Reviewed by
Tilman Schenk
University of Leipzig
The book is a compilation of papers presented at a workshop entitled "Recent Developments and Future Directions in Agent-Based Modelling in Populations Studies" held at the University of Leuven, Belgium, in September 2014. The workshop was part of a research project funded under the European Union's 7th Research Framework Programme (European Research Council Starting Grant). The research project investigated the consequences of a shifting gender balance (more females) among graduates of higher education on family formation, fertility, and separation. This might help to understand why a major focus of the book and presumably the previous workshop was put on decision-making models in family formation and fertility (part IV) as well as health and social support issues (part V). Contributors are researchers from universities and research institutions throughout Western Europe and Northern America.The book contains 17 papers grouped into five parts. As an introduction, the editors provide an insight on how two different perspectives on "the population" have shaped previous demographic research: One being a perspective dealing with population sizes, indicators describing it, and models pursuing to explain alterations in their values (termed the "macro perspective"). The second is focusing on individual decision making, heterogeneity and change (the "micro perspective") rather than average numbers. In line with similar lines of argument in other disciplines (Sun 2006), the editors go on to argue that agent-based models might help to link the two perspectives into a single paradigm (p. 7) and, by using the approach for simulation experiments, to build new theories of demographics (p. 12). The remaining two chapters of part I focus on general perspectives of demography as model-based research and the potentials to explore the role of social interaction in demography using agent-based modelling.
Part II is the largest section of the book compiling a total of six papers on the crucial part of work for any modeller: the design of the model itself, the analysis of its performance and ways to document it. As a start, chapter 4 introduces the software "JAS-mine" as a tool to build discrete-event simulations, fitting well to typical demographic research questions. The following chapters 5 and 6 are devoted to describing how various sources of empirical information can be utilised to inspire agent-based models in population studies. Regarding the analysis of model performance, chapters 7 and 8 introduce the reader to the methodology of meta-models. Meta-models will treat the micro-model of a process as a black box and then attempt to link its inputs and outputs by a statistical function, thus permitting an analysis of the model performance. This part is closed by a paper on ways to document agent-based models, with an overview of existing documentation of agent-based modelling (ABM) in demographic research and an introduction to the widely known ODD+D (Grimm et al. 2006; Müller et al. 2013) protocol.
The following papers discuss examples of models in demographic research, grouped into three parts. This selection of example studies addresses very specific cases; for instance, chapters 10 and 11 deal with modelling decision processes, on the examples of migration decisions and the decision to disclose alcohol abuse during pregnancy to medical personnel using a model approach inspired by game theory. Part IV combines three chapters on models of family formation and fertility, and part V another three papers on health, mortality, and support in old age. Some of the models discussed are rather data-driven, and their descriptions contain large data tables, maps and equation systems, while others have a more theoretical focus.
To sum up, one could say that the book provides a good introduction into what readers, who are familiar with demographic research, can expect from agent-based modelling for their discipline. It may facilitate a general understanding of what ABM is and what can be done and achieved with them regarding new research insights, answering old questions as well as posing new ones. The numerous examples of already existing applications demonstrate the flexibility of the approach and will also facilitate researchers unfamiliar with the method to gain access to it. However, it is not an introductory textbook that will guide the reader through a step-by-step process to build an own model.
Because the book brings together papers from a workshop in a (European) research project with a very specific focus, the chosen case studies in parts IV and V can by no means be representative for demographic research as a whole. They do however demonstrate the feasibility of the approach for a wide range of areas of interest within demography and should thus be understood as encouragements to address research questions in this discipline with agent-based models.
ReferencesMÜLLER, B.; Bohn, F.; Dreßler, G.; Groeneveld, J.; Klassert, C.; Martin, R.; (2013). Describing human decisions in agent-based models - ODD+D, an extension of the ODD protocol. Environmental Modelling & Software 48, 37-48.
SUN, R. (2006). Cognition and Multi-Agent Interaction. From Cognitive Modeling to Social Simulation. Cambridge University Press, New York.
Return to Contents
of this issue