I’ve been sent an English translation of an official body-by-body investigationStrangely, the English translation is more detailed than the Swedish original conducted earlier this year into Sweden’s murder rate for 2002. The conclusion: “A total of 95 persons fell victim to incidents of lethal violence in Sweden in 2002.” For a population of 8.94 million at the end of 2002, that makes for a homicide rate of 1.07 per 100,000 people per year. Not 10 per 100,000, as The Economist reported, and lower than Japan’s rate of 1.10 per 100,000 in 2002.

The Economist used old Interpol data for 2001PDF thanks to Jan Haugland, which has since been “corrected”. The old Interpol data showed 892 murders in 2001 and a suspiciously exact rate of 10.01 per 100,000Could it have been a data entry error?; new data shows 167 murders that year, with a concomitant murder rate of 1.87 per 100,000. Interpol has not yet published 2002 data for Sweden.

That’s quite an improvement. But putting both PDFs — old and new — side by side raises many questions. The new data is clearly wrong when it comes to counting totals. Both PDFs count the total number of crimes committed in 2001 to be exactly 1,189,393. But the new data is now missing 622,232 instances of theft reported in the old data. In the new data, the total for category 4, all thefts, is lower than some of its subtotals! The old data also doesn’t add up, but not so flagrantly. What a mess.

Meanwhile, the Swedish report accounts for an overreporting of murders of around 60% over the last decade:

This would reconcile Interpol’s own new rate of 1.87 in 2001, based on Swedish police statistics (red line), with the lower total of around 1.10 in 2002, based on a counting of actual bodies (blue line).

What we’re seeing, then, is a compounding of two errors. Interpol’s bizzare error, and then a systemic overreporting of murders in Sweden’s own police statistics.

From the chart it is clear that the divergence between the two lines becomes much larger starting in 1992. That’s when the police implemented a computerized case tracking system that was intended to solve cases, not give accurate crime figures, but from which statistics were culled nonetheless.

The result is overcounting. For example, murders committed abroad but reported in Sweden were counted. Conspiracies to commit murder that were not consumated but discovered were counted. Attempted murders were counted. Suspected murders that later proved to be accidents or suicides were counted. False murder reports were counted. Some murders were counted repeatedly:

One example of this phenomenon may be found in a case where there were two victims, but which was recorded as involving three victims; and where, in addition, the offence report was completed twice. This means that a total of six offences were registered, of which only two were correct. Furthermore, the two offences actually involved had been committed several years earlier.

Here is the breakdown in numbers:

Quite a cautionary tale, then. But it’s probably too late to combat the frisson of excitement that coursed through the conservative web when The Economist‘s chart unwittingly endorsed Interpol’s error. A typical reaction, from the conservative American news site NewsMax:

Sweden, supposedly the land of granola-munching socialist peaceniks, had 10 murders for every 100,000 people. Yes, Sweden is branching out and is no longer just Suicide Central.

Maybe Stockholm will take a cue from our Second Amendment and allow its citizens the right to defend themselves from its out-of-control population of greasy-haired blond killers.

The moral: How nebulous statistics can be, and also how dangerous it can be to draw conclusions from improperly vetted data.