Agent-based models to inform economic policies on migration

Large-scale agent-based models (ABMs) hold much promise in analyzing non-equilibrium, self- organizing behaviors of complex economic systems. Despite this potential, the existing methodological shortcomings hinder their wider application: Two major technical challenges are rigorous model validation and effective means for model-based stakeholder engagement. This study advances the state-of-knowledge in these two aspects. Using migration policy assessment in Austria and Russia as examples of application, the first objective of this study is aimed at improving ABM transparency and reliability through the development of a calibration protocol and validation via back-casting. The second objective is aimed at exploring an ABM combined with a gamified user- interface to test its usefulness in stakeholder deliberations on Austrian migration policy. This study will be carried out in two phases as joint research between the International Institute for Applied Systems Analysis (lead: Dr. Joanne Linnerooth-Bayer) and the Central Economics and Mathematics Institute of the Russian Academy of Sciences (lead Dr. Valery Makarov). In the first phase of this research, the two partnering institutions will jointly build prototype ABMs of the Austrian and Russian economies, respectively, building on their extensive ABM development. The models will incorporate an explicit representation of potential migration inflows and the capability to test migration policy packages. The study employs a variety of datasets, such as national accounting, census and firm-level surveys, to rigorously calibrate and validate the developed models. The models will be informed by a literature review and interviews with Austrian stakeholders in the topical migration policy issue. The prototype models will be further implemented on High Performance Computers to ensure scalability and fine granularity of the obtained results. In the second phase, the Austrian proponent will develop an interactive graphical user-interface that gamifies the dynamic representation of migration policy choice. The user-interface will be tested in a mock stakeholder setting to explore its usefulness for stakeholder dialogue and its potential for extending to a real-world policy exercise The research contributes to emerging literature on large-scale ABMs by being one of the first studies to calibrate ABMs to the historical trends of major economic variables. The study will also be one of the first to pilot the use of large ABMs in an experimental gamified stakeholder process. The scientific outputs of this study will hence be of interest to wide international audiences in fields such as computational economics, simulation games, stakeholder analyses, and migration policy.

The ABM2Policy project set out to i) advance the methodology of an agent-based model (ABM) as applied to the Russian and Austrian national economies to realistically analyse migration as an external economic shock, and (in Austria) to ii) explore the usefulness of the ABM combined with a gamified user interface to support Austrian migration policy-making discussions. Investigators from multiple disciplines (economists, modellers, mathematicians, demographers, political scientists, physicists, game developers, and linguists) worked together to create new methodological and applied innovations that moved beyond discipline-specific approaches. Taking account of the interactions of millions of agents representing households, the government, banks and firms operating in multiple markets, the IIASA research team analysed the economic consequences of migration scenarios in Austria, including GDP growth, inflation, unemployment and the budget deficit. Unique to the ABM, a search and matching mechanism was developed to match households' labor supply and firms' labor demand. The team focused on climate-driven migration from the MENA region due to extreme drought and locust infestations. Specific scenarios were developed with the EC (European Commission) Joint Research Centre. A model-based policy simulation, or Gamified Policy Exercise (GPE), demonstrated the usefulness of the ABM for informing policy deliberations in the complex migration setting. Building on a review of Austria's political party platforms and stakeholder narratives, the GPE focused on three controversial policy issues: i) assistance to MENA governments to help them cope with the crisis, ii) Austria's position on a new classification of 'environmental refugees', and iii) support for Mediterranean rescue efforts. The context differed from past migration policy issues in that Austrians, through their greenhouse gas emissions, have contributed to the climate risks driving migration. The GPE was based on a long-standing literature analysing political conflict in terms of distinct discourses, and, more specifically, in terms of worldview communities or plural rationalities. Each participant played the role of a parliamentarian with the goal of reaching a consensus on the above questions. The debate was aided by the ABM scenario results, as well as a motivating video and fact sheets. The GPE was played by 43 participants from IIASA, the University of Vienna and the University of Continuing Education, Krems, who submitted 23 post-simulation questionnaires. The results of the project went beyond its goals. Not only was the ABM capable of analysing the economic consequences of migration broken down by economic sectors and cohorts, but it also challenged the accepted view that ABMs are not equipped for economic forecasting; after calibration and validation, the model competes favourably with vector-autoregressive (VAR) and Dynamic Stochastic General Equilibrium (DSGE) models. The first-time demonstration of an on-line ABM-based Gamified Policy Exercise advanced the concept of experiential learning with participation of researchers and students of migration and geography.