VoxEU Column Development Frontiers of economic research Politics and economics

Crime and violence: Desensitisation in victims to watching criminal events

Governments in Latin America seemingly go unpunished at election times for high crime rates. This column examines whether the region’s high tolerance for crime is the result of ‘desensitisation’, with people reacting less to crime the more they are exposed to it. It finds that victims of crime become desensitised compared with non-victims, helping to explain tolerance to crime and a weak relationship between crime and happiness in high-crime areas.

Crime is one of the main social problems in developing nations (Jaitman 2017). It is particularly serious in Latin America, a region with 8.6% of the world’s population, but 36.5% of the world’s homicides.1 Moreover, Latin American citizens systematically rank crime as one of their main concerns (Latinobarómetro 1995-2015). Yet, their societies seem to show tolerance for crime. For example, many incumbent politicians are re-elected even as crime increases or remains constant at extremely high rates. Moreover, the relationship between crime victimisation and happiness looks weak (Di Tella and Schargrodsky 2009, Graham and Chaparro 2012). Our understanding of this puzzling mismatch, between actual levels of crime and concerns related to crime and violence, is limited.

A natural hypothesis in this context is that victims gradually become used to high levels of crime, so that perceptions of crime are not primarily driven by the actual amount of crime. This hypothesis follows ‘desensitisation’ (or ‘habituation’ or ‘adaptation’) phenomena, namely, the reduction in the response to repeat stimuli observed in humans across many settings (e.g. Thompson and Spencer 1966 on hearing habituation to sound, Rosburg et al. 2002, 2006, Sörös et al. 2001).

In a recent paper (Di Tella et al. 2017), we discuss an experiment that examines whether victims of crimes become desensitised to violence, as captured by watching footage of criminal events in real life. Participants were asked to watch footage of real TV news programmes, with half randomly assigned to watching crime-related scenes (treatment group) while the other half watched non-violent footage (control group). Subjects, who had different victimisation experiences, were monitored on standard implicit markers of habituation, including biological (cortisol and heart rate levels) and cognitive ones (Stroop-like and Raven’s Progressive Matrices tests).2

The expectation was that, in response to violent footage, participants that were previously victimised would exhibit small changes in cortisol, heart rate, and measures of cognitive function and fluid thinking (desensitisation) relative to control participants watching videos without violence. In contrast, we expected participants who were not previously victimised to respond to exposure to crime-related videos with significant changes in physiological and cognitive tests relative to participants in the control group.3


With the assistance of a recruiting agency, a sample of 160 individuals aged from 24 to 65 were invited, induced by a cash payment, to participate in the experiment at Universidad Torcuato Di Tella’s Neuroscience Lab (Argentina) during late 2015 and early 2016. The recruiting agency was instructed to follow a sampling quota scheme based on gender and socioeconomic status. For each participant, we obtained measures of the four markers of habituation (cortisol, heart rate, Stroop score, Raven’s score) before and after watching the videos. Assignment to the treatment and control footage was stratified at the socioeconomic level. Standard tests showed no statistically significant differences in pre-treatment markers of habituation, victimisation history, and sociodemographic characteristics between subjects exposed to the treatment and control footage. Moreover, no statistically significant differences were found between victimised and non-victimised participants regarding pre-treatment markers of habituation and sociodemographic characteristics either.

For the four markers of habituation we ran a series of regressions that exploit data variations within individuals. In doing this, we placed the focus of analysis on measuring the mean change between pre- and post-treatment values of the markers of habituation across different groups of participants (i.e. Control, Treatment, Treatment Victimised, and Treatment Non-Victimised). Here, we summarise our main results:

  • In terms of Control versus Treatment, the intervention video per se produces no significant effects on treated participants’ cortisol level and Raven matrices results. For heart rate, there is a significant reduction induced by watching the crime video. There is also a reduction in time Stroop induced by the crime video.
  • In terms of Control versus Treatment Victimised and Control versus Non-Victimised, our main results are derived from analysing the interaction of treatment and previous victimisation for two different definitions of victimisation – at the individual and household levels. In general, a very interesting pattern emerges. Individuals who have been previously victimised show similar behaviour to the control group when watching the treatment footage. Instead, treated individuals with no victimisation history react differently to the treatment. Overall, non-previously victimised participants show higher cortisol levels, higher Raven scores, faster Stroop, and a lower heart rate than the control group. These patterns are obtained both when victimisation is defined at the respondent level and when it is defined at the respondent’s household level.4


We measured biological markers of stress and behavioural indices of cognitive control before and after treated participants watch a series of real, crime-related videos (while the control group watches non-crime-related videos). Our data consistently reveal that victims of crime become ‘desensitised’ compared with non-victims. This evidence might help to understand tolerance to crime and a weak relationship between crime and happiness in high-crime areas, such as those in Latin America. Furthermore, our findings may also be helpful in explaining episodes where the share of violent crime grows as crime does – more violence may be necessary to induce desensitised subjects to surrender their possessions in a robbery.


Di Tella, R and E Schargrodsky (2009), “Happiness, ideology and crime in Argentine cities”, IDB Working Paper Series 112.

Di Tella, R, L Freira, R H Gálvez, E Schargrodsky, D Shalom, and M Sigman (2017), “Crime and violence: Desensitization in victims to watching criminal events”, Journal of Economic Behavior and Organization.

Graham, C, and J C Chaparro (2012), “The Linkages Between Insecurity, Health, and Well-Being in Latin America: An Initial Exploration Based on Happiness Surveys” in D Webb, and E Wills-Herrera (eds.), Subjective Well-Being and Security, Social Indicators Research Series No 46, Dordrecht: Springer, pp. 197-252.

Jaitman, L (2017), The Costs of Crime and Violence: New Evidence and Insights in Latin America and the Caribbean, Washington, DC: Inter-American Development Bank.

Latinobarómetro (1995-2015), Informe Latinobarómetro, available here.

Rosburg, T, J Haueisen, and B Sauer (2002), “Habituation of the auditory evoked field component N100m and its dependence on stimulus duration”, Clinical Neurophysioly 113: 421-428.

Rosburg, T, P Trautner, N N Boutros, O A Korzyukov, C Schaller, C E Elger, and M Kuthren (2006) “Habituation of auditory evoked potentials in intracranial and extracranial recordings”, Psychophysiology 43: 137–144.

Sörös, P, S Knecht, E Manemann, I Teismann, T Imai, B Lütkenhöner, and C Pantev (2001), “Hemispheric asymmetries for auditory short-term habituation of tones?”, in J Nenonen, R J Ilmoniemi, and T Katila (eds), Biomag 2000, Proceedings of the 12th International Conference on Biomagnetism, Espoo: Helsinki University of Technology, pp. 47–49.

Thompson, R F, and W A Spencer (1966), “Habituation: a model phenomenon for the study of neuronal substrates of behavior”, Psychological Review 73: 16–43.


[1] Authors’ calculation from the World Development Indicators for 2014.

[2] Our sample shows high victimization rates (25% of our sample reports having been the victim of a crime in the previous year), although it does not include victims of very severe crimes like rape, kidnapping or homicide.

[3] Experimental setup details available in Di Tella et al. 2017.

[4] Details on the interpretation of the direction of these effects as well as regarding statistical significance is available in Di Tella et al. 2017.

840 Reads