In big cities, research on geographical tendencies in crime incidents, along with mapping technology, have helped police departments target their patrols efficiently and cut crime where it happens. In small cities and rural areas, however, crime incidents show a different geographical distribution, so both technology and policing methods must be adapted for greatest impact. In a 2015 pilot project, researchers took a mobile phone-based mapping technology that was informing policing methods in the large city of Lahore, Pakistan and adapted it for use in Sargodha, a semi-urban district in Punjab province. This case examines the kind of human adaptations that technology needs to be scaled across geographical areas and be taken up by users, and it shows how data is informing a major shift in how Pakistan’s police deal with crime.
At dawn on the February 8, 2012 Dr. Ali Cheema was woken by his wife telling him that his newly bought car had been stolen.
Cheema, a Senior Research Fellow at the Institute of Development and Economic Alternatives (IDEAS), was shocked that this had taken place in his neighbourhood – the relatively quiet, affluent Defence Housing Authority near the Lahore University of Management Sciences (LUMS) where he teaches economics. “I always thought that I was in a really secure area,” he says.
Cheema called his insurance company to file a claim, and was surprised to receive a visit from the company’s senior vice president later that day. “I was a bit perturbed,” he says, “as I am prone to cynicism, so I thought oh-oh, they’re going to try to get out of it. But it turned out his own car had gotten stolen, just ten streets away from me.” The VP told Cheema that six cars insured by the company had been stolen in the area over a period of two weeks. “And they were all Toyota Corollas!” he adds.
What struck Cheema was the information gap this represented. “There were all these thefts happening around me, and I had no clue about. It’s an urban area, I’m not very connected to my neighbors. That got the researcher in me to thinking, hang on, information seems to be really important.”
Cheema discussed the incident with Zulfiqar Hameed, a long-time acquaintance whom he had instructed in economics at LUMS and was working as Senior Superintendent of Police Investigations, the number-two position in the investigation hierarchy in Lahore police. Hameed confirmed Cheema’s suspicion that certain neighbourhoods were more prone to crime. To that point, police practice did not take that into account. The two agreed that this represented the kind of opportunity they had often discussed, to use data and research evidence to improve police practice.
For decades, researchers and police officials around the world had debated whether crime happened in a random geographic distribution – in which case it made sense for police to patrol widely in police cars – or whether it was concentrated in certain neighbourhoods and on certain street corners, in which case police should focus on such ‘hotspots’. This was an important distinction, because it led to a number of related questions that had relevance for spending and strategy. For example, if police visibility deters crime then governments should invest in creating a broad presence, which is expensive in terms of both man hours and petrol.
Until now, the police leaders in Lahore and other cities across Pakistan had operated under the traditional assumption that crime occurred in a random geographical spread. But research in the US and UK had shown that, in those countries at least, crime incidents tend to cluster in hotspots, and random-patrol policing has no deterrence effect. (Click the button on the right to read more.)
Working with researchers is far from standard practice for police in Pakistan. “It’s very uncommon,” Hameed says. “As far as I’m aware, in Pakistan this was probably the first time police and researchers worked together on crime issues.”
To determine spatial distribution of crime in Lahore, Cheema and Hameed had to overcome certain hurdles. Pakistan doesn’t use specific street addresses or postcodes, which makes it difficult to pinpoint locations, and although the Lahore Police Department had a data-entry system in place, it did not geocode information. So the research team began in quite a low-tech manner: they handed out detailed paper maps to precincts and asked officers to mark crime scenes with pushpins.
Soon, they saw pins start to cluster in certain spots. But soon after that, the maps started to fray, the pins to fall out, and departmental attention to stray.
The team enlisted the help of Professor Sohaib Khan, director of the Technology for People Initiative (TPI) at LUMS. TPI collaborated with the Punjab Information Technology Board (PITB) to develop the tool that allowed the next stage of the research: an app that allowed for real-time digital record-keeping. When police arrived at a crime scene, they used smartphones to take a picture and punch in the answers to a number of questions. The app then sent the information along with the geo-stamped location to a central server. The TPI/PITB team developed an interface that automatically generated both spatial patterns and time trends from the data.
In March 2014, the team attended a Policy Dialogue event on Civil Service Reform in Pakistan, part of the Building Capacity to Use Research Evidence (BCURE) programme funded by UK Aid from the UK government and implemented by Evidence for Policy Design (EPoD) at Harvard Kennedy School in collaboration with the Center for Economic Research in Pakistan (CERP). Cheema and Hameed presented crime maps at an afternoon reserved for breakout sessions – small groups that gave members of the research and policy communities an opportunity to sit face-to-face, make connections, field ideas around a central theme.
The maps gave compelling visual evidence that crime was concentrated in neighbourhoods and within neighbourhoods around hotspots.
The breakout group compiled the materials and presented them to the general Policy Dialogue audience. Cheema says, “At that point, the tool hadn’t produced enough data for us to come up with a lot of analysis yet, but we felt like there was a story there. The question was, what do we do with this stuff?”
Describing how the Policy Dialogue audience reacted to the mapping tool and its data, Cheema says, “There seemed to be an appetite for it. People were very interested, quite intrigued.” To roll out the tool across all of Lahore would require further piloting, however, especially if it were to lead to the kind of changes that seemed to be in order. The Lahore Police Department had a shortage of officers, so wide testing would have to take place before it made any change in patrolling strategy.
Another question that emerged from the discussion was whether smaller cities and rural areas would show the same kind of crime hotspots as Lahore.
The BCURE initiative also supported a number of pilot projects intended to create ways for policy actors to utilize data and research evidence to improve their decision making. An idea emerged from the Policy Dialogue and subsequent discussions, to pilot the mapping tool in a small city. This would serve the dual purpose of further testing the tool and gathering data on spatial distribution of crime in another environment.
Cheema and his team identified a small city where the police superintendent was eager to have their help to use the tools of data and evidence to improve crime-monitoring. Before long, however, that official was transferred and replaced by one with little interest in the project. His scepticism of the tool and of data in general was evident, and the project ground to a halt. (For thoughts on the dangers – and unexpected benefits – of frequent transfers among evidence-minded policymakers, see the BCURE case on Tax Visualization).
The pilot was at risk of failing, yet an aspect of the BCURE agenda was to exploit existing relationships with governmental ‘evidence champions’. Zulfiqar Hameed was transferred and made Regional Police Officer of Sargodha, a region that included a city of 500,000 inhabitants as well as rural areas. To this point, there had been no attempt to analyse crime across spatial dimensions in Sargodha, and Hameed was keen test the system he had helped develop. Cheema saw this as an opportunity to test the system in a less urbanized setting in collaboration with a proven supporter.
The team had to adjust the tool for the setting. In Pakistan, districts are divided into police stations (the equivalent of precincts in the US), which are divided into several ‘beats,’ each overseen by a police officer. Prior to this research, the police station was the smallest unit at which crimes were recorded. In urban areas of Sargodha – as in Lahore – police beats were geographically small, so they served as a good basic level at which to aggregate data for mapping. In rural Sargodha, however, beats were far larger geographically, often containing several villages, and their borders less defined.
In the course of the BCURE pilot, the team mapped all 27 police stations in Sargodha district, defining boundaries of villages and beats. They entered into the map a year’s worth of crime data that had been recorded but not geocoded. They also trained officers in how to use the tool and developed with Hameed a plan on how to continue building capacity and institutionalize the tool into day-to-day practice. Finally, they produced initial reports on the spatial distribution of crime incidents.
Along the way they encountered some of the resistance that is common when trying to change practice in a department where employees are used to acting based solely on their own – often substantial – knowledge. Police officers knew their own beats, they said, knew where crimes happened, and shouldn’t be asked to change based on a machine-made map. However, Hameed performed an experiment that helped change their minds. He called a meeting with patrol officers and asked them to sit down with a map and mark the areas of highest crime. He then revealed the actual saturation of crime incidents in their beats as recorded by the mapping tool and showed the significant mismatch between their perceptions and fact. This meeting helped thaw resistance to evidence and analysis and led to greater openness to the crime mapping project.
A short-term result of the pilot project was that it changed decision-making regarding patrolling plans in Sargodha police. As the team had hypothesized, crime concentrations were much clearer in urban areas of Sargodha than in rural areas, and Hameed altered patrolling patterns accordingly. Also, the research team and their police partners observed that the mere act of mapping beats naturally created new rigor and accountability: delineating boundaries highlighted the need to re-evaluate the beats’ borders based on crime incidents and population density, while naming who was responsible called for rationalization of staffing decisions.
The project promised long-term benefits, too, in proving that the tool could be a useful permanent part of policing in Sargodha. With the completion of the pilot project, the researchers turned the tool over to the Urban Unit, to support its institutionalization in the Sargodha district with support from TPI and the computer wing of the Sargodha Police. The pilot showed that a built-in algorithm for capturing clusters was not relevant for rural use, so technicians are now writing the code for the app to be used widely and creating a data dashboard for Sargodha. Already Hameed has identified a way to deal with more dispersed crime in the countryside: he has installed tracking devices in police vehicles to allow dispatchers to swiftly send officers to rural crime sites.
An external evaluator of the BCURE pilot project said that, “It appears this mechanism was successful in achieving its stated aim, with respondents suggesting that data presented in visual form through the Pilot Projects was revelatory, challenged people’s preconceptions, and sparked enthusiasm amongst civil servants.”
Back at the 2014 BCURE Policy Dialogue, Cheema had presented three questions to the group of decision-makers in the breakout session. “I asked them, was this tool relevant to them, what kind of analysis would matter, and what could the police do if, for example, we found crime to be highly concentrated in certain areas?”
The answers he received were that the tool was relevant, and the evidence of hotspots could lead to a change in practice. Eventually, in Lahore the team showed that 7.5 percent of neighbourhoods were the site of more than half of the city’s crimes, and that these hotspots stayed constant over time. Such crime concentrations are not as dramatic as in American cities, but they are high nonetheless.
The question now turns to how the police will act on this evidence. “Now you know what the problem is,” says Cheema. “You’ve done the pathology work, now what’s the pill you’re going to dispense.”
It is a major step forward that Lahore is considering moving away from random patrolling and toward targeted policing. The choice may seem obvious, however it is one that was never entertained before the crime mapping project. Another policy response on the table is to invest in intelligence-gathering in areas of high crime saturation.
In Sargodha, while hotspots were clearly visible in urban and semi-urban areas, further analysis is needed to determine their degree. The policy response may be different, as options exist in small cities that don’t exist in large ones. Population churn and estrangement from neighbours kept Cheema from knowing about the wave of car thefts in his Lahore neighbourhood, but these factors are far less relevant in Sargodha. A neighbourhood watch programme might be more feasible.
Decisions have yet to be made and their effects measured. Yet there are more options on the table now, and they are being discussed at a deeper and more informed level than ever before. Hameed describes a “huge realization” of the power of data across Punjab police.
"There’s a sense that the old, conventional way of doing things is not really serving anymore, so officers now keep coming up with new ideas and new initiatives that are getting entrenched in the police. Three or four years ago, we thought it would be very difficult, but now it’s almost become a reality that in 760-odd police stations across Punjab, all police stations have at least a couple of computer-trained civilian staff who attend to people coming in, log in their complaints, send them feedback through SMS, and monitor the case through IT. Increasingly, all crime incidents are fed into the new software that was developed. So I don’t think this is going to go away, I think has actually reached a state where it will develop further."
Hameed’s sense was given very public expression on 11 March, 2017 when Prime Minister Nawaz Sharif visited the Central Police Office station to publically inaugurate a new policy called the Front Desk System across all police stations in Punjab, from crowded Lahore districts to the quiet Sargodha countryside. Now when citizens arrive at police stations they first encounter, not policemen, but civilians – generally young, computer-literate college graduates. For victims of crime this might mean a friendly face, but behind the scenes it means a far more data-rich environment where complaints are entered and tracked electronically and cases of non-response are automatically escalated to senior officers. Tracking of incidents has spread from the crime scene to the front desk, and the hope is that increased accountability will soon follow. ■
Text by V. McIntyre. Design by Gabrielle Hauray.