Predictors of Per Capita Complaints Against Police Officers

For this project we examine predictors of complaints of police excessive force. To keep the units of analysis consistent, we include only municipalities, not county or state level police departments.  The number of complaints will be adjusted for per capita population as will relevant predictor variables.  

Outcome Variable: Per capital police complaints.

Predictor Variables: GINI index, Frequency of mental distress, Frequency of physical distress, Excessive drinking, Violent crime in the city, Residual segregation, Adult obesity, Sexually transmitted diseases, air pollution, Ethnic composition (% White, % Black, % Latino), Ethnic composition of police (same percentages, if available), County level violent crime

Multicollinearity: If multicollinearity is above 3.0 VIF, those variables with multicollinearity will be reduced. 

Analysis: Analysis will be conducted with OLS regression.  If the outcome variable is not normally distributed, it will be normalized using square root transformation.  If this does not result in a normally distributed variable, we will switch to Poisson Regression.  No transformations will be made for predictor variables.  Deletion for missing data will be pairwise.