Why does a salad cost more than a Big Mac?

Nathan from FlowingData has published a post today which sets out a challenge to rework the graphic below, produced by the Physicians Committee for Responsible Medicine (PCRM) to present a comparison between Federal subsidies for food production and recommended nutritional servings.

The intention of this graph is to show just how out of sync the spending is with the nutrition recommendations, but this design completely fails to present this clearly to the reader. To start with you’ve got the 3D pyramids which make the task of accurately reading the values virtually impossible. You’ve then got the issue of the values within each pyramid being ranked by size rather than a convenient ordering of food types to allow at least some comparison. Furthermore, the food subsidies has meat and dairy separated from nuts and legumes, whereas the nutritional recommendations combine these into a single ‘protein’ group. I could go on with issues like the value labels on the left pyramid standing out in bold but the ‘serving’ labels on the right are not which creates more work for the eye, but I’ll leave it at that.

So how could it be redesigned? Here are a few alternatives that I’ve quickly put together.

The first two were done using Excel, the last one was done using Tableau.

The first task was to standardise the data into comparable units so the food type values now represent % of food subsidies and % of recommended daily servings. I then made the food type groupings consistent by merging the meat and dairy values with the nuts and legumes values for food subsidies. These graphs now provide an opportunity for direct, clear comparison between the two sets of values. Some further design thoughts:

  • I couldn’t decide between the column or bar approach. The food type labels were quite long so the space afforded by the bar graph for the axis labels appealed. In Tableau, the horizontal design failed to show up the small % values of some of the items.
  • I combined the title with the legend to save on duplication and space
  • I used Gill Sans font for text because it is nice and clear and I like it at the moment (particularly in its use on the classic Penguin book designs)
  • The numbers are in Calibri font
  • I used earthy colours green and orange because I felt they were consistent with the nature of the topic
  • The values are sorted in descending order based on the nutrition recommendation %.


foodieMarch 14th, 2010 at 11:27 pm

I’d stick with the pyramids.
Simply by comparing the colours on the two pyramids shows immediately that the foods that are most recommended are the least subsidised.
I am assuming that this simple message is the main point of the graphic, and that the pyramids give a reasonaby accurate representation of the true numerical values. (eg, that the orange section of the left pyramid does actually represent 73.8% of that form.)

My first impression of the article was that they were asking what could be done about the imbalance in the distribution of subsidies! It strikes me that the information provided tells me nothing about how the subsidies affect the price of the varying commodities. I also wonder if production matches the recommended proportions, or if the subsidies reflect the size of the respective markets. Does the US produce more animal than plant foods?
Many people are very familiar with the food pyramid for good nutrition, which makes the comparison with the subsidy pyramid especially powerful, while for a more scholarly article, bar charts would be more appropriate. What is not at all clear to me is whether a redistribution of subsidies would lead to a national diet that better reflects the recommendations.
It is not so much a change of graphic, as a deeper investigation of the information presented that I would like to see.

AndyMarch 15th, 2010 at 8:53 am

Hi Trish, many thanks for your comment (the first I’ve had since finally controlling spam problems!). You’ve raised some interesting and valid points. Rather than respond directly here, I’m going to incorporate my comments into a new post later today reflecting on this task and the reaction I’ve seen to the originals and reworkings like my own.

FfoegMarch 18th, 2010 at 12:17 am

this is great. i’d like to see a graph that shows the spending on meat and dairy separately from nuts and legumes. this is certainly part of the source of obesity problems in the U.S. what the government spends its subsidies on is what is cheaper for us to buy, by spending money on meat they are encouraging us to eat extremely disproportioned meals. we should push our legislature to reorganize our spending to properly fit our dietary needs

JojoUichancoMarch 23rd, 2010 at 9:25 am

The bar graphs are definitely better representations than the pyramids. The pyramids distort the information being represented. For example, meat and dairy are 73% and about 7/10 up the pyramid, but in terms of volume they look more like 9/10. This is the big problem with representing one-dimensional data with a three-dimensional graph.

How would you represent something that was 50%? That would be half-way up the pyramid. But clearly the upper half is smaller than the lower half of the pyramid.

MichaelApril 30th, 2010 at 8:42 pm

stacked bars or pie chart are one dimensional and show that these are percentages of a whole, which none of the other options do

MattMay 22nd, 2010 at 7:16 pm

With graphs 1 and 3, I would try flipping the sets, so that the “What they do” is on the left and the “What they say” is on the right. Perhaps the strongest cue from the pyramids is just how much money goes to livestock subsidies, and I feel like that is lost here. Further, I feel like the paired bars disjoint the comparison of sums that the pyramids do accomplish. With bars, it’s four pairs of points. With the differently ordered pyramids, the audience looks and says “They give money to meat, then grain, then veggies, but say we should eat grain, then veggies, then meat. That is not the same order” With the pairs-of-bars, the audience has to look, then in their head, split the eight bars up, regroup them, then reorder them to get the same effect.

The pyramids evoke the food pyramid metaphor, which increases recognition. A new graphic should still have as strong a connection to some societal fixture.

Maybe re-coloring the pyramids so the percentages are represented as volume rather than just height.