Non residents only

March 2020

This piece was triggered by Action on Empty Homes who continue to be very generous with their research data. Their latest work looks at second home ownership in London and the effect it might be having on the city's housing shortage.

The data tells a clear story of housing inequality but in the context of a housing shortage, I was keen to find an angle that focussed on homes that aren't in residential use. Second homes are often rented out so can still be considered part of the stock we live in.

This isn't true of AirBnB rentals, which are overwhelmingly for tourist use. Their effect is to suck homes out of London's residential housing supply.

Already armed with borough level data on empty homes, I found the additional data I needed on the brilliant Inside AirBnB website. It's a huge resource and shows the power of citizen data scraping to influence a debate.

When pulled together, we can see the amount of housing Londoners aren't living in...

Combining AirBnB's with homes that are long term empty adds up to over 70,000 homes which aren't in residential use. This seems crazy during a housing shortage.

AirBnB rentals proliferate in central London while empties are the bigger problem in outer London. Every borough has a different mix which means a tailored policy response is needed.

For wider context the Mayor currently estimates we need to build 66,000 new homes every year until 2025 to meet London's current housing needs.

We currently build less than half that.

While these homes don't represent a complete solution, they could clearly help.

Making this visualisation

It took me a while to settle on a solution to visualising this data.

How to compare two variables geographically, while maintaining a sense of volume for each? It clearly needed to be a map but a classic choropleth (filled map) wouldn't explicitly show volume. A bivariate choropleth? Too messy. I tried stacked bars on 3D map ala Kepler.gl but it just looked confused.

The solution I settled on was made in Tableau Public and it isn't a map - it's a scatterplot.

I used borough centroid latitudes and longitudes as the X & Y values on the chart to plot locations in a geographical context. The points are then sized to ensure they're fully visible and proportionate to the number of homes they represent.

I haven't seen this chart elsewhere so I'm calling it a "flying saucer map". Humans are generally bad at judging the area of circles, hence the supporting secondary bar chart.

Overall I think it works well but can you think of alternative viz styles? I'd be intrigued to know what other solutions the viz community might produce.

Data:

Tool: Tableau Public