Florian Evequoz / 2013-12-19

Living Cities, Ghost Towns
A tribute to Tulp's Ghost Counties

Permanently Occupied vs. Partially Vacant Homes in Swiss Communes
(1990-2000)

Canton - Year

0.5 home per population 1 home per population 2 homes per population
Permanently occupied Partially vacant

Context

Discovering the aesthetically wonderful Ghost Counties project by Jan Willem Tulp made me think about the recent "Lex Weber" vote in Switzerland, that was accepted by the people in March 2013. The Lex Weber is a law that basically limits the building of new secondary residences in Swiss communes by setting an upper limit: in communes where the proportion of secondary residences exceeds 20% of the total number of residences, building new secondary residences is forbidden. The results of the vote reflected the worries of alpine cantons losing their touristic development opportunities, as there was a clear separation between alpine cantons (mainly voting against Lex Weber) and urban cantons (voting for Lex Weber).

In the campaign, Alpine cantons were presented as having a high proportion of secondary residences and being the hostage of real estate developers who continued to build such residences without worrying about the nature and landscape. What the proportion of secondary residences actually was in the different communes was barely discussed in the campaign. So I wondered if I could find some data and put it in perspective.

Data

After some research, it turned out that the Swiss Federal Statistical Office does not collect information about secondary residences. The closest information I could find is number of partially vacant vs permanently occupied homes in each commune of Switzerland. The data is available only for 1990 and 2000. That was kind of a disappointment, but I think the data still offers a decent approximation of the real figures. A high number of partially vacant homes usually means that those are apartments on holiday rental. So I thought that would make a nice proof-of-concept and offer me an opportunity to practice with D3.

As my focus was to develop the visualization, I did not care about having the data for all cantons. Instead, I chose three Western Switzerland cantons with different profiles: Geneva, an urban canton; Vaud, a canton with a strong urban development around Lake Geneva, some smaller villages in the country and touristic destinations; Valais, finally, a sparsely populated canton with small towns, many villages and that mainly lives from tourism.

What does it show?

The profile of the three cantons is very different. In Geneva, most communes have a low number of partially vacant homes, therefore most bubbles are in the leftmost part of the chart. In Vaud, the situation is more mixed. Urban regions are still on the lefthand side, but there are a number of communes on the right. Those are communes with touristic development close to touristic resorts (e.g. Leysin, Villars, Château d'Oex) and many apartments or chalets on holiday rental. Finally, Valais has the most singular profile: many communes are far on the right, with more vacant households than population. Those are also touristic destinations (Blitzingen, Betten, Bellwald and other destinations in Haut-Valais, as well as the Crans-Montana or Veysonnaz regions). Switching from 1990 to 2000 shows how some regions have developed. The most noticeable one is in Valais, the commune of Riddes. In that commune, the proportion of partially vacant homes triples in 10 years. This can be explained by the development of the Mayens-de-Riddes resort town in the moutains above the village between 1990 and 2000.

About the visualization

This visualization is more artistic than analytical in nature. Concentric bubbles are not well suited to precise comparisons, and transient animations (when changing years) do not convey precisely the details of change over time. However, interesting trends can still be extracted from the data that may otherwise be hard to analyse, and the big picture (i.e. the difference between urban and alpine regions) is clearly visible.

The proportion of primary and secondary residences is mapped to the bubbles diameter. This is normally avoided, because it is generally admitted that we visually compare areas rather than circle diameters. However, in this visualisation, the bubbles actually show 2 correlated proportions, 2 slices of the same pie. If the proportion was mapped to the area, it would be very difficult to compare the area of the inner circle and the area around it that is in fact a ring, or dougnut. For example, imagine a 50-50 proportion. If the radius of the inner circle is 10, then its area is ~314 (pi*r^2). The area of the doughnut around should also be 314, therefore the doughnut outer-radius should be 14. This is represented in the image below.

Can you tell the area of the doughnut is the same as the area of the inner circle? I can't unless I pull out my ruler and compute the areas manually. Instead, when the proportion is mapped to the diameter, I think the situation gets better.

The main problem remains that one cannot compare effectively area of different shapes. But if we want to keep this visual representation, even though it is not a very effective way (from an analytical / perceptual perspective) to represent the data, I think mapping the proportion to the diameter gives a better overall impression.


Sources