Graphical inequity ruins the chart
If Clinton and Trump go to dinner, do they sit face to face, or side by side?

Depicting imbalance, straying from the standard chart

My friend Tonny M. sent me a tip to two pretty nice charts depicting the state of U.S. healthcare spending (link).

The first shows U.S. as an outlier:


This chart is a replica of the Lane Kenworthy chart, with some added details, that I have praised here before. This chart remains one of the most impactful charts I have seen. The added time-series details allow us to see a divergence from about 1980.


The second chart shows the inequity of healthcare spending among Americans. The top 10% spenders consume about 6.5 times as much as the average while the bottom 16% do not spend anything at all.


This chart form is standard for depicting imbalance in scientific publications. But the general public finds this chart difficult to interpret, mostly because both axes operate on a cumulative scale. Further, encoding inequity in the bend of the curve is not particularly intuitive.

So I tried out some other possibilities. Both alternatives are based on incremental, not cumulative, metrics. I take the spend of the individual ten groups (deciles) and work with those dollars. Also, I provide a reference point, which is the level of spend of each decile if the spend were to be distributed evenly among all ten groups.

The first alternative depicts the "excess" or "deficient" spend as column segments. Redo_healthcarespend1

The second alternative shows the level of excess or deficient spending as slopes of lines. I am aiming for a bit more drama here.


Now, the interpretation of this chart is not simple. Since illness is not evenly spread out within the population, this distribution might just be the normal state of affairs. Nevertheless, this pattern can also result from the top spenders purchasing very expensive experimental treatments with little chance of success, for example.



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

Fabio Machado

Very interesting, I didn't know so much of the health expenditure burden lay on the top 1%. I'm guessing it also has to do with them not carrying insurance and paying out of pocket for their healthcare?


@Fabio I would also guess that it has a lot to do with the fact that they go and get care, because they can afford to.

A lot of people cannot afford the care that they should be getting, both in terms of preventative care and regular checkups, and in terms of dealing with acute injury or illness.

Would also have to look at what portion of the spend is for elective or cosmetic purposes.

Tobias Gerken

I think this is largly an artifact of extremely high costs for some complicated treatments and lack of cost controls in the system. Last year my university health plan reported that they went broke due to five events that caused costs of more than 1 million dollars each...


I think that sometimes we connect "difficulty" with "speed"

I do agree that these charts can be difficult to interpret, but I wonder if that's because we assume they should be as quick to parse as a bar chart. If you commit the time to deciphering the chart, it becomes simple to interpret.

So: is it fundamentally "difficult" or is it just "time consuming"?

I like your remakes, btw.


What software to you use to generate your graphics?


Acotgreave: I do think speed of comprehension is an important attribute. I have also written about the "return on effort" concept: if there is a large enough benefit for the effort needed to understand the chart, then readers will be more willing to expense that effort.

I have explained this type of chart in the context of predictive modeling lift to many classes of students, which is why I know that many people have trouble grasping the concept. It doesn't help that the original doesn't have the diagonal line showing the reference level of equal distribution. I find that many students even have trouble grasping the reason why equal distribution (in the case of predictive models, that's random selection) leads to a diagonal line. The other difficulty is to grasp that the units are pre-ordered by some variable before being put into ten groups.

Elias: Those were manually created Powerpoint diagrams. And yes, you can make nice-looking charts using Powerpoint or Excel. If I needed code, I'd have used R for a custom design like this. Others will use Illustrator.


A challenging chart can be worth the effort but usability should be an important goal of any designer. A reader is at least as likely to misinterpret a difficult chart as he is to fail to interpret it at all. In this case the reader walks away with a mistaken impression, and never has a reason to decide whether to commit the extra mental resources to figure it out.

Thomas Ball

Having worked in the health care sector, the chart(s) are consistent with everything known about expenditures. Their distribution is extreme valued, largely due to the chronically ill. Not mentioned (unless I missed it) is the consistency of the distribution with Pareto's classic 80-20 rule. The chronically ill also tend to be older and poorer with significant shortfalls in insurance coverage. Medicaid and medicare are picking up the tab for a large proportion of this group. On an historical note, many people attribute the kink in the graph around 1980 in life expectancy over time to the advent of managed care.

The comments to this entry are closed.