Violence tends to be a very localized affair. Most areas have little violence, but very small and specific parts of a city tend to be very dangerous in comparison. In Chicago, the stability of these violent regions is interesting. The areas that were dangerous in the 1970s are, for the most part, the most dangerous areas of this decade. To visualize this, Andrew Papachristos mapped the homicide rates of Chicago neighborhoods in the last four decades:
To support to this hypothesis, I ran a simple correlation matrix of the number of homicides per police beat from 2001 to 2012. The result? There tends to be a moderate sized correlation (around .35 to .55). The more dangerous beats in 2001 tend to be the more dangerous beats in 2012, and the less dangerous beats in 2001 tend to be the less dangerous beats in 2012:
However, while this correlation exists, Andrew Papachristos also notes that rates of violent crime are going down in general. Here he plots the homicide rate in Chicago since 1965:
And while the data I have from Chicago's website only goes back to 2001, it can also be used to show a similar trend. Here, I plotted the total number of murders in Chicago over time since 2001:
According to Papachristos' data, 2001 represented a small spike in total homicides in a larger trend of decline. Starting in 2005, the total number of homicides seems to have stabilized at around 475 per year.
When this data is broken down to the beat level, it gets even more interesting. Here, the total number of homicides per beat are plotted in a given year. Each beat's point is slightly transparent, meaning that lighter dots are more rare numbers and darker dots are more common:
From this graph, there seems to be little change in the beats with 5 or fewer homicides per year. But starting in 2004, the frequency of very high homicide beats (7 or more) is reduced. That would seem to indicate that the reduction in total homicides in the early 2000s was more of a reigning-in-the-outliers strategy than a moving-the-mean one.
However, when you create histograms of the frequencies, you see a different picture:
While the number of beats with more than 6 homicides is decreased from 2003 to 204, there are also many more beats with 0 homicides in 2004, which might indicate that the homicide reduction occurred across the board rather than just in dangerous beats.