Misunderstandings of Multiple Systems Estimation: A Response to “On the Unreliability of Multiple Systems Estimation for Estimating the Number of Potential Victims of Modern Slavery in the UK” by Whitehead, Jackson, Balch, and Francis (2019)

Click to increase image sizeClick to decrease image size

Notes

¹ Human Rights Data Analysis Group (HRDAG) 2013, “Multiple Systems Estimation: The Basics,” Available from: https://hrdag.org/2013/03/11/mse-the-basics/

² Lum et al. (2013)

³ UNODC 2016, “Research Brief: Multiple Systems Estimation for estimating the number of victims of human trafficking across the world: Netherlands,” Available from: https://www.unodc.org/documents/data-and-analysis/tip/TiPMSE.pdf.

⁴ UNODC (2018a, 2018b, 2018c).

⁵ For example, at the top of page 5 of their paper about the inability to perform inference due to inclusion of the highest-order interaction, they note “Some form of assumption is necessary for any estimation to proceed”.

⁶ Furthermore, there are other mark-recapture models that could and should be explored; see for example, Bird and King (2018) and the examples on hidden populations discussed in Baillargeon and Rivest (2007) and Rivest and Baillargeon (2019), all of which explore the use of models based on a combination of heterogeneity, time/list and behavioral effects, and open populations. Such works are in the public domain, and have been formally recognized in statistical journals.

⁷ Whitehead et al. base some of their simulation studies and conclusions on literature and R packages that were not yet accepted for publication at the time of publication of their paper; see Silverman (2019) and Chan et al. (2020, 2019).