May 2014

Statistical issues in understanding domestic violence

Venue: Room A54, Postgraduate Statistics Centre, Lancaster University

Date: 22-05-2014, 2.30 - 5pm


2:30pm: Kicking off domestic violence - the football effect

Professor Allan J. Brimicombe, Centre for Geo-Information Studies, University of East London

Domestic violence is commonly associated with the consumption of alcohol and a link with major sporting events has been hypothesised, in particular in relation to football tournaments. The Home Office undertook to test the link between domestic violence in England and the FIFA 2006 World Cup but, as will be presented, this study has shortcomings. The BBC wished to test out the hypothesis in relation to the FIFA 2010 World Cup and the author was invited to undertake the analysis. With the FIFA World Cup again taking place this summer, this has continued relevance. In order to test the association between domestic violence and football, it needs to be based on an understanding of the temporal dynamics of domestic violence per se. This presentation will start with a statistical analysis of domestic violence and having set the context then move onto the details of analysing the association with the FIFA 2010 World Cup using Freedom of Information data from police forces and other corroborative data.

3:15pm: The use of formal and informal reporting mechanisms by victims of partner abuse: a latent class analysis of data from the Scottish Crime and Justice Survey

Dr Paul Norris, School of Social and Political Science, University of Edinburgh

It has long been argued that only a small proportion of partner abuse comes to the attention of the criminal justice system, yet existing literature is dominated by papers considering which factors influence the likelihood of an incident coming to the attention of the police. Using data collected as part of the Scottish Crime and Justice Survey (1011 victims between 2008-2011), this paper considers who victims choose to tell about their abuse. Latent class analysis suggests distinct groups of victims can be identified in terms of who they report their experience to. Only around 11% of victims report having told any formal agency about their experience, while approximately a quarter report tells friends of family, the remaining 60% of victims appear to have a low probability of telling anyone about their experience. The nature of reporting behaviour is considered as a function of those factors commonly believed, in the existing literature, to explain reporting abuse to the police. To some extent, different factors appear to drive the decision to tell friends and family compared to telling formal agencies. From a methodological viewpoint, this presentation will consider alternative mechanisms for including covariates in LCA models, and the Bayesian estimation options provided in MPlus 7.

4:00pm: Tea break

4:20pm: Measuring high frequency repeat victimisation in domestic violence – discovering the hidden victims and testing a theory

Professor Brian Francis, Professor Sylvia Walby, Dr. Jude Towers, Department of Maths and Statistics / Department of Sociology, Lancaster University

One recent and well-cited theory suggests that there are different types of domestic violence. (Johnson, 2008). These types are defined by the nature of the behaviour of one partner towards the other. In one type, "Intimate Terrorism" one partner attempts to gain control over the other . It is hypothesized by Johnson (2008) that this form of domestic violence is likely to lead to repeated acts of violence within a relationship, perhaps escalating in severity. Reporting some initial findings from an ESRC study, this talk therefore discusses the issue the measurement of high frequency attacks in domestic violence. It identifies that such attacks are hidden from official estimated incidents by a process known as capping. Introduced in the US National Crime Victimization Survey, its methodology was taken up by the British Crime Survey , now known as the Crime Survey of England and Wales. The intention of capping is to limit the maximum number of recorded incidents of a survey respondent to a fixed number, in order to avoid a small number of cases over inflating the estimated number of incidents. We investigate the effect of uncapping domestic violent incidents, and show that there is a small but consistent proportion of respondents who report a large number of violent attacks each year. We discuss how uncapping domestic violence increases substantially the number of violent attacks, and examine trends in uncapped figures. Finally , we discuss how uncapped data can be used to test Johnson's theory through appropriate mixture models.


Johnson, M. J. (2008) A Typology of Domestic Violence: Intimate Terrorism, Violent Resistance, and Situational Couple Violence. North Eastern University Press: Lebanon, NH.