Modelling Vulnerability to Slavery in Brazil (RADAR)

Slavery is very much present and prevalent in the 21st century, presenting challenges to both law-makers and law enforcement operations. Today, 40 million people are estimated as in some form of modern slavery across the globe. Correctly understanding the factors that render any particular individual or geographical region vulnerable to such exploitation is essential for the development of effective interventions and policy.

Currently, 369,000 people are estimated to be in some form of modern slavery in Brazil. Combating slavery mainly involves inspections of industrial operations suspected of involvement in human trafficking. However, there are concerns that current efforts are hindered by a rather limited federal response. Better understanding of the environments that render individuals vulnerable of falling into slavery is required.

In response to this, N/LAB has worked with private-sector data partners and the Rights Lab to model the environments which are perpetuating modern slavery in Brazil taking an inductive approach to elicit drivers of vulnerability in local populations.

  • Funder

    UoN, Rights Lab

  • Duration

    Jul 2019 – Jun 2021

  • Investigators

    Harry Marshall, James Goulding, Gavin Smith, Emily Wyman, Doreen Boyd

  • Partners

    Walgreens Boots Alliance

Project Description

Slavery (i.e. control of one person by another based on coercion, deception, violence or the threat of violence) is very much present and prevalent in the 21st century, presenting challenges to both law-makers and law enforcement operations. Today, 40 million people are estimated as in some form of modern slavery (including but not limited to involuntary servitude, human trafficking, and forced labour) across the globe. Correctly understanding the factors that render any particular individual or geographical region vulnerable to such exploitation is essential for the development of effective interventions and policy.

Modern slavery is defined as the recruitment, movement, harbouring or receiving of children, women or men through the use of force, coercion, abuse of vulnerability, deception or other means for the purpose of exploitation.

Despite the Brazilian anti-slavery infrastructure being one of the most developed globally, the number of individuals in some form of modern slavery in Brazil is estimated at 369,000. The country’s current approach to combating slavery mainly involves inspections of industrial operations, typically farms, factories and businesses suspected of involvement in human trafficking. However, there are concerns that current efforts are determined by the competence of regional or state governance, as opposed to the rather limited federal response. Better identification and understanding of the environments, conditions and motivating factors that render individuals vulnerable of falling into slavery is required.

In response to this, N/LAB has worked with private-sector data partners and the Rights Lab to model the environments which are perpetuating modern slavery in Brazil taking an inductive approach to elicit drivers of vulnerability in local populations.

Method

We combine and utilise 4 categories of data set to model vulnerability to modern slavery in Brazil. Modern slavery incidence data obtained from Smartab (), Brazil National Census Data (IBGE – ), and Mobility and ‘Neo-demogaphic’ data drawn from private sector data-providers.

Machine learning approaches were used to model modern slavery risk categories based on detection rates of survivors of forced labour for each municipality (see map below). We developed a new approach (see diagram below) to variable importance analysis that combines the advantages of Model Class Reliance analysis and Shapley Additive Explanations (SHAP) to enable extended explanatory analysis of vulnerability to modern slavery in Brazil, with insights not being restricted to a single arbitrary mechanism alone.

Results

Given the complexity of the problem, we achieved a highly accurate model, sufficient to support examination of the relationships identified. Use of MCR-SHAP demonstrated that in the context of modern slavery in Brazil merging a range of novel candidate covariate datasets can inform interventions creation. Key areas to focus anti-slavery efforts tend to have lower population densities, low rates of smart phone use, lower uptake of health insurance, lower mobility and high rates of economic security.

Associated Publications

Resources

Media, Blogs and News Stories

This is where associated blogs, news stories and anything related to the media can be posted.