Using alternative data to understand violent crime

Authors: Alex Sutherland and Lucy Strang, Researchers, RAND Europe

When it comes to dealing with crime, it seems as though everyone has an opinion about what needs to be done. Certainly, the criminal justice system must continually adapt to respond to and manage criminal behaviour. Yet crime statistics are an area where governments are not moving with the times. Official statistics for crimes are typically generated through reports to police, as they have been for more than a century in the UK.

This means that there is not a full picture of crime, which in turn means efforts to prevent or reduce crime might be misdirected. But there are ways to improve understanding of where and when crimes occur through combining existing, routinely collected data from other public services and even further afield.

Debates about how crime statistics are produced in the UK have recently been reignited. In particular, it is a matter of much discussion that police recorded crime data shows increases in crime, whilst independently collected survey data from the Crime Survey for England and Wales show consistent falls in crime victimisation. (Although if one takes the long view, the world is far safer than it ever has been before).

When a police crime report is produced, there are a number of stages involved: an incident occurs, is reported to the police, is recorded and appears in official statistics as a particular type of crime. However, there are many reasons why people do not report a crime to the police in the first place. For one, they might fear being victimised again, which is all too common for some types of crime. Furthermore, not all incidents end up as recorded as crimes by the police (‘hate incidents’ and ‘hate crimes’ are good examples of the distinction between ‘incident’ and crime).

So if there is a reliance on people reporting crimes to the police, this may not give the full picture. In fact, it seems very likely that the picture of crime is far from complete. The history of official crime statistics in the UK shows that what the police record is an incomplete picture of crime. Estimates of how big the ‘dark figure’ of crime might be vary, but the idea has been around for nearly 200 years. This is a sobering thought – many millions of crimes might not appear in official data. For example, online crimes are a recent addition to official statistics, but one should keep in mind that fraud has been in statute since 1275 AD.

While all crimes inflicts harm on society, violent crime is of particular concern and something that communities, the police and policymakers all wish to see reduced. So what is already known that might help us? For one, crime clusters disproportionately in place and time, something that has been researched extensively. This predictability about where crimes occur most frequently means that police can also direct time and resources to preventing crime very effectively. That sort of approach has been undertaken for years and there is rigorous evidence that targeting these “hotspots” prevents crime.

What might be done differently? One idea is sharing with the police the data on violent incidents that hospital emergency departments routinely collect. This concept, known as the ‘Cardiff Model for Violence Prevention’, inspired a recently completed RAND project for West Midlands Police that looks at whether and how ambulance callout data could be used to better understand violent crime. This approach uses the routine administrative data generated by ambulance dispatchers when responding to 999 calls – where an incident occurs, the time and the nature of the incident. This information can also be updated through GPS information provided by ambulance crews when they have responded.

Using calls for service (999 calls) covering a five year period, what was found was surprising. Up to 90% of ambulance call outs for violent incidents were not found in police records. This means that a substantial amount of violence is not being reported to the police, and that resources targeted at preventing violence might be ‘missing the target’ in some instances. Indeed, health data presents a very different picture to police data overall (Figure 1).

Although the number of incidents varies substantially between police and health data, it is the difference in trends that is starkest. Notably, whilst police recorded crime data shows a continual increase from 2013 onwards, ambulance callouts decline slightly during the same period. This tells us that violence should be monitored using multiple sources of data. The risk of not doing so is that incorrect conclusions might be drawn about what is happening to trends. What ambulance data do allow is police forces to look at their own data differently, to understand where gaps in knowledge might be, think about why those gaps exist, and engage in problem-solving activities.

Figure 1: Number of violence-related incidents in police, ED and ambulance datasets, January 2012 to April 2017

Source: Figure 4 in  Sutherland et al. (2017) .

Source: Figure 4 in Sutherland et al. (2017).

What other ideas could we consider? In another RAND study, conducted with researchers from Cardiff University, RAND are looking at whether it is possible to use publicly available Twitter data to predict offline hate crime in the United States. The real power with such studies that use public social media data is the sheer scale of the information available – but the task is to understand whether there is actually any practical utility in the approach. Although the idea of foreseeing crimes before they happen worked for Tom Cruise in his film Minority Report, predicting crimes accurately is very difficult (as other research has shown it might not actually be better than guessing).

Preventing violence is not solely a criminal justice responsibility, as recognised in the public health approach to preventing violence. In RAND’s project with West Midlands Police, ambulance data have now been incorporated into routine monitoring of violence by public health officials in the county. There is an ongoing global push to improve how research and information on violence is collated by public bodies.

As time goes on, alternative sources of data will increasingly inform the understanding of crime and criminal justice responses to it. Yet police forces already produce and use a lot of data – it is more about changing how data are used (a particular challenge given the cuts to front-line policing in recent years.

Yet merely presenting data to police does not necessarily result in changes to tactics or strategy. So the challenge is to highlight the benefits of using alternative data sources, but to also to be clear about what the police need to do in order that this can happen, and then, what they can do next.