It came from a UN report. INTERPOL was only the first of several data sources and I didn't want to crowd the graph. If you want to see everything: Linky
I can't really imagine a plausible way to establish confidence limits on this type of data. The major sources of error are almost certainly in classification and initial reporting of the crimes. The best you can hope for is that the error is equal in each country surveyed, and so the relative comparison ought to be valid if not the absolute values. I suppose you could collect the information several times and generate your interval from that, but such a procedure would undoubtedly produce a significant underestimate of the actual figure.
Edit: If you want to see some equally impressive homicide data it's all here.
[addsig]




