Almost two years ago I became a proud home owner in Scotland’s capital city of Edinburgh, and like many new home owners I began to wonder whether I could recoup some of the costs by renting out a room on Airbnb. But how much to charge?
Of course I could trawl through similar listings nearby, but I wondered if I would be able to use a more statistical approach, and whether I might learn something new in the process.
I wanted to find out what features are important for predicting price, and to start off, I had a hunch about a particular time of year…
The Edinburgh Fringe Festival
Each summer, Edinburgh is host to the international Fringe Festival, bringing comedy, theatre, music and thousands of tourists flooding to the capital. Exactly what effect does this annual event have on Airbnb rents, and what is the maximum I could expect to earn if I rented out as much of my house as possible during this period?
By pulling together datasets from Inside Airbnb, we can examine how rental prices are affected by the onset of the Fringe Festival.
Already it’s clear that August time (the month of the festival) corresponds to an uptick in rental price, as does new year, when countless ceilidhs and the Hogmanay street party draw people to Edinburgh. This data was compiled in March 2021, and something unusual appears to be happening in the future data, which may have something to do with the easing of lockdown restrictions within the UK, while holidays abroad remain fraught with uncertainty, pushing up demand for holiday lets. However, the same trend around August and January is still seen — which time of year corresponds to the greatest increase in price?
The graph below shows the effect of moving into and out of the Fringe period (going from 0 to 1 and 0 to -1 on the x-axis), and the corresponding percentage changes in average rental price (y-axis). If we take historical data only, we can see that the biggest jumps in price occur as the Fringe period (August) starts and ends, while the next biggest are (rather unclearly) annotated as being days around new years (2nd January).
By fitting a simple linear regression to the scattered data we can see that we should expect, on average, a impressive 15% increase in rental price during the Fringe Festival.
The full(er) picture
Clearly, it is not only the time of year that affects rental prices. To work out how much to charge for a listing on Airbnb the number of rooms for rent, the type of rent (full house/private room/shared room), and location must be taken into account.
Looking at linear correlation between features such as these and the price of rent gives many results that could easily have been expected. The number of people the listing can accommodate proves to be the most important, followed closely by number of bedrooms and beds. Unsurprisingly entire homes/apartments tend towards higher prices per night while private rooms tend towards lower ones, and Dean Village appears to be the area the most correlated with higher prices, while surprisingly The Calders comes out above Dalry and Fountainbridge. ‘Condominiums’ are worth more than apartments, hostels fare better than hotels, and the more bathrooms the merrier. However, low correlation on the whole means these comparative conclusions should be taken with a pinch of salt.
Predicting Price
But how much should I charge for a room in my house? How much could I expect for two rooms? Or what benefit might I get by adding a second bathroom? By training a simple linear model on information about the number of bathrooms, bedrooms and beds as well as the neighbourhood, the property type, room type, the number of people accommodated, and the minimum and maximum nights for booking, I can make a good guess.
Inputting the details of my house into the model, I get a suggestion of $106 (or, more usefully, £75) per night for a double room. Looking at nearby lets this does not appear to be at all unreasonable, with lets advertised for between ~£50 and ~£100. Adding in a fictional second bathroom nudges this up to £86 per night, which unfortunately does not seem enough to offset the costs of installation.
Interestingly, renting out the whole property only increases the predicted price to £171 per night, suggesting I may be marginally better off renting out each room individually, rather than the entire property as a whole.
Although there is clearly much more to the price than these simple details, and I haven’t taken into account the recent uptick since UK lockdown restrictions have started to be eased, my model serves as a useful tool to gauge how much money my rooms may be worth to total strangers, and therefore whether renting them out is worthwhile to me. If you would like to find out what your rooms might be worth, please check out my project on Github.
