Illusory disparity

The WSJ published a chart with the cheeky title of "Rich Get Richer" (reminiscent of the Economist).  The underlying data concerned one-, three- and ten-year returns for the buyout fund category.  For each return class, the overall mean and the means for the top and bottom 25% funds were depicted.

I won't go into the relevance of the title as I simply could not figure out how it connected with the data.  The following shows the original chart side by side with the junkart version.



Improvements include:

• Lines show the comparisons with a minimum of fuss compared with colored bars
• The overall mean return is placed in the middle of each line segment where it belongs, instead of being the first column
• The axis label, "annualized return", tells readers what is the performance measure
• Adding the word "funds" to "top quartile" and "bottom quartile" removes the possible confusion that those represent individual returns of the funds ranked at 25th and 75th percentiles, rather than the average returns of the bottom 25% and top 25% of funds
• The linear construct paints the correct picture that individual fund returns fall into a continuum

(Thanks to my students for some of these points.)

Reference: Wall Street Journal, Mar 3-4 2007.

Comments

[Tony Rose]

Great post! I think that far too many people go for what they think appears better (colors and bars) vs. what is more meaningful. Your junkart version has much more value than their chart. Keep up the great work!

Tony
www.supportanalytics.com/blog

[derek]

On a similar subject, this graph is inoffensive enough, but such a a waste to present only an average when the full set of observations would fit in the same space at the same resolution with zero loss of clarity.

Unfortunately it's one of those cases where you can't just extract the info from the original to make the "junk art" version. I've emailed the creators to see if they're willing to send me the background data so I can pimp it up a bit for them.

[Jon Peltier]

Derek -

That chart needs a box and whisker for previous cycles and an added point for the current cycle.

[derek]

Jon, that would be an improvement in the representation of a distribution, but I was hoping to be able to just show the whole distribution.

Sadly, the creators of the chart never responded to my email request for the data it was based on (data all public and the references named, but I lack time to collect it from the sources and face the heartache of exporting PDF tables etc.)

That's entirely their prerogative, I agree. But it's still a shame.