Bond yield makeover
Green panels

Popping the bubble, so to speak

Matt H., who authored the previous post, and Charles G. both pointed to a great example of how people like readers here can make a difference.   A bad chart got made over!

The Financial Times published a chart from a JP Morgan report, using ... you guessed it ... bubbles to illustrate the deep plunge in market capitalization of many banks.



Some readers at the blog were none too happy with the choice of bubble charts.   Among other things, the designer made the common mistake of plotting proportional diameters rather than proportional areas.  This is clear from looking at JP Morgan's bubble.

This chart exposes the weakness of bubble charts well.  Look at the top row of bubbles.  Most of them look so similar it is impossible to know, without spending much time studying the circles, which bank was hurt more.

Eventually, a bar chart was produced.  Felix Salmon linked to it.  I am not sure how the banks were ordered in this chart.  It isn't one of the two obvious dimensions, nor is it in alphabetical order.


In fact, neither the bubble nor the bar chart works well for this case.  What we need is the Case-Schiller style asset-bubble life cycle chart.  In order to interpret these changes in market caps, we need to know how big was the bubble, and then how steep was the consequent decline.  Take a look at our discussion of real estate bubble charts here.

Reference: "Bank capitalization chart of the day", Felix Salmon at and "Bank picture du jour", FT Alphaville, Jan 21 2008.


Feed You can follow this conversation by subscribing to the comment feed for this post.


Ordering is by percent decrease in market capitalisation. This is probably all that really needs to be charted, the actual capitalisation seems to be irrelevant, all we want to show is that there has been a massive decline. Maybe add the 2007 market cap to the label.


I did the bar chart, bit of a rush job at midnight so forgive its failings. As Ken says the order is decline in market cap, best to worst.

Ken's right that % decline might be all that is needed to chart, I quite like however seeing just how much value has been destroyed.

The chart really needs better labelling of the bars, I think some people think dark blue was the old market cap, when in fact it is dark blue plus light blue (and so wouldn't work if any of the banks had increased their market cap).

Jan Schultink

I agree that these bubbles are not the best visualization of market caps.

BUT maybe it is not necessary to be that scientific in this case?

I would "fit" all the blue bubbles in one big bubble (before situation) on the left and fit all the green ones in a "big/small" green bubble on the right (after situation).

To complete the effect stick a little needle in the left big blue bubble.

Not scientific, but gets the point across. PSSSS.


Matt - You can also consider a "bumps" chart, basically a line chart connecting the value in Q2 07 and the value in 2009. The slopes of the lines then give the scale of the decline.

The big issue of this chart remains the fact that different banks suffered declines at different times, and enjoyed run-ups of different scales and the original data set did not address these important factors.


I'm not sure that matters so much if we are looking at the banking sector as a whole. Lots of investors are interested in returns over the same time period, and the data does show that. The problem of the data in looking at the banking sector as a whole is it's only a subset of banks - you can't of course show the ones which have been nationalised (well you could...), but also missing are any Japanese banks, for instance.

Jerry Hughes

I've been reading with interest, and somewhat defensively, these discussions of bubble charts. Perhaps this isn't the overall stance of the authors, but the feeling about these tools seems overwhelmingly negative. While having only recently discovered InfoVis as a discipline, I have long been concerned about accurate presentation of data in business analyses and have been a great fan of bubble charts. So I've thought hard about why I love using them and so many in this discipline seem to hate them.

Maybe it's obvious, but I haven't seen it stated the following way. Bubbles are terrible, and almost always inappropriate, to use to display the primary dimensions of an analysis. This is for all the reasons commonly cited. However they can be terrific for displaying secondary categorization variables.

The best example of this is a scatterplot demonstrating some relationship between the x and y. Then use the bubbles to demonstrate some size metric to help show whether the relationship holds as well for big or small entities of analysis. The bubbles can also be color coded to show an additional categorization variable.

Another example is to use the bubble size to help prioritize decisions coming from diagnostic-type analyses. Think of the classic BCG growth-vs-profitability matrices. The core analysis helps direct attention to business units that are doing well or not so well; adding the bubble size helps prioritize attention and action.

I've posted some examples at I would love to see reactions here or there. Maybe bubble charts are more prone to malpractice than others, but I hope there's a way to develop guidelines to reduce user error rather than abandon this useful tool. Besides, with Google putting its weight behind the motion bubble charts (Gapminder), these seem unlikely to go away. Better to get ahead of the trend with good standards for their use.


Jerry - thanks for providing the other point of view. There is one use of bubble charts that I quite like, which is in charts that are "conceptual" where the point is not to have readers estimate ratios of areas.

I particularly don't like overlapped bubbles as much important information is typically obscured.

If you search for Gapminder, you'll find some old commentary on them. I love the interactivity and dynamic graphing but less impressed with certain choices of charts.


Kaiser, a bit of clarification is in order. You note:

"Among other things, the designer made the common mistake of plotting proportional diameters rather than proportional areas."

But haven't you defended this practice? Here's an earlier quote from you:

"some readers will respond to the areas while others will respond to radii ZBicyclist, for example, belongs to the first group while in this case, I find myself siding with the latter. When the bubbles/rings overlap, it is difficult to assess areas." This is from:

There's a nuance in your opinion I'm not getting. Why not "the area, the whole area, and nothing but the area, so help me Tufte"?

The comments to this entry are closed.