Overview
Spanish real estate market can be intense:
- Wages are one of the lowest in the E.U. (most common annually salary, statistical mode, is 16,490€. Around 1,100€ per month after taxes)
- High rate of unemployment.
- Job vacancies concentrate in the big cities.
- Incredible rise of the house rent prices in cities like Madrid and Barcelona.
Given the situation, room sharing is becoming the only choice for many employees that want to live close to their place of work.
Let’s try to explore a little deeper what is happening in the Spanish room rental market with the next interactive map.
You can explore the map in a separate page to see a cleaner source code
Read further to understand the data and visualization behind it.
Data
There is a mayor digital player in the Spanish room sharing advertising market, idealista.com, and lots of secondary players like Badi, Milanuncios, Spotahome…
I’ve extracted the data from idealista because I’ve already made a scrapper for this web-page: Dedomeno: A Spanish real estate (Idealista) python scraper.
I took a snapshot of Idealista room renting market for Spain on a random day in February. I used, like I did in Madrid Neighborhood monthly parking post, Scrapy (for scraping) and Django (for simple queries) python frameworks.
I used pandas library to clean the data in a DataFrame format, grouping by province and municipality code to calculate the median. I also added the official geocode m_code for each municipality so it could be compatible with the Spanish National Statistical Institute (INE) municipality code: m_code = int(province) + munic
# Panda DataFrame grouping by administrative boundaries
country province area zone munic m_code € median
ES 01 01 001 059 1059 287.5
02 001 001 1001 175.0
021 1021 600.0
04 001 031 1031 250.0
05 001 901 1901 200.0
06 001 002 1002 260.0
036 1036 225.0
02 01 001 003 2003 175.0
003 029 2029 125.0
039 2039 150.0
02 001 069 2069 155.0
003 081 2081 180.0
05 001 037 2037 175.0
03 01 001 137 3137 300.0
002 063 3063 200.0
003 138 3138 130.0
004 082 3082 250.0
006 047 3047 250.0
007 026 3026 200.0
030 3030 250.0
110 3110 275.0
...
003 032 48032 250.0
004 095 48095 315.0
06 001 065 48065 230.0
004 025 48025 220.0
075 48075 212.5
07 002 090 48090 300.0
49 01 001 275 49275 150.0
02 001 219 49219 150.0
50 02 001 165 50165 250.0
08 001 095 50095 170.0
252 50252 100.0
14 003 008 50008 300.0
17 001 297 50297 250.0
002 272 50272 225.0
003 089 50089 232.5
131 50131 230.0
004 219 50219 205.0
005 235 50235 250.0
288 50288 300.0
51 01 001 001 51001 290.0
52 01 001 001 52001 300.0
You can download the csv from: spain_muni_median_rent_room.csv
As you can see in the interactive map at the beginning of this post, there is a huge number of municipalities with no data.
For example Guadalajara province only has a few municipalities close to Madrid province where people offer a room to rent. This is because in rural areas the price of a full apartment is lower than renting a room in big cities.
Exploring the map further we can see a rare phenomenon; the highest room rental prices are on two small islands in the Mediterranean sea, part of the Balearic Islands: Ibiza and Formentera. This could be explained because there is a high demand for holiday apartments (specially with airbnb) and there are not enough houses available for the local people to afford them.
If you know how to read Spanish, take a look at this newspaper article called “No place to live in Ibiza”: En Ibiza no hay quien viva
In Ibiza we can check what is the sample size for each municipality, and for example in Sant Josep de sa Talia or Santa Euleria des Rius there are around 10 rooms for rent:
# Formentera (7024)
province area zone munic district neighborhood code
8943 07 04 001 024 None None 7024
Sant Josep de sa Talaia (7048)
df2.loc[df2['code']=='7048'].count()
geocode_raw 12
Santa Euleria des Riu
df2.loc[df2['code']=='7054'].count()
geocode_raw 10Visualization
Martín Gonzalez maintains a repository that provides a simple script to generate TopoJSON files from the Spanish National Geographic Institute’s National Reference Geographic Equipment vector data called es-atlas, that is inspired in Mike Bostock’s us-atlas and world-atlas. He also maintains with Lukas Appelhans Span, a small library to create modern Canvas maps with D3.
Lazy as I am, I just copied one of his examples, changed the legends, more detailed administrative borders, colors and made a small snippet directly in django console to change the rate variable:
for i in range(len(data['objects']['municipios']['geometries'])):
municipality_id = data['objects']['municipios']['geometries'][i]['id']
price = 0
try:
price = m.loc[municipality_id].price_raw
except KeyError:
pass
data['objects']['municipios']['geometries'][i]['properties']['rate'] = priceYou can download the json used in this example from:
Conclusions
As happened in last post, Madrid Neighborhood monthly parking, we have a poor quality sample size:
df2.count()
geocode_raw 16653
price_raw 16653
planet 16653
continent 16653
country 16653
province 16653
area 16653
zone 16653
munic 16653
district 15526
neighborhood 10213
code 16653
dtype: int64This is because there aren’t any final prices (16,5K rooms is a decent sample). We don’t have a way to be sure what the final price of that room was.
As we concluded in the previous post, we could avoid this issue if a snapshot were taken every day for at least one month and get rid of those items that have been on-line for a long period of time. This would avoid the highest items prices that are not going to be rented and also we could make our sample size bigger.
On the other hand, this map is far more understandable; it’s interactive, zoomable and it shows information of the name and median rent price.