[lbo-talk] Southern vs. Northern violence

[Miles Jackson]

Michael McIntyre wrote:
> Miles -
> You're right that you're still more likely to be a homicide victim in  
> New Orleans than in San Francisco.  But this was prompted by a  
> dispute between Carl and andie about whether there was some  
> "cultural" (for lack of a better word) factor peculiar to the south  
> that prompts higher homicide rates there.  "What everyone knows" was  
> invoked on each side - that homicide rates are higher in the south,  
> that homicide rates are higher in northern cities, etc.  Either way,  
> "what everyone knows" is wrong.  Three of the five cities with the  
> highest homicide rates are border cities (unadjusted: Baltimore, DC,  
> & St. Louis; adjusted: Baltimore, DC, & Kansas City).  But why does  
> San Francisco, of all places, have so many more murders than the  
> sociodemographic factors would predict?
> 
> MM

Two points.

1. When sociodemographic factors are controlled in a regression model as 
this researcher did, any variance these factors share with unspecified 
variables is also controlled for.  So if there is some cultural factor X 
has happens to be correlated with the sociodemographics, that local 
cultural factor is also held constant by this regression model.  Thus 
it's not true that the regression model is only controlling the 
specified sociodemographic variables.  Rather, the regression model is 
telling us that when we control these sociodemographic variables and 
many unspecific cultural/social factors, San Francisco is ranked #1.  So 
what?  I could easily create a regression model that "adjusts" the 
homicide rate to give any major city in the U. S. the highest rank. 
This is the important point: if you include different sociodemographic 
variables correlated with different unspecified variables in your 
regression model, the regression coefficients and the residuals will 
also differ (this is what social scientists call "the specification 
problem").  There is no reason to include only the sociodemographic 
variables that the researcher included; thus the "adjusted ranks" are 
arbitrary.

2. Looking at the homicide rates in few select cities is not an 
effective method for comparing homicide rates in different geographic 
regions.  Fortunately, our good friends at the FBI do aggregate homicide 
rates by region.  Here are the data for 2005, rates per 100,000:

South: 6.6
West: 5.8
Midwest: 4.9
Northeast: 4.4

The general pattern is consistent with Nisbett and Cohen's culture of 
honor hypothesis. --Interesting aside, I know Carl will love this: in 
one study, N & C exposed college students to an abusive insult (a person 
bumped the student in a hallway and called him "asshole").  Immediately 
afterwards, students who grew up in the South were more physiologically 
aroused, had higher cortisol and testosterone levels, and felt more 
hostile than students who grew up outside the South.  In sum: The 
culture of honor in the South encompasses not just people's attitudes 
but also their physiological reactions when their personal honor is 
impugned.

Miles