Some More Correlations on Mass Shootings in the United States

Read Part 1: Some Time Series Data on Mass Shootings

Estimating through end-May, count regression (negative binomial, maximum likelihood):

 

Figure 1: Mass shooting casualties (killed+wounded), left scale (brown), and mass shooting events (blue bar, right scale). Gray shading denotes assault weapons ban. Orange shading denotes Trump administration. Source: Mother Jones database.

We can analyze correlations using regression analysis (negative binomial for casualties, Poisson for events).

casualties_t = –1.72 – 0.72 ban_t + 1.21 trump_t – 0.14 biden_t +0.00001pop_t

Adj-R² = 0.022, SER = 32.6, NOBS = 389. Bold denotes significance at 10% msl, standard errors corrected for overdispersion (GLM option in EViews).

For number of mass shooting events, count regression (Poisson):

events_t = –5.27 – 0.55 ban_t + 0.55 trump_t + 0.014 biden_t +0.00001pop_t

Adj-R² = 0.100, SER = 0.565, NOBS = 389. Bold denotes significance at 10% msl, standard errors corrected for overdispersion (GLM option in EViews).

Population for April, May extrapolated using previous year’s worth of population data (in log first differences).