What Do Airbnb listings look like during pandemic?
Introduction
Airbnb nowadays has grown into one of today’s top communities for tourists. According to Airbnb, the company has over 5.6 million active listings worldwide in 100,000 cities with active Airbnb listings in more than 220 countries and regions.
COVID-19 impact
The global travel and tourism market is one of the worst-hit industries by the COVID-19 pandemic. Consequently, companies such as Airbnb are now coping with damaging effects of the pandemic. Given the alarming situation of COVID-19 and lockdown measures implemented by governments, it is interesting to see how people adjust their travelling behaviour. To measure the extent of travel activities, I use the number of listings getting their first and last reviews. I also extracted data on daily confirmed cases in the UK from coronavirus.data.gov.uk
As the first cases are being identified, we expect people would be more cautious due to increasing risk of infection.
In this project, we are going to look at the London dataset to answer the following questions.
1. How popular has AirBnB become in London? And how COVID-19 impacted the demand?
2. How the demand changed in 2020 in comparison to previous year?
3. How the number of active listings changed in 2020?
4. What London boroughs were rated best?
5. What are the most expensive areas in London during pandemic?
In addition, we will be predicting the prices of listings using XGBoost.
The dataset
The dataset used for this project is taken from Insideairbnb.com, an independent, non-commercial open-source data tool. The dataset was scraped on 6-Nov-2020. It contains all the data on London Airbnb listings that were active on that date. The dataset contains 76,619 unique listings. GeoJSON file of neighbourhoods of the city was also downloaded from Insideairbnb.com.
The data cleaning process is described in my Github.
How popular has AirBnB become in London? And how COVID-19 impacted the demand?
The number of listings increased over the years. We can see almost exponential growth in customer reviews on new listings which indicates a growth in demand
We can also observe the seasonal nature of the demand. There are peaks and drops every year indicating that certain months are busier than the others
We see a sharp decline in demand in the end of first quarter of 2020. The vertical line shows the date when WHO announced COVID-19 outbreak as a pandemic
How the demand changed in 2020?
There is a big difference in number of new listings between 2019 and 2020 from mid-March to September. The growth has not yet recovered to the level of the previous year.
Following pandemic announcement by WHO and local lockdown regulations, the demand has dropped significantly. The new listings started to appear on Airbnb a few weeks after new lockdown regulations come into effect and UK’s COVID-19 Alert level is lowered from Level 4 (severe risk, high transmission) to Level 3 (substantial risk, general circulation).
How the number of active listings changed during pandemic?
The figure shows that once pandemic has been announced and local lockdown regulations have been implemented, the number of active listings has dropped significantly almost reaching zero at the end of June.
Following new lockdown regulations where the rules on gatherings are relaxed and lowering UK’s COVID-19 Alert Level from Level 4 to Level 3, we observe the number of active listings increase a few weeks later.
The number of active listings started recovering at the end of July and peaked in October followed by seasonal decline. The number of properties getting their last reviews peaked in beginning of October followed by sharp drop.
Interestingly, alarming increase in number of new confirmed cases did not stop people from renting the listings. In fact, we see that when new confirmed cases started growing exponentially, number of active listings increased as well, reaching its peak in beginning of October.
How boroughs are rated?
People love their AirBnb. We can see that all boroughs, on average, are rated higher than 90%. The difference in ratings between boroughs is insignificant. Kingston upon Thames receives the highest location scores. City of London was rated the lowest.
What are the most expensive areas in London during pandemic?
Central areas are typically more expensive than outer London. However, it is interesting to note that are some expensive listings to the west of the city along Thames. It should come as no surprise that Kensington and Chelsea were found to be the most expensive areas in London. These neighbourhoods are home to the richest — celebrities, entrepreneurs, all live or have lived here. The average property price of a home in this neghborhood is about $1.4 million.
Predicting Prices
If you are a new host, it is important to get the price right, especially in a large city such as London where competition is high. It is also a challenge — pricing too high may drive away potential customers, whereas pricing too low would result in lost potential income.
In this project, we will use Machine Learning using XGBoost to predict the price for Airbnb listings in London. XGBoost has caught my attention as a widely used machine learning model among Data Scientists in industry. It is described as an optimized distributed gradient boosting library designed to be highly efficient, flexible and portable.
Data Preparation
Data cleaning process is described in github directory. In short, the missing data was imputed, and all categorical features were one-hot encoded, and data was split into train and test with the test size of 0.3.
The results are as follows:
Training RMSE: 30.82521036434503Test RMSE: 38.619985695474064Training R2: 0.7988038998561844Test R2: 0.6817755099212873Training Mean Absolute Error: 20.557898598877955Test Mean Absolute Error: 24.601560799028498
The result is not very impressive, but this is given that no feature selection was performed on the dataset.
Top 10 predictors of price (i.e., having the highest weight) were found to be:
1. If the Room type is private room
2. Number of bedrooms
3. Number of persons the property can accommodate
4. Number of bathrooms
5. If the Room type is shared room
6. Review scores related to location is 9/10
7. If the property is in Kensington and Chelsea
8. If the property is in Westminster
9. How many days are available to book out of the next 60?
10. Number of listings per host
Conclusions:
COVID-19 pandemic has significantly impacted both the demand in short-term rentals and active listings. Interestingly, alarming increase in number of new confirmed cases did not stop people from renting the listings. In fact, we observed that the highest number of listings was seen at the time when the number of confirmed COVID-19 positive cases was one of the highest.
Airbnb listings are typically scored high. We saw that all boroughs, on average, are rated higher than 90%.
Central areas are typically more expensive than outer London
Top predictors for price were found to be room type, number of bedrooms, how many people the property can accommodate and number of bathrooms
What do you think is the long-term impact of COVID-19 on Airbnb?
To see more about this project, visit my Github available here.
