Crime as a Myth to Be Exposed by Data
Crime is a subject surrounded by myths. Serious offences that receive particularly frequent media coverage trigger strong emotions in many peopleâespecially when the victims are perceived as defenceless and the perpetrators as foreign. A majority of the population therefore believes that crime has increased in the recent past. To the extent that this can be objectively established, however, the opposite is the case. With regard to important indicators of public safety (homicide rate; frequency of robberies, burglaries and motor-vehicle theft; violence in schools; and juvenile delinquency in general), the available trend curves for the last two or three decades clearly point downward. This does not contradict the fact that there are, in some respects, genuinely unpleasant new phenomena (e.g. forms of cybercrime made possible by technology, or the youth gangs in Vienna that have recently attracted much attention in Austria) and countervailing trends in specific areas of offending (such as sexual offences or hate crime). The latter, however, is due to the creation of new offences and to more careful detection and statistical recording. This long-term decline can be observed not only in Austria or Germany, but in most Western countries. There is still no explanation for it that is consistently accepted within the scientific community. Most academics assume that demographic shifts play an important role. This is plausible: in ageing societies there are simply fewer young people and therefore fewer young men, who, for as long as statistical records have existed, have shown a far above-average risk of offending in all cultures and at all times. Another important role is likely to be played by the considerable expansion of crime-prevention efforts, which successfully prevent offences at many levelsâfrom awareness campaigns in schools and technical security measures for homes and cars or video surveillance to reintegrative programmes supporting people who have committed offences.
At least as far as the prevalence of criminal offences is concerned, we are therefore by no means living in particularly unsafe times. For science, communicating this is a challenge. To a certain extent, distorted or at least selective perceptions of safety are psychologically unavoidableâthere need not be sinister plans on the part of corrupt media outlets behind them, even though there certainly are powerful political and commercial actors who have an interest in dramatising crime. This makes an objective and evidence-based view of the subject all the more important.
How Do We Measure Crime in the First Place?
It is not easy, however, to counter myths and exaggerations with a true, unbiased picture supported by empirical data. This applies, as already indicated, to correlations intended to explain causal relationships in temporal developments in crime or in the spatial distribution of crime rates. Because of the large number of possible influences on human behaviour, corresponding research findings must always remain model-based.
But the problem goes deeper: even the purely descriptive determination of the crime situation is a difficult undertaking. Social-science researchers have no privileged access to social reality beyond their methods. They have no magic crystal ball through which to see the true extent of crime. Measurements and counts of behaviour that may violate criminal law are inevitably dependent on perspective. In rule-of-law states, the criminal justice system is the only institution that can authoritatively decide whether an act is to be labelled âcriminalâ. It would therefore seem appropriate to use data on final convictions. Such data can indeed be highly informative, but they necessarily leave out cases in which no suspect could be identified. In addition, conviction statistics do not include proceedings discontinued by the public prosecutor because the existence of a criminal offence cannot be proven (empirically, this applies to the majority of all criminal proceedings). In the case of less serious offences that most likely actually occurred, the public prosecutor may also discontinue proceedings subject to conditions (so-called diversionâthis too is a quantitatively significant area), meaning that such cases likewise do not enter conviction statistics.
The latter statistics are therefore, although from an epistemological perspective they are the only statistics that reflect legitimate definitions of acts as criminal offences, of only very limited use as a âcrime barometerâ.
The police crime statistics are more commonly regarded as serving this function. They register cases closer in time to the event itself, namely after the initial completion of police investigations, and they also include offences with an unknown perpetrator. Since the cases have not yet been reviewed by the courts and the later outcomes of judicial proceedings are no longer incorporated into the police data collection, however, these statistics are nothing more (but also nothing less) than âsuspicion statisticsâ. Their content depends heavily on the populationâs willingness to report offences and, to a lesser extent, on police control activities (for example in drug offences). For offences that factually and unequivocally occurred and are almost always reported, such as motor-vehicle theft or burglary (the car is gone or the home has been broken into, and a report is needed for insurance purposes), police crime statistics provide a good approximation of the reality of crime. In the case of violent or sexual offences, by contrast, reporting to the police is the exceptionâthere are larger or smaller âdark figuresâ (unlike the colloquial term âdark numberâ, this is the technical criminological term) of offences of which the security authorities have no knowledge and which therefore do not appear in any official statistics.
How do we actually know to what extent offences are reported or not? To illuminate these âdark figuresâ, criminological researchers create their own data. When these are survey studies of experiences of victimisation using approximately population-representative samples, the results allow conclusions to be drawn about the extent of unreported crime: people are asked whether they have been victims of particular offences. If so, they are additionally asked whether they reported the incidents. A major advantage of such surveys is that they can ask not only about experienced victimisation or (as is sometimes done with young people) offences committed by the respondents themselves, but also about all kinds of contextual variables, including sociodemographic information, habits and lifestyle, perceptions of safety, attitudinal dimensions and personality variables. This makes it possible to test theories about victimisation as well as theories about the causes and perceptions of crime. Science thus attempts to measure and contextualise criminal activity independently of official crime statistics, which are not primarily produced for research purposes and are first and foremost records of the activities of law-enforcement agencies.
This type of data also has limitations, however. Apart from sharing the weaknesses of all survey data (we have to believe peopleâs answers, which can sometimes be strongly influenced by selective memory, social-desirability effects or a lack of understanding of the questions), victimisation surveys can only capture phenomena in which people are able to provide information about having been victimised. If respondents are not even aware that they have been victims (as in some cases of fraud), or if those harmed are legal entities (e.g. in cases of property damage) or society as a whole (for example in tax evasion or environmental offences), this method must fail to determine the âtrueâ extent of crime.
As is clear, therefore, measuring crime quantitativelyâwhich cannot be conceived independently of criminalisationâis no trivial undertaking. In circles within academic criminology that understood themselves as âcriticalâ, some seriously drew the conclusion for a time that the exercise should simply be abandoned. According to this view, crime was nothing more than a power-produced construct that was reified by seemingly objective numbers. The task of science was to expose this constructed character theoretically and, where empirical methods were used, through qualitative approaches. I was never convinced by this: if the problem of measuring crime is not addressed in a scientifically rigorous manner, the field is left to other actors who undertake the task with less scrutiny and transparency and who may be guided predominantly by a political or economic agenda. Data are produced and analysed anyway. What matters is being able to assess their evidentiary value, to interrogate them creatively and critically, and to use them in conjunction with other datasets so that robust findings about reality can nevertheless be extracted from themâand to put all of this up for discussion within a scientific discourse.
Correlations with Crime: Apparent and Actual Patterns
This brings us to the purpose of the âAs-if Science Jamâ: as the announcement puts it, the playful format is intended to make visible those mechanisms âthat also operate in real research, media reporting and politics: selective data selection, misleading visualisations, statistical tricks.â One way of doing this will be to pursue hypotheses that are as absurd as possible. Applied to crime data, this could mean establishing correlations with phenomena that quite obviously have nothing to do with the causes, conditions or contexts of criminal or criminalised behaviour. Relevant theories usually tell us which variables could really provide valid explanations of crime. Yet scientific theories can themselves merely reflect assumptions or ideological preferences. They are often themselves derived from data that were also collected according to more or less conscious preconceptions (or prejudices). If data production, selection and analysis are not free from ideological premises, then the identification of relevant patterns, correlations and results is all the more likely to follow criteria that do not always correspond to the ideal of objective science.
Criminology in particular provides abundant material for illustrating this. From todayâs perspective, one of its historical roots is even a textbook example of pseudo-science and erroneous pattern recognition. This root leadsâsignificantlyâto the prison: when imprisonment became established in Europe as the usual form of criminal punishment within what sociology calls âtotal institutionsâ, the need arose to organise prison operations. To do so, it was necessary to know certain things about the people being held there: When are they to be released? For which offences were they convicted? Do they have illnesses? Are they dangerous? Can their labour be used? Are relatives looking after them? Files on prisoners were therefore begunâdata were produced. At some point it became obvious that these data could be used not merely for administrative purposes, but also to make systematic, experience-based statements about criminals.
The Italian psychiatrist and prison doctor Cesare Lombroso (1835â1909), who is still frequently regarded as the founder of criminology, believed that the causes and forms of criminality could be identified through anatomical characteristics. Strongly influenced by Social Darwinism, he assumed that the criminal embodied an âatavisticâ type of human being, a throwback to earlier stages of development. Tattoos, which were widespread among prisoners, were regarded as evidence; Lombroso saw an analogy with peoples then considered âprimitiveâ, among whom tattoos were also common. He believed that different criminal inclinations could be read from particular skull shapes and body proportions. He regarded people with such characteristics as âborn criminalsâ. Although this crude biological determinism was criticised extensively even at the time and Lombroso was accused of flaws in his evidentiary methods (including inadequate control groups), he was a scientific superstar of his era, whose work continues to contribute to the belief that criminals can be physically identified. What is fascinatingâand dangerousâabout this view is that it makes it possible to exclude social causes and conditions of crime completely. The idea then suggests itself that, along with the identifiable potential and actual criminals, crime itself could be eradicated. It is no coincidence that Nazi eugenics drew on Lombroso. In his later years, when he became inclined towards spiritualism, he himself described the recognition of the bodily characteristics supposedly responsible for criminality as a flash of insight that came to him while examining the skull of a robber: âAt the sight of this skull I suddenly believed that I could see the problem of the nature of crime. An atavistic being, which in his person brings back the wild instincts of primitive humanity and the lower animals. Thus anatomically comprehensible were: the enormous jaws, the high cheekbones, found in criminals, savages and apes, the insensitivity to pain, (âŚ) and the irresistible desire for evil for its own sake.â
What criminology, insofar as it follows Lombroso in searching for individual causes of crime, still shares with him today is the fundamental question: What distinguishes criminals from non-criminals? From the 1950s onwards, psychosocial characteristics were predominantly brought into focus and biological explanations of the origins of crime were rejected, but criminobiological approaches have experienced something of a renaissance in the 21st century with the rise of modern neuroscience. The view that Lombroso was the progenitor of an empirical science of crime was, however, rejected early on by sociology with good arguments. In fact, a concentration on the offender research that he established so prominently obscures a second importantâand olderâroot of the discipline as an empirical, data-based science.
This root does not lead to the prison but to the curiosity of the modern stateâand thus to official statistics. In Napoleonic France and, at roughly the same time, in the British Empire, systematic numerical records of the activities of criminal courts began to be compiled at the beginning of the nineteenth century. These data were initially intended to serve efficient administration, but they were soon read as inventories of crime that could provide information about the moral condition of the population. It also seemed natural to analyse spatial distributions. It became apparent that these were not uniform. There were areas with higher and areas with lower crime rates. Interestingly, these distribution patterns remained remarkably stable from year to year. The idea suggested itself that the different social and economic conditions were responsible for producing these distributions. The Belgian statistician Adolphe Quetelet (1796â1874) put it this way: âSociety prepares the crime, and the guilty person is merely the instrument through which the crime is executed.â This was the birth of analyses that attributed crime not to characteristics of offenders but to social conditions. The variables to be explained were not individual acts of offending but crime rates in different geographical units, which could be correlated with all kinds of sociodemographic and socioeconomic variables, or even climatic variables (industrialisation, level of education, unemployment, average temperature).
As increasingly long time series of crime data became available over the course of the nineteenth century, it became possible to see that, despite often quite stable spatial distribution patterns, there were fluctuations: crime can rise and fall. This variability was also related to social and economic indicators. A number of statisticians found surprisingly strong correlations between crime and the cost of living, usually measured by grain prices. When food prices rose, the frequency of theft also rose. When they fell, less property crime was recorded. The reverse was true for offences involving bodily injuryâwhich was explained by the idea that people spent their disposable money on alcoholic beverages, which are known to encourage quarrelsomeness. I find these early social-statistical studies strangely fascinating: with only a very small number of variables, researchers at the time managed to reveal a plausible social mechanism for their eraâan achievement that many experts working with statistics, data analysis and quantitative social science probably still dream of today.
Later, of course, it was recognised that the relationships were not quite so simple (for example, they could be mediated by third variables) and that poverty does not necessarily cause crime. Once social policy and economic prosperity had freed large parts of the population from hunger and existential hardship, social reality also stopped doing researchers the favour of providing such powerful explanatory variables. The development of crime in particular is, as I already mentioned at the beginning of this text with reference to the still not fully explained decline in crime figures in the Western world, a subject about which relatively little is securely known. Apart from obvious turning points such as changes in the law (e.g. the decriminalisation of cannabis in Germany) or the COVID-19 pandemic, criminologists do not know why crime rates rise or fall without carrying out researchâwhich is often laborious and not very rewarding. Politicians know even less, despite their confident assertions. This uncertainty also creates room for a playful approach to dataâfrom which perhaps one or two things can nevertheless be learned.
Where Can I Find Crime Data?
Publicly accessible crime data that can be used without an excessive amount of preparation can now be found in countless places. I would like to recommend three data sources here that might be quite suitable for the As-if Science Jam:
Germanyâs Police Crime Statistics (PKS)
Germanyâs police crime statistics (see above regarding their evidentiary value) contain data on offences overall and on individual offences or offence categories, as well as additional variables. From 1993 onwards, time series for reunified Germany are available through the latest reporting year, 2025âwith 32 years, this is already sufficient for simple time-series analyses. The data are also regionalised for the slightly more than 400 rural districts and urban districts; in this way, correlations can be established with all kinds of socio-spatial variables. The data tables already provide frequency figures (offences per 100,000 inhabitants), so crime rates do not have to be calculated separately using population data.
Regionalised social data at district level: https://www.regionalstatistik.de/genesis/online/logon
UNODC Database on Intentional Homicides
The homicide rate is a popular dependent variable in criminology, particularly for historical and international comparisons. Unlike other indicators of violent crime, the dark-figure problem is mitigated hereâbodies are usually discovered, and established methods for determining non-natural causes of death have existed for a long time. In addition, mortality statistics provide a globally standardised data source. Although the data are not perfect and a certain proportion of undetected cases must also be assumed here (for example, forensic specialists frequently complain about inadequate autopsy practices), homicide rates nevertheless have the advantage of being much less dependent on nationally differing criminal codes and criminal-justice systems, as well as on the conventions and counting rules of the respective official crime statistics, which make international comparisons considerably more difficult.
The United Nations Office on Drugs and Crime (UNODC), a UN organisation headquartered in Vienna, has maintained a homicide-rate database continuously for many years:
https://data.unodc.org/datareport/hom-victim
The units of analysis here are countries, meaning that homicide rates can be correlated with all existing country-level data (OECD, World Bank, etc.).
German Victimisation Survey 2012 and 2017
The social-science data platform GESIS provides the data from the German Victimisation Survey, commissioned by Germanyâs Federal Criminal Police Office (BKA), for the 2012 and 2017 survey waves (see above regarding the evidentiary value of such surveys):