Thursday, April 8, 2010

Zeroing the Cash for Clunkers graph

Yesterday Brad Delong posted the government's assessment of the Cash for Clunkers program. No surprise - Obama's Council of Economic Advisers is calling the program a success. They were nice enough to supply this helpful graph:

But as someone in Delong's comment section pointed out, the graph is not zeroed. It presents 8 million cars where one expect to find zero. This makes the post-program dip look more severe, but more importantly, it raises the ratio of the successful boom to a misleadingly high level.

Here's the rest of the graph:


There's a word for this manner of representing data: misleading.

3 comments:

  1. The way data is presented makes a huge difference in perception, which is why it is SO important that you read multiple sources/studies on a subject before making a judgement. The funny thing, is that the graph you show to counter Obama's "misleading" depiction of the data is also misleading. Notice how the graph is in a heavy "portrait" orientation (height is greater than width). The correct graph should be square, because there are fifteen units for the amount of cars (y axis), measured over a fifteen month period (x axis). The portrait configuration "squishes" the graph inward, making the y axis look larger than it actually is.

    ReplyDelete
  2. It was a simple task of using Paint to fill in the missing units. Please keep in mind that the X axis is four month increments and the Y axis is millions of cars - I don't see any logical way to make one equal to the other.

    ReplyDelete
  3. Perhaps this is a poor reflection of my social life, but I was lying in bed this morning and I realized you make a very good point.

    A long, thin graph would accentuate any little bumps in the data. So keeping with my Paint solution, here's the same graph in a square shape:

    http://img25.imageshack.us/img25/6569/squaregraph.jpg

    ReplyDelete