Bernard Harcourt of the University of Chicago Law School is guest blogging at the Volokh Conspiracy this week. One of his topics is the very interesting graph below, which shows the institutionalization and homicide rates in the United States from 1934 to 2001. The line that is higher for the first half of the graph represents the institutionalization rate in the United States, counting those institutionalized in both prisons and mental hospitals. The other line represents the homicide rate.
At first glance, these data suggest -- or perhaps scream -- an intuitively plausible and simple relationship: higher rates of institutionalization, whether in prisons or in hospitals, result in fewer homicides. The correlation between the two is a staggering -.851. Harcourt, of course, goes beyond this first glance and examines the relationship in greater detail in two papers (here and here), but in part he concludes that studies of institutionalization and crime should use an aggregate measure of institutionalization tied to both prisons and mental hospitals, rather than just prisons alone. Prior studies omitted mental hospitals from the mix, despite their historical importance as a means of confinement.
One of the challenges for Harcourt's analysis is dealing with time-series data. My impression, which anyone is welcome to disabuse me of, is (1) there no strong consensus about how best to deal with time-series data and (2) the various solutions offered by methodologists change more rapidly than the solutions offered for many other problems. So if you have any time-series wisdom to offer, head on over to Volokh.