Lies, Damned Lies, and Infographics
I’m indebted to Chandler Howell of the Not Bad for a Cubicle blog for drawing my attention to this graphic which has appeared without credit in many blogs:
Then ask yourself: what does this picture tell you, and does it tell you the truth?
It’s a great graphic, but which statistics are actually important here? Do they receive the correct prominence? Are 307,655 deaths from cancer really that much more striking than 316,968 deaths from heart disease? And what about the 57 men who fell of a cliff or the 40 struck by lightning?
First of all note that the graphic uses absolute predicted figures. Will there really be exactly 21,289 deaths among American men from colon cancer in 2008? Extremely unlikely, as the year hasn’t ended yet. These are predicted figures with an error bound, so it is giving the illusion of precision.
Do the figures add up? The total here is 960,880. The US male population is about 148 million. There must be some deaths not recorded here. Indeed the CDC gives the male+female figure for 2005 as 2,448,017, so it looks like there are about 200,000 we haven’t heard about.
And what use are the figures? We can see (perhaps) that suicide is more common than homicide – something we won’t get from reading the papers. But unless we are male and know the male population of the US we can’t put these figures to a lot of use except if we are funding a federal program. Estimated percentages, such as those from the Center for Disease Control, give a much more realistic picture if we are trying to figure out what our personal risks are.
Changing to percentages isn’t enough, however. Ultimately the mortality rate for US males is 100%. It makes a great deal of difference whether that death is at age 2 months, 23 years, or 123 years.
Whether considering public policy or personal risk, timing is important.
I’m much more worried about something that may kill me tomorrow than something that may kill me when I am 100 years old.
It’s a clever graphic. I wish I had the skill to create something like it. But ultimately it fails because it doesn’t tell me anything useful.