Symposium: The Benefits and Limitations of Modelling the Risk of Modern Slavery

11 December 2018
Research Innovation

Laura Gauer Bermudez  | Director of Evidence and Learning, the Global Fund to End Modern Slavery
Shannon Stewart  | Senior Data Scientist, the Global Fund to End Modern Slavery

Pablo Diego-Rosell and Jacqueline Joudo Larsen present a model that uses a hierarchical Bayesian approach to estimate the risk of modern slavery at the national level using responses to household surveys conducted by Gallup. Based on previous work by the Walk Free Foundation, the authors identify five major categories of vulnerability risk factors: governance, access to social services, inequality, disenfranchisement and conflict. From these, they select 18 variables from a larger group of 157 to include in the model, considering both theoretical and practical implications. Given the cross-sectional nature of the data, variables could show both direct or inverse relationships with forced labour and forced marriage. Using this risk model, the authors then generate a predicted probability of modern slavery by country, where sufficient data is available.

Diego-Rosell and Joudo Larsen’s effort is significant and benefits the modern slavery community in several ways. The extensive explanation of statistical methods demonstrates a thoughtful approach to modelling modern slavery risk using existing survey data. As we seek to move the sector towards data-driven action, this commitment to quantitative rigour is particularly important. Risk modelling is necessary to inform the design and targeting of modern slavery prevention programs so they more efficiently and effectively mitigate risk. Further, the inclusion of psychographic assessments adds further benefit to the analysis, placing value on respondent perceptions in addition to standard demographic assessment metrics.

However, this approach also has a number of limitations. First, the model uses cross-sectional data, meaning the relationships between risk variables and outcomes may be bi-directional. For instance, higher negative experience index scores and difficulty living on present income are associated with a higher likelihood of being engaged in forced labour in the present model. It is quite plausible that engagement in forced labour has caused those outcomes for the respondent versus those characteristics being predictive of entry into forced labour. The authors recognize this limitation, which remains a challenge for predictive analytics within the social sciences more broadly.

In future work, researchers could benefit from selecting variables of interest that are more static and less subject to the potential inverse relationships that make the operationalization of these findings challenging. Further, due to the uncertainty around the causal direction of these relationships, significant limitations exist with respect to the authors’ extrapolation of findings from their risk model towards prevalence estimation. The modern slavery sector must take care when interpreting or extrapolating results from cross-sectional data, acknowledging the inherent uncertainty that comes with predictive modelling and continuing to craft improved models and refine methods.

The second limitation of the findings is the lack of potential application. Modelling modern slavery risk at the global level can be problematic if variables are prioritized for their suitability to standardization across national-level data sets, posing the risk that they be too generic to be actionable. Nuanced and highly contextualized risks factors exist that a set of indicators designed to be relevant globally may not be able to capture. The examination of HIV risk in a public health setting can offer parallels, as research has shown demographic risk factors to be highly variable by region. For instance, adolescent girls and young women in Southern Africa are proportionately at far higher risk of HIV infection than women and girls of the same age in any other location. Similarly, in this context of modern slavery, the authors’ standardized indicators across multiple countries/regions may lose the nuance required to be explanatory, particularly when aiming to address modern slavery within a specific geographic/industry nexus.

Another finding in the study that demonstrates this limitation is that education and youth development were found to be associated with lower levels of forced marriage. While this relationship is again subject to the challenge of bi-directionality, should further research establish these variables as predictors of forced marriage, they are also known indicators of a wide range of poor outcomes including violence victimization, disease and food insecurity. Such generic indicators can certainly support a broader global development narrative, but they are likely less insightful for modern slavery actors. Future efforts may want to consider engaging with end-users of a model to determine what type of data would be helpful for targeting programs aimed at preventing or reducing modern slavery; this approach would blend traditional academic and modern data analytic approaches in a way that could be highly actionable.

Despite these limitations, the analysis and results still warrant dialogue. While certain correlates are unsurprising—such as women being at lower risk of forced labour or higher levels of education being associated with lower rates of forced marriage—one finding in particular may be novel for the modern slavery community: individuals with higher scores on a Community Engagement Index have lower probability of being associated with forced labour and forced marriage. While this relationship is, again, subject to the limitation of being inversely correlated, it poses a unique suggestion about the potential power of social inclusion and community support as protection against modern slavery and warrants further investigation.

Overall, Diego-Rosell and Joudo Larsen’s work represents a valuable first step to systematically identify vulnerabilities to modern slavery. Their effort lays very important groundwork from which the sector can learn, modify and improve – particularly as we seek to model risk at a more micro-level, interrogate vulnerability within specific industries and geographies, and identify findings to compel legislators, policy-makers, business leaders and civil society to action.

This piece has been prepared as part of the Delta 8.7 Modelling the Risk of Modern Slavery symposium. Read all the responses here.

Laura Gauer Bermudez is the Director of Research and Development at the Global Fund to End Modern Slavery.

Shannon Stewart is a Senior Data Scientist at the Global Fund to End Modern Slavery.

This article has been prepared by Laura Gauer Bermudez and Shannon Stewart as contributors to Delta 8.7. As provided for in the Terms and Conditions of Use of Delta 8.7, the opinions expressed in this article are those of the authors and do not necessarily reflect those of UNU or its partners.

The Delta 8.7 Forum

Can New US Law Help Increase Financial Recovery and Reintegration of Survivors of Human Trafficking?

Professor Barry Koch, Dr Leona Vaughn, Sarah Byrne
Continue Reading

Gendered Understandings of Forced Sexual Exploitation

Ellie Newman-Granger
Continue Reading

Forced Labour Import Bans: What Does the Evidence Tell Us?

Owain Johnstone
Continue Reading

Gendered Understandings of Forced Sexual Exploitation

Ellie Newman-Granger
Continue Reading

Domestic Slave Labour in Brazil

Maurício Krepsky Fagundes
Continue Reading

Indigenous Peoples and the Anti-Trafficking Sector’s Blind Spot

Miriam Karmali, Krysta Bisnauth
Continue Reading