Professional Documents
Culture Documents
Samal Alina
Iliyas Tleubay
Islambek Safayev
Introduction
Airbnb was founded in 2008 and has been rapidly growing to become one of the
world’s largest marketplaces for accommodations with over 7 million listings worldwide
and around 750 million guests to date (Airbnb, 2020). Airbnb operates through a sharing
economy model, and helps travelers and visitors find an accommodation by connecting
them with local hosts. Users of the platform have a wide range of choices which differ in
availability of transportation, and most importantly price. Price is one of the most important
factors for consumer’s decision making, and it affects host profitability. Thus, in a highly
competitive Airbnb market, hosts should exploit effective pricing strategies to maximize
their profits.
an accommodation, and how hosts decide on pricing strategies. It can be assumed that
the price is determined not only by the characteristics of the place, but also by the external
factors such as its rating and reviews on the platform. Therefore, the hedonic pricing
model can be applied to investigate which factors affect the price to a larger extent. This
paper can be very useful for hosts to apply efficient pricing strategies, and for Airbnb to
help hosts decide on the price and prevent inefficient pricing. Moreover, it can also be
helpful for other housing marketplaces which operate through a sharing economy model.
The paper uses official data on Prague Airbnb housing, and uses a regression
analysis to explore how the both internal and external characteristics affect the pricing.
Prague was chosen as one of the most visited places in Europe with a wide range of
model with assumption and tests, and in Section 5, we provide results and interpretations.
Literature review
Gibbs (2018) has found that in the Airbnb sharing economy, size and location of
business. Gibbs (2018) also concluded that a number of reviews has a negative impact
on prices. The explanation for this is the assumption that by setting lower prices it is easier
for a host to attract more visitors who consequently leave more reviews. Regarding the
amenities, Gibbs (2018) claims that they have a substantial effect on prices. In his
analysis, he used parking, pool, gym as factors affecting pricing. In this paper, we use the
most frequently encountered amenities in the dataset which are parking, breakfast and
pets.
Ert (2016) has found that hosts’ photos positively affect prices by helping to
related data such information about hosts is important for the guests, because in sharing
economy trust is highly appreciated. In this paper, we excluded the variable for availability
of hosts’ profile photos because in the dataset almost every host had one. Regarding the
reviews, Ert (2016) has found that this variable has little variance and has ambiguous
effect which requires a meticulous interpretation. In this paper, we will treat this variable
carefully because we assume that the results can be similar to the above study.
Magno et al. (2018) has studied the factors affecting the price of accommodation
on Airbnb in Verona, Italy. He has found that professional hosts, who manage 2 or more
apartments simultaneously, charge prices which are 5.83% higher than non-professional
hosts. The author explains that professional hosts are able to improve their expertise
more quickly, and therefore apply more efficient pricing strategies (Magno, et al., 2018).
In this paper, we will also analyze the number of listings that hosts have simultaneously
and see what effect it brings in the case of Prague. Moreover, the author finds that the
price of the apartment is positively correlated with its size, and that the number of reviews
negatively affects the price with 0.26% decrease in price with each additional review. It
was found by other authors too, be Magno et al. (2018) explains that guests are aware
that hosts might manipulate the reviews by offering gifts to visitors in exchange for a
positive review.
accommodation rental in 33 cities on Airbnb, but his analysis did not include the city that
we analyze. His main findings are that a status of superhost positively affects the price
with 8,73% effect due to the status. However, he found that this effect is stronger in lower-
priced than in higher-priced listings (Wang, 2017). He also analyzes the effect of breakfast
provision and finds that its effect is negative. However, only 9% of hosts in his dataset
offered breakfast, and it might be because these hosts tried to make their offer more
appealing amid lack of other amenities. In our paper, we will also investigate how
We chose Prague as the city for our analysis because of the following reasons.
of listings compared to the majority of other cities in Europe. Turning to the date of listings,
we intentionally didn’t choose the most recent because COVID-19 outbreak significantly
minimum nights allowed, etc.). As the most recent pre-shock date, we chose listings as
of 29th November 2019, just before the first cases of the new virus in China. Overall,
there were approximately 15,000 listings in Prague, and we decided to eliminate some
irrelevant listings for the sake of focus on our analysis. Firstly, we excluded every other
type of accommodation except apartment. The reason is that analysis would be far more
Secondly, we excluded all listings, which were provided by hosts without a profile pic.
There are two reasons for that: hosts with profile pics prevailed and there is higher
probability of a reliable verified listing with a host picture. And finally, because of
restrictions of our software package, we randomly generated 1000 listings from the whole
pool of data. All the data were collected from the official Airbnb website.
There were dozens of features in each Airbnb listing, ranging from a short
summary of an offer to cancellation policy. In order to make our analysis as insightful and
precise as possible, we tried to select only the most relevant variables. In order to do that,
appropriate part of the city, period of staying and how many people can live in an
there was a lot of missing and ambiguous data on a size of apartments, we used the
apartments. Moving to the number of guests allowed, this variable consists of four
different dummy variables: whether the guests allowed equal 1,2,3 or more than 3. We
chose more than 3 guests as the base group because the majority of hosts allow 1,2,3
Moving further, there were also a lot of amenities that could be included. From
them, we chose the most frequent ones. Because TV, Internet and heating were present
almost everywhere, we didn’t consider these amenities. Three next most frequent
amenities were chosen as independent variables: available parking, allowance for having
guests usually investigate host characteristics and reviews of selected housings. Number
model. Basically, people choose things, which are positively recommended by similar
users. Number of reviews is used in the log form because of a wide range from minimum
number of reviews to maximum. There are also five independent variables that are
housing is instant bookable or not. The next two are about minimum nights: range
between 1-3 and exceeding 3 nights. The fourth variable is again a dummy variable, which
equals 1 if host identity is verified. And lastly, we included another binary variable which
indicates whether a host is a superhost, a term used by Airbnb for the most experienced
and best-serving hosts. We structured all variables into suitable groups and the
In this paper, the OLS regression model was used to study the factors that affect
the price of Airbnb listings. The co§re equation, which represents this model is:
In order to check whether dependent variables in the model are correlated with
each other, the preliminary bivariate test was conducted. The correlation matrix is
represented in Table 2. It can be seen from the table that the issue of multicollinearity
was not detected as the highest correlation is less than 0.36, which is considered
acceptable. Also, the variance inflation factor (Table 3) revealed that variables have a
moderate level of correlation. The only exception is that the highest VIFs are observed
for the minimum number of nights, which is not considered as a problem as they are
The last issue with the model was the presence of outliers in the data. There were
detected listings with the minimum nights above 1000, which largely differ from other
listings. It was assumed that hosts mistyped when entering the data. Therefore, it was
was formulated:
There were several assumptions about the effect of different factors on the price
of Airbnb listings. Firstly, it was assumed that the superhost status positively affects on
price as their special designation allows them to charge higher prices. Also, the number
of host listings may have a negative effect on price. As more listings and more sources
of income they have, it allows them to charge lower prices to attract consumers. The
number of reviews can be positively correlated with the price, as it plays an important role
in shaping a host reputation. Another factors that are correlated to the size of the
apartment, which are the guest capacity, number of bedrooms and bathrooms, are
expected to positively impact the price, as it implies additional spending for hosts. The
presence of amenities like parking, included breakfast and the allowance for pets also
can push hosts to charge a higher price. The dummy variable entire home was expected
to positively correlate with the price, as the private room is significantly smaller than the
entire apartment. And lastly, the requirement of minimum nights can affect the price in a
negative way, as larger number of minimum nights for rental implies the host to be less
concerned.
Results and interpretation
The results for the OLS regression are presented in Table 4. All the values of
regression are reported in the heteroskedasticity-robust form. It can be seen from the
table that the goodness-of-fit measure R-squared is equal to 0.3061, which means that
the model explains about 30% of the variation in Airbnb apartment prices. As the price is
taken in the log form, the coefficients in the table show the percentage change in price
Starting from the features related to the host characteristics, it can be seen that the
status of superhost allows asking for 10% higher prices. The variable is statistically
significant with the p-value of 0.027. The result is not unexpected as this price advantage
can be explained by their huge experience on the market. Moreover, the presence of
instant book feature and larger number of host listings lead to 7% higher prices. Both
variables are statistically significant. These results seem surprising. However, it can be
assumed that if an apartment has a feature of instant book, the following convenience for
guests requires a quick mobilization from its host, which results in a higher price. Turning
to the number of host listings, it can be considered that hosts with multiple properties can
improve expertise more quickly, therefore they charge higher prices. Then, the number
statistically significant. The explanation for this could be that hosts with many reviews
have a regular demand for their apartments, so as the opportunity cost of having an empty
apartment is high for them, they charge a lower price. The number of minimum nights
The explanation for this can be the specificity of data, as almost all of the hosts allow for
Looking at the characteristics of the apartment, the variable entire home seems to
be statistically significant with the p-value of 0, and with the OLS coefficient of 0.38, which
means that if the apartment is rent as a whole, and not as a private room, then the price
seems to be 40% higher. This result is quite expected, as there is a huge difference
between a private room and the entire home. Then, the effect of specific amenities like
parking, pets and breakfast appears to be different. It can be observed that the presence
of breakfasts and the permission to keep a pet is statistically insignificant. The explanation
for their insignificance can be the comparatively low costs for having these amenities, so
it doesn’t affect the price. As regards parking, the variable is statistically significant. The
observed that hosts that have a place for only 1 guest charge 40% less prices, while
apartments with the place for 2 and 3 guests charge 32% less. All variables are
statistically significant. Also, the additional bedroom raises the price by 17%, while the
additional bathroom raises the price by 9.6%. Both variables are significant. These results
are quite obvious as the increase of living space explains the price increase.
Conclusion
attributes are valued by Airbnb customers, and how they affect the pricing. Our research
has the potential to be helpful for hosts in determining efficient pricing strategies based
supported except for the instant booking feature, number of host listings, minimum nights,
reviews and the presence of specific amenities. The explanations for unexpected results
through the sharing economy model. It reveals some unexpected results which can be
Our research has some limitations. First, we were not able to include data on the
hosts’ time since started operating on the platform because the data was inconveniently
presented, and absent for some of the hosts. Second, we did not include data on location
because of the limited information. If the data is improved in the future, including these
variables will potentially explain more variation in prices. Finally, we used a dataset for
the specific day. Therefore, it does not reflect seasonal factors which can affect a demand
for renting. Using longitudinal data can make the research more comprehensive,
therefore, further research is highly recommended because of its importance for the
11, 2018). Colliers International. Airbnb in Europe: Key cities compared (2018).
Chris Gibbs et al., “Pricing in the sharing economy: a hedonic pricing model applied to
Airbnb listings.” Journal of Travel & Tourism Marketing 35, no. 1 (2017): 46-56. doi:
10.1080/10548408.2017.1308292
Eyal Ert, Aliza Fleischer, and Nathan Magen. "Trust and Reputation in the Sharing
Economy: The Role of Personal Photos on Airbnb." SSRN Electronic Journal (2015).
doi: 10.2139/ssrn.2624181.
Magno, F., Cassia, F., Ugolini, M. "Accommodation prices on Airbnb: effects of host
experience and market demand", The TQM Journal, Vol. 30 no. 5 (2018): 608-620.
doi:10.1108/TQM-12-2017-0164