Friday, 30 March 2018

Child Murder

Thinking about the dubious and misleading claims and assumptions surrounding the murder of trans people in New Zealand, and the extremely underwhelming statistics in the background, made me wonder: are there demographics which genuinely are more at risk of murder in New Zealand? There are many ways you could look at this: sex, ethnicity, income level and so on. I chose to look at age: specifically, the murder rate for under five children. High profile cases over the past ten or so years include Nia Glassie,  Cris and Cru Kahui and  Moko Rangitoheriri. Numerous media articles have drawn attention to the fact that New Zealand has one of the  highest rates of child homicide in the world with  sixty one children murdered over the past ten years.

The politics surrounding the media attention on these tragic deaths is murky. Lobby groups such as the Sensible Sentencing Trust and Family First skew the narrative in a conservative direction. There is a definite undercurrent of racism in the focus on the Maori families of the most high profile victims. Like many others on the left, I stayed away from the  March for Moko in 2016. I am uneasy, uncomfortable and unsure about the political motivations of many of the people who focus on these deaths. If I have learned anything over the past year or so in highlighting the problematic nature of gender ideology, it is this: we need to look carefully and honestly at exactly these issues. The ones that make us uneasy, unsure and uncomfortable.

So what do the stats look like for this group? I used the same police data for the 2007 – 2014 period, alongside Statistics New Zealand data and census data from 2006 and 2013. Carefully reading the tables revealed something I did not notice before: there is a distinction between ‘murder’ and ‘manslaughter’, and the statistics which describe age demographics are drawn from totals of these two categories. In what follows I will refer to “murder rates” when what I really mean is “murder and manslaughter rates”. There is of course a discussion to be had about how these two acts are distinguished, and the heated debate about whether the death of Moko Rangitoheriri was a case of murder or manslaughter is a clear example of the importance of this debate. Conflating the murder and manslaughter statistics is debatable and problematic, but I will not discuss this here for the sake of brevity.

A consequence of joining the murder and manslaughter statistics together is that the rates per 100,000 people are higher. My previous calculation using the same data (but just the murders) gave a homicide rate of 1.17. In what follows, I have calculated a combined murder/manslaughter rate for each year between 2007 and 2014. The rates look much higher than those I have seen quoted in sources such as Wikipedia (0.91): the median rate for the eight year period is 1.53, the mean is 1.60. In  this recent Stuff article, the homicide rate for the 0 – 19 age bracket is quoted as 0.75 per 100,000 people. My calculations for the murder rates for the under 5 age bracket are much higher, which makes me wonder about how many of the deaths were classified as ‘manslaughter’ rather than ‘murder’. (The police data does not show this breakdown).

The following table summarises my findings:

Anyone familiar with the pitfalls of statistical reasoning will look with some degree of wariness at this – samples of n=8 are basically useless, and calculating standard deviation for these tiny sets is a very rough and ready measure of the unpredictable nature of this data. We don’t know for sure that this data represents a clear trend – yet the fact that the murder rate for Under 5s is significantly greater than the general rate for four out of the eight years covered is fairly strong reason to suspect that there is. Caveats noted, here are my key findings:

  • In six of the eight years covered, the murder/manslaughter rate was higher for the under five subpopulation

  • The overall median rate for the period is 1.53. The child rate median is 2.59. The child rate median is 69% higher than the overall rate

  • The mean rate for the period is 1.60 per 100,000. The mean rate for under fives is 2.61. The child mean rate is 63% more than the overall rate.

  • 90% confidence interval for child rate is between 1.97 and 3.25

  • 90% confidence interval for overall rate is between 1.45 and 1.7

  • Taking into account the variability with 90% confidence intervals there is still a difference between the highest overall rate (1.75) and the lowest child rate (1.97). This difference is 12.6% of the highest overall rate.

There are very solid reasons for identifying the under 5 age demographic as more at risk of murder than the general population. Even if the rate for under 5s appeared to be consistent with the general rate of murder, this would be a shocking statistical fact: children this young simply should not get murdered, there is no morally acceptable rate except 0.00 per 100,000. 

No comments:

Post a Comment