Canadian Journal of Communication Vol 44 (2019) 343–350  ©2019 Canadian Journal of Communication Corporation

Data Corruption: The Institutional Cultures of Data Collection and the Case of a Crime-Mapping System in Latin America

Carlos Barreneche, Pontificia Universidad Javeriana

Carlos Barreneche is a full-time Assistant Professor in the Department of Communication at the Pontifica Universidad Javeriana, Bogotá, Colombia. Email:


Background  This article is a case study about a surveillance system deployed in a Latin American city that collects and analyses geocoded historical crime data in order to identify crime hot spots. 

Analysis  The case study focuses on the adoption of this technology by data collectors and the institutional cultures that mediate its workings. The article documents the conflicting adjustment strategies carried out by low-level police officers when the same crime data that they help to produce are operationalized as labour performance indicators.

Conclusion and implications  Drawing from scholarship in the field of critical data studies, this work situates the practices of data generation within institutional power relations to shed light on the particular politics at play in data-driven policing systems in the Latin American context.

Keywords Crime data; Critical data studies; Surveillance; Policing; Global South 


Contexte  Cet article présente une étude de cas sur un système de surveillance, mis en œuvre dans une ville d’Amérique-latine, qui collecte et analyse des données de criminalité historiques géocodées afin d’identifier les points chauds de cette criminalité.

Analyse  L’étude de cas porte sur l’adoption de cette technologie par les collecteurs de données et sur les cultures institutionnelles qui en assurent son fonctionnement. L’article décrit les stratégies d’ajustement problématiques mises en œuvre par les officiers de police de rang inférieur lorsque les mêmes données sur la criminalité qu’ils aident à collecter, sont utilisées en tant qu’indicateurs de leur performance professionnelle par l’institution.

Conclusion et implications  Sur la base des développements dans le domaine des études de données critiques, ce travail situe les pratiques de génération de données au milieu des relations de pouvoir institutionnelles afin de mettre en lumière la dimension politique présente dans la mise en œuvre de systèmes de surveillance. 

Mots clés  Données sur la criminalité; Études critiques sur les bases de données; Surveillance; Maintien de la sécurité; Sud global


In Latin America, surveillance is closely linked to insecurity and violence. It has mainly been publicly discussed as a technical problem, part of the strategies used for crime control, but not as a problem on its own (Arteaga, 2014). The public, fearful of the spread of violence, widely welcomes the adoption of all sorts of surveillance solutions in cities, which have the highest concentration of crime and violence in the world (Muggah & Tobón, 2018). The research for this case study was carried out in Bogotá, Colombia, a city representative of large Latin American capitals that is characterized by accelerated population growth, high levels of social inequality, high unemployment rates among young people, organized crime and illegal economies, high crime rates, and weak state institutions that elicit low public trust—factors prevalent in cultures of violence.

Instead of looking at the capacities of surveillance technologies, this study delves into the institutional cultures (McQuade, 2016) and agency of data collectors (Paterson, 2002) involved in the implementation of a crime-mapping system to assist policing work. The aforementioned system implements a model that approaches urban surveillance based on quadrants. That is, the city is divided up into streets and streets into quadrants. Police commanders, and the officers under their command, are then held accountable for the single quadrant they police. This is an approach framed within a strategy known as community policing, which is typified organizationally by decentralization—as responsibility is more widely delegated to officers—and by a geographic focus, whereby accountability is less temporal (the duration of the shift) and more location based. The approach to policing changes from being essentially reactive, taking action to address crime events, to proactive or preventive, carrying out actions to prevent crime. In order to try to anticipate what crimes might occur, where they might occur, and when, officers are offered statistical data on how crime has historically occurred within the limits of the assigned quadrant. This is a model of policing that relies on a comprehensive register of geocoded crime data.

While community policing emerged in the U.S. and the U.K. in the 1980s (Barlow & Hickman Barlow, 1999) as part of neoliberal programs that subjected public services that could be privatized into the managerial technologies of government (Stenson, 1983), predictive technologies in policing can be traced back to “tools for crime record visualization that were pioneered in the U.S. and Europe in the 1970s” (Pasquinelli, 2017, p. 284). Today, community policing informs the rationale for policing strategies across the world. At the same time, police surveillance increasingly involves the use of computational technologies for crime prediction.

There are a number of unintended social consequences related to databased surveillance systems (for racism, see Browne, 2015), and these are often particular to specific cultural and geographic contexts (Wood, 2009). Craig Dalton and Jim Thatcher (2014) have stressed that the production of data “is always the result of contingent and contested social practices” (n.p.). Following these premises, this research looks into local, situated, and everyday practices of data production in the case of the implementation of a crime-mapping system, in order to explore the power dynamics shaping the management of crime data in the Global South and their potential consequences.

The data for this case study come from in-depth interviews with one of the crime-mapping system’s designers and a small group of police officers, including one high-ranking officer in charge of IT management. The limited number of respondents is related to the culture of secrecy surrounding police work. The primary goal of the interviews was to understand how crime data was operationalized in software for crime control, and how police officers experienced such implementation in their daily work.

This article is organized into two sections. The first section introduces the case study and identifies a set of data practices that emerge as “technological adjustment” (Pfaffenberger, 1992, p. 282) strategies to the implementation of the mapping system. The second section discusses potential power asymmetries that might arise when hot-spot crime mapping creates certain representations of a city and its citizens.

Intentional errors in the bureaucratic machine 

In 2011, the police began to implement the Vigilancia Comunitaria por Cuadrantes program with the primary aim of improving the planning of police services. The program was supposed to help focus surveillance tasks more effectively through the identification of so-called crime hot spots. In the developing stage, the quadrants were designed as delimited geographical units sharing similar patterns of criminality, and a set of hot spots was identified and then validated by the police. In 2014, the crime-mapping system was set to assist daily police work (Fundación Ideas para la Paz, 2015).

Each police station was equipped with a geographic information system (GIS) that basically provided access to statistical information on crime (information on places, dates, times of day, crime modality, criminal profiling, and potential victims). This GIS allows for the identification and analysis of so-called hot spots of crime and other crime patterns, so as to provide the commander with a picture of where to focus police attention and how to plan actions and deploy resources (e.g., police patrolling). Moreover, the information system is fed by a database of crime reports made by citizens; victims are offered a map interface to situate where the crime occurred and register it as a geocoded data point that includes the incident report from police officers’ service record sheets, which is manually uploaded into the database. At each station, an analyst has access to all this information aggregated in statistics and map layers. The statistical data is analyzed and used to inform the commander for weekly planning. The commander uses the data to allocate tasks and goals for each officer.

In addition to informing operations, each incident reported within the quadrant is automatically also registered in a performance-tracking system linked to an officer, or a group of officers, who is then held responsible for addressing it. This is then weighted against the officer’s assigned targets to evaluate performance. Therefore, the data produced to track crime are also used to inform performance outcome measures.

When agents of surveillance are exposed to surveillance themselves, their attitude toward work and the tracking system changes (Goold, 2003). In some cases, officers are told they should not do anything that does not have an impact on the targets. Officers do not often engage in actions if a crime event is off the limits of their quadrant, even if they could do something, such as chase a thief.

Translated into performance metrics, crime data feed managerial systems of reward and punishment. Promotions or days off are granted as rewards, while tasks in urban zones that are considered less desirable are assigned as punishment. When police officers themselves are placed as data-collection agents, with a direct interest in the generation of data that may be used to audit their own work, these systems become subject to abuse. This sets up perverse incentives for the officer to produce certain data, or produce no data at all.

The next section of the article applies the “data friction” framework (Bates 2017; Edwards, 2010), which stresses the “socio-material counter forces … that slow[s] and restrict[s] data movements” (Bates, 2017), in order to identify particular officers’ data practices that function as strategies of “data friction” shape how and what crime data are collected. These data practices will be listed and briefly described through actual examples taken from the interviewees’ accounts.

Deliberate miscoding

This practice is particularly common in cases of murder. In a murder event, the incident is coded in the forms as the outcome of a street fight driven by intolerance. However, in the report’s description of the event there might be evidence that, for example, the victim’s body was tied up or presented signs of torture, making it apparent that no actual fight took place. In such cases accountability is deferred from the police to the citizens and their supposed lack of tolerance. This is a practice that may result in the creation of data silences, where data harmful to performance assessment may be less likely to show up in the statistics. 

Deterrence tactics to prevent citizens from reporting a crime

This practice is recurrent in cases of street theft. If, for example, someone’s mobile phone is stolen and the victim goes to the police station to report the crime, the police may send the victim to another station with the argument that the crime is not under their jurisdiction. If the citizen insists on submitting a report in that particular station, then the victim is asked for the phone’s purchase receipt as certification of ownership. If the victim persists, the officer may offer a form for lost objects. The victim may still state in the form’s observations that the mobile was actually stolen, but the report is made on a lost-property form. The whole process may take hours, and many people just give up on reporting, thus reducing the negative impact on the performance metrics of that particular quadrant or officer.

Deliberate delaying of data flows

There is a lack of interoperability among the crime databases of different law enforcement agencies, and even within different divisions of the police. This is a loophole that some officers conveniently exploit. For example, crime databases use different codes for crimes or citizens IDs. Thus, victims wanting to report a crime may be sent to the prosecuting authority, and since the databases are different, when the report finally arrives on the desk of a police officer, the officer has to manually type the crime report into the police’s own information system. According to the law, this has to be done within two months, an incentive to slowing down the flow of data. When the crime record finally shows up in the quadrant’s statistics, it is already too late to impact the performance indicators for the month. In this way, the incompatibility of information systems represents a case of “data friction” at the infrastructural level. 

Data manufacturing

The burden of administrative workload also affects the officers’ data practices.  Police officers have to fill in daily service sheets accounting for the routine work assigned to them. To deal with this burden, it is a common practice to photocopy these sheets, changing just the dates and even making up information. Moreover, the pressure to meet targets, such as the number of arms or drugs seized, sometimes also leads to manufacturing data. In some cases, officers are compelled to leave their quadrants, often to so-called red zones where such things are more easily found, so they can make them count as an achievement for their quadrant. Police refer to the unsaid things an officer does to meet targets as “the hidden curriculum.” This practice may well produce over-policing effects (e.g., increasing harassment) in certain quadrants.

Preventive and even predictive approaches to policing seek to visualize crime through its mathematization in order to control it more efficiently (Andrejevic, 2017; Pasquinelli, 2017). Consequently, the institutional adoption of these technologies of crime control is offered as an impartial and objective technical solution (Mantello, 2016)—an instance of what Evgeny Morozov (2013) has termed “technological solutionism.” However, as the motivations that drive police officers’ data-collection practices show, the crime-mapping system became a site for labour contention. Such practices—however dishonest—can be understood as subverting responses to contemporary neoliberal bureaucratic forms of government based on performance metrics (Beer, 2016), confirming findings in previous research on related surveillance apparatuses in the context of the Global North (McQuade, 2016).

Furthermore, the fostering of “data friction” through the deliberate tampering of data flows can be seen as ways to resist “technological regularization,” enhance workers agency and autonomy, and exploit the system from within to reappropriate the technology in more favourable terms (Bates, 2017; McQuade, 2016). Indeed, a few years after the crime-mapping system was implemented, officers hardly use it anymore because they simply do not trust the data they themselves have contributed to producing. Instead they use so-called “parallel statistics,” which they glean by listening to their radio devices, where they hear whatever events happen within the quadrant, whether they are reported or not. And since this information does not affect their performance assessments because it is not automatically registered (datafied), it flows freely. In this case, officers contest “technological regularization” through clinging to old modes and technologies for their daily work (Schafer, 2003). Helen Kennedy and Jo Bates (2017) acknowledge this occurrence, writing that data studies need “to acknowledge that datafication is experienced and called into question at the level of the everyday” (p. 704).

“Bad neighbourhoods” with “bad people”: The performativity of crime data

Looking at the conflicting ways crime data are generated for hot-spot policing is also important beyond their institutional sites of production insofar as such data call into existence certain urban geographies: the heat topographies of crime. Tobias Matzner (2016) proposed to focus on this performative dimension of data as a way to bring forward the “problematic consequences of surveillance procedures and the full scope of affected persons” (p. 197). In this light, the presence of noise, or “corrupted data,” in the mapping system is neither simply a problem of dysfunction or misuse due to human-induced error nor of data integrity (the objective veracity of data), since data are never an undistorted representation of the world (Boyd & Crawford, 2012; Gitelman & Jackson, 2013; Kitchin, 2014). It is important to take into consideration that urban crime data are actually used for several other purposes beyond policing, including the governing of cities at other levels: public policymaking, local planning, the allocation of public budgets, and investment in communities impacted by crime. Police chiefs also use crime data to back their security policies. Such important decisions might then be not only ill-informed by skewed data but end up indeed producing a certain image of the city (Kitchin, Lauriault, & McArdle, 2015) that could further reinforce existing patterns of social segmentation and deepen social inequalities.

In the case of Bogotá, the rate of some crime reporting is low. Robbery reports, for example, are low, particularly thefts of mobile phones and documents (wallets), crimes that disproportionally impact working-class people who commute by public transportation and on foot. This is partially explained by the fact that robbery has been naturalized in poorer neighbourhoods and public trust in the police is low—further adding to the data silences enacted by the officers’ practices—leading to the underreporting of such crimes. For crimes against private property, however, the rate of reporting is significantly higher because owners need to report in order to claim insurance. Noting that only more affluent people have access to insurance, what are the possible consequences of poorer neighbourhoods being underrepresented in police databases for property crimes but overrepresented for disorderly conduct types of crimes, such as drug possession, domestic violence, sexual assault, or so-called quality-of-life offenses? What are the consequences of richer neighbourhoods being overrepresented for property crimes? Arguably, this might end up in the investment of more police resources to protect private property in richer areas and more intervention on the private lives of citizens in poorer areas.

As the designation of hot spots accounts only for certain crimes more often associated with the poor (homicides, personal injuries, and various categories of theft of private property), those characteristic of the wealthy (e.g., theft of public goods) are made invisible. Such crimes, coded then as performance indicators for police work, end up legitimating the distribution of the police force (Barlow & Hickman Barlow, 1999). This has the potential to shape public perceptions of crime (Jefferson, 2017), police targets (criminal profiling), and what places should be protected along class lines. Since crime data are a byproduct of the interaction between citizens and the state, and these encounters are asymmetrical, the representation of social groups in systems driven by surveillance data does matter in terms of power relations. 


Although cases of meddling with information might be found in performance-management regimes in other sectors and across regions, cultures of corruption more often permeate public-sector institutions in the Global South, where such practices are easily naturalized as everyday routines for workers (as per the data practices documented above).

As long as public money is spent on surveillance systems fed by “corrupted data,” law enforcement institutions must be held accountable for the technical and ethical requirements necessary to produce fairer data in order to address potential social harms. Calls have been made to advance an agenda for data justice (Dencik, Hintz & Cable, 2016). Proposed solutions, such as open data approaches, are limited at best, as long as the public cannot have a glimpse into the messiness of data production and the social complexity of “data assemblages” (Kitchin & Lauriault, 2014). Likewise, totalitarian accounts of surveillance technologies power, and its privileged focus on matters of privacy, which proliferate in the theorization of accounts from the Global North, fall short in contexts where error, manipulation, and misuse are ingrained in the institutional cultures where these systems are adapted.

From the perspective of the Global South, this case study calls attention to the need for developing accounts of how institutional agents operationalize data in organizational cultures and social contexts characterized by pervasive corruption and inequality. Moreover, an exploration of the unintended social effects of faulty and unreliable data infrastructures prone to human meddling and bias may be a productive line of inquiry for a program of critical data studies from the South.


The authos would like to thank the Pontificia Universidad Javeriana for funding this research project, a case study which is part of the Ciudad de Datos: Datos Masivos, Ciudadanías y Gubernamentalidad.


Andrejevic, Mark. (2014). Big data, big questions: The big data divide. International Journal of Communication, (8), 1673–1689.

Andrejevic, Mark. (2017). Digital citizenship and surveillance: To pre-empt a thief. International Journal of Communication, (11), 879–896.

Arteaga, Nelson. (2014). Doing surveillance studies in Latin America: Social sorting in contexts of violence. Surveillance & Society, 13(1), 78–90.  [January 13, 2019].

Barlow, David E., & Hickman Barlow, Melissa. (1999). A political economy of community policing. Policing: An International Journal of Police Strategies & Management, 22(4), 646–674.

Bates, Jo. (2018). The politics of data friction. Journal of Documentation, 74(2), 412–429.

Beer, David. (2016). Metric power. London, UK: Palgrave Macmillan.

Boyd, Dana, & Crawford, Kate. (2012). Critical questions for big data. Information, Communication & Society, 15(5), 662–679.

Browne, Simone. (2015). Dark matters: On the surveillance of Blackness. Durham, NC: Duke University Press Books.

Dalton, Craig, & Thatcher, Jim. (2014). What does a critical data studies look like, and why do we care? Seven points for a critical approach to ‘big data’. Society and Space Open Site. URL: [January 13, 2019].

Dencik, Lena, Hintz, Arne, & Cable, Jonathan. (2016). Towards data justice? The ambiguity of anti-surveillance resistance in political activism. Big Data & Society, 3(2).

Edwards, Paul. (2010). A vast machine: Computer models, climate data, and the politics of global warming. Cambridge, MA: MIT Press.

Fundación Ideas para la Paz. (2015). Puntos críticos de inseguridad: Un tratamiento diferente a los problemas persistentes [Report No. 24]. URL: [July 2, 2019].

Goold, Benjamin J. (2003). Public area surveillance and police work: The impact of CCTV on police behaviour and autonomy. Surveillance & Society, 1(2). 

Jefferson, Brian J. (2017). Digitize and punish: Computerized crime mapping and racialized carceral power in Chicago. Environment and Planning D: Society and Space, 35(5), 775–796.

Kennedy, Helen, & Bates, Jo. (2017). Data power in material contexts: Introduction. Television & New Media, 18(8), 701–705.

Kitchin, Rob. (2014). Big Data, new epistemologies and paradigm shifts. Big Data & Society, 1(1),1–12.

Kitchin, Rob, & Lauriault, Tracey P. (2014). Towards Critical Data Studies: Charting and Unpacking Data Assemblages and Their Work. The Programmable City Working Paper 2. [Unpublished manuscript]. Pre-print version of chapter to be published in Eckert, J., Shears, A. & Thatcher, J. (Eds.). Geoweb and Big Data. University of Nebraska Press. [Forthcoming.] URL:

Kitchin, Rob, Lauriault, Tracey P., & McArdle, Gavin. (2015). Knowing and governing cities through urban indicators, city benchmarking and real-time dashboards. Regional Studies, Regional Science, 2(1), 6–28.

Mantello, Peter. (2016). The machine that ate bad people: The ontopolitics of the precrime assemblage. Big Data & Society, 3(2). URL:

Matzner, Tobias. (2016). Beyond data as representation: The performativity of Big Data in surveillance. Surveillance & Society, 14(2), 197–210.

McQuade, Brendan I. (2016). Police and the post-9/11 surveillance surge: Technological dramas in the bureaucratic field. Surveillance & Society, 14(1), 1–19. 

Morozov, Evgeny. (2013). To save everything, click here: The folly of technological solutionism. New York: Public Affairs. 

Muggah, Robert, & Tobón, Katherine Aguirre. (2018). Citizen security in Latin America: Facts and Figures. Rio de Janeiro, BR: Igarapé Institute.

Pasquinelli, Matteo. (2017). Arcana mathematica imperii: The evolution of Western computational norms. In M. Hlavajova & S. Sheikh (Eds.), Former West: Art and the contemporary after 1989 (281–293). Cambridge, MA: The MIT Press.

Paterson, Craig. (2002). “Street-level surveillance”: Human agency and the electronic monitoring of offenders. Surveillance & Society, 4(4), 314–328. 

Pfaffenberger, Bryan. (1992). Technological dramas. Science, Technology, & Human Values, 17(3), 282–312.

Schafer, Joseph A. (2003). From rhetoric to reality: How police officers view the implementation of generalized community policing. Justice Research and Policy, 5(1), 1–35.

Stenson, Kevin. (1993). Community policing as a governmental technology: Community policing as a governmental technology. Economy and Society, 22(3), 373–389.

Wood, David M. (2009). The “surveillance society”: Questions of history, place and culture. European Journal of Criminology, 6(2), 179–194.