Handle just finished reading William J. Stuntz’s The Collapse of American Criminal Justice which includes this chart of murder rate versus black population percentage:
The friendly professor, Handle explains, spends the first half of the book trying to explain it away, and the second half of the book proposing policies that fly in direct contradiction to it.
Addendum: The Amazon preview did not include the figure, but it did include this table:
I don’t particularly doubt that in the real world there may be a strong relationship of this sort, but “selected cities’ murder rates and black population 1950-1980″? What the hell? “Selected”? “1950-1980″? If I tried to present a graph of volume vs. weight for “selected biofluid samples from mollusca and echinodermata” I would likely be able to show a rather strong correlation that really does exist in the real world, but scientists would still rightly wonder what powerful laboratory psychoactive drugs I was on. Why is it sensible to just sorta average across the racial policy and welfare policy changes of the 1960s, and over the criminal policy changes of the 1960s and 1970s? Why is it sensible to use a selection criterion that’s too idiosyncratic to be describable in the caption, instead of something natural like “US cities over 500k population” or “US counties with population density over FOO”?
One cannot help but be reminded of Ancel Keys and his (in)famous fat-heart disease chart.