New England - ME, MA, VT, NH, CT, RI
Upper Atlantic - NY, PA, NJ, DE, MD
Middle Atlantic - DC, VA, WV, NC, SC
Gulf Coast - GA, FL, AL, MS, LA
Lower Midwest - OH, IN, KY, TN, MO
Upper Midwest - MN, WI, MI, IA, IL
Southwest - OK, TX, AR, NM, KS
Mountain - AZ, NV, UT, CO, ID
Great Plains - WY, MT, ND, SD, NE
West Coast - WA, OR, CA, HI, AK
You can argue that some of the states don't fit exactly, but if you want to re-run the calculations yourself, feel free and I'll be happy to post them here (I can't do this forever, I still have a day job). I arranged them by order of gun ownership (since that is the way everything else is done) and then looked to see if we have the rising trend that Hemenway's or mikeb hypothesizes or the chaotic mess that I postulate based on the higher datapoint sets.
|2006 Death Rate of Women (per 100,000)|
I think I found the sweet spot. The most important datapoint I see here is that the region with the highest gun ownership has the lowest homicide rate and a lower suicide rate and accident rate than all but the 4 lowest gun owning regions. What is so special about the Great Plains states? The first thing that pops into my mind is that there is only one major city in these states, Omaha. I would be willing to bet an entire year's salary that if I did this study further down to the county level (3000 datapoints), you would see one of two things. 1) The chaotic mess would truly look like radio noise, or 2) (which I think is more likely) The data would actually show the opposite is true: More gun ownership are in lower crime counties. Why do I say that, because I have a sneaky suspicion that the vast majority of violent crime (including gun deaths) occur in the cities, whereas the rural counties almost always have a higher percentage ownership of firearms. Unfortunately, this will just have to be a wish since the key piece of information, gun ownership by county, is probably never going to be found out.
One of the problems with using averages of several diverse input points is that you smooth out the highs and lows. This is exactly what has happened here, by choosing a ridiculously low dataset of 2 points (which I expanded to the full 5 it should have been) it appears that there is a correlation between gun ownership and death from guns. However starting at datapoints of 6 the correlation begins to have outliers, until at 51 datapoints, we have something that could be close to randomness. I ran a series of correlation lines in the spreadsheet for the 51 point dataset (I didn't want to bore you by including it here) and the R-squared values came out between 0.01 and 0.3. An R-squared value of 1.0 would indicate perfect predictability (correlation). So, depending on the line you draw, if you guessed what any state was, your likelyhood of choosing correctly or even nearly correctly would be between 1% and 30%. Not too many things less than that, except Vegas odds and the lottery.
Now, let me re-iterate again, I picked the groupings before I ran the numbers. Frankly, I think there are some things wrong with the groupings. DC should be with Maryland (I have stated my reasons before). Alaska and Hawaii shouldn't be with any other group (they are not similar to any regions). Nebraska doesn't really go with the other states, it is more similar to Kansas and Iowa. However, since I had to constrain myself in some way, those were the groups that I came up with. I am fairly certain that if you rearranged the regions using some set of reasonable rules (that doesn't have you gerrymandering), you would come up with the same results.
The last thing to look at with these regions is how "similar" they are (i.e. does it make sense to have them in the same region). For this I took the absolute difference between each state's rate and the average rate and then found out what the average ratio of difference there is. (By multiplying this value by 100 you would have the average percent difference. The goal would be to minimize all values.)
|2006 Death Rate Regional Decrement of Women|
Besides the New England region and the Great Plains region, the regions for homicides are clustered around .2-.3. Even New England and the Great Plains are only at .55 range. I would say this is pretty decent. If I had time (and someone was paying me), I could play around with the regions further to find the minimum. The other idea that this table indicates is that gun ownership does not affect homicides, suicides, and accidents in the same way (i.e. it isn't the cause). If it was the cause, we should expect to see all three numbers for a region fairly close together. So, in my hypothetical study that I was being paid for, I would actually have three different regions for the different death types.
Now I know that this was probably incredibly boring for the vast majority of readers. That is one of the points that I wanted to make. Any kind of statistical analysis is not going to be an easy read for your average college educated individual. If you have never had a statistics course, it is probably meaningless. Books and papers meant for the general public rarely have any kind of indepth statistical analysis. Their point isn't to provide you with all of the information. Their goal is to sell books or magazines and make money. Hemenway's table that mikeb provides is a perfect example of a completely watered down (dare I say doctored?) analysis meant to be spoon fed to the masses. Could you imagine ever making a major decision based on two datapoints?
Hmm, who should I marry? Well Jill has great eyes and knows how to dance, but Julie can play guitar and has a neat truck. Don't you think you should at least meet their parents, talk about life goals, finances, children, religion, etc? What mikeb is asking us to do is decide the issue based on two datapoints.
I have tried to be as honest as possible throughout this, I even published the information that could "discount" the gun case. I didn't fear it. No one should be afraid of facts. I've known for quite a while that guns aren't the cause of anything. Heck, guns don't even cause a range day to be good. They have no more influence than a rock. Today, I taught some young kids how to skip rocks. It was fun, not because of the rocks, but because of me and the kids. The rocks were just the tool that we used.
On to Part VII for the real fun!