I’d appreciate feedback on by gun deaths database and analysis. Comments, criticisms, suggestions, etc. are welcome.
With the recent Sandy Hook murders, NRA suggesting we arm teachers, others arguing stricter gun control laws, and possible new legislation on guns ownership, I’ve been reading up a bit on the issue. Among modern industrial nations, U.S. murder rates and overall gun deaths are an extreme outlier. There are case studies and correlational studies supporting a variety of conflicting theories. But correlation is not causation and these studies rarely statistically control for possible extraneous variables.
Additionally, these reports don’t release their raw data for further analysis and even if they did, databases often don’t include variables you would want to control on.
Thus, I decided to start my own State level database and include as many variables that others have suggested impact our high murder and gun related death rates. I chose State level data because that is where I could obtain the largest number of key variables. And using multiple regression I could go beyond mere correlation. So far the database includes the dependent variables of gun murders, all murders, percent of murders using a gun, suicides, all gun deaths, using a gun in a robbery, and using a gun in an assault. The independent variables include poverty rate, economic inequality (GINI index), percent college degrees, percent urban, percent white, gun control laws, and importance of religion.
Thus the database is State level (N=50) with most data coming from government sources (FBI crime data, poverty from Census, college degree from Dept of Ed). Gun control law ranking is from Open Society Foundations and 2000 ranking. While there have been State law changes since, it was the only ranking by State that I’ve found so far - suggestions welcome. Full references below.
Findings: The findings are below. The left column is the dependent variable - in this case the death and crime rates. The right column includes the independent variables in the order of their influence using step-wise multiple regression. They number range from 1 to 4 depending on how many additional variables are statistically significant (.05 level). The number in parentheses (#) after each is the adjusted R-square value or the amount of variance explained. The number associated with the second variable (if there is one) will always be larger as it includes variation explained by both that variable and the previous one. Subtracting the first value from the second will tell you how much the second adds to explaining the variance. Ditto for additional variables. The step wise analysis stops when adding another variable does not improve on predicting the value of the dependent variable.
Note: I inserted a “+” or “-” sign in front of the R-squared value to indicated whether the impact of that variable increased or decreased the value of the dependent variable. Thus for “Gun Murders” higher poverty rates and higher economic inequality (GINI) were both associated with higher murder rates. For “All Murders,” poverty rate increases all murders while percent white reduces all murders.
Dependent Independent variables include: poverty rate 2010/11 average, economic inequality (GINI), % own gun, White%, Urban%, college degree%, gun regulation score,
Gun murders Poverty rate (+.385), GINI (+.435)
All murders Poverty rate (+.478), White% (-.516)
Gun robberies White% (-.409)
Gun assaults Poverty rate (+.311)
Suicides % own gun (+.386), urban% (+.471), GINA (-.517), College% (-.589)
All gun deaths College degree (-.609), % own gun (+.692), urban% (+.719)
Guns as % murder GINI (+.179)
Conclusion: The most important factor for gun related murders is the State’s poverty rate and its economic inequality. Poverty is also the most important factor explaining gun related assaults. Race is the best predictor of gun use in robbers where States with lower minority populations have lower rates of robberies using guns. States with the highest rates of college graduates have the lowest rates for gun deaths, with gun ownership and percent urban contributing to more gun related deaths. The greater a State’s level of inequality, the more likely a gun is used in murder cases.
States with the highest rates of gun ownership have the highest suicide rates with greater urbanization having a contributing effect. Interestingly, greater inequality and college degrees are also contributing factors but they reduce the suicide rate.
Not shown in the above, is analysis of variables by region (Northwest, Midwest, South, West). For murder rates, gun murder rates, gun robbery rates, and gun assault rates, the South has rates from 50% to 300% greater rates than the other regions. The West has higher suicide than the other regions.
And evangelical friend suggested that I include a measure of religiosity. The State level measure I was able to find was a Gallop poll asking about the “Importance of Religion” in respondents’ daily lives.
Personally, I did not think this would have an impact. However, as you can see from the data below “Importance of Religion” is one of the better predictors. For gun murders, all murders, gun robberies, and guns as a percent of all murders, the importance of religion has a positive impact. That is States with a higher percent of their population claiming religion is important in their lives have higher murder rates. Important of religion was negatively associated with suicide rates.
Dependent Independent variables include: poverty rate 2010/11 average, economic inequality (GINI), % own gun, White%, Urban%, college degree%, gun regulation score, importance of religion (God)
Gun murders Poverty rate (+.385), God (+.459), GINI (+.515)
All murders Poverty rate (+.478), God (+.543), White% (-.577)
Gun robberies White% (-.409), God (+.465)
Gun assaults Poverty rate (+.311)
Suicides % own gun (+.386), urban% (+.471), God (-.579), College% (-.682)
All gun deaths College degree (-.609), % own gun (+.692), urban% (+.719)
Guns as % murder God (+.198), GINI (+.302)
Things to consider. First, the level of analysis is States, not individuals. N=50 or 49 for gun murders, robberies, assaults, and percent murder by gun as there is no data on those 4 variables for Florida. Second there are probably better measures for some of the variables. Most problematic is the gun regulation score. It is from 2000 and a composite score. Some States have changed their laws in the past decade and it might be more helpful to look as specific nuances of these laws. Finally, these data and analysis are no time series so we cannot see how changes in the dependent variables might impact the dependent variables.
A special note on suicide and percent urban. Numerous studies find the suicide is more common in rural than urban areas. In the data here, we see a positive impact from percent urban on suicide rates. That positive correlation is after controlling on percent of gun ownership. The bivariate correlation between percent urban and suicide rates is -.354 with a p-value of .011.
References:
All gun deaths (suicide, murder, accident)
http://www.statemaster.com/...
Murder rates
www.fbi.gov/about-us/cjis/ucr/crime-in-the-u.s/2011/crime-in-the-u.s.-2011/tables/table-4
Gun crimes 2011 (murder, burglary, assault, percent of murder by gun)
http://www.fbi.gov/...
Does not include Florida. Also Alabama is covered up on the summary page (xlx file) but you can get those values from the linked data sheets for the individual crimes.
Gun ownership by State
http://usliberals.about.com/...
Gun control law ranking (2000):
http://www.opensocietyfoundations.org/...
Poverty Rate 2010-2011 average
Census.gov
GINI index
http://en.wikipedia.org/...
College degree percent by State
http://www.ed.gov/...
Percent Urban
http://www.census.gov/... (Table 29)
Importance of Religion in your daily life
www.gallup.com/poll/114022/state-states-importance-religion.aspx#2