Airbnb and House Prices in Amsterdam — Part 1
This article explores Airbnb data and is broken down into two sections:
- Part 1 explores the data and looks for correlations with house prices
- Part 2 explores machine learning techniques to predict future prices
The gig economy is profoundly reshaping how we consume products and services. Unlike previously thought, we are not seeing a decline in employment within the same industry. Instead, as observed by HBR, there is actually an increase of hired employees. This correlation can be seen across the transport industry with the rise of Uber and also in the accommodation industry, after the emergence of Airbnb. In Uber’s case, traditional taxi firms have increased the number of employed contractors to remain competitive, ultimately improving business efficiency.
Despite some clear benefits, these firms have yet to come to a legal ‘equilibrium’; Uber’s license was just revoked in London, Deliveroo is being sued by riders, and Airbnb poses a threat to affordable housing — just to name a few. This article will explore the latter scenario by analysing these two data-sets.
More precisely, we’ll try to answer the following question:
How significant is the influence of Airbnb on house price increase?
It should be said that Amsterdam was the first city to sign some agreements with Airbnb regarding its regulation, which included a rental cap of 60 days per year. Bernard D’heygere, a spokesman for Airbnb, describes its relationship with Amsterdam as “our longest and strongest partnership with any city in the world”. Despite this strong bond, Airbnb has refused to disclose the identity of hosts who do not stick to the rules, on privacy grounds. According to Inside Airbnb, half of the listings do not respect the rules.
This article is written in plain English, however for the more technically inclined you can find a link to the Git at the bottom of the page.
1. Airbnb Data Overview
Let’s start by taking a global overview of the data, below is a price heat-map of all 15,181 Airbnb listings in Amsterdam. As expected the most costly areas are de Pijp, Centrum and Oud West.
The initial concern was that investors would purchase several apartments purely for Airbnb rentals and commercial operators would take monopoly of the platform. This would be a fair observation, however, given the graph below, it seems the 60-day cap has proven effective. What you see here, is the number of apartments that a host has on the platform. Note that the distribution is so dramatic that the y axis is best plotted on a logarithmic scale. 76% of the listings are from hosts with only one property on the market. When compared to London, where 41% of listings are from hosts with more than one rental, this can be considered a healthy distribution.
Research from the Los Angeles Alliance for a New Economy has shown that it takes only 83 nights per year to earn more on Airbnb that can be earned in a whole year of renting to a long-term renter. Acknowledging the different markets, this still validates the cap in the Dutch market.
Before moving onto property prices let’s take a look at the most common property types listed on the platform. Considering the aforementioned thesis, it would be reasonable to assume most listings would be private rooms, however, this is not true. The most common type of listing is ‘entire home/apt’, leading us to deduct that people are willing to move out in order to make that extra cash, or simply have an extra apartment that they are trying to fill.
2. Housing Price Overview
After a slow decline post 2008 crisis, there has been been a sharp increase in house prices. Unfortunately, the data is only available until 2015 but I can confirm the momentum has followed through.
For those lucky enough to have entered the market in 2012, there has been a healthy profit. Bear in mind that the graph below takes the mean of the whole of Amsterdam. Increases have been far greater (up to 20%) in gentrified areas such as de Pijp, and more recently in the North, where several public infrastructure projects have been taking place, notably a new underground station.
3. Correlations
In this concluding section, we make a comparison between the house price increase and Airbnb density. Note that 2015 is the earliest available Airbnb data-set available, however it’s reasonable to assume a similar density from the previous year.
After a brief visual analysis, it’s possible to see some correlation in the Jordaan area, however, this is mostly due to it simply being a popular area. Overall, there doesn’t seem to be any significant correlations. It’s just a great year for the property market in Amsterdam.
We could look further into calculating some correlations coefficients (Pearson’s: 0.446) but without taking external factors into account, these numbers don’t mean much. Note that the price difference is capped to 2,000 €/m², by doing so we simply say that anything above that threshold is classified as a high-growth area.
Whilst Airbnb isn’t the main driver, it does have an impact which we can break down into two factors. The first and most obvious is the increased cash-flow that Airbnb brings, which as economic theory would suggest, it increases prices. Then we have the nuisance level caused by the guests that counters the price increase. The effect of Airbnb on any given neighbourhood can be thought of as of one of these factors outweighing the other. Research from the University of Amsterdam (UvA) suggests the following:
“On average, house prices increase by 0.42% per increase in Airbnb density by 10,000 reviews posted in a 1,000 meter radius around the property in the period 12 months before the transaction date. An additional analysis shows that by 2015 the total value created by Airbnb for home owners in Amsterdam, via the house prices, is just over 79 million Euros.” — Vincent van der Bijl
Although, the above statement may be true, in a Bull market such as the one observed in 2014, the effects are negligible. It mostly keeps vacancy rates down, and puts money back into the pockets of the regular Amsterdammers.
I hope you’ve found this article informative, if so please clap to help share it to others. You can find me on twitter here.
Github:
Acknowledgement:
Thanks to my colleague Magno Sousa who helped me think this through.
References:
