Bad Graphs

While I’m picking on Brad DeLong, I want to criticize him for something that has nothing to do with economics–the very bad graph that he uses to depict current employment figures.

I think this is a classic example of “how to mislead with graphs.”  The decline in employment  as depicted here looks massive, running from almost the very top of the graph down to the very bottom.  The actual decline is approximately 8.2%,* but occupies over 90% of the vertical space of the chart, implicitly suggesting a decline about 10 times greater than reality.

To be fair, it looks like the real blame for this graph is the St. Louis Fed, but anyone who reproduces it shares in the blame. That’s not to say either the Fed or DeLong are purposely trying to mislead people, but the alternative, simple carelessness, isn’t highly respected, either. Here are some alternative presentations, using the same data.

If, instead of running the vertical axis from just the peak to the trough, we run it for the whole range of (mathematical) possibility, we get a representation that depicts the current downturn in employment as still a distinctly noticeable decline, but not a massive fall-off.

Or we could run the range from 50% to 70%, and get the following image that visually depicts a severe decline, but without suggesting wholesale catastrophe.

That’s the type of graph I recommend to my students. Obviously there’s no objectively correct choice of range to depict, but the choice of the graphmaker matters, and it’s easy to use the figure to try to manipulate the takeaway message. Here are two graphs showing the national debt in relation to the GDP. I like to show these to my students to get them to thinking about how the starting point influences the message. It shouldn’t be too hard to tell which of these is the liberal presentation of the data and which is the conservative presentation.


__________________________________________________________
*From a high of 63.4% to a low of 58.2%.

About J@m3z Aitch

J@m3z Aitch is a two-bit college professor who'd rather be canoeing.
This entry was posted in Politics in General. Bookmark the permalink.

4 Responses to Bad Graphs

  1. D. C. Sessions says:

    For that matter, civilian employment/population should be charted since Independence. For quite a bit of that time, it ran considerably less than 50%.

    We really have nothing to complain about today.

  2. Scott Hanley says:

    We really have nothing to complain about today.

    You’ll never stop me! Do you hear! NEVER!

    James, do you know what the historic range is for civilian employment? For a statistic that never reaches either 0 or 100, the sense of gradualism in the second graph would be equally misleading, especially if real-world effects are sensitive to small changes in that statistic. As another example, economists tend to make big deal over half-percentage point differences in GDP growth, so a 1-100 scale there would be a dreadful presentation of data. There’s no one “correct” presentation, either, which is also something students should learn.

  3. AMW says:

    As another example, economists tend to make big deal over half-percentage point differences in GDP growth, so a 1-100 scale there would be a dreadful presentation of data.

    Of course, the difference there is that the growth rate affects GDP exponentially over time.

  4. James Hanley says:

    Oh, I didn’t intend the 0-100 graph as a serious substitute, just a demonstration that it’s easy to make a graph that is misleading.

    I don’t know the historic range. That was something I thought of, as being relevant in helping determine what your parameters should be. Or absent that, perhaps use the depths of the Great Depression as the baseline.

    I agree that there’s no one correct presentation, which is what makes this process tricky. People should choose an approach that highlights their message–they just shouldn’t choose an approach that overhighlights it. It’s the great Goldilocks idea, not too much, not too little, just right. And like pornography, we can’t define it, but sometimes we know it when we see it.

Comments are closed.