Saturday, May 12, 2018

Is Even One Mass Killing By Guns Proof That Gun Control Does Not Work? Day Two of Fund-Raising Week.


Why would the shooting deaths of seven people (six of them apparently killed by the seventh in what is so quaintly called domestic violence) in Australia (far away) end up on the pages of  the New York Times?  And why were those same deaths trending on Facebook in the US?

The answer, my friends, lies in Australian history about gun control and what that stricter gun control history could teach Americans:

The deaths represent the worst mass shooting in Australia since 1996, when a gunman killed 35 people in Port Arthur, Tasmania. That event was the catalyst for a significant strengthening of the country’s gun laws an ambitious gun buyback program.

American proponents of gun control, including former President Barack Obama, often point to Australia’s strict regulations and few mass shootings as a guide to limiting such events in the United States. But in the hours after the shooting in Osmington, those who oppose such strict regulations pointed to the tragedy as proof of gun control’s limits — prompting many Australians to argue against that conclusion.

It's not just the limits of gun control that was pointed out in various Facebook posts.  Rather, many used this awful killing in Australia as proof that gun control doesn't work at all.

I get that when people argue politics they often use dirty weapons and illogical tricks and do not care.  Finding even one mass killing as proof that gun control does not work would be such an illogical trick, because the correct comparison would be to juxtapose the Australian and US historical statistics about mass killings, the availability of guns and so on, and then use those properly controlled comparisons to see how many excess mass killings by guns the US might have because of its much laxer gun policies*.

But it's not at all uncommon to find people, in general,  using individual anecdotes to try to prove something about statistical averages.  All an individual anecdote can prove, if true, is that at least one such event happened.  It tells us nothing about average propensities. 

Similarly, many people simply assume that their circle of friends, acquaintances and relatives are representative of all Americans, and that this allows them to judge the validity of various statistical averages which pertain to the whole country.  But very few of us have a random sample from the whole country among our acquaintances.

Many also ignore the "all other things constant**" part of various comparisons in studies and in how to interpret various news items.   Here's one example of that (scroll down), but it also applies much more widely.

For instance, when progressives and liberals want to see if people from different demographic groups are treated equally fairly in, say, jobs, the people to be compared should be similar in other relevant ways except their demographic group memberships.  That could include education and work experience, as one example, unless unfair earlier treatment has caused those to differ between demographic groups.

And this post (go down to 4.) discusses a few other fairly common (but interesting) interpretation problems.

------

*  Note the question I asked there.  If our question is about general gun deaths then the answer may be more complicated, because most gun deaths are not mass killings but individual murders or suicides.

** The ceteris paribus assumption of economics, say, where our goal is to get as close as possible to analyzing a question in such a way that the only variable allowed to differ is the one we wish to analyze.  We can literally hold other variables constant in laboratory studies and some audit studies, but in most observational studies the control is achieved (if only partially) through statistical methods of analysis.