News and Notes from PolicyViz - Issue #27

Dec 06, 2022

Hi all,

It's possible that I'm still stuffed from Thanksgiving! I was fortunate enough to be able to spend some quality time with family in upstate New York during a long weekend. And now it's the stretch run through the end of the year. In this week's newsletter, I have a new blog post I'm working on, a new podcast, and a book giveaway--so make sure you scroll all the way through!

Thanks,

Jon

DRAFT POST: Data Visualization Equity and Small Multiples

In one section of last year's Do No Harm Guide, we argued that a small multiples approach might be preferred to a single graph that shows multiple racial groups. We wrote the following:

In the end, we shifted our approach to use multiple smaller charts (known as panel charts, trellis charts, or small multiples) rather than plotting all groups on the same visual (figure 18). We felt that by showing each race and ethnicity individually, this design might better encourage readers to think about the specific needs and challenges facing each group. Having six charts to work with (one for each demographic group and one showing all groups) also allowed us to add the relevant location-specific average as a consistent benchmark.

This strategy of small multiples can, I think, help us get away from a “horse race” effect where the reader is likely comparing all values to the best-performing group.

Eli Holder got me thinking about this argument again this week after reading his blog post Unfair Comparisons: How Visualizing Social Inequality Can Make It Worse. There, Eli (who you can also watch on a previous episode of the podcast) reviewed a recent paper he and Cindy Xiong just published on how "visualizing social outcome disparities can create a deficit framing effect and that some of the most popular chart choices can make the effect significantly worse." In other words, instead of raising awareness about a topic through, data visualizations can actually make people's perceptions of inequality worse.

Holder argues that there are several ways to try to attack this particular problem--show more variability in your graphs (jitter/beeswarm charts rather than bar charts; also see this post by Ama Nyame-Mensah); don't simplify things that are inherently complex; be aware of how even 'accurate' data visualizations can misrepresent people and communities.

But back to the small multiples argument. Because I've been rethinking it a bit. Well, not rethinking it per se, but wondering if there's a nuance to the argument.

As an example, take this graph of poverty rartes. The reader might view the White group (in blue) as the goal or the target when, in actuality, the target is really a poverty rate of 0%.

Following the discussion in the Do No Harm Guide might lead the graph creator to break this one graph into multiple graphs, like this:

But that approach may not be appropriate in all circumstances. Take this graph of the suicide rate among young people across different racial groups? In this example, the point is to see how much higher the rate is for Native American youth—and its remarkable growth—relative to the other four groups (this graph from the second Do No Harm Guide report, which you should also check out). In this case, perhaps the eye not so much moves to try to make comparisons between groups, but instead focuses on the fact the pink line is an outlier relative to the others.

If you or someone you care for is in distress, suicide prevention and crisis resources are available at the National Suicide Prevention Lifeline in the United States at 1-800-273–8255. Many other countries have similar hotlines.

In other words, the argument around disparities is perhaps better shown when all of the groups are presented in a single view rather across multiple charts.

I'm still debating this in my head, but I think there is something here. Maybe this doesn't help resolve the issues Eli and Cindy found in their research, but maybe there are times when the "horse race" effect can be used to illuminate for good reasons.

I'd be curious to hear what you think. Reach out via email or hit me up on Twitter.

Please let me know if you have thoughts on this blog post. I post drafts here and then put them on my site about a week later.