Thursday, January 10, 2013

Critical Thinking Thursday: Statistics Mean Nothing

Today's Critical Thinking Thursday post is a bit late as I was busy learning an important lesson:  Always check the oven before pre-heating it.  I had left something in it previously and completely forgotten about it.  I didn't notice this until turning on the oven to pre-heat for a pizza....and copious amounts of smoke began billowing from the oven.  Oops.  We retreated to Buffalo Wild Wings until the smoke cleared -- hence the late post.

Anyway, today I want to talk about this graphic:

I saw this kicking around on Facebook and it's a beautiful example of the way you can twist statistics around to say basically anything you want.  So, let's go step-by-step and see how you can critically examine this, shall we?

  • First off, let's check that the figures are even accurate.  Since the graphic doesn't site a source for any of these numbers, we'll have to check them manually.  When searching, you'll want to find a website that doesn't have its own agenda (for obvious reasons), so we'll want something with some objective credibility.  
  • Let's take a look at that 195,000 figure first.  First off, the data is 10 years old.  Second, the actual causes of death are "failure to rescue, bed sores, postoperative sepsis, and postoperative pulmonary embolism".  So now we have that figure explained a bit more clearly.  
  • Now, the assault rifle figure.  It took some digging to find the primary source on this one, but it looks like it's from the FBI. (Interestingly, based on the FBI data, the figure quoted for knives is actually very low -- the figure quoted here is 1,694 for stabbing deaths).  
  • The 12,000 drunk driver fatalities appears to be from this site here, which offers an average (not the exact figure).  
OK.  Now that we've found the source of each figure, let's look at a few interesting things about how the data is compiled and represented, shall we?
  • By specifically choosing assault rifles, the graphic down plays the severity of gun violence in general.  According to the CDC, firearms account for 31,347 deaths.  Of these, according to the FBI, 12,664 were homicides.  The others could range from accidents (children handling guns, hunting accidents) to suicides.  
  • This means that gun homicide (in general) is actually more dangerous than drunk driving.  Someone with a different political stance could make a different graphic using exactly the same sources and come to that conclusion -- more gun murders happen annually than drunk driving deaths.  
  • Incidentally, according to the CDC, the total number of vehicle fatalities (not just drunk drivers) is 34,485, which is still higher than the total number of firearm fatalities.  Therefore, someone could, again, take the same data and make a new graphic showing that cars are more dangerous than guns (but the margin is much narrower).  
The point of all this isn't to take a side one way or the other about gun control.  The point is to show that this graphic -- which seems very straightforward -- presents its figures in a way that supports the point it wants to make.  This is the dangerous thing about statistics:  On their own, they don't actually mean anything.  The numbers have to be interpreted.  As you can see, different people can take the exact same data and extrapolate completely different interpretations from it.  

So the next time you're faced with a graphic or statistic, I invite you to take a few extra minutes to think critically about it before sharing.  Ask yourself:  Where are they getting these numbers?  Are they accurate?  Is the source reputable?  If you look at the data yourself, do you get a different interpretation?  How many other ways can those numbers be interpreted? 

No comments:

Post a Comment