Estimating conditional probabilities of historical migrations in the transatlantic slave trade using kriging and Markov decision process models

TL;DR

This paper introduces a novel two-step statistical method combining Kriging and Markov decision processes to estimate the likely origins of enslaved individuals during the transatlantic slave trade, based on historical conflict and transport data.

Contribution

It develops a new probabilistic approach to infer origins of enslaved people using spatial and transportation modeling, filling gaps in historical migration knowledge.

Findings

Visualizes conditional probabilities of migration paths

Provides probabilistic origins for enslaved individuals

Enhances understanding of intra-African slave trade routes

Abstract

Intra-African conflicts during the collapse of the kingdom of Oyo from 1817 to 1836 resulted in the enslavement of an estimated 121,000 people who were then transported to coastal ports via complex trade networks and loaded onto slave ships destined for the Americas. Historians have a good record of where these people went across the Atlantic, but little is known about where individuals were from or enslaved \textit{within} Africa. In this work, we develop a novel two-step statistical approach to describe the enslavement of people given documented violent conflict, the transport of enslaved peoples from their location of capture to their port of departure, and---given an enslaved individual's location of departure---that person's probability of origin. We combine spatial prediction of conflict density via Kriging with a Markov decision process characterising intra-African transportation. The results of this model can be visualised using an interactive web application, plotting estimated conditional probabilities of historical migrations during the African diaspora. These results help trace the uncertain origins of people enslaved in this region of Africa during this time period: using the two-step statistical methodology developed here provides a probabilistic answer to this question.