Brazil: Maps of future crimes

City maps that plot estimated risks of homicide. A statistical tool that can do just that has been proposed for the city of São Paulo, Brazil, by research conducted at Brazil’s National Space Research Institute (INPE). The project comes from doctoral research carried out by Eduardo Celso Camargo who has adapted a geostatistical method that eliminates data instability and imprecision for the purposes of looking into potential crime.

Concerned with the impact of violence over public health, Camargo sought a method that would make it possible to map the risk of violence accurately reflecting the fact that levels of violence are unequally distributed in the city, a map useful for also violence prevention strategies.

Camargo used data from the Improved Mortality Data Program (Programa de Aprimoramento das Informações de Mortalidade - Proaim) available for the municipality of São Paulo, generating a homicide risk map for the town. “The database provided me with the homicide rate, that is, the total number of homicides divided by the size of the population.

The difficulty is that the homicide rate is unstable. In areas where the population is small, one or two homicides have a high impact on boosting the homicide rate, while in an area with high population density, a few extra homicides will not affect the homicide rate.”

Adopting the kriging technique from geostatistics, (more specifically: binomial bi-kriging, a technique that estimates a continuous risk surface associated to the event under study) the researcher was able to do away with instability, and this allows the technique to be used in other cases. “The idea is to map rare events, events that have little probability of occurring.”

Despite the initial progress, the study must also account for settings where cut off figures are used in evaluating event probability, which is especially relevant in planning. To solve this matter, Camargo resorted to simulation. Simulation is used to generate a large set of risk maps.

By combining the two techniques, the method may be used in any locality to measure any number of rarely occurring events, such as, for example, mapping other types of crimes, or even to estimate the incidence of cancer in a specific area.

Application

According to Camargo, the great challenge facing his method is application. “My proposal is academic. To convert it into a law enforcement tool we would need to make it operational, that is to say, we would need to develop software with a simple, user friendly interface.”

Although his work has been published, (in Brazil’s Cadernos de Saúde Pública last July and is available at the Scielo site), Camargo has not been approached by anyone in the field of public security interested in implementing his model. “We know the Brazilian police forces face problems such as the lack of infrastructure. To make the method applicable we would need to have information transfer from databases into the tool, so as to then generate a map. For that to happen, we need interested people.”

In Camargo’s view, the method could be very useful in the fields of public health and safety. He proposes the incorporation of other crime causing factors to make the tool even more efficient.

 
This could mean variables would be taken into account such as an income concentration index, school drop out rates, population growth information, unemployment among youths, among other data, that would make the estimates more precise.

Police in search of precision

According to the coordinator of Analysis and Planning of the São Paulo Public Security Office, Túlio Kahn, (photo) the Inpe researcher’s tool is more relevant for strategic rather than for operational matters. “The police need absolute figures. We need to know where, in which street, a certain crime is taking place”, he said.

According to Kahn, to adopt a rate as input data for mapping is good for pointing out trends or causes. “We need to have real time information on crimes to know where to send our officers,” he said. The São Paulo police force uses the Digital Crime Report (RDO) tool currently. According to Kahn, this allows officers to identify the precise location of each crime, and this helps their operations. The system is currently used by approximately 60% of the police force in the state.

RDO data allows the police force to plan ahead, police precincts draw up weekly plans, and battalions bi-weekly, and the Public Security Office looks ahead quarterly.

Kahn adds that the Office is familiar with the study and has already recommended it for those who work with crime risk and rates. “It is an interesting academic exercise, in that it points out correlations and causes, but the police needs more,” Kahn said.
 

Initial findings

Initial findings used aggregate homicide rates of the 96 districts of the city of São Paulo. Results point that the average homicide rate fell by approximately 27% from 2002 to 2004.

Taking the central region of the city, Camargo's study noted a homicide hot spot around the districts of Sé, Brás and surroundings. In 2002 the area had an average 128 deaths per 100 thousand inhabitants, dropping to 116 in 2003, and rising again to 124 in 2004.

Still according to the study the areas of Jardim Paulista, Alto de Pinheiros, Morumbi, Tatuapé and Perdizers, adjacent to the city center are all less susceptible to crime. These areas also have more infrastructure and have higher income.

From Comunidad Segura:

São Paulo's Human Rights Map

Compstat: IT contributing to Public Security

Read Further:

Camargo's research, (In Portuguese) 

Translated by Lis Horta Moriconi