Symposium: Modelling the Risk of Modern Slavery | A Response
Estimating the prevalence of modern slavery is important for understanding the scale of these crimes. Despite being at an early stage of development, prevalence estimates are a critical factor in encouraging the interest of governments and funding bodies and galvanizing action. Our approach to estimation was developed with these overarching objectives in mind, together with the need for a clear, replicable method that will only be strengthened as more data becomes available.
Contributors to this symposium have raised several important points about the risk factor model we presented, including the suitability of the model for prediction, data gaps in developed countries, the choice of Bayesian over frequentist inference, reliability of prior information, usefulness of global data and data-sharing.
Professor Silverman notes that the risk factor model is good for explanation but raises concern over its use in prediction. As we noted, this work is at an early stage and the data gaps for developed countries have been acknowledged. While a greater degree of precision is a priority in future iterations, we do not agree that there is no utility in the current country-level predictions. In the example given of the United States, knowing that the number of victims could be as high as 4 million is a substantial improvement on current efforts to estimate national prevalence.
There is certainly some way to go in improving precision in the model but to do so, further sources of data are required. We don’t claim this is the best possible model, rather that it is the best possible model based on available data. In fact, generating estimates of the prevalence of modern slavery has revealed gaps in the data available at the regional level. Walk Free Foundation continues to invest in national surveys, refining measures of vulnerability, and seeking alternative forms of measurements in developed countries through techniques, such as multiple systems estimation, to ensure substantial improvements can be made to future models. In addition to this, research on specific sectors or regions will add a great deal of valuable information to models such as ours.
The decision to adopt a Bayesian approach over a frequentist one was questioned by Dr Gleason, who noted that researchers in the field would not find the prior information on modern slavery reliable. Although frequentist inference is a more familiar among social scientists, we used a Bayesian framework primarily for computational reasons. The available data deals with rare events that in a frequentist approach may lead to singularities in matrix inversions. A Bayesian approach can be successful even when there is complete separation in logistic regression. Besides the computational advantages, a Bayesian approach also allows us to incorporate basic prior knowledge about the prevalence and distribution of modern slavery.
For example, few would suggest that all possible risk values are equally likely. We agree with Dr Gleason that there is scant prior information on modern slavery, which motivated the decision to assign independent weakly informative priors for model intercepts and regression coefficients, using a t density function with 7 degrees of freedom and scale 2.5. Future iterations of this modelling approach will be able to incorporate prior information using the Bayesian approach.
Bermudez and Stewart warn that models which are based on variables chosen because of their availability across national datasets may lead to findings that are “too generic to be actionable”. This is a valid assertion, but a necessity given this work is undertaken for the purpose of measuring vulnerability and estimating prevalence of modern slavery across 167 countries for the Global Slavery Index. Identifying global level risk factors helps shape the overarching framework for policy responses. Vulnerability to modern slavery is affected by a complex interaction of factors related to the presence or absence of protection and respect for rights, physical safety and security, access to the necessities of life such as food, water and health care, and patterns of migration, displacement and conflict. At its most basic interpretation, this level of analysis confirms that modern slavery cannot be addressed in isolation but that it should be addressed alongside other fundamental rights issues highlighted in the Sustainable Development Goals.
There is value for modern slavery actors in identifying that, for example, education and youth development are associated with lower levels of forced marriage. The success of long-term solutions in this field essentially rests on driving systemic change and building resistance among vulnerable populations. This finding reinforces the need to focus on access to education among vulnerable populations – an entirely actionable finding for frontline organizations and an intervention that many already incorporate into their anti-slavery programming.
Having said that, we acknowledge that global models of risk can only go so far. Our own findings demonstrate the added value that regional analysis brings. In developing and testing the models, regional variation was found for “business ownership”; this was a significant predictor of forced labour in South Asia and Sub-Saharan Africa and thought to be due to the greater number of necessity entrepreneurs in developing countries. As noted above, research on specific sectors or regions will add substantially to our knowledge about modern slavery and in doing so, lead to improved modelling.
Finally, the replicability crisis in the social sciences has led to increased focus on transparency and this issue was raised by Professor Silverman. The sharing of data underlying global estimates is complicated by multilateral partnerships, which require a joint decision on sharing data, and the need to maintain the trust developed with governments and non-profit organizations. Assurances that data will be anonymized and protected are not always enough.
Notwithstanding these pragmatic considerations, Walk Free Foundation is committed to transparency. In the case of the study that is the subject of this symposium, the data, code and other relevant files were shared with an independent statistician for review, and a detailed technical paper was made available via the SSRN Electronic Journal. More broadly, our methods are developed and refined with an expert working group, pre-briefings on methodology are given to interested parties, a detailed methodology paper is published in the Global Slavery Index, and a great deal of our data is made freely available.
The contributors to the symposium highlighted aspects of the analysis presented in “Modelling the Risk of Modern Slavery” that are important areas for refinement, and we are grateful for their careful consideration of this paper. While significant advancements in the measurement of modern slavery have been made in a relatively short period, this is very much a field in the earliest stages of development. Ultimately, the analysis under discussion sits at the forefront of estimation in this field. As a result, it comes as no surprise that the areas for refinement that we and the other contributors have identified are not unlike those encountered in the initial stages of measurement in other fields. The health sector, for example, faced a paucity of data with demand for better data growing throughout the 1980s and 90s, leading to the adoption of sample surveys as the primary tool for understanding the extent of health status, risk factors and responses, particularly in developing countries.
Even in the health sector, a field that many think of as data-rich, there remain challenges that are reminiscent of those we face in measuring modern slavery, including the inadequacy of country-reported data, the need to fill data gaps, and to ensure that there are independent and objective assessments. The speed with which we have encountered these issues and taken steps to address them in our field is encouraging, as is the increasing level of genuine collaboration – shown through the establishment of Alliance 8.7, the development of joint global estimates, and creation of data platforms such as, Delta 8.7 and The Counter Trafficking Data Collaborative. Such collaboration is critical to ensure the end of modern slavery.
This piece has been prepared as part of the Delta 8.7 Modelling the Risk of Modern Slavery symposium. Read all the responses here.
Jacqueline Joudo Larsen is Head of Research for the Walk Free Foundation.
Pablo Diego-Rosell is a Senior Consultant at Gallup.
This article has been prepared by Jacqueline Joudo Larsen and Pablo Diego-Rosell 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.