The same ruler for everyone: improving trafficking estimates

Current guidelines for measuring the prevalence of trafficking are inadequate. Improving the accuracy of trafficking estimates will require comprehensive, standardised guidelines which have been rigorously tested in the field.

Donors increasingly call upon their grant recipients to conduct survey research in order to estimate the number of victims of trafficking in a target region or industry. Donors use these estimates to plan resource allocation, inform programme design, and engage with other governments. They also look to these estimates for empirical proof that an anti-trafficking programme is reducing the prevalence (the number of detected victims) of trafficking. However, most organisations do not include prevalence of victims as an indicator of success because there are no comprehensive, standardised guidelines which academics or researchers employed by these organisations could feasibly use to implement prevalence estimate methodologies. This means researchers develop survey instruments and methodologies that, although context-specific, do not benefit from systematic field-testing, are difficult to broaden out for general use and do not lend themselves to prevalence comparisons. Those prevalence estimates that have been published so far have faced criticism and scepticism. Existing guidelines are not inclusive and are challenging for smaller organisations (a considerable proportion of those engaged in fighting trafficking) to implement.

Current metrics

The Global Slavery Index (GSI) provides national and regional estimates of the number of people in modern slavery.[1] It receives considerable media attention and its figures are frequently cited by organisations and governments. The most granular estimates that the GSI provides are at a national level. In order to provide a useful metric of success for an anti-trafficking programme, the programme would need to be implemented at that same level. However, most anti-trafficking organisations usually implement programmes at a sub-national level, focusing on a handful of municipalities and often a specific type of trafficking. It would be difficult to demonstrate using the GSI data that any such anti-trafficking programme contributes to a decrease in national prevalence, given the many other forces at play on a national level. While the GSI continues to refine its methods and is likely to be able to produce sub-national estimates in the future, due to the immense expense of compiling the index the exercise is only carried out approximately every two years. For an organisation to use it as a baseline the organisation would need to coordinate its intervention with timing of GSI data collection, which is not always possible. Thus for national or regional programmes, the GSI figure can only provide indications of trends.

In 2012 the International Labour Organization (ILO) published survey guidelines[2] which contain detailed steps for national governments to implement a survey on forced labour, including definitions, assessment tools, ethical considerations and guidance on data analysis. Although useful for governments looking to obtain the full picture of forced labour in their country the guidelines are limited. The ILO itself called these guidelines “a starting point” and we believe it is now time to expand them to include additional methods targeting a wider research audience. The International Conference of Labour Statisticians, which is convened by the ILO, published guidelines on the measurement of forced labour in 2018 that largely focus on randomised household or establishment surveys.[3] However, most organisations or government agencies focused on trafficking have limited budgets and cannot afford to implement these. Consequently, many organisations end up using referral-based sampling methods (such as snowball sampling), which have been proven in other fields to better identify members of a hidden population. Other means of collecting data to measure prevalence are generally implemented in an ad hoc manner and rarely have a direct connection with a programme.

Legal definitions

A common statistical definition of trafficking has eluded researchers, in part due to the existence of different and sometimes overlapping international and national definitions. For example, international law uses the definition of forced labour in ILO Convention 29, requiring both involuntariness and the threat of a penalty; instead the Palermo Protocol[4] defines trafficking as requiring an act, means and purpose. Despite the technical language, forced labour and labour trafficking both refer to compelling (and profiting from) the labour or services of others. While the ‘means’ of force, fraud or coercion and the ‘purpose’ of compelling individuals for labour services might be the same across both definitions, confusion for measurement arises around the ‘acts’, such as recruitment, transportation, harbouring or receipt, which are not part of the ILO’s definition. This raises key questions for researchers: whether, for example, scientists should measure forced labour by the ILO’s standards in order to demonstrate the prevalence of trafficking; and whether scientists need always to include the ‘act’ in surveys.

Another difficulty is applying international norms at a national level. Under the Palermo Protocol, movement is not required to constitute trafficking. However, some national governments have enacted domestic anti-trafficking legislation that does require some form of movement. These varying definitions can be an obstacle to researchers comparing national administrative data.

Forced migration and trafficking

Movement makes it more difficult not only to measure vulnerability and prevalence within a population, and to compare results with other studies but also to reach vulnerable populations within migration flows. Many studies attempt to measure prevalence of trafficking by interviewing returnees, who can be more accessible than workers in destination countries or those who are in transit. However, research with returnees must take into account several factors that may limit the study. For example, exploited individuals may have small social or isolated networks and if using a referral or network-based sampling method, the resultant prevalence estimate will probably be too low. That research may have identified how many returnees themselves were subject to exploitation but not have captured the total trafficked population in the migration flow.

Additionally, forced migrants often do not use formal channels but instead move irregularly and may use smuggling networks. Forced migration routes may differ in each scenario, exacerbating the measurement difficulties of referral-based sampling and severely limiting the ability to conduct traditional household or establishment surveys.

Common methodologies

Trafficking researchers have used a wide variety of methods to calculate prevalence estimates. The most traditional survey sampling methods select a random sample of households or establishments from a complete list, such as a government census. Using their survey questions, they attempt to identify respondents who may be victims of trafficking. Because everyone in the population had the same likelihood of being included in the survey, the researchers are able to generalise the results to a larger population. Hidden populations, however, who are often obscured from these lists, make this problematic and random sampling is therefore likely to under-report trafficking victims.

Snowball sampling entails asking respondents to report on what they know of the trafficking experiences of members of their social networks. With enough waves of sampling the results are seen as being representative of the general population, even though not every member of the population had an equal probability of being selected. However, snowball sampling in anti-trafficking studies tends to start with a known group of people who are connected to victims or who are victims themselves. Consequently, the final estimate is likely to over-report prevalence of trafficking in the general population. The true prevalence estimate is likely to be somewhere between the estimates produced by random and snowball sampling.

Multiple systems estimation is based on lists of trafficking victims detected and recorded by local authorities. It requires a country to have at least two lists from different sources with a minimum of around 80 victims.[5] This technique is most frequently employed in high-income countries that have strong data systems.

While these methods provide a tremendous insight into the prevalence of trafficking, no single method provides a complete picture of the extent of the crime or of the victims’ experiences. Each method has advantages in relation to evaluating certain types of crimes, environments or populations and, correlatively, each has its drawbacks. The chief difficulties have centred on researchers’ lack of awareness of situations in which individuals are particularly vulnerable to being trafficked and the frequent inability or unwillingness of respondents to identify themselves as having experienced trafficking. This may be because of fear of stigma (no matter how strong the assurances of confidentiality); because respondents may not know that they are victims of trafficking; or because they fear retribution from their trafficker.

Another problem is the reliance on primary data for measuring programme success. Organisations increasingly see the value and importance of strong data collection and management practices, but it will indeed take a sea change before best practice principles are integrated into every organisation. If data collection and management practices were stronger, these could be heavily relied upon to validate estimates or even to develop estimates where there is insufficient budget to carry out new data collection.


This leads us to make three recommendations that we feel would help drive better data collection and management and lead to more accurate estimation of trafficking prevalence:  

Develop comprehensive standardised guidelines: Given that no methodology provides a fully comprehensive estimate, guidelines are needed in order to help groups determine which method to use based on demographic factors, type of trafficking and their budget and timing constraints. These guidelines should be evidence-based, meaning that there needs to be further testing of each method. They must focus on returning the most precise estimates possible, in order for researchers to be able to confidently detect the effect that a programme has had on prevalence. Survivors should be consulted in the development of the guidelines and in the implementation of each methodology.

Utilise an impact evaluation methodology: Conducting and comparing prevalence estimates at the beginning and end of programme implementation can show an increase or a decrease but only through setting up a control group can a change in prevalence be attributed to the impact of the programme. While this impact evaluation methodology does not have to be used for every programme it should certainly be used for new programmes and to inform decisions about whether to expand existing programmes.

Involve new academics: Trafficking is an interdisciplinary topic that would benefit from the focus of a greater number and variety of researchers. When looking for partners, organisations should target emerging scholars and academics not customarily engaged in the trafficking arena, possibly from fields such as criminology, sociology, social work, economics, demography and public health. Their fresh perspectives can unearth new insights and help make much-needed progress. Combatting trafficking requires a scientifically rigorous, interdisciplinary response that does justice to the experiences of the victims.

Joshua Youle

Abigail Long

Program Advisors, Everwatch Solutions Corporation and Cherokee Nation Mission Solutions, Contractors, US State Department, Office to Monitor and Combat Trafficking in Persons

The views expressed in this paper are those of the authors and are not an official policy nor position of the US Department of State or the US government.


[1] The Walk Free Foundation (2013) Global Slavery Index

[2] ILO (2012) ‘Hard to see, harder to count’

[3] ILO (2018) ‘Guidelines concerning the measurement of forced labour’, 20th International Conference of Labour Statisticians. Geneva, 10–19 October 2018

[4] UN General Assembly (2000) ‘Protocol to Prevent, Suppress and Punish Trafficking in Persons Especially Women and Children, supplementing the United Nations Convention against Transnational Organized Crime’ (known as the Palermo Protocol)

[5] Johansen R (2019) ‘UNODC’s use of Multiple Systems Estimation (MSE) to assist countries in measuring human trafficking and reporting on SDG indicator 16.2.2’


Opinions in FMR do not necessarily reflect the views of the Editors, the Refugee Studies Centre or the University of Oxford.
FMR is an Open Access publication. Users are free to read, download, copy, distribute, print or link to the full texts of articles published in FMR and on the FMR website, as long as the use is for non-commercial purposes and the author and FMR are attributed. Unless otherwise indicated, all articles published in FMR in print and online, and FMR itself, are licensed under a Creative Commons Attribution-NonCommercial-NoDerivs (CC BY-NC-ND) licence. Details at



facebook logo Twitter logo RSS logo email.png

Forced Migration Review
Refugee Studies Centre
Oxford Department of International Development
University of Oxford
3 Mansfield Road, Oxford OX1 3TB, UK  +44 (0)1865 281700