Saturday, January 15, 2011

Why Gun Control is for the Innumerate: Statistics as Suppressing Fire


State Firearms Death Rate per 100,000 Overall Homocide PER 100K Percent of Homocide w/Gun Brady Score AW Ban  Black
Hawaii 2.6 2.61 19.4 42 1 3.2
Rhode Island 3.4 2.41 61.5 45 0 6.4
Massachusetts 3.6 2.64 58.1 54 8 7.1
Connecticut 4.2 2.6 54 53 3 10.4
Iowa 5 1.36 47.5 14 0 2.8
New York 5 4.61 57.8 50 8 17.2
New Jersey 5.2 4.51 58.8 73 8 14.5
New Hampshire 5.5 1.39 30.8 9 0 1.4
South Dakota 6.1 2.34 31.3 4 0 1.2
Minnesota 6.5 2.22 66.4 15 0 4.7
Maine 7.6 1.37 52.9 11 0 1.2
Nebraska 7.9 2.89 63 8 0 4.6
Illinois 7.9 6.1 75.2 28 0 14.9
Vermont 8.1 2.58 18.8 8 0 1
Washington 8.3 3.06 56.1 17 0 3.9
North Dakota 8.6 1.41 50 4 0 1.2
Wisconsin 8.6 2.79 64.5 10 0 6.2
California 8.8 6.67 72.2 79 10 6.6
Delaware 9 2.05 64.3 21 0 21.1
Ohio 9.6 4.52 52.7 11 0 12.1
Oregon 9.8 2.51 53.3 17 0 2
Colorado 10.3 4.41 58.3 15 0 4.4
Kansas 10.3 4.5 60 7 0 6.2
Utah 10.4 1.9 49.1 0 0 1.4
Virginia 10.5 5.23 69.4 17 0 20.2
Indiana 10.6 5.06 69.6 6 0 9.2
Pennsylvania 10.6 5.24 71 25 0 10.9
Michigan 10.8 6.36 71.5 23 0 14.2
Texas 10.9 6.07 64.8 9 0 12
Maryland 12.1 9.37 74.2 52 1 29.7
Florida 12.2 5.08 62 6 0 16.1
North Carolina 12.3 6.23 59.5 19 0 21.6
Idaho 12.7 2.15 56.7 2 0 1
Missouri 12.8 6.15 68.8 4 0 11.5
Oklahoma 13.2 5.28 57 2 0 8.1
Georgia 13.2 6.87 64.5 8 0 30.2
South Carolina 13.3 6.86 67.7 10 0 28.2
Montana 13.7 3.24 71.4 4 0 0.7
West Virginia 14.2 3.75 50.8 4 0 3.7
Kentucky 14.2 5.7 62.7 2 0 7.9
Tennessee 14.8 5.95 63.1 8 0 16.8
Wyoming 15 2.17 27.3 10 0 1.4
New Mexico 15 8.88 50 4 0 3.1
Arizona 15.1 6.28 72.4 2 0 4.4
Arkansas 15.2 6.4 62.7 4 0 15.8
Nevada 16.3 7.37 64 9 0 8.3
Alabama 17.5 5.61 64.3 16 0 26.3
Alaska 17.8 5.63 45.9 2 0 4.2
Mississippi 18.5 7.83 70.9 6 0 37.2
Louisiana 20.2 12.74 79.5 2 0 32.1

This is a table showing gun deaths per 100,000 across each U.S. state and comparing it to other factors, including overall violent crime, certain demographic factors, and the Brady Center's "score" of the state's gun control laws, with a higher score meaning more restrictions on a citizen's second amendment rights.

This data is this table tells a story. It's the story of a country where cultural habits and beliefs differ significantly from state-to-state and where those cultural habits and beliefs manifest in different ways, including beliefs about the efficaciousness of gun control.

It also tells us that gun control is a feel-good measure with little empirical basis.

Analyzing the Data


One of the favorite targets of gun control supporters are so-called "assault weapons", which supporters believe are somehow less safe than "non-assault weapons".

In reality, a true assault weapon is a selective fire weapon firing a medium range cartridge, such as the AK-47 or M-16. Civilian ownership of selective fire weapons has been strictly controlled since the National Firearms Act of 1934. What nutso gun control supporters call "assault weapons" are actually just self-loading pistols, shotguns and long guns, occasionally cosmetically similar but functionally distinctive from their NFA counterparts.

But I digress. Let's just give the wack-jobs the benefit of the doubt for a moment and attempt to test their AWB (assault weapon ban) theories empirically.

Theory: AWBs save lives.

Hypothesis: If AWBs save lives, then there should be a negative correlation between a state having an AWB and the overall deaths per 100,000 from firearms.

Operationalizing the Test: The Brady Campaign provides a scoring system that compares the restrictiveness of each state's guns laws on a total scale from 0 to 100, with 100 being the most restrictive. In the case of AWB, Brady scores states from 0-10, with 10 being the most restrictive. For the purpose of this test I am going to perform a Person's r test to look at the correlation between Brady's AWB score with a given state's death per 100,000. If Brady is correct the safest states should be those with the most restrictive laws and vice versa.

Findings: I conducted a regression analysis in Excel to test the covariance between overall deaths per 100k and AWB scores. The overall findings, for all fifty states, was -.39, which suggests a medium negative correlation.

However.

The .-39 dosen't tell the whole story. While the number for all fifty states suggests that there is some connection between a state having an AWB and a relatively low number of gun deaths a deeper examination of the numbers reveals a problem. When we examine only the 10 states with the lowest number of guns deaths, the correlation drops to just -.15. If we expand to the 20 states with the lowest number of firearms deaths, we get a correlation of -.25. .25 is considered the very low end of covariance, and .15 is considered an ink blot test, so those numbers tend to indicate little connection between AWBs and guns deaths. Besides, shouldn't states with a higher AWB score be ranked higher? Shouldn't the state with the highest score (California) be ranked #1 (in lowest overall deaths) as opposed to barley cracking the top 20? Furthermore, shouldn't more than 50% of the states in the top 10 have AWBs in place?

What this suggests is not that an AWB saves lives (note that the two states with the lowest number of gun deaths score low on Brady's AWB score), but rather that, for whatever reason, some states that happen to have lower gun death rates also have some other factor that makes them more likely to enact an AWB or perhaps support gun control more broadly. But as you can see from the chart above, 50% of the states in the top-10 score 0 on the AWB scale. Except for Hawaii, you'll notice that the 9 of ten states with the lowest gun deaths are also in the Northeast or Upper Mid West, a topic I will return to later.

For now, let's turn to a broader question about whether or not their is evidence for gun control in general reducing gun crime.

Theory: Gun control saves lives.

Hypothesis: Other variables held constant, there should be a negative correlation between the percentage of homicides committed with guns and a state's Brady score.

Operationalizing the Test: The chart provides us with two measures of gun homicides, one is the homicide rate per 100k and the other is the overall percentage of homicides committed with firearms. Notice that in many states the rate of gun homicides is less than 60% and in some states its as low as 18%  (Hawaii) which suggests that even if all firearms were wished away, many states would still have a relatively high homicide rate in comparison to other OECD countries. Because of this relatively high non-gun homicide rate, I'm going to compare the overall Brady score both to overall homicides and to the percentage of guns used in homicides. If the Brady wackos are correct, gun control might not cut a state's homicide rate, but should at least reduce the percentage of homicides committed by firearms.

Findings: There is no link between gun control and a state's homicide rate or rate of homicides committed with firearms. None.

The Pearson's r between a state's Brady ranking and overall homicide rate is: -.02

The Pearson's r between a state's Brady score and percentage of homicides committed with firearms is: .08

The Pearson's r between a state's Brady AWB score and the percentage of guns used in crimes is .06

Conclusions: There is no evidence gun control, at least gun control as it is defined and measured by the Brady Campaign, reduces either the homicide rate or even the percentage of guns used in crimes.

So why do people in some regions of the country seem to think gun control is a good idea?


Conclusion:

The bottom line is that the evidence supports the position that gun control, at least as defined and measured by the Brady Campaign, has 0 effect on homicides committed with firearms. If measures such as AWBs mattered, we would see the evidence in state-by-state homicide rates. Instead, we see that just as many gun friendly as non-gun friendly states are among the top-10 safest and that the state with the most Brady points is 18th lowest in overall gun deaths.

No comments: