Modelling the Mafia

By Nigel Gilbert and the GLODERS team

“Cape Town: Long Street business owners – who lifted the lid on an apparent security racket – fear for their lives and those of their doormen and patrons if they refuse to pay R1 500 in “protection services” each month … “I was once approached by a company that wanted me to use their services but when I refused my windows were broken days later,” he added. Davidson said he paid R1 518 every month for consultation to a security company – Lifestyle Entertainment Security Services – that doesn’t provide any security.”

News item from the Cape Argus, 24 June 2015.

How can one gain an understanding of how extortion racket systems like the Mafia (but also many other lesser known criminal organisations, not only in Italy, but also in most of the developed nations of the world) actually work?  Conventional research methods have only a limited use.  Clearly it would be highly dangerous to engage in ethnographic or participant approaches, interviews with members could be somewhat difficult to arrange and are unlikely to yield much, and survey methods are equally inapplicable.  What we do have is a mass of documentary evidence, taken from court transcripts, journalist accounts and even autobiographies of the criminals (Militello, 2015).  Nevertheless, this still leaves the question of how to put all this material together to reveal a picture of how such organisations work, and how they are changing in what is an ever-shifting environment, as the police develop new methods and what is profitable alters.

The GLODERS (Global Dynamics of Extortion Racket Systems) research project, funded by the European Commission as part of its 7th Framework programme, is directed towards understanding Extortion Racket Systems (ERSs). ERSs, of which the Mafia is but one example, are spreading globally from a small number of seed locations, causing massive disruption to economies. Yet there is no good understanding of their dynamics and thus how they may be countered. ERSs are not only powerful criminal organizations, operating at several hierarchical levels, but also prosperous economic enterprises and highly dynamic systems, likely to reinvest in new markets.  If legislators and law enforcers are to be successful in attacking ERSs, they need a much better understanding of the dynamics of ERSs.

As well as collecting data about ERSs, the GLODERS project is developing a computer model of extortion racket systems, using an approach called agent-based modelling (Gilbert, 2008). Having extracted ideas about the behaviour involved in extortion from documents and other sources and formulating them as a set of ‘rules’ about what people do in which situations, computational ‘agents’ are programmed with those rules.  Then the agents are set to run within a computer generated environment, rather similarly to the way that virtual agents act within computer games, except that in the GLODERS computer models, the agents act out theories about how extortion rackets work.  The agents can be of different types:  extorters, victims, police, the public and so on.  The advantage of this kind of computer modelling is that you can check whether your theory about extortion racket systems is coherent and complete, and you can try ‘virtual experiments’ to see the consequences of, for example, proposed specific counter-measures (Troitzsch, 2015).

Consider the Mafia in Sicily as an example of what such modelling can show.  Until the middle of the 20th century, the Mafia was successful because of the strength of a set of traditional norms among the population who accepted that the money paid out to the Mafia (called ‘pizzo’) was the price that had to be paid for protection services, and that one should not cooperate with the legal authorities.  Italian state laws introduced from the 1980s to counter the Mafia had only limited success.  While many leading members of the Mafia have been imprisoned, the new laws were less successful in motivating the public to report instances of extortion or to resist supporting Mafia dominated enterprises.  It seems that policies to change social norms, though collective action (e.g. Libera), distributing leaflets and stickers (e.g. Addiopizzo), and supporting victims might be more effective.  These are the kinds of policies that the GLODERS model can investigate.

The idea of building computational models to examine the effect of different policies is now being applied in many areas in addition to criminology.  There are models being developed to help recommend future renewable energy policies, models to test policies for minimising the spread of epidemics, and models about regulating changes in the use of land, to mention just a few.  In fact, this is becoming such an important adjunct to policy making that we are starting a new Masters programme on Computational Policy Modelling here at Surrey in October 2015.


Gilbert, Nigel (2008) Agent-based Models.  Sage.

Troitzsch, Klaus G. (2015) Distribution Effects of Extortion Racket Systems. In: Fréderic Ambard, Francisco J. Miguel, Adrien Blanchet, Benoît Gaudou, Editors, Advances in Artificial Economics, Cham, Heidelberg etc.: Springer 2015 (LNEMS vol. 676), pp. 181-193

Militello , V.,  A. La Spina, G. Frazzica, V. Punzo, A. Scaglione (2014), D1.1 Quali-quantitative summary of data on extortion rackets in Sicily.  GLODERS report.


Please note: Blog entries reflect the personal views of contributors and are not moderated or edited before publication. However, we may make subsequent amendments to correct errors or inaccuracies.