Reviewed by
Piergiuseppe Morone
Unitelma-Sapienza, University of Rome
As a heterodox economist, when I first received this book I was captured by its title: “Agent based models of the economy”. Indeed, the book delivers what it promises in the title: a passionate call to use agent based models (ABMs) in order to better understand the economy. The book is divided into two parts: part I (shorter) covers concepts and methods related to ABMs of the economy and part II (longer) is about ABM applications.The three chapters composing the first part of the book do not really address theoretical issues associated with modelling (as one would think when reading the subtitle “From theories to applications”), but rather provide the conceptual framework within which economics meets agent based modelling, how this tool can contribute to enhance our understanding of complex societal phenomena and how ABMs can provide more realism in modelling the economy when compared to the neoclassic reductionist perspective. This is followed by a thorough methodological discussion on a specific class of ABMs. While interesting and inspiring, this part of the book might have benefitted from a discussion on some limitations of ABMs as well as on some hot and debated issues among modellers. Just to mention a few: the need to define a shared protocol, the existence of different ABMs’ validation approaches, the strong dependence on parameters, and the consequent need for robustness checks (e.g. model docking), etc.
Some of these issues are briefly mentioned in the book; however they probably deserve more attention. In particular, there are two issues which are introduced but not fully developed. The first one is complexity - if ABMs are indeed more appealing than dynamic stochastic general equilibrium (DSGE) models, which have the technical complication of ABMs but not the flexibility of these modes, it might not necessarily be the case for a simple mathematical model grounded on a good and novel idea. Also, where should modellers strike the balance between simplicity and descriptivism when designing a complex model?
The second issue stems directly from the complexity aspect and is linked to the assertion that what makes a model good model are good ideas. More specifically, what I’m missing in this book is a discussion on the debated problem of “garbage in – garbage out” issue: as ABM is such a flexible tool it makes it a lot easier to just model the wrong thing in the wrong way.
The second part of this book proposes an array of applications of ABMs going from human capital developments to interbank payment systems; from health insurance systems to broadband developments; from coordination failures to optimisation of fiscal systems. The issues covered in this second part of the book are so disparate that a malicious reader might think that this is just a collection of models developed by the authors and assembled without a common thread. I am more inclined to see it as a playground for apprentice modellers, indeed useful to get, at a glance, an idea of what can be done with ABMs in economics.
Summing up, I would say that this book is a learning tool, which introduces readers and apprentices into the magic world of agent based models of the economy! It does so in three ways:
1) It tells readers why ABMs are superior to standard neoclassic modelling (the argument developed could be simply put as follows: the world is complex and mathematical models are unfit to capture properly all its complexity);
2) It provides a “tutorial-like” introduction to ABMs by means of an exemplification of SLAPP (Swarm-Like Agent Protocol in Python);
3) It provides the sceptical reader with a handful of examples (applications) of what can be actually done with ABMs.
After reading this book you will either love ABMs or hate it; in either case you will have learned something. I would recommend this book to master students and young researchers (PhD students), as they will learn the most from it.
Return to Contents
of this issue