© Copyright JASSS

JASSS logo ------

Party Competition: An Agent-Based Model (Princeton Studies in Complexity)

Laver, Michael and Sergenti, Ernest
Princeton University Press: Princeton, NJ, 2012
ISBN 9780691139043 (pb)

Order this book

Reviewed by Víctor M. Eguíluz
Instituto de Física Interdisciplinar y Sistemas Complejos IFISC (CSIC-UIB), E07122 Palma de Mallorca, Spain

Cover of book One of the standard problems addressed by agent-based models and social complex systems is the competition of opinions, cultural traits, rumors, and innovations. Typically, in these cases, a (large) set of agents adjusts their states by interacting with other agents and the mass media. A variety of models of social impact, bounded confidence, and voter-like models have already looked at this. This book offers a different perspective. Given a distribution of political opinions in a population, here the question is how political parties compete to get the largest support from voters, which support the party that is closer to their opinion.

To do so, the book presents an in-depth investigation on an agent based model developed by one of the authors (Laver 2005). It is structured around three parts. The first part (chapters 1-4) discusses the usefulness and the need for agent-based models to address multiparty competition. The second one (chapters 5-7) introduces the model. The third part (chapters 8-12) explores extensions of the model with a chapter where the model is compared to real party systems. Interestingly, the book is complemented with supplementary online material, including the NetLogo computer codes that allow everyone interested to re-run (and possibly further develop) the various model extensions explored in the book.

In the first part, after an introductory chapter, the authors review a set of spatial models. In this case, given a d-dimensional space, which represents policy positions and a distribution of voter positions, the problem is to choose the position of each party policy. This static problem can be mapped by a Voronoi diagram. The competitive setting can be analyzed in the Voronoi Game, where each party sequentially revises its policy given the position of others. The authors recall that, in a multidimensional space, this problem cannot be analytically treated (page 24) and thus motivates an agent-based model approach. Chapter 3 introduces three basic decision rules used by parties to decide new policy positions: Sticker - keep the same position, Aggregator - move to the centroid of its supporters, Hunter - vote seeking strategy. Chapter 4 gives a detailed account of the methodology of computer simulations.

In the second part, the basic model is explored with the competition of parties using the same rule, i.e., all are Stickers, Hunters or Aggregators. The main findings are as follows: in a population with symmetric (two dimensional) distribution of voter positions, voters are best represented by Aggregator parties; in an asymmetric population, the larger the number of vote-seeking (Hunter) parties, the better the representativeness of voter positions is. Chapters 6-7 introduce the competition of parties using different decision rules, and an endogenous birth and death of parties. The set of rules is also increased to allow two new vote-seeking strategies and new parameters to control, for example, the speed of adaptation and the exploration-exploitation trade-off. The interesting result is that the success of vote-seeking rules critically depends on parameter values.

While chapters 8-10 continue exploring various settings of the parameter space, chapter 11 offers a calibration of the model. Indeed, here the model is confronted with real party systems, in particular with party systems of European countries from 1989 to 2002. The expected path to calibrate the model would be to compare the outcome of the model in the last election year given the values of the previous years (not only the electoral shares but also the position of parties and voters). However, the authors suggest a different perspective and use each available data to calibrate the model and extract the distribution of parameter values capable of matching observed data. In this way, it is possible to estimate to what extent a given party used a vote-seeking rule, whether a leader was charismatic and popular, or what was the most likely political position for new parties. How important were leaders, such as Blair, González and Schroeder, for the voter support of their respective parties? You will find an agent-based answers in this book.

To sum up, the book is an excellent starting point for everyone interested in party competition and provides food for future agent-based research. Among them, it is worth to indicate the endogenous evolution of voter opinions.

* References

LAVER, M (2005) Policy and the dynamics of political competition. American Political Science Review 99, 263-281


ButtonReturn to Contents of this issue