Carlo Gatti – NSfK ResearchSeminar 2021

Sex crimes and actuarial prediction in Nordic countries. A gaze from predictive policing studies

Carlo Gatti, PhD-student, Faculty of Law, University of Turku


This study intends to set an anchor point between the thorny matter of sex crimes in Nordic countries and my personal background as a researcher in predictive policing.

I locate this bridge-topic in the wide and long-lasting reliance on predictive techniques informing the prevention of sex offences, traditionally regarded as a laboratory of actuarial calculations from the Chicago school onwards.

The contribution that I try to offer from my position is an overview of the actuarial methods currently deployed in this field by the Nordic countries and a critical analysis of their theoretical background. Acknowledging my lack of specific training in sex crimes, I have no pretension of taking part in the specialist debate, nor am I assessing the predictive power of the devices, as not only a specialist experience, but also a set of notions from fields other than law and criminology – in particular clinical psychology – would be required for a serious analysis.

It will be up to the specialists to estimate whether a renewed sensitivity is needed and, if so, to what extent new technical solutions are workable to better reconcile crime prevention goals with the sore spots that I try to bring to light.


While seeking a thematic parallelism with my main research field, important differences need to be preliminarily clarified.

First, actuarial prediction in sex offences is limited, contrary to predictive policing, to a recidivism risk-assessment.

Secondly, while predictive policing relies on algorithmic technology, sex crimes prediction only represents a manually replicable prediction or, at most, a machine scoring systematized by pre-established grids of factors.

Thirdly, predictive policing is not conceived a priori for one class of crimes.

Finally, recalling that predictive policing splits into two main categories, person-based and place-based systems, the similarity with actuarial calculations of sex crimes recidivism is limited to the first type and only with this restriction in mind it makes sense to reflect on possible analogies.

Methodology and content development

The argument builds on the reconstruction of the actuarial protocols currently adopted in the Nordic countries and the modalities by which they are deployed by the different correctional agencies. Contrary to the secrecy that normally shrouds predictive policing algorithms, the mapping of the actuarial tools employed to profile sex offenders does not face this obstruction.

Nevertheless, the instruments in use in a given time and place often follow non-institutionalized praxis and enquiries to scholars, law enforcement agents and actuarial tools developers turn out an inevitable step.

The data so collected are placed into a synoptic grid laying the ground for an individualized description of the different tools around two basic traits: the type of risk-assessment at work and the illustration of the predictors.

DENMARK RSVP/SVR-20 and STATIC-99R are the most applied tools for assessment and/or treatment of sexual offenders
ICELAND HCR-20 (violence threat assessment, mostly focusing on domestic violence).
NORWAY STATIC-99R used to select participants to one of the treatment programs according to the risk-level.
SWEDEN Generic tool RNR-A for all SPPS clients.

STATIC 99R/STABLE 2007 sometimes complement the RNR response.

ERASOR for adolescents.

The subsequent critical analysis centres on the definition of a common background and proceeds from two major perspectives: the challenging of a prototypical selective idea about the offender and the criminal behaviour, and possible conflicts with basic guarantees of the rule of law.

Concluding discussion

The point of my critique is not an irreflexive rejection of any recourse to predictive techniques. It is rather an appeal for caution on four main points that, in my view, unify the different experiences and deserve a deeper problematization.

First, the underlying idea of sexual deviancy, which adjusts the dangerousness of identical conducts according to whether they take place or not within a stable and heterosexual partner relationship.

Secondly, the impact modulation of non-sexual crime record on the risk-assessment for the mechanical application of schemes borrowed from situational doctrines, deeply rooted in the socio-economic deficit paradigm.

Thirdly, the preponderant role of static variables for their tendential collision with the rehabilitative function of punishment.

Fourth and last of this short review, the problem deriving from the use of logistic regression in such a context where normative and cultural elements are pivotal in the definition of the predicted outcome.