Tuesday, September 15, 2015

Statistical Evidence for a Ferguson Effect

Robert VerBruggen, who edits RealClearPolicy, looks for a "Ferguson Effect" in the recent homicide data for 59 of the sixty largest U.S. cities, data we linked to three days ago.
Advocates of a "Ferguson Effect" say that intense criticism of police over the past year has emboldened criminals and cowed officers.

The protests are heavily focused on race, so if there's a Ferguson Effect, cities with high black populations should have bigger increases in homicide — the protests presumably embolden white criminals less, and intimidate officers less when they're dealing with white suspects.

It's striking that, with one exception (Memphis), every single city where homicide declined had a black population under 31 percent. (Jacksonville had a tiny decrease and is 30.7 percent black.) Inversely, every single city except Memphis with a black population above that mark saw homicide increase.

Details: p=.035, R-squared .075.
When I was a number cruncher, my rule of thumb was R-squared equalled the effect size as a percentage. Horseback analysis: post-Ferguson has seen a 7.5% average increase in homicides for cities with large black populations. Not huge, but significant because black lives matter. The Pearson r = 0.27 where a perfect lockstep correlation = 1.00.