Thursday, 22 July 2010

How to lie with a Graph

One of the irritations of the Andrew Sullivan's Daily Dish, aside from the new post every 10 minutes clogging up the Reader, is the fact that comments are not allowed. In this instance "chart of the day" demonstrates how to lie with a graph, and I would love to make this comment over there.


This would look VERY different on a logarithmic scale as it shows incomes of an order of magnitude's difference. It starts at the bottom of a boom i n1979 and it stops at the top of a boom in 2007. I wonder what happened between then and now?

Finally of course, who cares how many millions have the top 1%? They've always been mega wealthy. The top 1% is, by its nature self-selecting. The people who were in the top 1% in 1979 may have done extremely badly, and are now only in the top 20% and new players have joined the ultra rich. This is distorted by selection bias not prevalent in larger cohorts.

Comparing a top 1% with the middle 60% is not therefore comparing like with like. A graph showing each of the deciles' incomes over that period would be informative. What would that show? The answer is that you cannot see on this graph, because the scale is of insufficient resolution to answer it, but it appears that the bottom 80% are all going roughly the right way at roughly the same rate. But as charts as dishonest as this one are all you ever see from leftie blogs, then I suspect the 10 lines representing 10 income deciles do not show the picture the inequality-obsessed left want you to see.



No comments:

There was an error in this gadget