Lies, lies and statistics : A lesson in reading charts.
Wednesday, July 21st, 2010i was reading the EPA's National Inventory Report for Greenhouse Gas Emissions the other day. Some may say that I should get out more but these reports are real pageturners.
Anyway, rather that talk about the report, I'm just going to mention one particular chart that I saw which piqued my interest. On pg 39, we have Figure 2.4 Vehicle Numbers and Population 1990-2008.
What we have is the red bar chart showing the increase in car ownership and then the line showing the increase in population over the same period. So far, so simple.
Now at a first reading, the eye sees the level of vehicle ownership increasing at roughly the same rate as the increase in population. Good on you, Ireland, nothing to be reprimanded there.
The tricky bit though is that the population line uses a different scale, the scale on the right, the one that starts at 3 million and goes up to 4.5 million. Now you can argue that different scale is needed to make some sense of the chart, but the scale chosen is curious in that it makes the two trends, vehicles and population seem pretty much equal.
Now if we really wanted to compare trends then we should use a comparable scale, for example the % increase over a given time, i.e. if vehicle ownership goes up by 10% in the same time as population ownership goes up by 10% then we can safely say that nothing strange is going on.
So lets make a chart doing just that
Now that chart seems to tell a whole different story. Between 1990 and 2008, population increased by roughly 25% while vehicle ownership increased by a relatively whopping 140%.
So, caveat emptor, you've just got to watch out for what someone is trying to tell you in a nice chart.
Remember, the numbers never lie, once you have access to the real numbers. Its the use of the numbers that brings the spin.