Saturday, April 25, 2009

And now, the reveal

Well, here's my interpretation. Each element is 787 units wide. The height is proportional to the fraction of the money going to the specific element. This way, you can see both the proportion that each element is in relation to the other elements and to the budget as a whole simultaneously.

Colors were adjusted to give each color the same visual weight, in accordance with Goethe's theory of colors.

The data dot is a section of the SVG which contains the data used to create the chart, in CSV format.

It's my first time, so please be gentle.


  1. Interesting.

    Doesn't need to be so wide.
    If you made the text bigger proportionally, it could be legible in the small format first visible in your current example.
    Not currently clear whether th e labels apply to the color above or below. Centering them in their color would fix that.

  2. Actually, there are reasons for each of those decisions.

    The chart must necessarily be that wide, as it is creating an implicit comparison between the size of the entire recovery package (the width) and the size of a specific element in that package (the height of each bar), as well as allowing comparisons of the size of different elements to each other.

    The text is that size because that is the largest size that can fit comfortably in the Other line, and I wanted to place the labels directly onto the chart to make the connection between label and data stronger. If I were to scale the font, that would create three variables for the text on each bar: font height, text length, and text area. This would create three conflicting visual cues for the same data, each with a different lie factor. Thus, based on text length, the visual weight of something like Education and Training ($53 B) set in 53 pt Johnston would appear to be about 72% more significant than Health Care ($59 B) set in 59 pt Johnston, despite the fact that it's actually less significant. Keeping the font size consistent ensures that the viewer does not misinterpret the wrong property as conveying information.

    The labels are placed using the bottom of the bar as a baseline mainly as a design choice, and I certainly welcome criticism of any design choices, as they can be more easily changed than the properties chosen for statistical reasons.

    (Which is not to say that I don't welcome criticism of the other properties as well)

    In defense of the current position, however, I will say that placing the labels in the center can create some problems, as it effectively bisects the bars and creates visual clutter. Bisecting the bars may make some of the smaller bars appear relatively smaller as well. What's more, there should be no confusion as to which label represents which bar, since nothing sits above or below the chart, as you would expect if the label did not refer to the bar in which it sits.

    But all in all, those are some good suggestions, and I thank you for your helpful advice.