Exploring Gaps in Identification: Estimating the Prevalence of Sex Trafficking in Sacramento County

ABSTRACT

Human trafficking is a hidden crime, and lack of empirical data on the scope of the problem and the extent to which it is under-identified limits efforts to disrupt trafficking and meet the needs of those who have been exploited. In addition to understanding prevalence, formulating effective prevention and response efforts requires more information on the nature of people’s lived experience with human trafficking. In this multiple-methods study, we estimate the prevalence of sex trafficking and the extent to which it has been identified by law enforcement and service providers in a single county and explore what contributes to the gap in identification. We use multiple systems estimation to estimate prevalence and assess the extent of under-identification and use semi-structured interviews with people with lived experience to explore their interactions with those tasked with identifying and responding to trafficking (i.e. law enforcement and service providers). The findings suggest that only about 10% of individuals who experienced sex trafficking were identified, and both law enforcement and service providers are missing opportunities to intervene. We conclude with recommendations for policy and practice and for future research. This article addresses U.N. Sustainability Goal 16 to Promote just, peaceful, and inclusive societies.

KEYWORDS:

• sex trafficking
• prevalence
• Commercial sexual exploitation
• multiple systems estimation
• human trafficking

Introduction

Estimating the prevalence of sex trafficking has been a longstanding challenge, and reliable estimates do not exist in most communities. However, it is widely recognized that human trafficking is a hidden problem, and the true number of individuals who have experienced trafficking is much higher than those who are identified by law enforcement or service providers. For example, while the National Human Trafficking Hotline identified 11,500 potential trafficking situations in 2019 (National Human Trafficking Hotline, 2023), only 1,830 suspected trafficking cases were opened by federal law enforcement (U.S. Department of State [DOS], 2020) and only 1,883 were reported in the Uniform Crime Report (U.S. Department of Justice [DOJ], 2020). The lack of knowledge about both the scope of the problem and the extent of under-identification hampers efforts to develop effective prevention and intervention strategies. Without empirical data, efforts to disrupt trafficking or meet the needs of people who have been exploited will be driven by an inaccurate understanding of the magnitude of the problem, potentially resulting in either too little intervention or too much.

Although interest in accurate prevalence estimates has increased in recent years (e.g., Executive Order 13903, 2020; National Academies of Sciences, Engineering, and Medicine, 2020), estimating the scope of the problem alone is not sufficient to inform an appropriate response. Beyond understanding the size of the problem, it is critical to understand the nature of people’s lived experience with human trafficking in order to formulate effective prevention and response strategies. Additionally, although trafficking is understood to be under-identified, more information is needed on the magnitude and causes of under-identification. Individuals who have been trafficked may not self-identify, and distinguishing sex trafficking exploitation from other sex selling experiences can be difficult for law enforcement, first responders, health care providers, laypersons, and others in a position to help (Farrell & de Vries, 2020; Kulig, 2022). Understanding the experiences people have had in their interactions with these systems is crucial in guiding any strategic, coordinated approach to improving identification and better serving people who have experienced sex trafficking.

There is a dearth of research on prevalence, and some scholars have argued that micro-level studies, focused on a specific target population in a clearly defined geographic region, are more likely to provide valid prevalence estimates, richer information about the experiences of individuals who have experienced sex trafficking, and more actionable recommendations (Barrick & Pfeffer, 2021; Weitzer, 2014). We build on these recommendations by conducting a multiple methods study in a single county (Sacramento County, CA) to understand the scope of sex trafficking, the extent to which sex trafficking is being identified, and what may contribute to gaps in identification.

To address these questions, we used a participatory action research (PAR) approach, which included a strong practitioner-researcher partnership and survivor engagement, in order to do the following: 1) collect administrative data on individuals who have been identified as having experienced sex trafficking and (2) conduct interviews with individuals who have experienced exploitation in the commercial sex industry. The results from these complementary data collection and analysis strategies are used to generate recommendations for improving efforts to respond to sex trafficking.

Prior Research on the Prevalence of Sex Trafficking

Numerous factors make human trafficking prevalence estimation difficult. Human trafficking is difficult to define and operationalize for research purposes. Most crimes, such as burglary and sexual assault, are defined by individual incidents; however, human trafficking may involve a series of incidents over time. Yet, there is no standard threshold for determining when these events become trafficking (National Academies of Sciences, 2020). Although the Palermo Protocol established an international definition of trafficking in persons (UN General Assembly, 2000), differences in national and state definitions remain. Variation in the legal definition of trafficking can impact the availability, nature, and scope of data that is collected that may capture human trafficking victimization (de Vries & Dettmeijer-Vermeulen, 2015). Moreover, individual research studies often differ in how trafficking victimization is operationalized that may be misaligned with established legal definitions.

Beyond definitional differences, human trafficking is often a hidden crime, and those being exploited may be hard to reach with traditional survey sampling methods (Franchino-Olsen et al., 2022; Gauer Bermudz et al. (2021). For example, the National Crime Victimization Survey, an annual household survey conducted by the Bureau of Justice Statistics is one of the primary ways we know the extent of crime in the U.S., and it does not include questions about trafficking victimization. Despite challenges associated with using traditional survey sampling to access individuals who have experienced trafficking, some scholars have successfully used surveys to reach victims of trafficking in the broader population, including among adult samples and to examine the risk factors associated with these experiences (e.g., Dank et al., 2021; Kulig, 2022; Zhang, 2012).

Although the identification of trafficking victimization by law enforcement and service providers has improved over time, there is a recognition that additional training is warranted (Farrell, Dank, de Vries, et al., 2019; Franchino-Olsen et al., 2022). Thus, official crime figures from local law enforcement agencies and the Uniform Crime Reporting system represent undercounts of victimization (Durgana & van Dijk, 2021; Farrell, Dank, de Vries, et al., 2019; Tueller et al., 2021).

Despite the challenges in measurement, researchers have generated trafficking prevalence estimates in numerous ways over the past 20 years. A recent scoping review of human trafficking prevalence estimation studies identified 25 studies globally that generated an estimate of sex trafficking between 1999 and 2020 (Barrick & Pfeffer, 2022). These studies varied greatly in terms of cultural setting, geographic scope (local, national, global), age (i.e., minors and/or adults), subpopulation of focus (e.g., homeless youth, students), data sources used (e.g., administrative data, surveys), and sampling and estimation strategies (e.g., probability, respondent-driven sampling, and MSE). Depending on the method used, prevalence estimates are provided as either a point estimate or range (e.g., 650–1,000 minor victims (Farrell, Dank, de Vries, et al., 2019)) or as a proportion of a specific population (e.g., 3.5% of students in grades 7–12 (Edwards et al., 2006)). This makes it difficult to compare estimates across studies or make assumptions about whether an estimate in one community or in one population can be extrapolated to others.

Among the rigorous sex trafficking prevalence studies conducted in the U.S., most have focused on youth. A few studies included administering surveys to samples of youth using traditional probability samples (Edwards et al., 2006; Greene et al., 1999; Martin et al., 2020). The strength of these studies is the use of large, representative samples of youth. However, the studies did not focus on the same populations or use the same trafficking definitions. Populations of focus included junior high students, high school students, and runaway and homeless youth; trafficking definitions varied from trading sex for money, drugs or something else of value (Edwards et al., 2006; Martin et al., 2020) to survival sex (Greene et al., 1999). These methodological variations may explain, at least in part, the resulting disparate prevalence estimates ranging from 1.4% to 9.5% of youth. Recently, Kulig (2022) conducted a national-level survey of at-risk adult women in the U.S. and found that over one in five experienced sex trafficking during their lifetime. Additional research with adult populations is needed to assess the generalizability of these findings.

Other scholars have used sampling designs developed to identify hard-to-reach populations, such as respondent-driven sampling (RDS) and other link tracing methods that rely on peer recruitment. These techniques are useful when sampling frames are not available and traditional survey techniques are not feasible given time and cost constraints. Studies using these methods also exhibited population and definitional variation, ranging from current commercial sexual exploitation of children in New York City (Curtis et al., 2008) and Nepal (Jordan et al., 2020), current prevalence of minors working in Kathmandu’s adult entertainment sector (Dank et al., 2019), and lifetime sex trafficking among sex workers in Muzaffapur, India (Vincent et al., 2019). These methods typically yield an estimate of the number of victims rather than a percentage of the population (e.g., 3,946 youth in New York City), which limits the value of comparing estimates across studies on very different populations and social contexts.

A more frequently used technique uses administrative data rather than existing or new survey data. Multiple systems estimation (MSE) evolved from capture-recapture methods in ecology (Cormack, 1968) and requires two or more different samples of trafficking victims from administrative data (e.g., law enforcement or social service providers) and examines the extent to which individuals are captured in more than one list to generate an estimate of the size of the total population (see Baillargeon & Rivest, 2007, and Goudie and Goudie, 2007 for summaries of MSE research). Many of these studies have used administrative data on presumed or identified victims of either sex or labor trafficking to estimate the total number of trafficking victims in the U.K. (Bales et al., 2015), the Netherlands (Cruyff et al., 2017; Van Dijk & van der Heijden, 2016), Ireland (United Nations Office on Drugs and Crime [UNODC], 2018a), Romania (United Nations Office on Drugs and Crime [UNODC], 2018b), Serbia (United Nations Office on Drugs and Crime [UNODC], 2018c), Australia (Lyneham et al., 2019), and individual cities within the U.S (Phillips, 2017; Bales et al., 2019; Farrell, Dank, de Vries, et al., 2019). One of the primary limitations of these studies is that the estimates do not distinguish between individuals who were trafficked for sex or non-sexual labor. Although this provides a more holistic view of the extent of trafficking in a community, it makes it more difficult to identify where and how resources should be allocated to respond to the issue. Only two recent studies have used MSE to develop prevalence estimates specific to sex trafficking, both focused on minors (Farrell, Dank, de Vries, et al., 2019; Tueller, et al., 2021).

Sex trafficking prevalence studies have focused on various, sometimes narrow, populations (e.g., homeless youth) and were not all aimed at producing a total prevalence estimate for all individuals who experienced sex trafficking in the study site (i.e., some focused on only minors and others combined sex and labor trafficking). The field needs additional prevalence research to improve our understanding of the magnitude of the problem and how it may vary across communities or populations.

Information Supplementing Prevalence Estimates

Estimating the number of people who have experienced sex trafficking is critically important in gauging how large the response may need to be (e.g., capacity of anti-trafficking advocates to reach all of those in need, levels of funding needed). However, the estimate alone does not provide any context or nuance around the nature of exploitation, recruitment and entry into commercial sex, networks of victims and traffickers, and experiences with law enforcement and service providers. These issues are all central to determining what type of response is needed to prevent individuals from being trafficked in the first place, identify individuals who are being victimized, and provide the proper services and supports for individuals after trafficking has occurred. Yet, nearly half of the existing sex trafficking prevalence studies in the U.S. have focused exclusively on generating the prevalence estimate (e.g., Anderson et al., 2019; Bales et al., 2020; and Tueller et al., 2021).

Fortunately, others have also explored additional topics (Curtis et al., 2008; Edwards et al., 2006; Farrell, Dank, de Vries, et al., 2019; Greene et al., 1999; University of Georgia, 2020). This has included a few studies that ran additional quantitative analyses on the correlates and risk factors of involvement in commercial sex (e.g., Edwards et al., 2006; Greene et al., 1999; University of Georgia, 2020) and one that used semi-structured interviews to better understand the characteristics, experiences, and service needs of the New York City population of children who were commercially sexually exploited (Curtis et al., 2008).

Only one study has explored issues surrounding the identification of victims by law enforcement and service providers. Farrell et al., (2019) generated prevalence estimates as part of a larger effort exploring how local law enforcement agencies classify human trafficking cases in three communities. The study involved examining human trafficking case files, interviewing law enforcement and other stakeholders about how trafficking cases are identified and reported, and using MSE to estimate the number of individuals who experienced trafficking in each community in order to assess the degree to which law enforcement data capture the population in a community. They found that both law enforcement and service providers struggled to identify human trafficking cases. Law enforcement faced challenges in disentangling human trafficking from other related crimes (e.g., prostitution) whereas service providers often lacked standardized assessment protocols for identifying trafficking. In one study site, about 29% to 45% of minors who had experienced sex trafficking were identified by law enforcement or service provider records. In another site, only 14% to 18% of individuals who experienced human trafficking (sex or labor) were identified by law enforcement or service provider records. This study lays the foundation for better understanding both prevalence and identification issues among law enforcement and service providers. Additional research that incorporates is needed to illuminate why individuals who have experienced trafficking may not come in contact with these types of agencies or organizations or, if they do, may not disclose the exploitation they have experienced.

Most individuals who have been trafficked for commercial sexual exploitation are not identified as such. It is critically important to both understand the scope of sex trafficking as well as the lived experiences of those who have been exploited and why they may (or may not) present as victims to law enforcement and service providers. Collectively, this information can be used both to assess the magnitude of the problem and resources needed to respond, as well as to inform the development of strategies to improve the identification and response to individuals who have been victimized. To this end, we used MSE to estimate the prevalence of sex trafficking to which it is identified, and conducted interviews with people who experienced sex trafficking to explore the gap in identification.

Methodology

Study Design

The current study sought to address three primary research questions:

R1.

What is the prevalence of sex trafficking in Sacramento County?

R2.

To what extent are law enforcement and service providers in Sacramento County identifying people who have experienced sex trafficking?

R3.

What contributes to the gap in identification of people who have experienced sex trafficking?

R3a.

What is the nature of their experiences with law enforcement?

R3b.

What is the nature of their experiences with services in the community?

To address these questions, this study used a practitioner-led and survivor-engaged framework that was grounded in a participatory action research (PAR) approach, which is particularly well-suited to studies on hidden, marginalized, and stigmatized communities (Jumarali et al., 2021). PAR is characterized by ongoing and meaningful involvement by community members who are affected by the research (Reason & Bradbury, 2008; Stringer, 2014), which can help overcome traditional sampling and recruitment challenges with hard-to-reach populations (Gerassi et al., 2017; Martin, 2013). This study involved a practitioner-researcher partnership and included ongoing and meaningful involvement of individuals who have experienced sex trafficking through a Survivor Advisory Council (SAC). The SAC provided input on all aspects of the study (e.g., research questions, instrumentation, and recruitment protocols), conducted interviews, and engaged in the analysis and dissemination. This study was approved by Sterling IRB (IRB ID#8984). The study was also protected by a certificate of confidentiality issued by the National Institutes of Health.

The use of multiple methods was needed to address our research questions. We used MSE, which relies on multiple sources of administrative data, to estimate the prevalence of sex trafficking in Sacramento County and the extent to which individuals who have experienced sex trafficking are under-identified. Additionally, we interviewed people with lived experience in Sacramento County using a respondent-driven sampling design to gather more contextual information about their experiences with law enforcement and service providers to explore the under-identification of sex trafficking.

Administrative Data

MSE uses the overlap in multiple incomplete lists of individuals to estimate the total number of individuals who are members in the lists. “Overlap” means that individuals could be observed in more than one list (e.g., they had interactions with law enforcement and with one or more service providers). “Incomplete” indicates that no one list exhaustively documents all people who have experienced sex trafficking. These methods arose from work in ecology, in which animals were captured, tagged, and recaptured and the total population size was estimated for the recaptures. As an alternative to “recaptures,” multiple lists with overlap are typically used in human populations. MSE is often applied to human research when usual research methods such as random sampling are unlikely to work as expected. It is one of the primary methods used to estimate the prevalence of human trafficking and has been successfully applied in numerous previous studies (e.g., Bales et al., 2020; Bales et al., 2022; Chan et al., 2021; Cruyff et al., 2017; Sharifi Far et al., 2021; Silverman, 2020; Tueller et al., 2021; Whitehead et al., 2021).

Thecurrent application looked at lists of individuals identified as having experienced sex trafficking from 2015 through 2020 in Sacramento County. Our research team identified local agencies and organizations that might have administrative data with information about people who have experienced sex trafficking, including victim service providers, community health providers, law enforcement and other criminal legal – related agencies, and others. We then approached these various organizations about this study and inquired about whether their records were such that they could provide a list of individuals they know to have experienced sex trafficking. Of the 14 agencies approached, nine (three criminal justice agencies, three direct service providers, two child welfare agencies, and one health care organization) provided lists of unique identifiers representing individuals known to have experienced sex trafficking from 2015 through 2020.

Maintaining the confidentiality of individuals on these lists was paramount. Rather than requesting or collecting any personally identifiable information about sex trafficking victims from agencies that were eligible and willing to contribute information to this study, we asked that agencies instead provide a list of victims using meta-attributes to create a unique identifier in lieu of providing names or other identifying details. These unique identifiers were created using selected initials and digits from birthdates that could not be linked back to individuals. As with prior human trafficking prevalence studies in other geographic regions in the United States (see, for example, Farrell, Dank, de Vries, et al., 2019), the unique identifiers were sufficiently detailed to allow for matching across administrative lists. Data sharing agreements were negotiated individually with each agency and organization that provided data. The data were deidentified prior to sharing with the analytic team. It is also important to note that the requests for data came from a member of our research team who is herself the director of a local community-based advocacy organization, a trusted and known figure in the local anti-trafficking community. This was critical to gaining access to this sensitive data.

The identification of sex trafficking victimization varied across the data providers and the variables they had stored. Law enforcement agencies were instructed to extract all cases investigated as sex trafficking as well as cases of prostitution and pimping or pandering that included indicators of trafficking in the incident report (e.g., threat of or actual physical or nonphysical harm, demeaning or demoralizing the victim, disorienting, and depriving the victim of alternatives). Because of the variability in level of detail provided across incident reports, if there was evidence of trafficking in the incident report, the identified victim was coded as having experienced sex trafficking (i.e., they did not need to have multiple indicators present). The other data providers included all cases of sex trafficking based on their agency’s definition of trafficking and its trafficking screening procedure. The screening processes varied in rigor; they included checklists of trafficking indicators, caseworker screening, Commercial Sexual Exploitation – Identification Tool (CSE-IT), and self-identification. Variation in screening protocols may result in differing definitions of trafficking or thresholds for trafficking across the agencies and organizations that provided data. MSE cannot account for these differences, which is a limitation of this approach.

MSE Analytic Strategy

The first step when conducting MSE is determining whether the data represent an open or closed population. An open population is one that members can enter and exit during the study period (sometimes multiple times); it requires repeated observations of individuals with the same capture method. A closed population is one for which individuals were considered members at any time during the study and requires multiple independent lists like those described in the prior section. The analyses reported herein used closed population models.

MSE estimators build on several assumptions (Hall, 1977; Hook & Regal, 2000). Some can be assessed if the necessary data are available and future studies should endeavor to collect data needed to assess assumptions more fully. These include that data about list inclusion is not lost (this is possible, but unlikely for the current data with modern data collection automation and data backup); the act of being on a list does not change overall mortality rates (likely given the small capacity of services relative to demand, although as services expand they may be observed to significantly reduce mortality); and that individuals on the list mix randomly with those not on the lists (which likely in a highly mobile society like Sacramento, although data could be collected to assess geographic clustering of recipients primary domicile or work locations near service locations).

Other assumptions are less amenable to assessment as discussed by Hook and Regal (2000; see also references therein). The primary assumption that is likely always violated is that those on and not on a given list are equally likely to appear on another list (or homogeneity of list inclusion probabilities). Hook and Regal nonetheless found that under violations of this assumption, the coverage of confidence intervals is generally acceptable.

Another strength of this study is that with three or more lists, inclusion heterogeneity can be modeled using interactions between lists (Darroch et al., 1993; Hook & Regal, 2000), noting that these interactions do not have a direct interpretation and are included solely for addressing heterogeneity in list inclusion. The current study looked at two-way, three-way, etc. up to the nine-way interaction terms among the nine lists. The recent advances of Chan et al. (2019b) provided a robust method for model selection. It is necessary to note that with larger numbers of lists, sparsity between lists (up to and including non-overlap of lists) has challenged model selection, though this has been addressed by Chan et al. (2019a).

MSE analyses started by creating frequency tables containing all unique patterns of appearances in the nine lists described in the prior section. These patterns and their interaction terms were modeled using the estimate population.0 function of the SparseMSE (Chan et al., 2019a) R package (R Core Team, 2021), and the stepwise procedure of Chan et al. (2019b) was used to select the set of interaction terms that best fit the heterogeneity in inclusion between lists.

Semistructured Interview Data

In addition to collecting administrative data, this study team also interviewed a sample of people who experienced sex trafficking in Sacramento. The interviews were intended to gather additional information about the nature of sex trafficking and experiences of those impacted in the county. The population eligible to participate in interviews included English- and Spanish-speaking adults aged 18 or over who had engaged in commercial sex under the direction of a third party in Sacramento County within the past five  years. If a participant met the participation criteria, we further probed about whether they had given some or all their money to a third party in an effort to prioritize interviews with those who had been exploited. Interviews lasted between 30–60 minutes.

The interviews covered a range of topics, including nature of commercial sexual exploitation, participants’ recruitment into sex trafficking, the networks of people who both experience sex trafficking and their traffickers, experiences with services in the community, strategies for navigating an exit from sex trafficking scenarios, encounters with law enforcement, and the impact of COVID-19 on the commercial sex market.¹ The results reported here are focused on participants’ experiences with law enforcement and services in the community to explore why sex trafficking is under-identified.

Recruitment and Data Collection

Interview participants were recruited using RDS, which relies on social link tracing to sample hidden or hard-to-reach populations. The RDS recruitment process is designed to use the social networks connecting members of the target population (Crawford et al., 2018). We initially planned to use the interviews to generate a second prevalence estimate, which requires sampling from the broader population of individuals who sell or trade sex (i.e., people who have worked independently and were not forced to give some or all of their earnings to a third party). However, we decided it would be more beneficial to prioritize interviewing victims and survivors to better understand their lived experiences and needs. Their perspectives are critical in developing recommendations to improve the response to commercial sexual exploitation and sex trafficking that will be effective in addressing their needs.

Although the interviews were not used to estimate prevalence, the use of RDS allowed us to reach a larger, more representative population of individuals engaged in commercial sex under the direction of a third party than would have been accessible solely through individuals already known to the practitioner partner. Because this study was community-driven and sought to address questions that could be used for local impact, the interview protocol was developed in collaboration between the research team, the practitioner partner, and SAC. In developing items, existing instruments used to better understand human trafficking were consulted (e.g., Swaner et al., 2016; Zhang, 2012).

For this study, we began with a number of seed participants whom we believed had robust social networks of other people involved in commercial sex in or around Sacramento County. Each of these participants was paid $50 for participating in a semi-structured interview. At the end of the interview, each participant was given three coupons with unique referral codes to share with other people in their social networks who met the eligibility criteria for the study. Participants were asked to recruit up to three other people to participate in the study using these coupons and informed that they would receive an additional $25 for each person who completed an interview. Individuals who recruited the maximum of three people were given an additional $25. Those recruited participants then recruited others from their social networks to participate in the study. In this way, RDS uses existing social networks to identify and recruit members of the target population for inclusion in the study. While similar to snowball sampling, RDS uses an incentivized process that also limits the number of people recruited by each study participant in an attempt to ensure broad representation across the target study population. Participants were then given the option of scheduling an interview. Initially, all interviews were planned to occur face-to-face, but to observe COVID protocols and to increase flexibility for participants we also offered virtual interviews via Zoom. Participants were asked to bring their recruitment coupon or to provide their recruitment coupon code at the time of the interview.

Before we began an interview, we collected verbal consent from participants in an effort to maintain anonymity. Interviewers either read or played a video of the informed consent form to participants and offered an overview of the study, procedures, contact information for further questions about the project, and a copy of the consent information. From September 2021 through March 2022, we conducted 159 interviews. The vast majority were conducted face-to-face; fewer than 10 were conducted via Zoom. All interviews followed a semi-structured interview format, and with the permission of participants, interviews were recorded.

Qualitative Analysis Strategy

Recorded interviews were transcribed using an automated transcription service. A team member reviewed the transcripts for accuracy, cleaned and edited any inaccuracies, and redacted any identifying information. The de-identified transcripts were then uploaded to QSR NVivo 12, a qualitative coding software for coding and analysis. The approach we used to qualitatively analyze the data applied well-established methods (MacQueen et al., 1998; Miles & Huberman, 1994). We developed an initial deductive coding strategy, with the interview guide serving as a template for the code list. Inductive codes were added to the coding structure based on emergent themes from the interviews, such as children, a theme we did not include in our interview protocol, but which emerged in nearly one-third of our interviews.

Generally, the coding system was organized around two main objectives: (1) describing the nature of sex trafficking exploitation of participants and (2) better understanding recruitment and entry into commercial sex, networks of victims and traffickers, and experiences with law enforcement and service providers. Research team members met regularly to review and revise the codebook during initial qualitative analysis. Additionally, to ensure interrater reliability, before individual coding began, all members of the research team involved in qualitative analysis coded several interviews simultaneously and then met to discuss any discrepancies and to establish processes for areas of disagreement. We developed a series of analytic memos to explore each theme in depth. Qualitative findings are organized around these themes.

Results

We present the findings by research question. We start with findings from the administrative data, which address R1—what is the prevalence of sex trafficking in Sacramento County, and R2—to what extent are law enforcement and service providers in Sacramento County identifying people who have experienced sex trafficking. Then, we present findings from interviews, which address R3—what contributes to the gap in identification of people who have experienced sex trafficking.

What is the Prevalence of Sex Trafficking in Sacramento County?

Key to measuring prevalence using multiple systems estimation (MSE) is identifying how many times an individual appears on more than one administrative list provided by agencies that regularly interact with the target population. Nine lists include both minors and adults who experienced sex trafficking between 2015 and 2020. Although it was known administratively that lists contained minors, there was not sufficient data about birthdates and ages to disaggregate groups who were always minors from 2015–2020, who were ever minors from 2015–2020, or who were never minors from 2015–2020. Hence, a separate youth prevalence was not estimable. Table 1 summarizes this overlap, or the number of times individuals appeared in the lists. As expected, nearly all individuals appeared in only one (82%) or two (14%) of the nine lists; at maximum, two individuals appeared on five lists.

Table 1. Number of Lists in Which Minors and Adults Who Experienced Sex Trafficking Appeared.

# of Lists	Frequency	%
1	1,115	82.23
2	190	14.01
3	43	3.17
4	6	0.44
5	2	0.15
Total	1,365	100.00

Table 2 shows the number of individuals who appeared on each list, which ranged from 24 to 629 individuals.

Table 2. Frequencies by Data Source.

List	#	%
A	153	11.28
B	94	6.94
C	139	10.25
D	629	46.39
E	272	20.06
F	134	9.88
G	31	2.29
H	24	1.77
I	182	13.42
Total	1,658	100.00

For the closed population model fit to all nine lists of minors and adults who experienced sex trafficking, the estimated population size was N = 13,079, with a 95% confidence interval (CI) of 9,637 to 17,953. In other words, about 13,000 minors and adults were trafficked for sex at some point during the period 2015–2020 in Sacramento County. This does not mean 13,000 individuals are currently being victimized. It is also important to note that the length of time that individuals were trafficked varies, such that some may have been trafficked for a couple days during this period and others for the entire six-year data collection period. Because of the variation in length of trafficking exploitation, we cannot use this estimate to determine the number of people who are victimized in a single year or day.

To What Extent are Law Enforcement and Service Providers in Sacramento County Identifying People Who Have Experienced Sex Trafficking?

Perhaps even more important than the estimate of the number of people who have been victimized, is the extent to which they have been identified in the community. As shown in Table 1, the data included 1,365 unique people who experienced sex trafficking that were identified by law enforcement and/or service providers. Comparing this to the estimate of the total population (13,079) suggests that nearly 90% of the population has not been identified by law enforcement and service providers. In other words, there are 9.6 times more people who have experienced sex trafficking than have been identified and recognized as such, clearly indicating a need to better identify and serve people who have experienced sex trafficking. Through in-depth interviews, we were able to explore why the rate of identification among law enforcement and victim service providers is so low.

What Contributes to the Gap in Identification of People Who Have Experienced Sex Trafficking?

Our multiple methods approach allowed for further examination of the gap between the estimated total number of people who have experienced sex trafficking and those who had been identified and recognized as such by formal sources of help including law enforcement and victim service providers. To further understand the gap identified in our quantitative analysis of MSE data, we turned to the interview data. The interview protocol included questions about perceptions of and experiences with both law enforcement and victim service provider organizations and the barriers to accessing help from these sources.

As shown in Table 3, the interviewee sample included a group of individuals diverse in age, race, and ethnicity. The mean age at the time of the interview was 36, with a range from 18 to 64. The mean age when participants first sold or traded sex was 20, with a range from nine to 49. Respondents were asked to describe their race or ethnicity. Because the question was open-ended, responses varied greatly from individuals responding with one race or multiple specific races, reporting only Latinx ethnicity but no race, reporting specific heritages (e.g., Hebrew, Filipino), or responding “biracial” or “multiracial.” For purposes of reporting, we used the U.S. Census definition and classified individuals as being of a single race, if only one was reported, or of two or more races, if multiple races or multiracial was indicated. Similarly for ethnicity, if an individual identified as being Hispanic or Latinx, or reported Spanish-speaking lineage, they were coded as Latinx. Because we did not explicitly ask whether individuals were Latinx or not, those who did not identify as such were assumed to be non-Latinx. Our measures of race and ethnicity rely entirely on how individuals identified themselves. A large number of interviewees identified as Black or African American only (47.5%), followed by white only (12.7%) and Latinx (9.3%). Very few respondents identified as American Indian/Alaska Native Hawaiian/Pacific Islander (n = 4). The race and ethnicities reported by study respondents do not mirror those of Sacramento County. In particular, we found that Black individuals are overrepresented in our study sample (47.5% compared with 13.6% of Sacramento County residents) and white individuals (12.7% compared with 59.3% of Sacramento County residents) and Latinx individuals (9.3% compared with 18.9% of Sacramento County residents) are underrepresented (U.S. Census Bureau, 2021).

Table 3. Interviewee Characteristics.

Characteristic	#	% or mean
Age at interview	18–64	35.8
Age first engaged in commercial sex	9–49	20.4
Gender
Female	149	94.3
Male	8	5.0
Transgender	1	0.6
Race
White only	20	12.7
Black or African American only	75	47.5
Native American, Alaskan Native, or Hawaii or Pacific Islander only	4	2.5
Asian only	0	0.0
Other race only	11	6.9
Two or more races	15	9.5
Latinx	17	9.3

Most interviewees identified as female (94.3%), despite intensive efforts to include individuals who identify as male or transgender (e.g., we recruited male seeds and allowed one seed to recruit more than three individuals because they were well-connected with this group). The demographics of participants in this study are not representative of the entire population of people with lived experience of commercial sex in Sacramento County – they represent only the individuals that we were able to recruit through our RDS strategy.

Experiences with Law Enforcement

Study participants were asked about their experiences being stopped, cited, or arrested by the police, including whether they were screened for trafficking during these interactions. More than half of respondents reported having had an interaction with law enforcement, almost all of which included an arrest or citation for prostitution. In some of these interactions, interviewees were not recognized or treated as potentially having experienced sex trafficking but rather were mistreated and exploited. These experiences shine a light on why people may not feel comfortable disclosing their exploitation to law enforcement. One interviewee recounted an experience of being degraded and even spit on by an officer in the holding cell:

I ended up going to the jail, they put us all [in] this holding cell and this one officer comes and he was like, “Look at my collection of whores” and is like calling names and shit. And I’m hella close to the, um, the bars, and I was like, “What did you call me?” And he was like, “I called you a fucking whore.” And he spit in my face. This is a cop. (Female, 31)

A few respondents reported police asking for sexual favors, including soliciting them for sex. As one participant described,

So instead of asking me like, am I a victim of trafficking or anything like that, I get in one of the unmarked cars with my little handcuffs and the guy was like, “Well, we can resolve all of this right now. Um, you know, we can hit this corner.” Basically, he wanted to trade sex for my freedom. I told him, “No,” he got mad. (Female, 31)

Nearly half of interviewees who had been arrested or cited for prostitution reported having been screened for trafficking by law enforcement. However, the screening did not always result in their identification as having experienced sex trafficking or in being connected with services. Three broad themes about screening emerged: (1) many participants reported having lied during screening to protect their traffickers; (2) screening did not always prevent people from being arrested or guarantee services; and (3) when resources were provided, they were not necessarily used.

Several interviewees reported being screened (sometimes regularly) but said they would lie to protect their traffickers, saying things such as, “but I say he was just giving me a ride, but they did ask” (Female, 31) and “I never let the police know that he was like my pimp, just that he was hitting me” (Female, 24). A couple interviewees said that law enforcement knew they were pimped or trafficked, but they still would not tell on their traffickers. For example, “They knew it and they wanted him so bad and I wouldn’t tell on him” (Female, 55).

A few respondents reported that while screening for trafficking victimization occurs, it does not result in any help. For example, “They did [screen]. Yes they did. I do remember that. They did ask you that, but they don’t do anything about it … and they don’t give you no papers for resource or nothing” (Female, 44). Some recounted incidents when they avoided jail because of the screening but were still booked and fingerprinted. It is unclear how these individuals would be captured in law enforcement records – as an individual charged with prostitution or as an individual who experienced sex trafficking. Other interviewees were provided with resources; however, only some used them at the time. One participant described just throwing away the list of resources.

Interviewees’ experiences with law enforcement suggest that failures to identify someone as having experienced sex trafficking occur in a few ways: (1) individuals are afraid to report their victimization to law enforcement because of previous negative encounters; (2) police fail to screen for trafficking or, even worse, abuse or exploit them; and (3) police screen for trafficking but those who are screened are not ready for help and evade detection to protect their trafficker.

Experiences with Services in the Community

Although law enforcement represents one mechanism for identifying people who have experienced sex trafficking, a variety of community, social services, and health care providers may also interact with this population. Study participants were asked about their awareness of and experiences receiving services in the community. Their experiences suggest that many never accessed services for a variety of reasons. Some interviewees said they had not heard about services or providers, and some reported that it had not occurred to them that help might be available. One person explained, “No, that’s not something I thought about. I never thought like, oh, I could go and get help somewhere. There would be a resource to help me. I never thought that” (Female, 31). Some respondents described being in “survival mode” and not thinking about seeking out formal help.

It was just, “Get the fuck up out of here. Run away.” But it was never like, “Look for help.” You know what I mean? I mean, it’s awesome that you guys [Practitioner Partner] have all these programs now and all that, but that was never something that crossed my mind. (Female, 33)

Similarly, another interviewee’s experience exemplifies how extreme self-reliance can be a barrier to seeking help, which is consistent with some prior research (Ijadi-Maghsoodi et al., 2018; Labouliere et al., 2015).

I never knew about no services or programs … . I’m the type of person that I never expect nobody to do nothing for me ’cause nobody never did nothing for me … . So, I never asked nobody about a program or looked for a program. (Female, 27)

Of those aware of local service providers, many respondents said they had not tried to access services. Some respondents were unsure how; as one respondent stated, “Nobody ever schools you about help” (Female, 63). Some interviewees did not think they needed help, and some reflected that they were not ready or in the right mind-set to engage or take advantage of services at the time they were available. A similar barrier that respondents identified was a lack of understanding of their trafficking situation (e.g., not knowing what trafficking is, not recognizing their own experience as trafficking, not understanding the severity), which made it hard to see services as useful.

I never even thought about the situation being this way … . I didn’t look at it as a problem. I say I looked at it as a weird situation that I was in, but I didn’t, you know, you wouldn’t think you[r] girl being a pimp you know or anything like that?… It’s like, wow. Mind blowing. Is kinda like, I never knew, but I did know kind of [the] situation. (Female, 27)

Other respondents were aware of and tried to seek services but reported that it can be hard to get into programs (e.g., due to restrictive eligibility criteria) or find immediate assistance. For example, one interviewee – describing herself as single and without kids or substance use needs – said:

I didn’t qualify for anything … . Things have to get really, really bad for you to get any type of help. Like, not trying to, like, put a level of how bad things get on life, but you have to get to those points to get help. (Female, 22)

Participants also described scenarios reaching out to providers, but their calls were not answered or returned, or they had to call multiple times. One interviewee described her experience calling an organization about two  months before the interview: “I contacted them as well and they told me they’d put me in their case management, but no one’s ever reached back out” (Female, 27). Several respondents also said they were told they would be put on a waitlist or that they were denied program enrollment. For some, this response deterred them from seeking help in the future. For example,

I got discouraged with [organization], and they’re supposed to be, like, the top kind of thing. When they denied me, it kind of discouraged me from looking anywhere else because, man, if they’re gonna deny me of that, who else is gonna, they’re not gonna want me either, you know? So, it kind of discouraged me. (Female, 44)

For others, they were deterred not by direct experience but by hearing about these negative experiences from friends or other people in their network.

Interviewees' descriptions illustrated the various circumstances that prevented them from receiving the services and assistance they needed from community providers. Many individuals were either not aware of or chose not to access services. However, even those who were actively seeking help may have had difficulty getting in touch with a service provider or been denied entry into a program. This not only prevented them from receiving timely assistance, but it also may have discouraged them from seeking help in the future. These were all missed opportunities to better identify individuals who have experienced sex trafficking provide them with the support they needed.

Discussion

Understanding the scope of sex trafficking is important for developing adequate and strategic responses that are aligned with the magnitude of the problem. Micro-level studies, focused on a specific target population in a clearly defined geographic area, are likely to be the most feasible for calculating robust estimates and may be more useful in developing specific and actionable recommendations. We have built on and expanded prior U.S. prevalence estimation research by estimating the prevalence of sex trafficking among minors and adults in Sacramento County and exploring reasons for under-identification.

Approximately 13,000 minors and adults were trafficked for sex in Sacramento County at some point during the period 2015–2020. This analysis further suggests that there are nearly ten times more people who have experienced sex trafficking than were identified by law enforcement and service providers. This demonstrates that a significant gap exists between how many community members are affected by sex trafficking and how many are currently being identified and served. Law enforcement frequently fails to screen people for trafficking, and in some situations, they mistreat individuals involved in commercial sex. Even when identification occurs, it does not always result in help. Although community service providers represent another line of protection for people who have experienced sex trafficking, individuals are not always aware of services or encounter challenges when they try to access them. Both law enforcement and service providers are missing opportunities to help people in the moment when they most need it. These actions (or lack thereof) further deter individuals from reporting their victimization or seeking assistance in the future. We need strategies that improve identification of sex trafficking and intervention.

Recommendations for Policy and Practice

Law enforcement and the criminal legal system are frequently tasked with leading identification and intervention efforts (de Vries et al., 2023; Farrell & Pfeffer, 2014). One of the most important ways law enforcement officers can make a difference is by treating all people involved in the sex trade with dignity and respect. Treating individuals in this industry with anything less can cause those who have been exploited to avoid reaching out to law enforcement for help. Just like individuals with an addiction, it may take multiple contacts before an individual is ready to seek or accept the assistance they need. If people who experience sex trafficking are demeaned, this may make them reluctant to seek help from law enforcement due to lack of trust. Treating people with dignity and respect can help establish a foundation of trust for them to seek help when ready.

Law enforcement should also consider every contact made with individuals involved in commercial sex as an opportunity to screen for sex trafficking. Moreover, regardless of the outcome of the screening or the law enforcement officer’s personal assessment, services and resources should be offered to meet their identified needs. Although a particular moment may not be the right time for them to seek help or leave due to a variety of personal reasons, offering services on every contact can help establish a trusting relationship and educate people on what resources are available for them should they decide to seek assistance. Further, these efforts should not be limited to human trafficking specialty detectives as it is patrol and other front-line officers most likely to make contact with individuals (Farrell & Pfeffer, 2014) who are then, ideally, able to refer suspected cases to more specialized officers with expertise on human trafficking-related exploitation.

This study also supports the perspective that law enforcement should not have primary responsibility for responding to sex trafficking. Prior research finds that a significant portion of human trafficking case referrals received by law enforcement come from other community-based and systems-based community agencies (Farrell et al., 2019) and that there are inherent challenges related to taking a criminalized approach to addressing sexual exploitation at all (de Vries et al., 2023). Strategies to remove barriers to community services must also be prioritized and addressed. First, providers need to increase and vary dissemination efforts to improve awareness of what services are available and how to access them. In addition to formal awareness-raising campaigns, word of mouth can be a powerful dissemination tool. It is important to recognize that people not only will hear about positive service experiences, but they may also hear misinformation and about negative experiences. It is harder for service providers to establish a reputation as a trusted community resource than as an organization that is not helpful, not accessible, or unavailable.

It is also critical to reduce obstacles to access services. When a person is finally in a position to reach out for help, the response should not be “Call back” or “We can put you on a waitlist.” For many, having to repeatedly reach out for assistance quickly leads to a dead end, and they will not receive the critical help they need. Providers that are at capacity or cannot meet someone’s immediate needs can help them connect to another provider with availability if there are strong partnerships and coordination among local providers. Providers must always follow up when they say they will. People who have experienced sex trafficking understandably often feel mistrust toward people. Not following through can exacerbate that feeling and reduce their likelihood of trying again.

Recommendations for Future Research

Policymakers are increasingly calling for better understanding of the prevalence of sex trafficking to support decision making on prevention and intervention strategies, including resources needed to tackle this issue. Clearly, additional human trafficking prevalence estimation research is needed. Although some scholars strive for global or national estimates, smaller-scale studies offer several advantages. First, collecting the amount of data necessary to generate a reliable prevalence estimate is an enormous undertaking. Large, representative samples are needed for survey or interview approaches. These are challenging at the local-level and scaling up to the state- or national-level is likely cost prohibitive because specially targeted recruitment strategies may be needed to access different populations. Although some school-based surveys include indicators for trading sex (e.g., AddHealth, Minnesota Student Survey), they do not capture the most marginalized and at-risk populations, such as runaway youth. The same would be true if trafficking indicators were incorporated into the National Crime Victimization Survey.

Except for some countries with robust and linked human trafficking data repositories (e.g., Cruyff et al., 2017; UNODC, 2018a; UNODC, 2018b; UNODC, 2018c), conducting MSE at a large-scale is similarly challenging. A challenge inherent in MSE, which is perhaps even more serious in large-scale studies, is the variation in how trafficking victimization is defined by the different data providers (e.g., short screener, full assessment tool, self-identification). Another limitation of MSE is that there may be certain subpopulations who do not make contact with any of the organizations providing administrative data. Moreover, the MSE would ideally include data from all agencies and organizations that interact with people who experience sex trafficking. A topic that has not yet been studied in the MSE literature is how sensitive MSE estimates are to the inclusion or exclusion of different lists. One likely reason impeding this type of research is that most studies that use MSE draw on a very small number of lists (a notable exception is K. Bales et al., 2020, who reported using eight lists with a total n = 172). A comprehensive simulation study is needed to evaluate MSE estimate sensitivity to list inclusion, exclusion, and patterns of overlap between included or excluded lists.

In addition to being more feasible, another benefit to micro-level studies is their ability to inform local or state response. Because crime is not evenly or randomly distributed across communities, it is unclear how policymakers and practitioners could turn a national or global estimate into actionable recommendations. However, understanding the scope and extent of under-identification in a community can be leveraged for additional funding for service providers or specific law enforcement trafficking personnel or units. Even more beneficial is collecting data that tell us more than the number of people victimized or proportion of a population that is victimized. The prevalence estimate alone does not provide any context or nuance around the lived experience of those who experience sex trafficking, which is critical to determining what type of response is needed to prevent individuals from being trafficked in the first place, identify individuals who are being victimized, and to provide the proper services and supports for individuals after trafficking has occurred.

There are numerous methods available for estimating prevalence, and it is beyond the scope of this paper to weigh in on the strengths and weaknesses of each; that has been done elsewhere (e.g., Barrick & Pfeffer, 2021; Global Fund to End Modern Slavery, 2021). However, we cannot overstate the importance of collaborating with practitioners and engaging people who have experienced sex trafficking in this type of research. Adopting a PAR approach has numerous benefits for both the research study and the community. For researchers, this approach works particularly well on topics that involve hidden, marginalized, and stigmatized communities, like sex trafficking. Without community engagement, researchers may struggle to secure buy-in from the community or even develop successful strategies for outreach. For the community, participating on a research team can provide opportunities for leadership roles and other professional development. It also allows them to be directly involved in identifying practical and effective recommendations because they are intimately familiar with what has already been done and what might not work. In this type of research, it is essential to recognize that people with lived experience have important expertise that researchers do not.

Acknowledgments

We would like to thank the members of our Survivor Advisory Council, who provided critical input at all stages of the project and helped generate the recommendations for policy and practice presented here: Sawan Vaden, Charity Harmon, Antoinette Smith, Makayla Hines, Ashley Green, Breana Baker, Dominique Roseborough, April Grayson, and Kayla Baumann. This project would not have been possible without them.

Disclosure statement

No potential conflict of interest was reported by the authors.

Additional information

Funding

The work was supported by the California Department of Justice .

Notes

¹ The interview protocol is available upon request.