Visualizing Homelessness

This poster presentation asks the question: How can data visualization techniques facilitate new analyses of homelessness?
The project originates with a partnership between the Edmonton Pipelines research group at the University of Alberta, Canada, and Homeward Trust, a “community-based, comprehensive housing organization that provides leadership and resources toward ending homelessness in Edmonton [Alberta].” 1Since 1999, Homeward Trust has produced a bi-annual “Homeless Count” – a one-day snapshot of homelessness in the city. The 10 “Counts” produced to date, alongside the city of Calgary’s, represent the largest number of homeless counts conducted in any Canadian city. The data are collected manually (by volunteers), compiled into spreadsheets, and a hard copy research report is produced. Importantly, this report then becomes a key tool for policy makers, social service agencies, and all levels of government.
Homeward Trust approached the Pipelines team in 2012 to ask for assistance in producing alternate format reports, in order to more fully use and understand the data they were collecting. They were most immediately interested in mapping their data, which fits with the expertise of the Pipelines team, but over the course of our collaboration, they’ve become increasingly interested in how they can use other forms of visualizations to fulfill their mandate.
This poster is the result of our collaboration. It includes: 1) an outline of the methodology used by Homeward Trust to produce “The Count,” including their rationale for it its known limitations; 2) a number of experimental visualizations of the data, including heatmaps and infographics; 3) a series of arguments about what we can learn about homelessness based on these visualization techniques; and 4) some proposals for future research, including supplemental data gathering techniques and historical comparison, in order to more fully avail ourselves of the possibilities that the visualization techniques make possible for community-based research and action.
To date, Homeward Trust has relied on conventional communication techniques to share its results. It produces a narrative report of approximately 40 pages that provides a narrative summary of the data, and a series of graphs and tables (bar graphs, pie charts etc.) that conventionally represent demographic data. The content of the graphs and tables are also conventional given the dataset: what percentage of homeless people are men? Are aboriginal? Are within certain age ranges? Certain attributes are also cross-referenced, so the count provides an account of, for example, gender broken down by age or by type of homelessness (but not both). Overall, it is a robust piece of work produced with very few resources, but the written report is the only public record of the research and the data.
The visualizations that our team has produced are able to ask very different questions and produce different results. By producing heatmaps, for example, we are able to ask whether particular age or ethnicity groups tend to gather in particular areas and not others. In identifying those areas, we can begin to hypothesize what particular affordances a specific location might offer to a particular demographic. Similarly, by geo-locating the data and overlaying it with known social service agencies, we’re able to analyze whether homeless people tend to congregate around those services designed for them. If so, which ones? Again this allows us to hypothesize about why that might be the case (or not, as the case may be). Given the extent of the dataset, we can also track these dynamics over time to see how these populations have moved over the course of more than a decade. We can even make inferences about whether particular forms of urban renewal and development push homeless populations out of their traditional comfort zones.
These serve as a few examples of how specifically geo-located visualizations can assist us in interpreting the data. Beyond the geo-spatial models, however, we are also producing infographics, both static and dynamic, as further means of facilitating access to the data. The static infographics are being purposefully designed as educational tools to be used by Homeward Trust in their outreach activities with the broader population. For example, the infographic that displays the basic demographic information deliberately uses human figures as icons in order to remind its audience of the humanity of the subject at hand; similarly, a second infographic asks its users to "how many homeless people are like me?" and allows users to input their own age, gender, ethnicity etc. and to see themselves reflected in the homeless population. Lastly, a graph database, currently under construction, will allow us to visualize many more relations and combinations than the static infographics are able to portray, and will facilitate exploration of the data by both service providers and the general public.
At the first stages of this project, we are limiting our approach to one year's data (2012). The next stage will be to broaden our reach to the historical data sets and to begin to formulate comparisons across time.
References

1. Homeward Trust. 2012 Edmonton Homeless Count. www.homewardtrust.ca/images/resources/2013-01-22-11-53FINAL%20%202012%20Homeless%20Count.pdf