Graph Makeover: Educational Attainment by Race and Ethnicity

The US Bureau of Labor Statistics' May 2018 report on Race, Economics and Social Status contains the following stacked bar chart. Take a moment to look at it.

• Do any meaningful patterns jump out of you?
• Can you tell which race has the overall highest educational attainment?
• The next highest?
• Do you see any educational trends that are true across the races?

Source, (Public domain)

Now take a moment to look at this graph that I made of the same data. What do you see now?

I made three major changes to the original graph to produce this redesign.

Change 1: Use a Diverging Color Palette

First, I replaced the discrete color palette (blue, black, orange, etc.) with a diverging color palette. Discrete color palettes are best for categorical data--that is data that has no intrinsic order, such as different countries, or different hospitals. However, educational attainment categories are ordinal data: Less than high school is the least amount of education, High school graduate the next, and so on. However, blindly coding educational attainment using a single color gradient scheme looks like this:

That might be a bit better in conveying less education (the lighter color) to more (the darker colors), but this graph really doesn't tell us much. It might even be worse than the original, because the segments now bleed into each other.

Instead of using a gradient color palette, I used a diverging palette: two hues that get darker the more they move away from a common center point. In this case, I used Associates degree in college as the start of the center point, because many higher paying jobs require at least some form of associates degree. This also splits the 6 attainment categories in half, leading to this stacked bar chart:

Here, you can begin to see differences among races, simply by focusing on the orange vs. green masses. However, you can still focus on individual segments. For example, you can see that the Asian race has the longest dark green segment. You can also easily compare the lengths of all dark green (or dark orange) bars because they all share a common starting edge of the graph. With some difficulty, you can compare lengths of the bars in between, such as the light orange bars.

Change 2: Sort Race From High to Low Attainment

Sorting Race from high to low educational attainment eases your ability to recognize patterns across the races:

Notice how it is easier now to see how the orange portion grows as you scan down the chart, while the green portion shrinks. You can also see that the individual segments don't quite follow the same pattern. For instance, the dark green bars do not smoothly decrease from top to bottom.

Change 3: Align Associates Degree Segments on 0%

Finally, I turned the stacked and sorted bar chart into a diverging bar chart by aligning the start of each Associates Degree segment on 0%. Now you can see how the bars shift as your eyes move down the graph:

Different Graphs for Different Patterns

The diverging bar chart is better than the original, but it still doesn't allow us to easily see some types of patterns. The beauty of information visualization is that we can use multiple graphs. Here I've combined the bar chart with a parallel coordinates plot. Each line represents a race, but instead of a numerical X axis, the X axis uses the ordinal educational attainment levels.

The parallel coordinates plot reveals a somewhat common pattern among the races: there is an increase up through some college, a major drop for associates degrees, another bump for Bachelors, and then a decrease for higher degrees except for multi-race, which goes up. Notice, however, that the two graphs above convey different senses about this data. The lines show trends across races quite well, whereas the bar chart clearly shows the shift in college-level educational attainment.

The live version of the dashboard is interactive, such that clicking on a Race in the bar chart highlights the race's line in the parallel coordinates plot:

Clicking on an educational attainment level, highlights its segments in the bar chart and its points in the parallel coordinates plot:

You can try out the live version on my Tableau Public site where you can also download it.

Or download the data file and workbook on GitHub

To see more of my work on information visualization check out my blog @toddrjohnson