Skip to content
 

An Economist’s Guide to Visualizing Data

Stephen Jenkins wrote:

I was thinking that you and your blog readers might be interested in “An Economist’s Guide to Visualizing Data” by Jonathan Schwabish, in the most recent Journal of Economic Perspectives (which is the American Economic Association’s main “outreach” journal in some ways).

I replied:

Ooh, I hate this so much! This seems to represent a horrible example of economists not recognizing that outsiders can help them. We do much much better in political science.

To which Jenkins wrote:

Ha! I guessed as much — hence sent it. And I’ll now admit I was surprised that JEP took the piece without getting Schwabisch to widen his reference points.

To elaborate a bit: I agree with Schwabish’s general advice (“show the data,” “reduce the clutter,” and “integrate the text and the graph”). But then he illustrates with 8 before-and-after stories in which he shows an existing graph and then gives his improvements. My problem is that I don’t like most of his “after” pictures!

In just about every case, Swabish’s advice is reasonable and his graphs improve on the originals. But I just don’t think his versions represent best practice. And, in an influential journal, you’d like to demonstrate best practice.

Here’s an example. Before:

Screen Shot 2014-03-11 at 3.43.54 PM

And after:

1b

The small scale and blurriness are my fault; something happened in my cut-and-paste, so please don’t blame Schwabish for that.

In any case, yes, the second display is better, but in addition I’d label the y-axes and, most obviously, I’d get rid of those heavy gray horizontal lines. I’d also put tick marks on the x-axes, especially for the two graphs in the upper row, also he seems to have forgotten to put y-axes on the two graphs on the right. As it is, the four subgraphs seem to merge into each other. You really need some visual cues to separate them.

My other problem with this paper is its lack of ambition. In each case, an existing graph is redrawn with only slight changes. But what is really needed in economics, I think, is a larger sense of the importance of graphical discovery. The excitement of visualization is not conveyed in this article at all. Rather it all seems like a boring application of certain principles of graphics design.

I think the article is well-intentioned and may do some good. And the topic is important. But I agree with Jenkins: Schwabish should’ve taken a broader perspective. This was a bit of a wasted opportunity.

22 Comments

  1. K? O'Rourke says:

    Kipling’s quip “In the land of the blind, the one-eyed man is king” – should not be taken as a complete dismissal but rather mostly regret of lost opportunities?

  2. Nate says:

    I tend to go to the Economist’s graph blog a lot, just to peruse (http://www.economist.com/blogs/graphicdetail) for a baseline of good graphing technique. They do an excellent job with graphs conveying lots of information in a simple and clear way. Plus, I like the styles. For a lot of the web graphics, they remove the blue background you see in the print magazine, so the graph looks a little crisper. I feel like this page should be “The Economist”s Guide to Visualizing Data”

  3. MP says:

    I think he is excluding the y-axes on the right because he believes them to be redundant. But in another chart (fig 6A), he splits a line chart with 4 lines into 4 line charts, all showing the same thing but with one line emphasized bold. That seems far more redundant to me.

  4. Hadley says:

    A pale gray background with white grid lines would help with many of those problems ;)

    • Andrew says:

      Yup! Another issue in graphics, which is not discussed by Schwabish or for that matter by me in most of my writings, is that if you make a lot of graphs, and they all look the same, they’ll get boring fast. Red State Blue State has hundreds of graphs but they start to blur in the mind, as they’re just the same set of maps, scatterplots, and lineplots over and over again. One of the great virtues of R is that it can be used to make graphs in all sorts of different styles, but one problem with many users, even experienced users such as myself, is that we tend to use the same templates repeatedly, giving our work a bit of a boring and uniform look.

      • Anonymous says:

        Do you mean changing the style, like the colors used (red state, blue state, green state?), or changing something like lines to bars? I can imagine changing either could create some confusion and make comparing different plots difficult. I do understand having plots that look a lot a like can get boring, but depending on what is changed I could imagine the changes having negative consequences.

    • We should get Andrew to do a rundown of why he doesn’t like the default ggplot2 output. I believe the white-on-gray is one of the things he dislikes the most, as well as how close together the faceted graphs are, where the titles of the facets go, how many times the axes are repeated, etc.

      I disagree with Andrew on many of these points. I’m also not convinced there is a “right” answer.

      Is there any research on the tradeoffs in plotting? Usability experiments are notoriously tricky to formulate, but I’ve seen them done well in the past, at least for spoken-language understanding interfaces.

  5. Stuart Horwitz says:

    Andrew Gelman, in your blog on “The Ethics of Lying…” you mention a Lutheria Gelman who I believe is the grand daughter of Ethel Holzman Gelman and Max Gelman and the daughter of Moses Gelman. Ethel Gelman is my great great grandmother from her first marriage to Moses Horowitz-Zuckerstein. I am making a family tree of Ethel’s descendants and was wondering if you are her great great great grandson. Sorry to leave you this message here, but I did not know how to contact you otherwise. Best regards, Stuart Horwitz

    • Andrew says:

      Stuart:

      Aunt Lucy was indeed Grandpa Moses’s daughter, but Moses’s parents were not named Ethel and Max, so I’m guessing this is a different family.

      • Phil says:

        Andrew, I don’t follow your logic. The following statements could all be true:
        1. Lutheria=Lucy is Moses’ daughter.
        2. Lucy’s grandparents were Ethel and Max.
        3. Moses’ parents were not Ethel and Max.

        Tell me you didn’t fall victim to the Smurfette fallacy on this!

      • Stuart Horwitz says:

        The Moses Gelman in my family was married to Ida Korchin in Milwaukee Wisconsin on the 24th of Dec 1910. They moved to Brooklyn New York and lived on President St in 1930. Moses and Ida had four children: Bertha (born in Wisconsin), Woodson (born in Illinois), Lutheria (born in Illinois) and Robert (born in New York). Moses Gelman was born in 1890 in Russia. Ida Korchin was born in 1891. Does any of this correspond with your family? What were the names of your grandpa Moses’s parents?

        • Andrew says:

          Stuart:

          Yes, that’s right. Except that Uncle Woody’s name was Woodrow, not Woodson, otherwise I think all these details are correct. I don’t remember what were the names of Moses’s parents. Maybe they were indeed Ethel and Max?? I recall the mother’s name as maybe being Jenny, but maybe it’s my memory that’s wrong here.

          • Stuart Horwitz says:

            I just looked again at the 1930 census and I was mistaken, the name was indeed Woodrow and not Woodson. Thanks for clarifying that. Do you [or any of your relatives] have any pictures of Moses, Ida or their kids? I’d like to add them to my file. Also do you have the names of their kids children so I can add them to the family tree? I have a picture of Moses’s mother [if this is really the right Moses]. If you are on Facebook, there is a page for Descendants of Abraham and Rebecca Horwitz, where you can get the picture of Ethel Gelman. [Abraham was her son from her first marriage to Moses Horwitz. They got divorced and she remarried Max Gelman.] It is a closed group but I can invite you to join [just friend me at "Stuart Horwitz"]. Otherwise, just send me an email at stugongju@live.be

  6. Steve Sailer says:

    Creating a good graph always takes me two or three times longer than I thought it would.

  7. […] Scientists losing data at a rapid rate – An Economist’s Guide to Visualizing Data (og Gelmans indvendinger) – Seven deadly sins of contemporary quantitative political analysis – Programmatically download […]

Leave a Reply