Bilateral Migration Estimation via Agent-Based Modelling

Xinyi Kou , University of Manchester
Arkadiusz Wisniowski, University of Manchester
Natalie Shlomo, University of Manchester
Eduardo Fe Rodriguez, University of Manchester

The study of population has always been at the centre of public policy and planning due to its vital role in human society. Among all population components, international migration is the most challenging element to estimate and it plays a key role in population change. To gain a better insight of migration behaviours, we propose a microscope simulation approach, agent-based modelling aligned with microsimulation applied to estimation of international migration. The aim of this study is to disaggregate migration data by characteristics of agents, and to impute missing migration patterns. In our migration module, we focus on the connections between two countries and their immigration and emigration. The behaviours of individuals in both countries are simulated simultaneously. The model also takes aging, births and deaths into account. Our migration decision-making framework is based on logistic regression with individual characteristics such as age, gender, employment status, nationality and income. Further work includes social networks modules, and it is assumed to alter behaviours in a non-linear way. Instead of empirically determining all measurable parameters in the model, we propose to estimate the parameters in a Bayesian way, making use of available qualitative and quantitative information. We then utilise a Gaussian Process Emulator to calibrate and analyse the impact of selected model parameters on the outcome, which is estimated emigration and immigration rates. The proposed method introduces the possibility of disaggregation of migration by characteristics of agents and provides the ability to generate multi-level agent-based scenarios with some aspects of empirical demographic reality.

See extended abstract

 Presented in Session P1. Postercafe