Measuring violence : what goes on back stage

Studies of violence do not necessarily begin by counting the number of cases of homicide, but when some “particularly odious” crime is committed, there is a rush on available statistics on the subject, so as to place the peculiar event in a numerical context. There is also the idea that because homicide is such a serious matter, the information about it is more complete and more accurate. A comparison of the figures made available by various sources and examination of the categories used to analyze them does not corroborate this postulate.

Acts, attempts and victims

French law prescribes the same punishment for attempted homicide as for “successful” homicide. Now, police statistics did not make any such distinction until 1988. In 1991, 1,259 of the 2,614 cases counted (48 %) were attempts, whereas the proportion of attempts was 51 % in 1988. Since we have no knowledge of this proportion before 1988, there is no way, barring the assumption of a sort of golden rule of constant ratios, of determining what is behind the variation in the number of homicides (including attempts) in police statistics over the long term, with a rise from 1,387 in 1972 to 2,614 in 1991 (an 88 % increment) for all of the following headings : “gang killings”, “heinous homicide” (accompanied by robbery), “non-heinous homicide” (for other motives), “infanticides” and “poisonings”.

Does this inclusion of attempts account for the gap between the figures published by the Police department and the cases of “homicide or intentional injury of another person” counted by the INSERM (National institute for public health and medical resarch) in its statistics on the medical causes of deaths ? In 1991, for example, the INSERM found 625 deaths of this type whereas police statistics reported 1,355 killings. As we see, statistical sources are still far from coherent, because of differences in conceptualization and in data-collecting methods. But there is also the definition of the unit of reckoning, which is the act in police statistics and the victim in medical statistics. The Police department report for 1991 presents a detailed analysis of 1,600 “acts” of deliberate homicide or attempts reported to the central bureau of the investigating police. Surprisingly, we learn that these 1,600 homicides produced 2,007 victims, 953 of whom were simply victims of attempts. That gives us 1,054 deaths, if our calculations are accurate. Unfortunately, we are not told how many of the 1,600 “acts” were attempts, so that we cannot calculate the average number of deaths per successful homicide. This is enough to comfort people who are convinced that they are not being told the truth about the increase in violence, and accuse the experts of doctoring the statistics.

Defining homicide

These statistical difficulties stem partially from the very definition of homicide. If we restrict ourselves to the legal aspect, the French doctrine establishes a distinction between voluntary homicide (deliberate manslaughter and premeditated murder) and unintentional homicide, which pertains to the action itself (a gun is fired intentionally or unintentionally). The former, in the strict acceptation, excludes deliberate assault and battery having caused death without the intention to deal it, which represents a third, hybrid category. In this latter case, the action is deliberate, but not the outcome (the gun was shot deliberately, but the person did not intend to kill). These distinctions are delicate, and the charges may change on the way from the police to the court. The only indisputable definition would be the one behind the penal statistics, and they show 532 convictions for voluntary homicide in 1991 and 224 for deliberate assault and battery having caused death. The decision handed down by the jury has then put an end to all legal debate over the nature of the act. And yet, even at this level, there is a considerable shadowy area : certain homicide cases are sent to a “correctionnel” court1 because they are unintentional, whereas some people contend that they are serious offences susceptible of “cours d’assises”. Homicide attempts receive similar treatment, and in case of legitimate defense the verdict may be acquittal. When the author of the homicide is not identified, the legal status of the case remains somewhat vague, following dismissal resulting from an unsuccessful investigation. Last, when the author of the act is found to have no criminal responsibility (in a demented state when the crime was committed), one is tempted to consider that homicide did nonetheless occur.

Simplification and further details

There is nothing neutral about the decision to count – or not to count – assault and battery having caused death as homicide. Whereas the number of truly intentional homicides are definitely on the rise in police statistics, as we have seen, at the same time these assault and battery cases have declined from 706 in 1972 to 217 in 1991. And what about infanticides, which have also dropped from 140 cases in 1972 to 36 in 1991 ? Should we include them, despite all the doubts one imagines as to their definition ? This leads to a situation in which both extremes may contend to be right. A restrictive, or strict definition of homicide would only consider those cases recorded as such by the police. By this count, the number of homicides has more than doubled between 1972 and 1991 (going from 1,247 to 2,578). A broader definition, and one which is perhaps less sensitive to the variations between the categories, would add infanticide and assault and battery having caused death, thus arriving at an increase of only 35 % (from 2,093 in 1972 to 2,831 in 1991).

— Hold on, we are consulting the computer...

Or again, if sensationalism is desired, we might note that it is the “heinous homicides” – that is, homicides or attempted homicides attended by robbery – that have increased most, since the number of acts has tripled (going from 152 in 1972 to 447 in 1991) ; but in 1991, six out of ten of those homicides were attempts !

There is no such thing as a figure that is truer than another one, and specialists will tend to view these different presentations as complementary. The mass media require simplification, however, and therefore a short message. Can we accept the evacuation of the complexity of findings, for the sake of clarity ? More cogent still, can we restrict ourselves to the framework provided by police statistics, especially if we consider – to return to our starting point, homicide as a symptom of violence within society – that another form of more or less conscious violence produces ten times as many deaths in traffic accidents as may ever be counted by statistics on intentional homicide. But those deaths will never be included in statistics on crime or offending, even if they were caused by a drunken driver. It just depends on how you look at it ?

Bruno Aubusson de Cavarlay
September 1993

1 . In France, moderately serious offences (délits) are judged by tribunaux correctionnels, whereas major offences (crimes) are given a jury trial by a cour d’assises.

Penumbra, 2001 June