Global Trafficking Prevalence Data Advances the Fight against Trafficking in Persons
DOI: 10.14197/atr.201217811

Global Trafficking Prevalence Data Advances the Fight against Trafficking in Persons1

Courtland Robinson, Casey Branchini and Charlie Thame

Response to the ATR Debate Proposition: ‘Global Trafficking Prevalence Data Advances the Fight against Trafficking in Persons’

Please cite this article as: C Robinson, C Branchini and C Thame, ‘Global Trafficking Prevalence Data Advances the Fight against Trafficking in Persons’, Anti-Trafficking Review, issue 8, 2017, pp. 157—160,

We defend the proposition that global trafficking prevalence data—when gathered using validated methods and presented with sufficient detail on study design and data analysis—can advance the fight against trafficking in persons. Space does not allow us to engage in a full review and critique of existing data and methodologies, but we subscribe to the view that the field in general suffers from ‘“epidemiological anaemia”—lack of primary data collection based on sound sampling procedures’.2 We would add another condition: demographic disorder—unsystematic use and interpretation of population data. Until that situation improves, and it can best do so through systematic application of qualitative and quantitative ‘microlevel research’,3 we recommend that existing global prevalence data be presented with clearer caveats and used with due caution.

Estimates of the global prevalence of human trafficking have varied widely, as has the credibility accorded them. The 2012 International Labour Organization (ILO) global estimate of 20.9 million4 people in situations of forced labour (including human trafficking) is commonly cited and among the most respected estimates (though by no means universally so).5 On the other end of the spectrum is the Global Slavery Index, published by the Walk Free Foundation, which estimates 45.8 million people in situations of ‘modern slavery’ (including human trafficking) worldwide.6 Gallagher describes the GSI estimate as based on ‘a mysterious, inconsistently applied methodology, a raft of unverified assumptions, and multiple, critical errors of fact and logic’.7

Despite the challenges of measuring human trafficking, and the rather unreliable global estimates available at present, we agree with Sheldon Zhang that ‘we should not abandon macro-level estimation just because it is full of problems’.8 Global figures are helpful for advocacy purposes, for allocating resources, and for tracking global, regional, and national trends. We offer several recommendations for improving prevalence estimates of human trafficking:

  1. Start small. Weitzer articulates several advantages of ‘microlevel research’: estimates may be more reliable (‘because of the limited parameters’); they provide ‘richer insights regarding actors’ lived experiences’, whether these be migrant workers, brokers, employers, or local authorities; and they may provide useful data for identifying risk factors and for targeting interventions, including prevention, protection, and/or prosecution activities.9
  2. Clarify terms, objectives, and contexts. Any measure of human trafficking needs to identify the terms and definitions used and how they are operationalised for estimation purposes. If there is a local law defining trafficking, that should either be incorporated, or an explanation provided as to why it is not deemed appropriate. As the UN Trafficking Protocol is the prevailing international instrument and definition, we recommend its use in measurement so that data can be compared across sites. Finally, study objectives and contexts need to be clarified in order to set out organisational priorities (and possible biases) and local factors that may affect measurement.
  3. Triangulate and validate methods. A range of both qualitative, quantitative, and mixed methods are available to estimate prevalence, including household surveys, surveillance, and institutional registries. Similarly, probability and non-probability sampling options abound, though not all are equally suited to measure prevalence among a given population in a given locale, or equally valid for purposes of statistical inference and extrapolation. A direct measure of either point prevalence (cases in a population at one point in time) or period prevalence (cases identified during an interval, often one year) is most commonly done by taking a sample of a population. Because human trafficking is a crime, trafficked persons may be hidden or hard-to-reach, thus necessitating the use of adaptive sampling methods, like capture-recapture10 or respondent driven sampling.11 These methods, along with various approaches to model population estimates using single or multiple sources of data, need to be triangulated and validated in diverse field settings to learn which yield the more accurate results.
  4. Share data and results, critique, and collaborate. As human trafficking measurement proliferates (and, we can hope, improves), it is critical to share data and results and to critique (through the peer review process, workshops, and conferences) methods and findings. Disagreements abound in the field of anti-trafficking but that makes it all the more imperative to collaborate in field work and discovery.
  5. Iterate. No one should pretend that the process of improving the measurement of human trafficking will be straightforward or simple. Donors need to ensure that research is well-supported and researchers need to ensure that studies are both rigorous and helpful for programme and policy decision making. Everyone needs to ensure that each new study learns from what has been done before and informs what comes next.

Courtland Robinson, Ph.D., is Associate Professor, Department of International Health, Health Systems Program, Johns Hopkins Bloomberg School of Public Health, and core faculty in the Center for Humanitarian Health. Email:

Casey Branchini, MHS, is a Ph.D. candidate at the Johns Hopkins Bloomberg School of Public Health. Email:

Charlie Thame, Ph.D., is a lecturer in the Faculty of Political Science at Thammasat University. Email:


1 Some of these ideas were advanced, though in a different context and with a country-specific focus, in a report presenting results of a stakeholder analysis of anti-trafficking activities in Thailand. See C Robinson, C Branchini and C Thame, Anti-trafficking in Thailand: A stakeholder analysis of Thai Government efforts, the U.S. TIP Report and rankings, and recommendations for action, Johns Hopkins Bloomberg School of Public Health, Baltimore, Maryland, 2016, retrieved 30 June 2016,

2 S X Zhang and L Cai, ‘Counting Labour Trafficking Activities: An empirical attempt at standardized measurement’, Forum on Crime and Society, vol. 8, 2015, pp. 37—61 at p. 39.

3 R Weitzer, ‘New Directions in Research on Human Trafficking’, The ANNALS of the American Academy of Political and Social Science, vol. 653, issue 1, 2014, pp.6—24 at p. 15.

4 International Labour Organization, ILO Global Estimate of Forced Labour: Results and methodology, ILO Special Action Programme to Combat Forced Labour, Geneva, 2012, retrieved 30 June 2016,

5 A Gould, ‘From Pseudoscience to Protoscience: Estimating human trafficking and modern forms of slavery’, Second Annual Interdisciplinary Conference on Human Trafficking, 2010, Paper 6, p. 36.

6 Walk Free Foundation, Global Slavery Index Report 2016, Perth, 2016, retrieved 29 June 2016,

7 A Gallagher, ‘The Global Slavery Index: Seduction and obfuscation’, openDemocracy, 4 December 2014, retrieved 30 June 2016, See also A Guth, R Anderson, K Kinnard and H Tran, ‘Proper Methodology and Methods of Collecting and Analyzing Slavery Data: An examination of the Global Slavery Index’, Social Inclusion, vol. 2, issue 4, 2014, pp. 14—22.

8 G Kessler, ‘Why you should be wary of statistics on “Modern Slavery” and “Trafficking”’, The Washington Post, 4 April 2015, retrieved 30 June 2016,

9 R Weitzer, p. 15.

10 E B Hook and RR Regal, ‘Capture-recapture Methods in Epidemiology: Methods and limitations’, Epidemiology Review, vol. 17, 1995, pp. 243–264.

11 D D Heckathorn, ‘Respondent-driven Sampling: A new approach to the study of hidden populations’, Social Problems, vol. 44, no. 2, 1997, pp. 174-199.