Monday, March 27, 2006

Lower crime in Australia

The recent Australian Crime, Facts and Figures 2005 showed a widespread reduction in crime across Australia over the past decade:

  • Homicides were lower in 2004 than since 1996 and have fallen each year since 1999.
  • Robberies are at their lowest since 1997 and have fallen each year since 2001.
  • Motor thefts have fallen continuously since 1998 while ‘unlawful entries with intent’ have fallen since 2000.
  • Arrests for drug offences have fallen 24% since 1995/96, in part due to drug law reforms.

An exception to this happy picture is that assault figures, which comprise the vast majority of violent crimes, and particularly recorded sexual assault figures, have risen steadily since 1996. 82% of sex assaults are on females (mostly girls aged 10-14 years) but boys aged under 10 years made up 33% of victims in this age group. Interpreting these sex and assault statistics is difficult. There are low rates of reporting such crimes to police so rising trends in recorded crime might reflect mainly increased incentives to report.

The big issue: Why the improvement?

Is the fall due to the substantial improvement in the economy over the past decade with greatly reduced unemployment? To better policing? Or, as suggested in an earlier post, is it due to changed abortion laws that reduced unwanted children 20 years back? As I wrote in that post:

….Steven Levitt and Stephen Dubner in Freakonomics … show ….The main factor explaining the reduction in the crime rate in the 1990s was the Roe v. Wade ruling in the US Supreme Court that extended legalised abortion to the entire US. The …evidence provided in Freakonomics strongly suggests that it was unwanted children who tended to become criminals and murderers. Abortion law reform reduced numbers of such children…..'

Andrew Leigh confirms similar findings for Australia here. This evidence is less clearcut than that for the US since there was no specific date from which access to abortion became legal but liberalisation did tend to occur at the end of the 1960s with the murder rate peaking 20 years later as with US experience. In the 1990s when the law could be expected to impact on numbers of unwanted children the homicide rate fell.

Like Leigh, the Wikipedia discussion of abortion in Australia dates the introduction of liberalised abortion laws in most Australian states to 1969/1970 or 35 years ago. The Australian abortion statistics are poor but numbers of abortions rose strongly from 1984-1996 and have continued to have risen steadily in recent years. This is consistent with sharply-reduced crime now.

As I stated in my earlier post I am not a keen supporter of abortion - as a parent I have an irremovable, intense, negative block on the idea of killing anything approximating a young 'child' - and am unhappy with recent Australian trends as described. But I am (i) not sure it is my business to impose my views on others on this, (ii) unsure of the effectiveness of prohibitions anyway, and, (iii) must identify some positive social consequences of abortions on crime.

The determinants of recent crime trends, and the role of abortion laws in particular, would make a good PhD topic for an empirically-minded economist with interests in social issues. Interested takers?


Rabee said...

So what do you make of

and the reply

There is an industry in economics revolving around lurid empirical research with hardly any scientific substance. Is this a product of this industry?

conrad said...

The figures are still ridiculously high if compared against many countries in Northern Europe and Asia.

An even simpler explanation to the reduction in crime is that the age distribution in Australia is getting older, and that older people create less crime overall. (It would be useful to see if that is true of the exceptions mentioned -- I seem to remember there is a report hanging around from 1996 that has those figures).

The economic conditions have also been getting better for high crime groups (young males), hence reducing the amount of crime this group causes. Whilst this may have traded off with hidden unemployment amongst older males, this group doesn't cause nearly as much crime given the same SES, hence the trade-off may create a net negative in the crime rate.

D. Prentice said...

Hi Rabee

Is it the subject matter or the approach that makes it lacking scientific substance? While I haven't seen the papers that you refer to one way some of the work of Levitt (like his AER paper on crime) can be interpreted as getting better measurement of the direction and size of effect - related to approaches (possibly) first used in labor economics (Card and Krueger on minimum wages for example). It is important that this work is about (and just about the) measurement of specific effects so cannot be used to estimate consumer surplus/mark-ups etc like say contemporary structural empirical io (so would you see this work as being of more scientific substance?) But as far as it goes, the methodology when applicable can represent an improvement over earlier approaches. The IV approach in the crime paper is well motivated and seems to improve on the earlier approach.

It can be combined (for example I supervised a student who estimated production functions, but used a natural experiment type approach to identification to set up the set of regressions to estimate - which seemed to work out well.

I think the natural experiment type approach (when you can use it) has some advantages over the earlier IV approach in that it probably leads to more careful consideration of exogeneity than the earlier approach.

It is also possible I haven't seen the more lurid applications of this approach though as they would be in literatures (or job market papers?) that am not up with.

hc said...

Rabee, As I understand it Foote and Goetz identified an error in the Donohue III and Levitt (DL) paper which DL accept was an error. DL in their response however claim that, after dealing with the error, the original results hold. Its a bit hard to read the links you offer to both papers so I'll reprint them as a separate post above and invite comments.

I'll try to work through the articles to make a response myself. I'd be very interested in what David Prentice thinks.

I think Conrad's suggestions would definitely be worth incorporating as explanatory variables in the analysis of crime.

conrad said...

Actually, now I look at the population pyramid :

it definitely shows that there is a bulge around 35 (Vietnam babies I presume).

I seem to remember the figures for age and crime can be found in the ethnicity and crime report if you are really interested on the Aus.Inst.Crim. site. You can see how much is caused by young males!

I don't remember them having the kiddy fiddling figures, but I believe that is one type of crime that is not well correlated with SES status and the like, so increases in kiddy sexual assault crime are not going to tell you much about the reasons for other types of crime anyway.

Another reason I suspect that a lot of the effect is just due to to aging is that crime has also been reducing in France (and other European crime hotspots) for quite a few years (despite what one would believe from the "be-tough" government) -- I believe since the mid 90s if I remember correctly.

However, although abortion was legalized in France at a similar time as Australia (1970), they had additional social problems that Australia doesn't have, which would have created really large amounts of low SES groups for years after Australia reduced their creation.

Despite this, the crime rate has fallen, despite social conditions still being poor for many. SInce the fall began a bit earlier than Australia and the population is a bit older, I can only attribute it to people being less motivated to commit crime due to factors like age.

Further cross-country comparisons might be useful in this respect to see whether the same pattern holds.

hc said...

Conrad, That's a fantastic ABS graphic. Particularly if you highlight the surplus of males or females in the pyramid and trace out the evolution. There are going to be quite a few randy young men without partmers as we move towards 2050 and many partnerless older ladies.

Testing for age-structure effects on crime iusing cross country regressions could presumably be carried out in a way similar to testing the impact of abortion laws in different US states. Its a reasonable hypothesis.

I suspect the aging effects are less in the US where fertility has remained high. But the crime rate there has fallen dramatically in recent years.

hc said...

Conrad, I wonder if the story isn't simpler than the Donohue, Levitt tale. Australia has enjoyed an expansion for 14 consecutive years, there is record low unemployment and, yes, factors like population aging and legal abortions have a role.

By the way Foote and Goetz recognise that having ferwer young makles will reduce the crime rate. In fact they say that is the only effect that D-L are picling up. D-L say such effects account for 40% of the reduction in violent crime. See the post that follows this for the gory details.

conrad said...

I'm sure its partly due to economic growth too, as that has kept the unemployment rate of young males down, who are the ones causing all the crime (and especially crime in Australia -- if you saw the OECD report that came out yesterday).

Alternatively, there has been a reduction in crime in many countries where they have had much less growth, different social conditions, and different policies (like France, excluding late 2005 and 2006 :).

I don't see any great similarity in the policies across countries (or economic conditions), but you still have the same effect, which is why I suspect that a fair chunk of the reduction is due to things that are not due governments finding good solutions (although there are sure to be examples where it has helped), but just shifts in demographic patterns (like unemployment moving to older males and people just getting older overall).

In fact, if you didn't get such a reduction from the aging of the population, it would be very surprising, since it would suggest that for some reason there is a qualitative shift in crime patterns. Thus a null effect in this situation should really be taken as poor government policy.