A Financial Data Analysis of Airbnb
Travel has stopped, and with it, Airbnb bookings. 68% of Airbnbs have zero bookings until August, and the average host has already lost €14K — Can Airbnb as we know it survive?
Amsterdam has a troubled history with Airbnb. The city has been lacking affordable housing for years and also just housing, in general. The Dutch capital is tiny, constricted by water, and struggling with crippling building regulations. This drives up rental prices in the private market, which reaches on average two-thirds of amsterdammers’ monthly income. The arrival of Airbnb took up rental listings out of the local market, but also provided a life-line for many Amsterdammers struggling to meet their rent obligations. After twelve years, Airbnb has undeniably become a staple of Amsterdam life, with millions of tourists swarming the canal streets and looking for an authentic local experience only Airbnb can provide.
With the disruption of short-term travel due to the current pandemic, the Airbnb host experience and thus Amsterdam has completely changed. I will analyse datasets from Airbnb listings in Amsterdam, estimate the relative change in bookings and revenue, estimate rental/mortgage costs for Airbnb hosts renting out their second homes, and answer the question if Airbnb can (or not) survive in Amsterdam.
The Data
The project Inside Airbnb has been collecting listing data for years from the platform, cleaning and structuring the datasets and making them publicly available. For this project, I am analysing the datasets that were collected on the following dates:
- February 4th 2019 vs. February 14th 2020
- March 7th 2019 vs. March 13th 2020
- April 8th 2019 vs. April 16th 2020
- May 6h 2019 vs. May 8th 2020
Since the tourism to Amsterdam fluctuates according to months and seasons, I will be comparing the yearly change for each month between February and May of 2020 vs. 2019.
The nightly, weekly and monthly prices columns are not numeric, so we need to clean them of string characters and transform them into numeric type columns:
How much have bookings dropped?
According to the data from Inside Airbnb, there are currently 19,278 listings in Amsterdam as of May 8th 2020. However, in February of the same year, there were 19,700 listings, which means 422 listings were taken down between February and May of this day, probably due to covid-19 restrictions. However, if we only looked at this statistic, we would be underestimating the problem by about 10 thousand hosts.
Many hosts have decided to make their listings unavailable for the foreseeable future instead of outright removing them, which makes sense since their listings are be full of good reviews that increase the credibility of their hosting, and thus bookings. Here’s how we can query that data:
We get that 14,129 listings are completely unavailable for the next 30 days and 13,037 are unavailable for the next 90 days. This means that the 68% of Airbnbs have a 0% booking rate for at least the next three months.
The Inside Airbnb datasets contain a host ID for every host on the platform. To make the comparison as accurate as possible, I will clean the datasets to only take into account hosts that were on the platform by February 2019 and have remained in the platform since then. Later on, we can use this sample to extrapolate the estimates to all the Airbnb hosts in Amsterdam.
We will be calculating the percent change in bookings for each host (or in other words, the relative change in vacancies). That means that we will compare the vacancy rate of each host in 2019 to their vacancy rates in 2020. This is a great feature of the data, because we don’t need to make any generalisations and can really take into account the experience of each individual host, comparing the booking rate that they should have had this year to the booking rate that they actually had.
First we need to standardise any zero availability days in 2019 to one since we will be calculating the percent change i.e. dividing by the 2019 numbers. We also need to transform the zero availability days in 2020 to 100% vacancy dates (as we discussed earlier) since hosts are making their Airbnbs unavailable on purpose due to travel and hosting restrictions. This is the code:
The graph below gives us a good idea about how many bookings the hosts are losing due to the pandemic alone:
The blue bar shows that vacant days for the next 30 days (from May 8th to June 8th) have increased on average more than 2000% compared to their 2019 numbers. For the next 90 days (May 8th to August 8th), vacant days have increased almost 6000% on average. Usually, Amsterdam Airbnbs don’t stay available for long, especially during the tulip season of March-June, when millions of tourists come to the netherlands to see our world-famous tulips, making these numbers even more disconcerting.
How much money are hosts losing in revenue?
There are two ways of answering this question:
- Calculating the difference in bookings of 2020 vs. 2019
- Taking into account the growth estimates for Airbnb bookings from 2019 to 2020.
Airbnb relies heavily on travellers being able to trust hosts, which can be a deciding factor when tourists decide whether to stay at a hotel or an Airbnb. Therefore, if a host has been on the platform for more than a year, we can safely assume that their bookings would likely increase year-to-year as they grow their reviews and ratings.
Furthermore, if we didn’t consider growth, we would probably be underestimating the financial hardship that Airbnb hosts are going through, since they probably assumed a certain growth rate as well when deciding to make capital expenditures into their Airbnbs — such as renovations, new furniture, hiring staff — and personal expenses they planned out for this year.
According to these two articles, Airbnb growth seems to average between 25% and 30% year-on-year: Skift and Reuters. Here, I will adjust the relative vacancies by 25% (meaning this year, hosts should have had 25% more bookings than they had last year).
This is the code adjusted for growth for the potential revenue hosts would have had in 2020 if the pandemic hadn’t happened:
We multiply the mean result by the number of hosts in our dataset that have been in Airbnb for more than a year (13,964) and we can graphically produce the lost revenue per month, for the next 30, 60 and 90 days counting from when the data was collected (specified on the start of this article):
This graph shows a very saddening impact of covid-19 in Amsterdam. Although we all know that touristic industries are suffering, it is always harsh to discover exactly how much the pandemic has affected them.
This graph shows both realised and unrealised losses. Let’s break down exactly how much Airbnb hosts have lost, and have much they will very likely lose over the next 3 months.
The Realised Loss
To calculate the realised loss (the loss of Airbnb revenue that has already happened) we can look at the numbers for the next 30 days of bookings for the months of February, March, April and May. According to the datasets and our calculations, we obtain the results:
Amsterdam Airbnb hosts have lost more than $197 million euros between February to May. That’s on average €14,517 euros per host in Amsterdam.
The Unrealised Loss
Calculating the difference between the May lost revenue and the forecasted loss of revenue for the months of May, June and July, we get that hosts still stand to lose almost €94 million euros in the months of June and July, which turns out to be on average €6,896.14 per host.
Extrapolating Results
If we consider the hosts that have been on Airbnb for more than a year as a sample for all the hosts of Airbnb in Amsterdam, we can extrapolate the estimates we produced and obtain a more realistic figure for how much Amsterdam hosts are losing in revenue due to the pandemic.
We calculated in the start that there are about 19,278 active listings on the Amsterdam Airbnb. If we multiply the revenue loss of the average host in our sample by the total hosts in Amsterdam, we get the figure:
An estimated €280 million euros of realised loss and €133 million euros of unrealised loss over the next two months.
What Does This Mean For Amsterdam Hosts?
When we query the dataset for how many listings a host has on Airbnb, we find out that most hosts (about 13,394 hosts) only have one listing, and out of those only 1,952 hosts list private rooms. This means most hosts with only one listing (11,442 hosts) are listing entire apartments, which are probably either bought as an investment on rented additionally to the hosts’ primary home.
Why is this important?
When hosts invest in real estate considering the high profit margins of offering that state on Airbnb, they probably budgeted their mortgage payments or rentals payments higher than they could afford them if Airbnb suddenly was not an option anymore. This includes not only monthly costs, but also expenses on renovations, high-standard appliances, furniture and decoration.
This is especially worrying if the rental is not the primary home of the host, since the initial expenses and risks of decorating a room in your apartment and renting it out whenever possible as a non-essential source of income are considerably lower than taking a 30-year mortgage or signing a year-long lease on an entire secondary home, furnishing and decorating it in a hotel-like standard.
Estimating Rent Prices of Airbnbs
There are two certain things in life: Taxes and Rent. Even though Airbnb vacancies are up 6000% compared to last year, the Netherlands has not put forth any type of rent freeze for renters during the coronavirus crisis. In fact, the yearly rent increase usually done on July 1st is set to go on unbothered. However, the government has agreed to freeze evictions in the case where the household has lost their primary income.
In most cases however, Airbnb profits do not count as primary income in the Netherlands and the Airbnb houses in question are not even the owner’s primary residence, making it extremely difficult to avoid an eventual eviction.
In my other article Machine Learning and Real State, I trained a Random Forest model that predicts rental prices for Amsterdam apartments with a 94% average accuracy. Why not use it in this project as well to have an estimate of how much these hosts probably pay for their secondary Airbnb homes? We are assuming they are renting these houses instead of owning them, however mortgages in Amsterdam are usually very close to the rental price, if not even lower. Therefore, our estimate would be somewhere between realistic and pessimistic, considering a few hosts could own multiple houses without paying either rent nor mortgages. This the code for using the model for predictions:
Predictions
After querying the dataset for only hosts that offer their entire houses/apartments on Airbnb, the model calculated rental predictions for more than 10 thousand listings. Calculating the profit margin using this code:
We get that the average profit margin in amsterdam is about 182%, which means hosts are almost doubling what they pay in rent by using Airbnb. We can interpret this in two ways: hosts are doubling what they pay for rent, or if they own a house, Airbnb is double the revenue hosts would receive if they rented the apartment in the local market instead.
However, if Airbnb is not an option anymore for these hosts (at least in the short term), how much is their liability every month to mortgage payments or rent obligations? If we sum the rental estimates for the Airbnbs in this sample of 7,343, we get a rent liability of almost €70 million euros over the last 4 months (February to May) where Airbnb hosting has been severely restricted. This works out to be about €7,592.73 euros per host on average.
If we extrapolate again for the 15,352 hosts that offer an entire apartment or house in Airbnb, the total liability increases to €145 million euros. The real question now is if the average Airbnb host in Amsterdam has at least €7,592.73 idle in the bank that can serve as a buffer throughout this crisis. If we assume the tourism crisis will continue at least until the end of July (as backed by our vacancies data for the next 90 days), this liability increases to about €11,389 per host. If not, hosts could get into serious trouble by missing mortgage payments or paying rent, and consequently lose their investment, effort and collaboration with Airbnb.
According to a 2017 article by Captain Finance, the average Dutch household has about €20,000 in savings. Assuming the same ratio can be said about Airbnb hosts, and additionally that they haven’t already spent those savings to own or renovate their Airbnbs, then we can conclude that most Amsterdam Airbnb hosts can survive until the end of July.
If €20,000 is the average limit for hosts, our estimation is that most Airbnb hosts have savings until the middle of November of 2020. After that, it is unlikely that hosts will have enough cash in the bank to continue to keep up with mortgage payments and rental obligations of their Airbnb investment apartments and houses.
The Airbnb Fund
Airbnb also announced a fund to help Airbnb hosts that suffered a lot of cancellations between May and June, refunding the average host about 12.5% of what they would have received. This might provide a life-line for hosts and extend the time they can survive without bookings.
The average revenue loss for the months of May and June is €10,541 per host in Amsterdam. If Airbnb refunds 12,5% of that, it will be €1,317.62 per host. Against the average Airbnb rental estimate of €1,898 a month in Amsterdam for the hosts that list entire properties, that gives them about 2/3 of an extra month of cash, which extends our estimate to early-December.
However, there are reports that the Airbnb Fund has not been very useful and hosts are receiving way less refunds that the values calculated here, and it’s not clear how the fund payments are being calculated instead.
So what is the answer? Will Airbnb make it?
According to the estimates produced in this article, the average Airbnb host in Amsterdam can and will survive at least until mid-November without any bookings. If the pandemic hinders travel for more time than that, it is likely that Airbnb will irrecoverably lose most of their hosts as apartments get foreclosures and hosts lose their subletting.
It is also very possible that with the negative outlook of Airbnb, hosts turn to the local market again to rent their apartment short-term until the pandemic is over — as before the pandemic they could make twice more money on Airbnb than renting to Amsterdammers. However, it is questionable whether that strategy can work considering many have lost their jobs, and many more are in fear of losing their jobs, and a non-renewable temporary rental with an inflexible end date may not be in their best interest at the moment.
It could be beneficial in the short term for Airbnb if its hosts turned to their local markets and rented their apartments for longer periods of time while the world works on shaking off this crisis, but due to the very strict tenant laws in the Netherlands, it would be very difficult for those hosts to get back to the platform when it’s safe to do so, which could be just as bad for Airbnb as irrecoverably losing those hosts for other reasons.
On the other hand, if hosts decide to stay vacant until the pandemic is over, they will have to storm through a significant loss in their savings and possibly risk losing their apartments if people don’t feel safe traveling to cosmopolitan cities by the end of the year.
Conclusion
Airbnb can survive in Amsterdam if their hosts remain loyal and the pandemic is over before the end of the year. With news of vaccine trials and European countries easing lockdowns, there is definitely a place for optimism concerning the platform. However, even with relaxed restrictions, it may be even more difficult to know exactly when people will be comfortable travelling to big cities again for both tourism or business. In any case, our estimates show that Airbnb hosts can actually hang in there for quite a few months, but maybe not enough to see the rise of tourism again. Other solutions, such as renting the apartments at the local market, will have to be weighted by the hosts so that they can extend their life-line just enough to see bookings rise again.
Good luck to all of you in these uncertain times, and I truly hope we all come out of it better, stronger and wiser. Thank you for reading.
