This week’s data investigated the important issue of evictions in the city of San Francisco, in California. As the area has become more and more of a hub for tech companies, it has been more difficult for people of lower income levels to stay in their rented homes, as landlords try to make their properties appealing to the higher income workforce that those companies attract. The dataset contained data on evictions carried out from 1997 to 2019, and included the when, where and why of each eviction notice. Here is the visualisation given to the community for improvement, highlighting the number of evictions for the top 10 neighbourhoods over time:
Before I get into what I made, it is necessary to give yet another shout-out to Andy Kriebel, a.k.a. @VizWizBI for the inspiration that he give me and I am sure many other participants with his livestream on Monday. It was clear when the favourites came out that his approach was well received, and that we all learned a lot from watching him work.
Personally, I decided to adopt a similar approach to his, and focus on the questions when?, where?, and why?
Last, but not least, I allowed the viewer to use any graph to filter the others, because a Tableau dashboard would not be a Tableau dashboard without at least one “apply to all” filter. I believe, through these three graphs and the filtering ability, that the viewer is now able to have a better overall understanding of the dynamics behind eviction.
*While writing this I became curious whether this was an impression or an actual fact, so I used a SQL query to have a look into the dataset on data.world and in fact it turns out that December, November, January and February indeed have the lowest number of evictions! You can run the same query here, but you’ll need to have an account. Query is:
SELECT MONTH(file_date) as month, COUNT(eviction_id) as evictions
FROM sf_eviction_notices
GROUP BY MONTH(file_date)
ORDER BY count(eviction_id) ASC
While the first two charts are quite straightforward - two bar charts, really doesn’t get any more better than that - the hexbin map was fun to learn and fun to build. I learned how to make it just this week through Andy’s livestream, since even though I had seen them a few times in visualisations, I had never asked myself how they had been made.
These maps are based on rounded-off version of geographical coordinates, and they are useful for representing geographical areas without having to generalise by country or state, but also without having to treat each separate coordinate set as its own data point, which is often way too crowded. If you want to learn how to make a density map/hexbin map, either have a look at this simple tutorial I found on the subject, or have watch from minute 19 of Andy’s livestream.
The only thing I am a bit annoyed about with this solution is that, since the coordinates are rounded off, I could not add the neighbourhood each area belonged to in the tooltip, as some areas belonged to more than one and I could not find a way to properly show all values in the tooltip instead of that dreaded “*”. Now that I type this though, I am thinking I maybe could add a neighbourhood filter, either as a filter on its own or as a graph… These visualisations are never really done, are they!?
This week, I tried to take it easy, between the stress from last week’s vis and the fact I have been ill since last Friday now. So, when it came to designing this vis, I kept it simple: monochromatic, white background, basic overview of the data with classic modes of representation. I did not expect it to generate attention, and yet it was picked as a weekly favourite - and this is not the first time this happens! I guess this only taught me again that less is more, and that clarity trumps all. But most importantly… I can’t wait for more geographical data to make more density maps!