March 18, 2009

Center for American Progress Charts

I am a big fan of Edward Tufte, who is an information scientist and Yale professor. Tufte is a student of communicating (and mis-communicating) through charts. He writes about the effectiveness of charts and he critiques the effectiveness of Powerpoint. In that vein, he once testified before Congress that a factor in the launch decision leading to the Challenger disaster was a misguided reliance on Microsoft Powerpoint.

I mention Tufte because I saw a slide today from the Center for American Progress (on Der Geis's blog) that made me wish Tufte was available to critique it.

Tufte says that if we would become good communicators we must understand the techniques of willful miscommunication - for instance, to understand mass communications, the study of propagandists is informative. Tufte suggests that in visual communication nobody is better at miscommunication and misdirection than magicians, who can convince the audience through visual cues that the magician has done the impossible.

The chart I came across, from the Center for American Progress, has a bit of sleight-of-hand and I thought I'd analyze it. Here's the chart, which claims to show how different groups score on a "Progressive Index":


This chart is problematic in a few essential ways. First consider the scale of values, presented on the left and repeated below (I've rotated it to the horizontal for sake of analysis:


The scale shows a minimum value of 0 and a maximum value of 400. The two sets of crossbars are meant to communicate that the scales are discontinuous- the actual range of the displayed data is not the dramatically overbroad 0 to 400; the range of datapoints is from 160 to 247, or a range of 87 points. If charts rely on understanding specialized symbols, it's responsible to communicate the meaning to the audience rather than assume that they'll discern the meaning.

If you wanted to make a chart that communicated accurately, you'd include actual values on the axis, and you'd be explicit that the scale was not continuous - and your chart might look something more like this: (changes in red)


Tufte would say that every dot on a chart, every pixel of ink should be communicating something. If a dot is not communicating something valid then it might be an obfuscation, a magician's trick of misdirection. Look at the faint blue vertical lines connecting the data points to the names of the associated groups. Those lines are completely unnecessary; the chart could be designed with the text above and below the data points, or along the bottom axis. These lines have the effect of exaggerating the perceived visual range of the data points - in fact, the sweep of the artificially extended labels conveys that the higher datapoints are actually beyond the 0 to 400 range in the margin, when the opposite is true.

The numbers associated with the datapoints are presented in a heavier font than any other data on the chart, but they needn't be; the values should be communicated by the scale on the left. What these visually heavy datapoints, artificially displaced by the faint blue vertical lines, do accomplish is to call the eye to see the shape presented below:

Wow, there's some dramatic differences in that chart! The lower range of the visually significant shape is equivalent to a value below 0 on the scale. The upper range of the visually significant shape is well above 400, and closer to the equivalent of 500 - literally off the charts.

With a continuous axis on the left, and the datapoints shown without the exaggerating lines and labels, the chart should look something like this:

Credit (revised chart using original data values from Center for American Progress)

I must admit to a bit of bad practice of my own, in presenting the chart with a left-side scale of 0 to 400. I only used the upper bound of 400 because that's the number presented in the original chart.

When comparing the two presentations of the same data, I think that some of the techniques used in the original tend to exaggerate the differences among the groups, and the redesigned chart tends to show a fairly gentle slope and relatively modest differences between groups.



Why would anybody go to this much effort other than to advance their agenda or business case; in this situation, possibly both. The nobly-named (or perhaps Orwellian-named) Center For American Progress is a DC think tank, and like all think tanks it has an agenda, an audience, and a business goal.

1 comments:

Pittsburgh Conservative said...

That is really interesting. It's my understanding that the centerpiece of Colin Powell's WMD presentation, the reconnaissance photos of Iraq's "mobile germ warfare labs", were specifically selected because they were evocative of the aerial shots of the missiles in Cuba and would tap into Americans' memory of the threat posed by the Soviets back in the 60's.

I participated in a strategy session where we dissected a set of financial numbers. One of the financial analysts was a whiz with Excel, and as we sat there he cranked out 7 or 8 different graphs off the same set of data, each one leading to a different conclusion. It was kind of scary. Our response was to downplay the data since it was so clearly malleable to all of the particular points of view on our issue.

Post a Comment

Comments and Feedback? Love that stuff. Please leave your thoughts in the box below--