-Frenzel, 2001. another exploiter. Or, they go back to their original exploiter. Victim cooperation is our biggest stumbling block.
In some sites, prosecutors acknowledged having knowledge about other victims in the community who were victims in cases that went forward to prosecution. These victims either refused to cooperate with law enforcement or provided initial statements and then refused to follow up with prosecutor requests for interviews.
Although the quantitative data from cases did not support interviewees' statements about the importance of victim cooperation, we found that the arrest of a victim positively and significantly predicted the filing of both human trafficking and other crimes in state human trafficking cases. In state-level cases, law enforcement may actually be using arrest to coerce victims' cooperation. Victims were arrested in 59% of all state-level cases. As described above, all of the human trafficking charges studied here involved acts of sex trafficking. Thus it is not surprising that when human trafficking victims were arrested, the arrest was for a prostitution-related offence. In some cases, the charges were dropped against a victim when they provided information to law enforcement about their trafficker. Despite concern that juries may find victims who faced criminal charges to be less credible witnesses, human trafficking cases were more likely to be prosecuted when the victim was arrested in the sites we studied.
The influence of victims being arrested on state prosecution decisions was supported by data from qualitative interviews. In particular, police indicated that they often had to arrest sex trafficking victims because there was not a safe and secure place to house victims, particularly minors. They described victims as 'evidence' that needed to be secured and stabilised. The victim services provision most often cited by prosecutors was secure, specialised and long-term housing for domestic minor victims of trafficking. If shelter was available at all, it usually consisted of a youth shelter or shelter for victims of domestic violence that was unsecured. Much more often, victims were arrested or sent to juvenile detention as a mechanism to keep them in a secure facility long enough to get them to cooperate.
Law enforcement officials explained that arresting victims was necessary to get them to 'flip' and provide information that could lead to successful prosecution of pimps and other individuals who may be part of a larger trafficking network. As one prosecutor explained, 'You can't get there [a trafficking charge] without breaking a few eggs…at some point in time you've got to be willing to charge some of these girls with prostitution, or charge some people at a lower level to move up.' Despite the advantages of securing victims in locked facilities identified by prosecutors, the subjects we interviewed were knowledgeable and concerned about the potential for arrest or detention resulting in long-term victim harm.
In line with research on prosecutorial charging decisions for other sensitive crimes, we found numerous 'extralegal sources of uncertainty' 37 that reduced the likelihood of a human trafficking charge. For example, charges were generally more common in cases involving female victims. Contrary to the expectation from the literature and the fact that under federal law and across most states minor sex trafficking cases do not require prosecutors to prove force, fraud or coercion, we did not find a statistically significant effect of minor victims on the likelihood of prosecution for state human trafficking crimes or other crimes. Suspect race and adult age had little relationship to prosecution decisions after controlling for important legal factors. 38

Discussion and Conclusions
This research provides a preliminary, exploratory examination of how state human trafficking offences are being charged in a sample of US counties. Although many of the findings will not surprise prosecutors who are familiar with human trafficking, they provide some empirical support for a host of concerns reported through anecdotal accounts. One of the most notable findings was the fact that state prosecutors utilised human trafficking charges in only one-fifth of the human trafficking cases reviewed. Instead, a majority of human trafficking perpetrators were charged with state promotion or compelling of prostitution offences or prostitution offences. Prosecutors interviewed for this study were often the first in their state to prosecute a human trafficking case using state antitrafficking laws. State and local prosecutors were often operating on their own with little or no source of legal guidance. We also found that many state prosecutors are unaware of their own state's human trafficking laws. Further, no state prosecutors in our sample charged a case with labour trafficking. These findings suggest more work is needed to educate state and county prosecutors about human trafficking laws and how to utilise them effectively. Training to support prosecutors in successfully developing cases to charge offenders with new human 37 Albonetti, 1987, p. 311. 38 Research on prosecutor decision making for other crimes suggests that victim race is an important factor in predicting prosecutor decisions to take cases forward to prosecution (see: Spears and Spohn, 1997, Davis, 1998, Frohmann, 1997. We could not reliably measure race of the victim because this information was often missing from police reports or prosecutor records.
trafficking offences is critical to the effective implementation of these new laws. Training should include information about state human trafficking statutes including summaries of human trafficking case law and legal strategies that have been effective in securing prosecutions in other jurisdictions.
Although human trafficking is a new crime, this exploratory study identified many parallels between charging decisions in these new types of cases and patterns of charging that have been established in other types of crimes. For example, legal factors, particularly the existence of evidence, are critical to explaining variation in prosecutor decisions to pursue criminal charges. There are also important ways that human trafficking cases are distinct from other crimes. For example, victim cooperation is not independently associated with prosecutors' charging decisions in this sample of human trafficking cases, as it is in numerous studies of sexual assault and domestic violence cases. It is possible that victim cooperation alone was insufficient to bringing human trafficking cases in the absence of strong corroborating evidence. Interviews with prosecutors confirmed the need for advanced law enforcement training to foster gathering the type of evidence necessary to support human trafficking prosecutions. It is also possible that the positive impact of victim cooperation is lost in the basic, dichotomous coding scheme where a case was classified as having victim cooperation or not. Victim cooperation is varied and changes over the course of a criminal prosecution. Further research is needed to refine the collection of information regarding the quality and depth of victim cooperation. The present research did confirm that securing or, even more problematically, coercing a victim's cooperation through arrest or threat of an arrest is a primary driver of state-level human trafficking prosecution. This finding is problematic because detaining victims can re-traumatise and further harm vulnerable victims. Victims need both short and long-term shelter that will keep them safe from retaliation from their traffickers and provide them opportunity to meet their own restoration needs. Because traditional housing strategies for victims such as group shelters or residential placements may be ineffective for meeting the needs of human trafficking victims and keeping them safe, police often rely on arrest and other less favourable forms of housing, such as secured detention in treatment facilities in an attempt to secure victims.
Contrary to our expectations, indicators of human trafficking did not predict charging of human trafficking offences. In other words, the apparent strength of the case relative to the elements of the offence that must be proven did not seem to influence whether or not trafficking charges were filed. Additionally, prosecutors were less likely to file other, lower criminal charges such as pimping or promotion of prostitution when cases contained more indicators of human trafficking. It may be the case that evidence of human trafficking actually disrupts the established calculations of the likelihood of conviction that prosecutors utilise when deciding whether to prosecute a case. These findings support the notion that when faced with uncertainty prosecutors are reluctant to utilise new human trafficking laws.
The existence of physical and corroborating evidence strongly predicted prosecution in the studied human trafficking cases. Human trafficking cases necessitate the acquisition of corroborating evidence to help support a victim's statements at trial. However, this examination also highlighted that almost one-third of cases did not have physical or corroborating evidence. Absent strong corroborating evidence, prosecutors may be forced to abandon prosecution or agree to a plea to a lower-level offence, which will spare vulnerable victims from the pain of testifying. Acquiring this important corroborating evidence necessitates training law enforcement in human trafficking investigative techniques (which may differ significantly from the investigative routines of traditional vice units), proactive collaboration between the police and prosecutor to guide the collection of evidence necessary for prosecution, and the allocation of resources to support investigations.
The deficiencies in physical and corroborating evidence necessitate better support and preparing of victims, as their testimony, inevitably critical to a prosecution, is even more important when physical and corroborating evidence is unavailable. Once they have identified victims, police, prosecutors, and victim service providers should commit to long-term support for them. Required services include health, mental health, education, job training, and most importantly secure housing. Since human trafficking prosecutions are often very lengthy, a corresponding long-term victim support plan will increase participation of victims as witnesses and, thereby, the number of successful prosecutions. Even with improved victim support, prosecutors recounted serious challenges to securing credible victim testimony that may be endemic to the human trafficking victimisation experience. These limitations necessitate training prosecutors about the impact of trauma and violence on victim behaviour, and providing techniques for presenting evidence at trial, even with a victim who may be perceived as less than fully credible. This study also found, unsurprisingly, that charges were more likely to occur when there was evidence to support victim testimony. Future research should take this examination one step further and examine what evidence in particular lead to successful convictions, allowing practitioners to orient themselves and investigations to gather that type of evidence. We also found that victims of human trafficking are frequently arrested, and this arrest is associated with prosecution of human trafficking of the offender. Future research should also examine what happens after a victim is arrested. For example, are charges dropped in exchange for testimony as qualitative interviews suggest? Are victims cared for by a victim services provider following arrest? Further, we have posited that the high percentage of victim arrests has occurred because victims are initially being identified as offenders rather than victims, but future research should again go further and confirm if these victims were prosecuted and, if so, with what law they were prosecuted.
There are many challenges to the successful prosecution of new human trafficking crimes. As state and county prosecutors become more adept at bringing human trafficking cases forward to prosecution and as states affirm human trafficking convictions through the appeals process, we should expect to see routines developed that support the prosecution of human trafficking cases. Additionally, state laws have improved significantly since the first generation of human trafficking prosecutions. Many states have amended their trafficking laws to provide state prosecutors with the legal and procedural tools needed to prosecute human trafficking cases such as lower burdens of proof, safe harbour provisions and restitution. Additionally, states have expanded training for law enforcement and mandated statewide task forces to hold those responsible for enforcing new trafficking laws accountable. Although human trafficking cases may continue to frustrate prosecutors because of the many challenges endemic to this particularly nefarious crime, proper support and training, established case processing routines, and experience in prosecuting these new crimes will decrease the conditions of uncertainty that impede human trafficking and facilitate justice being served for victims.  Monica J DeLateur is a dual-degree doctoral student in the School of Criminology and Criminal Justice, and law student at Northeastern University. Her research focuses on sentencing predictors in federal human trafficking cases and the potential use of conspiracy to prosecute third-party websites for federal human trafficking.
Colleen Owens is a Research Associate with the Urban Institute's Justice Policy Center where she directs several national and international research projects on human trafficking-spanning eight countries and five continents. She currently serves as co-Principal Investigator leading a National Institute of Justice-funded study to examine the organisation, operation and victimisation of labour trafficking in the United States, and three U.S. Department of State-funded projects including an impact evaluation of a trafficking victim service provision programme in Cambodia, a process evaluation of an anti-trafficking awareness raising initiative in Brazil, and a project identifying promising practices in anti-trafficking prosecution, prevention, protection, and partnership in Italy, Colombia and Taiwan. Email: COwens@urban.org Stephanie Fahy is a graduate of Northeastern University's School of Criminology and Criminal Justice. Dr Fahy has extensive experience conducting research on human trafficking, including a national study of law enforcement responses to human trafficking funded by the National Institute of Justice and another NIJ-funded study that examined the challenges faced by state and local prosecutors investigating and prosecuting human trafficking cases. Fahy's dissertation focused on understanding criminal justice officials' perceptions and treatment of domestic minor sex trafficking victims in a state with a safe harbour law. Email: sfahy@pewtrusts.org