Legal Beagle by Graeme Edgeler

I now have this data, following contact by the Ministry of Justice after my retraction (and Nikki Macdonald’s excellent work in the Dominion Post) was published, and the Ministry apologised for falling short of the high standard they set for themselves, and offered to provide comparable data if I still wanted it.

Well, that’s encouraging. Well done you and them.

Okay, so no evidence of deterrence. The onus is now on advocates of the policy to justify it. If reasonable people can see no evidence that the policy is necessary, support for it will ebb.

Equally, we don't seem to be getting evidence that restorative justice works, either. Intuitively, one suspects it ought to, and I've long been a supporter of the notion. Perhaps there's a problem with implementation?

Since justice must be seen to be done for the public to be confident that the justice system is performing a public service, how do we get people satisfied with the performance of the system? Seems to me the next step is formal incorporation of victims' rights, and an operational charter for the system that requires it to include victim participation in the design of outcomes...

justice must be seen to be done...

...ergo, it must also first be done, to be seen.

Okay, so no evidence of deterrence.

Some evidence of deterrence. The reduction in strike recidivism is greater than the reduction in strike offending.

Some, but so little that it is effectively negligible. Politics being a numbers game, the advocates of deterrence ideology would have to be able to cite significant numbers to win the public debate by presenting the policy as validated by testing. As a proportion, the improvement seems likely to be measured as such a tiny percentage of the whole that the adherent who employed a statistician to measure it would take such fright at the result as to lose confidence in the feasibility of announcing it to the public.

Some, but so little that it is effectively negligible. Politics being a numbers game, the advocates of deterrence ideology would have to be able to cite significant numbers to win the public debate by presenting the policy as validated by testing.

Well, what type of evidence would convince you?

Valuable research Graeme - cheers.
Many variables indeed in terms of measuring reducing recidivism efficacy.
I think third strike sentences are the real kicker. And the example you note is a great reflection of how inequitable the third strike effect is. That prisoner receiving the mandatory maximum 7 years prison for that particular indecent assault; there is no doubt but for the strike legislation and even with that defendants previous conviction list that there is not a judge in the land who would have sentenced him to more then 3 years. And if there were, that would undoubtably have been reduced on appeal.

I'm with the general public in this respect: same standard as for restorative justice. We just need to see a significant percentage improvement from both policies to make them seem credible. When it isn't forthcoming, we get the impression that policy-makers are operating on blind faith - or using smoke & mirror tactics.

A reduction of a few hundred in a population that size is likely to mean that the difference is not statistically significant. As you stated above that there are many other factors that could also account for the difference.

You could share the data with a statistician and get them to run a few tests to see how different they really are. But they might just end up telling you what you've already surmised.

We’re now at the level where alternative explanations become more likely.

For the lack of significance of the policy- that you have to argue about how and why it might be compared with other things that have been going on, I think it illustrative to look at Australian state comparisons-

http://www.abs.gov.au/ausstats/abs@.nsf/0/44E7066A7D173A95CA256BDC0012241F?opendocument

If you assume that Australia, as a federal aggregation, is setting a generally similar approach among all the states that call themselves Australian, and then look at the amount of variation that crops up between states, to make any strong statements about the effectiveness of anything, that effectiveness needs to be high in relation to the kind of variation we see within a country kindred to New Zealand.

There was a report in the paper recently about how the number of drink driving convictions (or maybe arrests) has gone down. So, yay!

But I happen to know that there has been less resources put into policing in this area. So basically there is no way of knowing whether the decrease is due to people drink driving less or it's the same rate as before but that people are not being caught.

Getting a further strike only counts if the person is caught. If the police are putting more resources into, shall we say burglary, then a person with a first strike may continue to commit crime but may not be caught.

And that leads us to the fight clubs in prisons and Serco turning a blind eye to it (at the least). How many strikes were not counted because Serco didn't want their inmates arrested/hospitalised because it would have screwed up their statistics.

So, it was before this law about 2% of serious offenders would re-offend in a serious way, and of those, 4% (or 0.07% of the original) would offend for a third time. Or you're just more likely to get it pinned on you, or whatever.

And aside from judges not entering certain convictions since the law change because they would be manifestly unjust, such as a 2nd or 3rd strike conviction, and people accepting lesser sentences rather than face the risk of such a strike, let's assume that the 30% reduction in re-offending could be real and note ...

That's reduced "serious crime" by half a percent. On top of the eight percent it dropped by anyway, across the same time period. Because there's less lead in the petrol when the current bunch of 20-somethings were kids, or the police got their funding cut, or whatever else has changed.

Which isn't nothing, it's like 30 less serious crimes, in five years, or 1 every couple of months. So it is possibly something, like a 1 in 700,000 reduction in your chance of having a serious crime committed against your person every year, or at least a reduction in the chance of someone being sentenced as such after the fact. Plus or minus an unknown amount.

--

But I guess if you get shot by a repeat violent offender while he's on parole, that's great headlines for weeks, and if you die of diabetes like everyone else, then it's not headlines at all, so sugar taxes are bad and no parole for all sorts of people even if it does nothing to help is good. Because "good" is whatever protects the government from headlines.

Some evidence of deterrence.

Just eyeballing those numbers I doubt that there is any statistical significance in the change.

What is far more interesting and less reported is the rationale behind the 3 strikes was to prevent the enormous problem of recidivism

Over 10 years, of 10765 "first strike" offenders, a mere 171 re-offended!!! That argues there is no need for any legislation targeting re-offending at all.

Over 10 years, of 10765 “first strike” offenders, a mere 171 re-offended!!!

You can't add these together. A bunch of the pre-first strike serious offenders will have reoffended since three strikes was enacted. In fact, nearly every single one of them could have re-offended with a strike offence, and it would count as their first strike, and not show up in this comparison.

nearly every single one of them could have re-offended with a strike offence

Um no.

For that to have happened you'd have to argue that "first strike" crime rate dropped to near zero in the second five years and the only people committing first strike crimes were those who'd committed them in the previous five years.

So yes you can absolutely add those numbers together. And even if you treat the two five year samples independently there is clearly no real evidence of recidivism in either sample.

So yes you can absolutely add those numbers together.

You really can't.

Is someone who committed their first armed robbery on on 1 January 2010, who committed a further armed robbery on 1 January 2013, after they were released someone who has re-offended? I say they are.

Under your addition theory, they are a person who didn't re-offend pre-three-strikes, and committed a wholly new offence, counting as a first-time offender post-three strikes. Why do you exclude this person, who committed two armed robberies three years apart, both within the 10 year period you are measuring from the group of 171 people who re-offended?

Um no.

For that to have happened you’d have to argue that “first strike” crime rate dropped to near zero in the second five years and the only people committing first strike crimes were those who’d committed them in the previous five years.

I do not say that that is what happened. In fact, I have other data showing it didn't. I am merely pointing out that the data we have in this post does not show this hasn't happened, and if that had happened, the data I have here would look identical. Because of this you cannot add the numbers together to get overall recidivism rates.

From your post

In the five years prior to three strikes, 5517 people were convicted of an offence where that conviction would have been a ‘first strike’ had three strikes been in force at the time, and 103 were convicted of an offence that would have been a ‘second strike’.

So no I'm not excluding that person who committed two armed robberies within three years. Your own numbers say that only 103 people fit into that category.

Treating the two five year periods completely separately you have 103/5517 two timers and 68/5248 two timers.

It is highly unlikely that those numbers are significantly different and neither proportion is indicative of a problem with recidivism.

There might actually be some data that shows some problems with recidivism but you have not presented it in your post and nothing in your post provides any evidence that the law has done anything. You'd need year by year data and a proper statistical analysis of the variability in the year on year data to say anything of value about the effects of the law.

BTW that's the only data that really should be being discussed if you are interested in rates of recidivism, by summing over five years you have obscured variability which makes much of what you discussed meaningless.

But since this is a political law rather than any attempt to improve the justice system the whole discussion is moot - you may as well talk about selling lost luggage.

So no I’m not excluding that person who committed two armed robberies within three years. Your own numbers say that only 103 people fit into that category.

Yes, you are. I have provided the numbers for:

offending that would have been a second strike from Jun '05 - Jun '10;
offending that was a second strike from Jun '10 - Jun '15.

I have not provided the numbers for offending that would have been a second strike from Jun '10 - Jun '15 but was not because what would otherwise have been the first strike would have been between Jun '05 - Jun '10.

I have not provided the numbers for offending that would have been a second strike from Jun '10 - Jun '15 but was not because what would otherwise have been the first strike would have been between Jun '05 - Jun '10.

Ah I see. So your thesis is that some (presumably large) number of first strike offenders in the '10-'15 tranche were in fact second strike offenders. This would imply in increase in offending from the two strike rate seen '05-'10 that is hidden.

Kind of like the rush to buy before the price goes up.

Or the rate could be pretty much static and the law change could have had no effect at all.

Really if you want to do an analysis you need year by year data from which you can extract variability and then you might be able to determine if any given year or group of years was statistically different. But frankly looking at what you have presented so far I'd say you'd be wasting time because the numbers are so low and so similar that it's highly unlikely you will find any difference that correlates with the law change, but I am not a statistician, I just use stats.

Meanwhile I remain utterly unconvinced by your data that there was any real problem with recidivism that required the law in the first place.

Essentially you are asking whether a politically motivated law change had any measurable effect on a tiny percentage of recidivist criminals.