How Dataviz Can Unintentionally Perpetuate Inequality: The Bleeding Infestation Example

Sometimes, whether we know it or not, the choices we make when we visualize data can reinforce and even perpetuate racial disparity and it’s time that we talk about it. The lull of the computer monitor and the belief we are just working with numbers can make us lose sight of the fact that there are people behind this data who have given of themselves, sometimes unwillingly, and that we have a responsibility to them when we visualize.
There’s a lot to say on this topic (maybe because it hasn’t been discussed out in the open very much). But I’m going to start here with an example from my colleague, Vidhya Shanker (she’s Director of Research, Innovation & Program Evaluation at Catholic Charities of St. Paul and Minneapolis), and share other examples in future blog posts.

The Case of the Bleeding Infestations

Maps seem to be especially prone to misrepresenting people in disadvantaged situations, particularly as we get ever closer to being able to pinpoint individuals’ locations geographically. Vidhya shared with me a map of concentrated poverty in Minnesota (a factor extremely comingled with being a person of color), where individual participants were marked by red dots. When presented to the actual participants living in these areas, they were not stoked. Instead, they felt like they were perceived as a threat, and the little pixels made them look almost like an infestation on an otherwise subtly colored map.

This is an approximation of that map (the original having too much embarrassing identifiable information on it) created by Mark Herzfeld, Senior Program Evaluator at Ramsey County and former colleague of Vidhya.


Mark redesigned the infestation map in two ways so it was less visually insulting.


Mark’s second redesign, where individuals aren’t mapped but rather the poverty rates are identified at the census tract level, may be the best solution because it aggregates the issue rather than identifying individuals. It also doesn’t use angry red, which Mark found on this other map of the area, and which he said looks “like gaping wounds sucking all the money out of public coffers.”



The original designers of these maps likely didn’t intend to offend anyone (well, let’s just assume that’s the case). Let this be a learning opportunity for the rest of us to be ever mindful of our choices in colors and visuals so that we can maintain respect for the individuals that provide us with their personal data.

See the other posts in this series: Inequality Part 2 and Part 3


  1. I think this is a great post Stephanie, and I look forward to the second part. I have done quite a bit of work with maps over the past couple of years, often using this kind of data. Somewhat recently I was engaged in producing a series of maps related to crime, neighborhood conditions (i.e., abandoned buildings, dangerous buildings, foreclosures etc), socioeconomic characteristics, and race. These were being presented to stakeholder groups and were very well received. We were trying to facilitate a resident driven process to identify neighborhood level issues related to crime that could build a community response. At one point I was driving home from one of these meetings, sort of at the apex of information giving in that I had just presented my last set of maps and each neighborhood had probably 20 or so individual maps to look at. After the conversation though I started wondering if we had given too much too fast? Were we going to unintentionally facilitate the formation of conclusions on the part of residents that they will believe are supported by “data” and therefore true, because they saw it in a map and things like poverty, race, and crime seemed to be correlated? These were troubling thoughts, and your post brings me back to them.

  2. Very nice post and it illustrates several issues in applied social cartography. When we map the social we often map the structure of our societies and specific groups are visualised in context – and inequality is the norm in most societies. Secondly, as noted, we need to take the audience with us i.e. build their capacity for coming out ahead in these processes by improving their numerical and spatial literacy – if they want those skills. Thirdly, the value of a smoothing function or density maps is sometimes undervalued! Fourthly, and lastly, specificity has its place but it can also feel like being targeted.

  3. This “heat map” technique is so useful in so many ways (technically, this is a choropleth map). Years ago, I worked with a university client to relocate 3 campuses in a large metro area. The interesting twist in the problem was that faculty had the (somewhat incorrect) impression that the core student segment was the college bound transfer student, when in reality it was the career-changing adult learning. Using a consumer segmentation model, we identified and plotted by zip code the 7 consumer segments where household concentrations were highest, overlaying them with current campus locations and drive times during schedule hours. The message was crystal clear, and a major impetus to the sale of 3 existing properties and relocation of campuses to much more accessible, high growth locations. Enrollments grew about 15% over 2 years.

  4. I read this post when you first posted and still (unconsciously) created maps using a red color scale to depict the percentage of Hispanic/Latino households in the project’s study area. This was (gasp) for a data viz presentation at the AEA Conference – actually right before your TLDR presentation. So, of course – someone noticed and brought this to my attention after my presentation. I was a little horrified, because I remembered this blog post and was disappointed that I hadn’t even thought of it when designing my maps. When asked how I chose my color scheme, my mind drew a blank, but of course, after the fact, I realized why I fell into the red scale. I used ColorBrewer. I wanted a sequential scale of 4 data classes that was both photocopy safe and colorblind friendly. If you try to specify these conditions yourself, you’ll see that unfortunately the only color scale that ColorBrewer will pop back to you is a scale of oranges to reds. Anyhoo, I learned my lesson – friends, be weary of “photocopy safe” priorities when deciding your color scales.

    1. Bless your honesty, Heidi. Crazy that ColorBrewer only ends you with reds. I had thought (and experienced) that reds were the worst for being print friendly. Keep at it. Maybe talk about your lessons and show your improvements next year.

  5. Very interesting. I have plans to create an overlay of clusters of public housing with disadvantage users of public hospitals, and also children in out of home care with areas of overdue immunisation. If the ‘point’ display has this issue that you have raised what would you suggest in highlight clusters on to show a relationship between datasets? (very much a beginner with visualisation).

Leave a Reply

XHTML: You can use these tags: <a href="" title=""> <abbr title=""> <acronym title=""> <b> <blockquote cite=""> <cite> <code> <del datetime=""> <em> <i> <q cite=""> <s> <strike> <strong>

This site uses Akismet to reduce spam. Learn how your comment data is processed.