Distinguished Professor Sylvia Walby, Dr Jude Towers and Professor Brian Francis of Lancaster University developed the method and used it to analyse data from the Crime Survey for England and Wales over a 20 year period from 1994 to 2013/14.
They found the rate of domestic violence has been increasing since 2009. The rate of violence against women has also been increasing since 2009. But violent crime against men is still decreasing.
The rate of violent crime overall has started to increase since 2009, driven by the changes in domestic violent crime and in violent crime against women.
These findings contradict the official view that violent crime in England and Wales is continuing to fall.
This increase is concentrated among ‘high frequency’ victims - those who experience multiple attacks.
The research finds that domestic violent crime and violent crime against women have been increasing since 2009. Before this, these forms of violent crime had been falling since the mid-1990s. The change point, from decrease to increase, coincides with the start of the economic crisis.
The researchers compare trends based on the number of victims, capped crimes, and all reported crimes. This new analysis finds that ‘high frequency’ victims, rather than all victims, are key to the increasing rate of violent crime in England and Wales.
This new methodology rejects ‘capping’ and uses all reported crimes without increasing volatility. Official methods of estimating violent crime using this data cap the yearly number of violent crimes against any one individual at five, despite around 5% of respondents reporting a greater number than this. Capping is a widely used statistical technique designed to reduce year-to-year volatility when examining change over time, but this method can introduce significant bias.
The new method manages volatility through an alternative statistical technique of three-year moving averages. This results in the same level of volatility but without introducing the bias caused by capping. The research analyses trends using segmented regression which allows the identification changepoints.