how not to graph trends over time

Talking Points Memo has a slide showing President Obama’s approval rating 2011-2012:

Putting the theorized causes of opinion shift between the two lines is pretty, but misleading since at least to me it implies that these events caused the difference between the two lines, not the change in the overall rating.

More misleading–though in a direction that undermines the graph’s thesis–is the manipulation of the y-axis so it ranges only from 40 to 55. The graph’s claim (that the President’s approval is “Nearly back to where it started”) is sort of true, but truncating the y axis makes it look false, since the Gallup line has fully 33% of the y axis to go before matching its peak!

Author: andrewperrin

University of North Carolina, Chapel Hill

4 thoughts on “how not to graph trends over time”

  1. The graph is stupid for the first reason you cite. But in principle there’s nothing wrong with restricting the axes to the range of the data as opposed to, e.g., requiring it go from zero to 100 percent. How you do it depends on what you think counts as meaningful variation in the data.


  2. And in thinking about the “volume” of variation, wouldn’t we rather need to see margins of error for each of the plotted lines? Seems like for much of the spring and early summer the two lines could be essentially the same.


Leave a Reply

Please log in using one of these methods to post your comment: Logo

You are commenting using your account. Log Out /  Change )

Google+ photo

You are commenting using your Google+ account. Log Out /  Change )

Twitter picture

You are commenting using your Twitter account. Log Out /  Change )

Facebook photo

You are commenting using your Facebook account. Log Out /  Change )


Connecting to %s

This site uses Akismet to reduce spam. Learn how your comment data is processed.