top of page
Writer's pictureGeoff Russell

Bolt's confounding problems

The Advertiser today featured Tory Shepherd at the top of the page taking on racists pretending to be protectors of free speech and Andrew Bolt at the bottom. My hyperlink is to a slightly edited Herald Sun version of the same story, you'll have to subscribe if you want the 'Tiser version. Shepherd points out that when a down trodden group wants justice, those at the top defend all manner of bad behaviour by appealing to the right to free speech. Bolt is in her sights, but he's a little different. He tries using numbers to smear aboriginal people and culture.


During the last couple of weeks, everybody has heard about 434 aboriginal deaths in custody

over the past 30 years. Bolt rightly points out that these are not murders and that they are not disproportionate. Aboriginals are not more likely to die in custody. Fair enough. Bolt claimed that this fact was "buried". Obviously not very deeply and not by Bolt's favorite whipping boy, the ABC, because I heard it on the Insiders last Sunday (7th June). What Bolt did bury was the fact that indigenous Australians are between 15 and 26 times more likely to be in custody in the first place. Ignoring this is almost as shameful as the fact itself.


But I want to discuss Bolt's far more libelous claim about more indigenous Australians are killed by other indigenous Australians than by non-indigenous Australians. The Institute of Criminology report containing the data is here. Contrary to popular belief, data doesn't speak for itself, it needs context and explanation. Is Bolt claiming that blacks are intrinsically more violent than whites? It certainly sounds like it. If he wasn't trying to say that, then he should have explained clearly why the data don't necessarily imply this.


Statisticians talk about "confounding". It's a magic concept that really should be taught in primary school. Who knows ... maybe it is. I haven't checked.


Warning ... the following paragraph contains lots of hypothetical sentences. Don't assume they are true or quote out of context!!


Suppose people who eat lots of red meat get more lung cancer. The claim is analogous to a

claim like "people with black skin commit more homicides", or "people who carry matches get more lung cancer". These kinds of things are called correlations. A is correlated with B just means that more of one is associated with more of the other. Scientists are interested in the causal relationships between things and correlations are constantly getting in the way and causing no end of problems.


A confounder, and there can be more than one, is a set of things which are causal, but ignored for some reason. In the case of red meat and lung cancer, the rate of smoking might be a confounder; the actual cause of the lung cancer. Similarly, high indigenous homicide rates may be caused by some complex combination of confounders ... like poverty, unemployment, alcoholism, poor nutrition, an abusive upbringing and others that remain to be uncovered.


Whatever are the causes of those high relative rates of indigenous homicide, you can be pretty certain that skin colour isn't among them. The murder rate in Australia overall has been declining for decades. It's currently about 0.8 per 100,000 people per annum, compared to over 5 per 100,000 per annum in the US. Even predominantly white California has a murder rate lower than the national average but it is still over 4 per 100,000 per annum. White Americans kill people, and not just one at a time. If Bolt is worried about murders, then calling for permanent border closures for US citizens (of all colours) would be a good call.


The percentage of indigenous Australian's as a ration of the population has been rising during recent decades ... from just under 2 percent in 1996, to about 3.3 percent today. Is this driving the decline of murder rates? If you are thinking that this is just an irrelevant correlation, then I'd be betting the same.


Are indigenous homicide rates high? They are certainly higher than they should be on a per-capita basis. One important question is why? And the other, probably more important question, is what can be done about it. We can probably go a long way towards fixing the problem, even if we don't entirely understand it. All of the things that we know work with non-indigenous Australian's will probably work with anybody. Satisfying employment yielding a reasonable stable income is probably at the top of the list. So lowering an indigenous unemployment rate which is close to double the non-indigenous rate seems a useful target.










0 comments

Recent Posts

See All

Kommentare


bottom of page