In light of the little flurry of comments here on whether the murder of Anna Lindh proves anything about Swedish society, I decided to dig up some facts. While I am usually of the opinion that facts get in the way of a good debate, I suspected that in this case looking up some statistics might just lead to some enlightenment.
Specifically, I’ve been wanting to compare two indicators for Sweden and the US: incarceration rates and murder rates. America’s numbers are conveniently pored over in the current edition of The New York Review of Books, but I got a whiff of border-line data massaging going on in that article. Still, an interesting read, if one to lean into.
I’ve found a British Home Office document [PDF] that serves my purpose. It lists both indicators for a whole range of interesting countries. All I did was put the two series on perpendicular axes and voila, my scatter chartData is for the year 2000.:
Now, what does this chart tell us? Well, obviously that the US has both a higher incarceration rate and a higher murder rate. But we expected that. Lets look instead at the policy options available to each country’s government:
The orange lines represent the trade-off each country can expect between incarceration rates and murder rates. For example, if the US were to let half its prisoners go, it might have a homicide rate of 10 per 100,000 people per year, or nearly doubleThese are just ballpark figures, of course.. Now, let’s assume that both countries are happy with their spot on their respective policy lines P’ and P”. If you now look at the slopes of the lines drawn from the origin to each country’s position, you can clearly see Sweden is willing to tolerate more murders per prisoner than the US.
What if we implemented American policy preferences for crime fighting on Sweden’s level of crime? That would bring Sweden to point A on the chart. Incarceration rates would rise from the current 64 per 100,000 people to about 160. The murder rate would drop from the current 2.06 per 100,000 people per year to around 1.5. For a population of 9 million, that would mean forty-five extra people would be alive each year, but 9000 more people would be in jail.
If Swedish preferences for crime fighting were implemented in the US, we’d be at point B. US incarceration rates would drop from the current 685 (in 2000) to around 320, and the murder rate would rise from 5.87 per 100k to around 10. For a population of 290 million, it would mean 12,000 more people die each year, though there would be 1 million fewer prisoners.
There is of course another possibility: That people’s preferences are fundamentally the same, but that they are expressed differently depending on the crime rate. As the crime rate rises, people’s attitudes quickly harden, and the government’s stance on crime toughensOther interesting factoids from the report; Sweden’s homicide rate is a little higher than the EU average of 1.70 per 100k. Stockholm’s homicide rate, at 2.97 per 100k, is a little higher than the EU city average of 2.48. New York City’s homicide rate is 8.77. Pretoria wins hands-down with 41.1:
This would mean that Sweden and the US both have crime policies that are well suited for their environment. Swedes moving to the US do, over time, become tougher on crime, while Americans moving to Sweden tend to soften up. This approach lets both the Swedes and the Americans off the hook when it comes to toughness on crime. And it would let Sweden’s crime policy off the hook for Anna Lindh’s murder. But it does not explain why the US has much higher homicide levels in the first place.
Two problems with your very prettily drawn analysis. You assume that there’s a perfect correlation between the level of incarcertation and the murder rate. On an allied point, you also assume that prison are only populated by murderers. Both assumptions are silly.
But just to show I don’t only pickpickpick, I have a question, too. Who’s that country in Chart No. 1 with the massive incarceration rate AND the high murder rate? Would suck to live there.
Clearly, there are quite a few assumptions going on here. You point out two. It would have been nice if I had been able to find prison populations of convicted murderers, and I’m sure the info is out there somewhere. But as it is, I’ve used total prison populations as a proxy. Probably the US has a larger percentage of non-murderer prisoners due to its war on drugs.
Another assumption: That the murder statistics have the same criteria (they do not; some countries count euthanasia, some do not; Sweden counts initial homicide reports, even if later they turn out to be accidents or suicide.)
The policy lines also assume a certain correlation between the two indicators. It would be nice to see this chart over time — how countries’ different eras of tolerance and toughness translate into movement on this chart. It might help resolve which comes first: more homicides or higher prison populations: In other words, do countries move around in a clockwise or counterclockwise fashion on the chart, over time.
Oh, and that country in the top right is Russia. One country is off the charts, literally: South Africa has a homicide rate 54 per 100k, and a prison population of “only” 385 per 100k.
Statistics of murder in Sweden has been systematically wrong for several years. The Brottsförebyggande Rådet – Council for Crime Prevention – discovered this a few months ago.
Basically, misleading numbers for murder was picked from the police computer system and not checked for relevance.
Error sources included: murders commited abroad; deaths first thought to be murders, then found to be natural deaths; cases where there were five suspects for a killing resulted in five murders in the murder statistics, etc.
I wrote about it – including links to the press releases from BRÅ – in a comment on the blog Förvetet a few months ago – see
http://www.freiholtz.se/forvetet/2003_07_01_arkiv.shtml#105726053350391709
Aha! Maybe that explains why none of the current OECD or Word Bank data show crime statistics for Sweden since 1998. I was beginning to think there was a cover up:-)
Instead of treating the United States as one country
why not treat it as fifty? All of these statistics
are available on a state-by-state basis, so it should
be doable.
The states of the united states have different laws,
different prison policies and different murder rates.
If one is going to speculate on what is causing
what, or what is correlated with what, this should
give a lot more to go on.
Lumping it all together may actually be disguising
what is going on. I’d be curious what state of
the United States is closest to Sweden, and what
the similarities if any there are between the
two regions.
A pseud(o)-economist writes: what twaddle. For all of the reasons you’ve pointed out and then some. Where the hell do you get your pretty orange lines from? –not from anything in the data. The figures you quote based on them aren’t ball-park, as you suggest, they are ball-lox.
To try to work out what (if anything) the data might tell you about the correlation between the US reducing (or seeing a reduction in) its prison population/capita to Sweden’s level and resultant homicide rates, take the equation:
homicide rate = A + B*(prison population)
Run a regression analysis to find the value of B. Then calculate:
‘new US homicide rate’ = ‘old US homicide rate’ – B*(‘old US homicide rate’ – ‘swedish homicide rate’).
Of course, as you could tell from looking at the graph, the data suggests that a lower US incarceration rate would be associated with a lower (not higher) crime rate. The idea that higher incarcerations lead to higher murder rates is a plausible one, but I doubt anyone would say it was the major cause of high crime (anyone?). What the data may suggest (if anything) is what you say in your last paras –countries with more crime lock more people up. The data can’t tell you much more than that. (The fact that the data is crappy is actually less important as long as its random crappiness (i.e. that the error in the data for countries with large reported murder rates is about the same as for small muder rate countries)).
i have to agree with charles… what’s with the curve of the orange lines?
is this some sort of typical policy curve i would have known about if i went to international policy graphing school?
as for the data, or the comparisons… do you really expect to pick up on the swedish v. american pov’s on murder v. incarceration in this manner?
clearly such an analysis tells you something, without telling you anything. it is only telling you what you set it up to tell you in the first place.
what about-
economic state of each country for time of data collected? % of inmates from specific income groups and how that group factors into the income distribution of respective countries? % of murders committed in various styles of crime (murder as a result of a sunlight starved insanity v. someone hitting the rear bumper of the victim’s suv v. gang related v. robbery v. domestic v. etc.)? typical environmental conditions at times of murders in respective countries. education level of murderers in each respective country? % of murders taking place in urban areas v. population living in urban areas.
without a more diverse point of view you might as well prove that in comparison to america, murder in sweden is largely about the consumption of lutefisk.
i expect a proper report on my desk by friday.
I agree with Charles, too, apart from he got his second equation wrong:
‘new US homicide rate’ = ‘old US homicide rate’ – B*(‘old US prison rate’ – ‘swedish prison rate’).
Also, I’m intrigued by the possibility of the lutefisk link. Perhaps the repeated viewing of rotted fauna puts people off creating dead humans?
Charles, you doofus, you get it exactly wrong. If it’s any consolation, my first attempts to understand the data were along your lines, but the mistake I made (and recovered from) which are still wallowing in is treating the orange lines like an indifference curve and the black line as the line along which policy options are available. It’s of course the other way round.
It is fiendishly hard for a country’s government to move the country towards the origin, where utility is maximized (no prisoners and no murders). It’s utopia there, clearly, and impossible to get to. Yet for some reason some countries are closer than others. But I am in no doubt that if you open the jails, the murder rate would jump. Look at the extremes here too. If you let loose all prisoners, your murder rate would be astronomical. If you emprisoned all of society, there would be nobody left to be a murderer. Ergo, my curves are ballpark correct. Go ask a real economist down the hall. I dare you.
“If you emprisoned all of society, there would be nobody left to be a murderer”
ahhhhh- such comfort in the orwellian utopia at the end of the y axis.
comfort in the darwinian (mad maxian?) utopia at the end of the x axis as well.
Stefan– My point is that you have absolutely no basis in the data on which to draw your pretty orange lines. None. Not a one basi. Not even a hemidemibasi.
*All* that your data suggests is that countries with more crime also appear see higher incarceration rates. If you chose to make the theoretical statement that ‘changing incarceration rates cause changing crime rates’, then the data you present can help you answer the question ‘if my theory is correct, what is likely to happen to S crime rates if I changed the US incarceration rate to the Swedish incarceration rate.’ You come up with the couterinuitive result that crime would go down a lot. This type of result is usually taken as a reason to revisit the theory.
Let’s try a different one, then: ‘changing crime rates cause changing incarceration rates.’ Then the result is that lower crime leads to lower incarceration. Sounds less silly, really, doesn’t it. Of course you could also argue that some third factor (culture, whatever) leads to both a higher crime rate and a higher incarceration rate. The data wouldn’t dispute that, either.
However, that is all you can learn from the data. You can learn nothing about indifference curves between crime and punishment. These could be horizontal or vertical and you couldn’t tell from this data. Your curves (and the resulting numbers you have the gall to derive from them) are ballpark correct like its ballpark correct to say Matthew is a socialist lesbian pygmie piledriver-driver.
Oh, and while we’re at it, peoples’ indifference curves regarding crime and punishment and actual outcomes, as pointed out by Matthew in post one, don’t necessarily have squat to do with each other. So the assumption that you could move along your indifference curve and see the actual outcome suggested by it is a related set of bunk.
Indeed, the only theory your actual data would support (rather than have nothing interesting to say about) is the idea that if people said “I’d trade more prison population for a lower crime rate,” and then locked more people up, they’d *actually get a higher crime rate as a result*. I’m not saying that is what would actually happen, I *am* saying that the data you’ve got can’t say diddly about any other potential outcomes.
But I should rephrase the equivalent ballpark-correctness of your indifference curves to either the real indifference curves or the actual impact of changing incarceration rates on crime: its equivalent to the ballpark correctness of saying Matthew wants to be a socialist lesbian pygmie piledriver-driver. We’ll never know for sure (with this data), but it seems mighty unlikely that its right.
And, in the end, nothing about these graphs and comments answers the question: why did a social welfare state generate the murder of the most prominent member of the social welfare state?
It figures that a bunch of Swedes would immediately draw comparisons to the murder rate of their enemy, the United States. Murder is, according to Swedish welfare state thinking, a product of the social instability of a free-market society. But that only brings us back even more intensively to the question: why did this murder happen in a social welfare state?
From the discussions on this blog, it seems that most Swedes are simply clueless. But I can think of at least one possibility: that the social welfare state is a failure.
Thank you indeed, Markku, for bringing us back to the point. For surely this was it. Apparently, it always is.
I’m glad you concur.
The Economist reports erroneously that Sweden’s murder rate is higher than that of the US.
As an econometrician, I will have to go out on a limb (not very far, mind you) and concur with some of the above comments that your analysis is biased and has absolutely no statistical significance. There are several problems with your “regression” but one of the most important is that you have insufficient data. A single year’s worth of data is going to tell you what you want it too, regardless of what that is, but it won’t be even close to showing the “true correlation”.
Second, should you lack the necessary time series data that would be required to analyze any hypothesized correlation, use of cross-sectional analysis could work. That is, should you choose to run a regression with this data, you would do so using Y[sub-i]=f{X[sub-i], e}, thus resulting in one regression line for all i countries. The clearly wrong assumptions as to how to show correlations between two data sets.
Also, it is asanine to assume that because there is a correlation between 2 variables that there is causality. This is a common mistake that most people with limited knowledge of statistics make. For example, I recently did research on the United Nation’s Human Development Index and in running several models proposed as alternative social welfare indices, I had a little fun with my regression and found that there was (after selecting at random 50 of the world’s countries and when paired in alphabetical order) a very strong correlation between the HDI and auto sales in the United States by state in 1997. Now, does this mean that because more autos were purchased in Alabama that the quality of life in Albania will shoot up? No. That would indicate that three’s company on your 23rd chromosome. What does it mean? Statistics, unless done under the strictest of regulations, can and most times do provide inaccurate, biased, and illegitimate results and conclusions. I suggest you stick to Hamiltonian and Eulerian mathematics, maybe even branch out further into Knot Theory. That was a very enjoyable read.
Anders, this was me having some fun. Charles knew that.
this web page is very complicated