Computer simulation models can help explore the complexity of migrant journeys, but require specific, dedicated data on migration decisions. To make simulations relevant for policy, what insights can be gained from ethnographic studies and psychological experiments?
To be useful as policy aids, and to be able to answer causal questions (what if…), the underlying simulation models need to be informed by detailed information on the mechanisms driving human decisions. However, the availability of such detailed data is limited, so specific information required for modelling often needs to be collected anew. For the European Research Council project “Bayesian Agent-Based Population Studies”, we have collected such information through ethnographic interviews and psychological experiments.
The findings from the ethnographic work, based on interviews with Syrians and Afghans in the UK and selected EU countries, focused on factors identified as relevant for shaping migrant journeys and trajectories to Europe. We identified four such factors, which were interacting in sometimes unexpected ways.
The first key aspect was knowledge: better-informed migration decisions were more planned and less subject to unforeseen factors. The second aspect was migrants’ capital, and especially social capital: their dynamic social networks, and ability of networking. Thirdly, personal circumstances and characteristics also mattered, such as gender, health, or family situation.
Finally, we identified another crucial factor in migration decisions and outcomes: chance. Chance largely drives the uncertainty of routes and journeys. In many cases, migrant destinations were dynamic: not perceived as the end-points of the journeys, but possible intermediary stages – stepping stones – with further options being intentionally kept open.
In the experimental work, our aim was to produce ecologically valid results – findings that can be generalised to the real world – by asking the participants to follow a first-person migration story, making several decisions. Early results are promising, but call for a need for caution. It is difficult to mimic complex reality in controlled experiments, and to identify clear effects of different decisions.
Computational models can help create laboratory conditions for testing different policy solutions and develop more robust policy or operational responses, especially for large and rapidly-changing flows. However, such models cannot rely exclusively on data about the migrants (who?), migration processes (what?), and their features, spatial (where?) and temporal (when?), no matter how detailed. To be useful for what-if questions, the models also need to include information about the underlying reasons for migration decisions (why?), and mechanisms of decision making (how?).
Here, despite the promise of new forms of data, including digital traces from mobile phones or social media, data-driven approaches alone are not sufficient. Honest policy advice based on simulation models needs to include multi-perspective insights from across social and human sciences, and to recognise and be open about the limits of the relevant knowledge. From the users’ point of view, this implies acknowledging that some aspects of migrant journeys will always be uncertain and specific to particular contexts and situations, as confirmed by our studies.
Acknowledgement: Work funded by the European Research Council grant 725232 Bayesian Agent-Based Population Studies (www.baps-project.eu). All views and interpretations are ours.