Tuesday, August 04, 2009

What does the decline in homicide rates look like?   posted by agnostic @ 8/04/2009 08:40:00 PM

Steve points us to a brief review by Steven Pinker on the decline in war and violence. Focusing just on homicide rates, what exactly does that mean -- a decline in violence during modern times? It is impossible to have a solid feel for the observation Pinker wants to explain without seeing time series data on homicide rates (one of which he includes in his TED talk on the same subject). The pictures come from Manuel Eisner's review article in the British Journal of Criminology.

This is required reading (only 20 pages) for anyone who wants to understand crime, and especially changes in crime -- changes in the overall rate, differences across regions in the decline, differences in the decline across social classes, etc. If you don't have access to it, it's one of those rare articles that is worth the one-time price of $28 -- or just request it from one of your friends or colleagues who does have university access.

Below the fold, I've included the pictures for all countries that Eisner found data for, along with a brief remark on the trend for each country. The vertical axis is homicides per 100,000 population and is on a logarithmic scale (so that the visible changes are by orders of magnitude). Also note that the recent decline in crime since the early-mid 1990s may not be easily visible in these pictures, given that Eisner's article came out in 2001 -- not very long for the reversal to jump out of the graphs.

First, England:

Increases during the High Middle Ages, decreases sometime starting in the Late Middle Ages or Early Modern period.

Netherlands and Belgium:

Decreases starting in Early Modern period.


Decreases starts as late as the 17th C -- Scandinavia being one of the last parts of Western Europe to become civilized.


Apparent increase during High Middle Ages, decreases starting in Late Middle Ages or Early Modern period.


Barely visible change during 18th C, while steady decline only starts in 19th C -- Italy having lacked a strong central state until then. Article says that Northern Italy shows a much earlier decline than Southern Italy (no surprise).

Also notice the presence of cycles about the overall trend. Just because there were recurring crime waves and abatements of crime waves during the 19th and 20th centuries -- see here for the US, or see the Scandinavian graph above -- should not distract us from the clear downward trend going only a few centuries farther back. Any account of rises or declines must deal with all of these patterns, making it impossible to generalize the narrow hypotheses for the 1990s decline in crime -- there were no cell phones before then, the trend since 1500 has been toward less corporal punishment and harsh sentencing rather than more, and so on.

What we would do is write down a system of differential equations that claimed how two or more groups of people interacted with each other -- say, "criminals," "law-abiders," and "police" -- and fool around with them until they produced a solution that would show cycles or oscillations around an overall downward trend. The interactions between these groups of people are what real historical causes are made of -- not the sudden introduction of some technology or law (or sudden disappearance of some technology or repealing of a law).

I'm up for a math modeling jam session if anyone else is. I remember seeing ODE models from ecology where one species replaces another, although the values oscillate around the upward trend of the winner, as well as around the downward trend of the loser.

Labels: ,