Home > Applying Political Science, blogs > More Adventures in Bad Chart Making

More Adventures in Bad Chart Making

I still think Andrew Gelman does this kind of thing better, but it’s hard not to comment on the following chart from the Washington Post’s Sean Sullivan:

oimgThe larger point that Sean is trying to make is important: There has been a significant decline in the production of federal legislation. That said, this chart is horrible. Among the many problems:

  1. It is incredibly problematic to include the current Congress (113th) in the chart. Doing so exaggerates the trend in an unfair way. Every other Congress has been completed. We haven’t even made it to the end of the first session of the 113th Congress.
  2. The chronology is reversed. Reading left to right, we travel backward in time from the current Congress to the 105th Congress (1997-98). Time series should go forward.
  3. Only Congress nerds (and I am happy to be one) know the dates for each Congress. Everyone else has to rely on tables like this one. If you are going to make a time series, use commonly known labels for the time periods.
  4. Why separate the two sessions of each Congress instead of stacking them? The overall trend would still be there (if anything it would appear stronger), and people would still be able to see that most of the action has come in the second session of each Congress.
  5. Don’t stagger the axis tick labels. If the labels won’t fit horizontally, rotate them.
  6. Why start the y-axis at 20 instead of zero? Is the difference that large? Starting the axis at 20 because that’s what Excel defaulted to is just lazy.
  7. Label your axes in ways that people will understand. What does “Public laws enacted” mean? Why not just say “number of laws?”

Here’s another stab at the chart using the same data:


  1. ezra abrams
    December 16, 2013 at 7:29 pm

    Y axis
    the *only* real way to evaluate a chart is to first have some experts decide what the point(s) are, and then show the chart to members of the target audience, to see if and how long it takes them to figure out he point
    having said that, my opinion is that the Y axis has to start at zero is a shibboleth, without any factual basis
    the point of a chart is to make a point; you may or may not need an y axis 0 start to do it.

    PS: naomi robbins is best place to start, not the overpraised tufte
    PPS: widely scattered thru the literature are people who have tried the test I suggest; many of them are in public health, trying to figure out what graphics help people make good choices in healthcare

    • December 16, 2013 at 8:51 pm

      To be sure there is no a priori reason why the y-axis ought to start at zero. In many cases it is legitimate to start the axis at some other value. My complaint in this case wis simply that, given the data’s range, the difference between 20 and 0 is so small that it seems silly to start the y-axis at 20 and not 0.

      My personal preference for good chart making is Jessica Hagy (http://thisisindexed.com).

  1. No trackbacks yet.

Leave a Reply

Fill in your details below or click an icon to log in:

WordPress.com Logo

You are commenting using your WordPress.com 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

%d bloggers like this: