How to kill a good report by using Excel

My brain is hurting this afternoon, because someone decided to use a stacked bar chart for two figures in the brand-new Georgetown Center for Education and Workforce report on employment and salary data for different undergraduate degrees, written by Anthony Carnevale and Ban Cheah. Here is the chart on unemployment data for different clusters of disciplines:

Stacked bar chart -- note how this chart stacks percentages for different populations.

Stacked bar chart — note how this chart stacks percentages for different populations.

Stacked bar charts only make sense if adding the quantities make sense–that is, if each set of bars refers to the same population and also if sums are cumulative in some sense. So, for example, with a cohort, bar charts would make sense if you are piling up bars representing the proportion who finish high school, those who in addition finish two years of college, and then those who also finish an undergraduate degree. In this case, the bar chart is attempting to combine quantities for different populations (new graduates, more experienced graduates, and masters-degree recipients), and where there is no cumulative meaning to the sum of the unemployment percentages. While I don’t think this data requires a chart–something this simple really can be displayed as a table without pretty pictures–those who go for bar charts should use unstacked bars, such as the following:

 

A reconfigured bar chart, to eliminate the stacked-bar problems in the original chart. This is still not a great way to display the data.

A reconfigured bar chart, to eliminate the stacked-bar problems in the original chart. This is still not a great way to display the data.

Because of an earlier, erroneous version of the file (even more erroneous than the stacked-bar chart displayed above), I know that the stacked bar chart was created using Excel or another spreadsheet program, and someone evidently chose the pretty stacked bar icon. Ouch. This is a good report, and because it was clearly designed to be accessible for the general public, it really is a shame that the figures were formatted to be confusing rather than edifying.

If you enjoyed this post, please consider subscribing to the RSS feed to have future articles delivered to your feed reader.

2 responses to “How to kill a good report by using Excel”

  1. CCPhysicist

    Brilliant analysis, but you forgot to credit the Lumina Foundation and the Bill and Melinda Gates Foundation for supporting this figure (made with Gates’ software?) and the arguably worse one for earnings on page 8, nor mention that this doesn’t increase my respect for Georgetown!

    Notice that your version also shows information that is hidden in the report’s version: There is one case where experience trumps post-grad education and a couple where the difference might not cover the additional cost plus opportunity cost of the post-grad degree.

    Even your version suffers from the limitations of the software you used, since the best way to show that particular comparison is to have the bars slightly overlap as if stacked in 3D.

    I also don’t like that the display is alphabetical rather than grouping the majors into the broader categories used elsewhere (albeit inconsistently) in the report or ranked by the fresh-out-of-school number so they are easier to compare.

    PS – Your “industrial arts” columns are incorrectly coded.