US police are cutting crime by using risk-based policing to forecast where incidents are most likely to occur. The strategy relies on Risk Terrain Modelling, a method of analysis that identifies correlations between specific street features and criminal activity. The approach has cut gun crime in Newark by 35%, car theft in Colorado Springs by 33% and robberies in Glendale, Arizona by 40%.
Results & Impact
Risk Terrain Modelling (RTM) helped reduce gun crimes by 35% in Newark, vehicle theft by 33% in Colorado Springs and robberies by more than 40% in Glendale, Arizona. Atlantic City also cut violent crime by 20% between January and May 2017 using the method.
Rutgers Center on Public Security, local police forces
The Risk Terrain Modelling Diagnostics Utility (RTMDx), developed by professors at Rutgers’ Center on Public Security, allows police forces to more accurately predict where crime will take place, then allocate officers accordingly. Users first upload a digital map of the area they wish to study in a shapefile. (Shapefiles bring together multiple file types, storing different types of geographical data, such as building types and street features, in a single location.) After uploading a map, users log the locations of past incidents of crime. Potential buildings or street features identified as hotspots - such as banks, liquor stores or streets frequented by gangs - become visible. The RTMDx software produces a table listing the risk values of each street feature, revealing their correlation with a type of criminal behaviour. The map displays the parts of the city where crime is most likely to take place. RTM can also be used to account for differences between crime rates at different times in the day. For example, the effect of a bar on crime will be very different in the morning to the evening. As such, different maps can be produced for police operating on different shifts.
Risk Terrain Modelling is currently used in more than 30 countries and US states.
Cost & Value
The software is free to download and use, but online training courses can cost about $750.
Running since 2013
Although Risk Terrain Modelling can be accurate in predicting where many types of crimes - such as shootings or robberies - are likely to occur, this is unlikely to be the case for all criminal behaviour. Risk Terrain Modelling's effectiveness depends on a clear correlation between specific street features and criminal activity. Without a clear link between geospatial features and crime, Risk Terrain Modelling's predictive capabilities are limited, as it is likely that some forms of crime will have lower correlations with street features than others. Another challenge faced by police is access to data, as obtaining the necessary crime statistics can be time-consuming. For example, Atlantic City authorities lacked an efficient way of tracking the locations of crimes when implementing Risk Terrain Modelling, making the process more difficult.
Risk Terrain Modelling is used in more than 30 countries - including France, Canada, Australia and Italy - and in over 30 states in the US.
Risk-based policing is slashing crime in cities across the United States by empowering police departments to deploy their officers more effectively.
Risk-based policing focuses on the relationship between geospatial features and incidents of crime. It uses a method of analysis known as Risk Terrain Modelling (RTM) which predicts where crime is most likely to occur by identifying correlations between building types, street features and locations of past crimes.
“What we are trying to model is the way in which these features of the landscape intercept and interact or co-locate to create behaviour settings that are most suitable for crime,” said Joel Caplan, Associate Professor at Rutgers University, who helped develop RTM modelling.
RTM has been widely adopted by police departments due to the Risk Terrain Modelling Diagnostics Utility (RTMDx). A software program developed in 2013 by Caplan and Leslie Kennedy at Rutgers University, it can be installed on any computer and automates the analytical process of RTM. RTMDx identifies correlations between a type of crime and physical structure. The program scores the risk of an incident taking place near any type of structure, such as a bar, alleyway, liquor store or broken streetlight. It also highlights the most vulnerable areas of a city, where multiple high-risk buildings intersect, on a digital map.
In Atlantic City, RTMDx reduced violent crime by 20% in five months without a corresponding increase in rates of arrest. This was possible because the program revealed that more than 30% of crimes occurred in just 2% of the city, particularly in the vicinity of convenience stores and laundromats. Having identified the critical hotspots, meetings were held between police officers, city authorities and citizens to understand why crimes were clustered around them. It emerged that gangs were using the laundromats and corner stores to sell drugs and cigarettes. Using this information, Atlantic City police launched an intensive community engagement strategy, through which they check in regularly with high-risk businesses in the most vulnerable areas, sometimes on an hourly basis.
In the cases of Glendale, Colorado Springs and Newark, RTM showed police that they could significantly reduce crime by focusing on between 1% and 5% of each city. Like in Atlantic City, intensified community engagement strategies were key to the police’s success. In Newark, patrols initiated a check-in system for businesses in the highest risk locations, whereby the on-duty manager was required to sign a form certifying that officers had visited.
In Glendale, police patrols were redirected to focus on seven types of buildings with the highest risk values, as well as hand out flyers and organise community meetings. It was found that mobile phone theft was prevalent. The phones were frequently being stolen near convenience stores that allowed customers to trade in their mobiles for cash. Police prompted stores to move the location of the phone return kiosks to make them more visible. They also implemented a more effective system of surveillance.
In Colorado Springs, police focused more heavily on tracking driving violations and deploying registration plate recognition software in the target area. They also implemented community engagement strategies.
RTMDx is a computer program that requires two main inputs. To run RTM analysis, users must first identify previous incidents of the crime they are investigating and where they occurred. Once that information is logged, the second step is to determine which structures are most likely to be associated with the type of crime. For example, if investigating robberies, a user might wish to track all liquor stores, pawn shops, banks and other outlets. These are entered into the program by typing in their addresses.
RTMDx requires users to insert a map of the study area in the form of a shapefile. Shapefiles bring together multiple file types, storing different types of geographical data in a single location, and can be read by Geographic Information System software platforms.
Once crime type and building types have been logged, RTMDx runs a series of probability tests to determine how a type of structure affects the probability of a given crime occurring nearby. The system first uses Bayesian information criteria calculations and then conducts regression modelling to identify any spatial correlations.
RTMDx produces a table showing the correlation between each structure and type of crime, as well as a digital map outlining the most vulnerable areas of the city. Each building type is listed with a relative risk value, which represents the risk of a crime being committed nearby. One is the lowest relative risk value and, although there is no maximum figure, each score is designed to reflect the risk relative to other structures. Building risk values are then weighed against each other to identify high-risk intersections within the city, and logged on the digital map.
“For every single place it asks, ‘Are you within a half a block? Are you within two blocks? Are you within three blocks up to the threshold of this particular feature? And has a crime occurred within that area as well?’” said Caplan. “It basically models the entire landscape for every different risk factor up to every different distance, and determines whether or not crimes are occurring more often in places defined by these risk factors – or the combination of these risk factors.”
“Essentially what the system is doing is reverse engineering offender preference,” said Caplan. “Ultimately what we are assessing is the likely preference of the individual who committed that crime across all the people who made similar decisions and chose those same locations, to see what attracted them there and not to somewhere else.”
The granular nature of the system allows authorities to develop a highly detailed picture of very specific areas. Users determine the level of analysis they want RTMDx to perform using factors like minimum and maximum distance from one building to another. The minimum analysis unit is frequently half a block and a common maximum is three blocks. The map is broken up into cells matching the minimum analysis unit, with each cell evaluated independently. This granular approach allows the system to identify patterns of high correlation around specific intersections of buildings and crime throughout the study area.
RTM’s real advantage is its predictive capability, allowing police to remain one step ahead. By focusing on the geospatial characteristics of locations where a crime has taken place – such as a broken streetlight – RTM reveals other parts of an area that are likely to be vulnerable, even if no crime has occurred there.
Although hotspot mapping tracks where crimes have taken place, its value is limited due to the non-static nature of crime.
“It has been demonstrated through research that crime moves – that crime doesn’t always occur in the future where it has done in the past,” said Caplan. “Risk terrain modelling takes into account previous patterns of crime, as well as locations that are high risk for crime even if it hasn’t occurred there already.”
RTM maps can also be updated to account for shifting patterns of crime and to test previous maps’ accuracy. In Atlantic City, a new risk map is produced each month to take into account the effect new patterns of policing have had on crime locations. RTM can also be used to account for differences between crime rates at different times in the day. For example, the effect of a bar on crime will be very different in the morning to the evening. As such, different maps can be produced for police forces operating on different shifts.
The platform can be downloaded from the Rutgers Center on Public Security Website free of charge by anyone and installed on a user’s computer. Training is provided for city agencies through the Rutgers Center on Public Security. This can take the form of in-person classes at agency premises or Rutgers University, or online webinars and tutorials. Guides are also available to download. Although the software itself is free, some courses require payment. An online webinar complete with software provision and video demonstrations costs $750.
A new, more user-friendly version of RTMDx will launch in September 2017. Complete with a guidance wizard, it is hoped this version will reduce the amount of training police forces require. It will also allow users to enter data using more file types, and allow risk terrain maps to be exported to Google Maps.
(Picture credit: Wikipedia Commons)