Keywords

1 Introduction

Several terms including sharing economy, collaborative economy, platform economy, and access economy, among others are currently used to describe a phenomenon that started approximately in the early 2010s. The basic principle of this phenomenon is that individuals offer products and services to other consumers and to firms using online platforms. Bulchand-Gidumal and Melián-González [1] state that to really be part of the sharing economy, individuals must have a high degree of independence when deciding what products or services to offer, what prices to set, and what characteristics to offer.

In the case of tourism, the use of these initiatives has been named collaborative tourism; in the case of lodging, peer-to-peer accommodation and sharing accommodation. The latter terms are used in this paper. Sharing accommodation takes place when individuals offer short-term lodging to tourists using platforms such as Airbnb, Homeaway, and CouchSurfing, among others. In all of these cases, individuals decide what to offer and when to offer it, set the price (if there is a price), and choose the amenities or complementary services that they will offer with their properties.

Much research has been conducted on sharing economy. This research has generally focused on the reasons for participating in the sharing economy [2], the global problems of this type of economy [3], the type of work that is generated [4], and whether sharing is a better option from a sustainability perspective [5], among other topics. Research has also looked more specifically at the case of sharing accommodation, as will be explained later in the literature review.

A common claim made by those who support the sharing economy is that it empowers individuals [6]. From the point of view of those making the offers, this refers to the ease with which anyone can be a supplier of products and services. Specifically, sharing accommodation has been used as an example of this possibility. Nevertheless, this assumption has not been addressed enough by current research. As a result, it could seem that sharing accommodation is not subject to the same market rules that are considered relevant for the success of traditional accommodation options (i.e., hotels, apartments).

This study deals with the issue of the possibility of anyone becoming a supplier of sharing accommodations, by analysing three different perspectives. First, the relationship between the average income in neighbourhoods where listings are placed and three key variables that measure hospitality success: price, user satisfaction, and demand. Second, the attributes of amateurs in comparison with professional hosts. Third, the main complaints that guests mention regarding their hosts.

2 Theory

2.1 Collaborative Tourism and Sharing Accommodation

Collaborative tourism refers to the tourist dimension of the sharing economy [7]. Like the tourism industry, collaborative tourism involves different types of activities. Examples of these activities include experiences such as gastronomy, guided tours, leisure activities, car-pooling and car sharing, and accommodation. Along with transport, accommodation accounts for the greatest portion of both sharing economy and collaborative tourism. Therefore, sharing accommodation and platforms such as Airbnb, are well known and do not require a great explanation. As a typical sharing economy platform based on Internet technology, Airbnb allows individuals to temporarily rent their properties to consumers. Thus, Airbnb and other digital platforms behave as a digital marketplace that facilitates agreements between two independent parties [1].

According to Airbnb’s data, the platform provides more than five million listings in over 81,000 cities and 191 countries [8]. In short, it has become very popular among travellers. The success of sharing accommodation draws on a number of factors including Internet technology, consumer cost-savings, household amenities available, and the potential for more authentic local experiences [9,10,11].

Much research has been conducted using data from Airbnb. Studies have addressed topics including motivations for using the sharing economy [12], the role of electronic Word of Mouth (eWOM) and authenticity in the intention to repurchase [13], price determinants [14], the impact of sharing accommodation on the hotel industry [15], and race discrimination against hosts [16].

Most of the research in the sharing economy has analysed the consumer point of view. In this sense, one topic that is currently under researched is the factors that can facilitate or prevent the success of sharing accommodation from the provider’s side. Wang and Nicolau [14] have analyzed the relevance of different variables in the price of Airbnb properties. They studied property distance from the city center, some properties’ amenities and the host’s status (i.e., host vs. superhost). Their results showed that the farther properties are from the city center, the lower their price. Real beds, bathrooms, bedrooms, free parking, and free WiFi also influence listing prices. Additionally, being an Airbnb superhost positively impacts prices. According to these findings, it is not easy for some property owners to be successful within the sharing accommodation. In other words, an owner whose property is located far from the city center, without real beds, and without free parking, will have to set lower prices which will lead to less income.

In addition to the variables identified by Wang and Nicolau [14], there are two other factors that could affect demand guest satisfaction, and the price that owners can charge for their listings. The first of the aforementioned factors is the area in which the listing is placed, which has been shown to influence client satisfaction [17,18,19,20]. The second factor is the type of host, that is, the extent to which the host can be considered a professional of the hospitality or alternatively an amateur. Regarding the type of host, extant research has considered the superhost status on Airbnb. This is a status that Airbnb awards to those hosts who meet certain criteria: a certain number of bookings, no cancellations from the host side, good ratings, and fast response times. However, more research of the role of the host has been demanded [19].

In this sense, hosts can also be classified by the number of units they offer. Some hosts offer only one property, which could be their own house that they rent out when they are not there because they are travelling; others offer their second homes, which are available most of the year; and others may rent out units purchased specifically to use as rentals to earn some additional income. Other hosts offer several properties, that is, two or more units combining one or more of the previously mentioned cases. Additionally, there are professional hosts. These hosts manage properties for others by taking care of check-in, check-out, cleaning, bookings, and maintenance of the unit, usually in exchange for a percentage of the total income. Professional hosts may also be simply owners of a significant amount of properties.

The number of properties that a host runs can influence variables that are relevant for sharing accommodation guests. For example, the more properties a host manages, the less time available for guest encounters which, in turn, could impact guest satisfaction. This is, hosts that manage more properties could provide less authentic experiences, which has also been proven to be a determinant of guest satisfaction, and one of the key factors behind the use of peer-to-peer accommodations [11, 21]. In this sense, research has shown that while guests value room attributes in the case of hotels, in the case of peer-to-peer accommodation it is the relationship with the host what is valued the most [22].

Regarding the neighbourhoods in which the units are placed, and to the researchers’ knowledge, there are no studies available. Some studies do indicate that guest frequently highlight the neighbourhood in where the unit is located in their comments [20]. The present study examines whether the income level of the neighbourhood in which a property is located influences the price of the property, the demand for the property, and guest satisfaction. It is assumed that a higher income level could positively impact these variables because it is associated with both better properties and settings.

2.2 Research Questions

Based on the findings of the literature review in the previous section, three research questions are presented:

  1. 1.

    Are neighbourhoods equivalent in the sharing accommodation (i.e., are prices, demand, and satisfaction with units independent of the neighbourhood)?

Research has already shown the relevance of location, understood as distance to public transport or to the city centre, to the listings’ prices in sharing accommodation. This research analyses if other variables, such as average neighbourhood income, are also relevant to units’ prices as well as to their demand and guest satisfaction. This will allow understanding if there are restrictions to the areas in which guests are interested, thus limiting the possibility of anyone participating in the sharing economy.

  1. 2.

    Are there differences between types of hosts (amateurs vs. professionals) with respect to demand, guest satisfaction, and pricing?

  2. 3.

    What are the main guest complaints in the sharing accommodation in relation to the host?

As previously commented, the role of the hosts has been shown to be capital in sharing accommodation. These two research questions will allow diving in further depth into this issue. On the one hand, the impact of the type of host on the performance indicators will be analysed. On the other, guest complaints. This analysis could reveal a demand for very specific host behaviours and attributes that maybe not everybody will be able to provide in the service encounters that take place in sharing accommodation.

3 Methodology

In January 2018, data from all the available listings on the Airbnb platform in Spain were automatically downloaded. In order to analyse properties that were active, only those properties with at least one guest review are considered. Some properties were eliminated because they had errors in their prices. Once the faulty ads were removed, the initial sample was composed of 98,075 listings. These listings included entire apartments and houses (74,722), private rooms (22,819), and shared rooms (534). In order to analyse a homogenous sample, only entire properties with a maximum capacity of six guests were studied. Thus, all the listings in which six or less guests can stay are considered. This last restriction was imposed because calculating the price per person in larger properties was challenging. Table 1 outlines the sample data.

Table 1. Sample data
Full size table

In the sample, the average property rate per night was 78.93 euros (std: 43.28) and the average client rating was 4.75 (std: 0.27) on a scale from one to five. The guest satisfaction rating scale consists of five stars with the following values: 1; 1.5; 2; 2.5; 3; 3.5; 4; 4.5 and 5. Apart from the general satisfaction with the unit, a specific measure of guest satisfaction with the location of the unit was also used.

The researchers calculated the number of properties that each host was offering and then classified the hosts as amateurs (those renting just one unit, accounting for 57.47% of all the properties), semi-amateurs (those renting two or three units, accounting for 23.75% of properties), semi-professionals (those renting more than three units but less than 10, accounting for 12.55% of properties), and professionals (those renting 10 or more units, accounting for 6.23% of properties). Previous research [23] had reported a much higher amount of amateurs (single-listing hosts), but it is believed that it was due to the sample used in that research and to the market dynamics.

Regarding the neighbourhoods, data from the Spanish Tax Agency was overlaid to the sample. This data is aggregated at the zip-code level and includes information such as the average income within the zip-code. Based on this information, the quartiles were used to divide the sample in four groups: low-income neighbourhoods, median-to-low-income neighbourhoods, medium-to-high-income neighbourhoods, and high-income neighbourhoods.

Airbnb does not provide the number of bookings that each unit has had. However, there is a proxy measure that can be used to determine the number of bookings: the number of reviews that the property has received. It is true that this measure can present some problems. For example, long bookings only generate one review and not all users review. Nevertheless, it is believed that reviews can be used as a good estimate when determining the demand a property has.

Finally, to analyse guest complaints content and identify items causing dissatisfaction, a random procedure by which 1,000 comments were downloaded which had global ratings of less than 4 was implemented. In other words, comments with global ratings of 3, 2, or 1 stars were downloaded. The constraint that the comments had to be in Spanish or English in order to allow processing by the researchers was implemented.

Considering a comment with a three-star rating as negative on a scale from 1 to 5 can be questionable. However, the reality is that, at least currently, most ratings on platforms such as Airbnb are very high [24]. This research confirms this phenomenon, as the average property rating was 4.75. As will be explained later, all comments were rated 4 or 5 for most properties, and 3-star ratings usually included comments that described negative experiences. Comments with a rating of 1 or 2 were very rare.

Content analysis was applied to the downloaded comments. After analysing these, two independent researchers independently proposed categories for the guests’ complaints. After a discussion, both reached consensus about the final categories. Subsequently, the same researchers independently classified the downloaded comments into the categories with a 94% agreement level.

In summary, the independent variables included were the income average of the neighbourhoods and the type of host. Dependent variables were price, guest satisfaction, and demand. Finally, the control variables taken into consideration were the capacity (maximum number of guests allowed in the property) and the starting year in Airbnb (2009 to 2017). The data were analysed using one-way ANOVA. Calculations were made with STATA v14.

4 Results

The following sections present the results of the three analyses that were completed: at the neighbourhood level, at the host type level, and at the guest satisfaction level.

4.1 Results by Neighbourhoods

The first research question sought to establish whether there are differences in property prices, demand, and guest satisfaction between neighbourhoods, depending on the average income of the area. As previously noted, the neighbourhoods were divided in four groups based on the quartiles of average income in the neighbourhood.

First, the analysis found that there were differences in average price per person across the different neighbourhood groups. While the average price per person in the low-income neighbourhoods was 19.27 euros, it was 20.79 in the higher-income ones. In other words, there was a difference of 1.52 euros per-person. It must be considered that, for a one-week rental for four people, this per-person difference can amount to more than 40 euros. Thus, properties in the higher-income group charge 8% more than similar units in the lower-income one. In order to confirm that these differences were statistically significant, an ANOVA test was run, which found that the differences were significant (p<0.001 using Bartlett’s test of equal variances). In the Bonferroni tests, all the pairs were significant except for two: medium-to-low vs. medium-to-high-income neighbourhoods, and medium-to-low vs. low-income neighbourhoods. In these two cases, no significant differences were found.

Second, guest satisfaction was analysed in relation to neighbourhoods. In global guest satisfaction, some differences between groups were found, but these differences were not consistent across all the cases. Thus, the researchers decided to use guest satisfaction in relation to the location instead of overall satisfaction. With regard to location, satisfaction increases consistently from lower-income neighbourhoods (4.72) to higher-income ones (4.78). According to an ANOVA test, these differences were statistically significant globally and across all group pairs (p<0.001 using Bartlett’s test of equal variances, and Bonferroni test for pairs comparison).

Lastly, regarding the demand for each unit, the number of reviews that a unit receives depends heavily on the year in which the property was added to the Airbnb platform. This relationship is negative: the later the unit was added to the site, the less reviews it has received. Thus, in this case, the ANOVA was performed by year. The results can be found in Table 2.

Table 2. Average number of ratings per unit by neighbourhood and by year
Full size table

As Table 2 shows, properties in higher-income neighbourhoods systematically get more reviews, and thus have greater demand than properties in lower-income areas. That is, the higher the neighbourhood income, the more reviews the property has received.

4.2 Results by Type of Host

Regarding the type of host, they were classified in four types: amateurs, semi-amateurs, semi-professionals, and professionals. The objective was to establish whether there were differences in guest satisfaction, prices, or demand depending on the type of host.

Regarding prices, it was found that as hosts manage more properties, prices rise (19.34, 20.07, 20.91, 21.04 euros per person per night per host group, respectively). That is, on average, professional hosts tend to charge 9% more than amateurs do. Of course, this price increase could be the result of a number of characteristics of the properties, including the amenities they offer, or other variables, unrelated to the hosts. For this variable, the differences are globally significant (p<0.001 using Bartlett’s test of equal variances). The same is true pairwise, except for the difference between the semi-professional and the professional hosts. In this case, the price difference was found to be non-significant.

Results regarding guest satisfaction indicate that as hosts rent more properties, user satisfaction goes down (4.77, 4.74, 4.71, 4.67). Globally, these differences are significant (p<0.001 using Bartlett’s test of equal variances). Pairwise, all the differences between groups are significant.

The results for the demand for each unit depending on the type of host can be found in Table 3. This analysis is divided by year.

Table 3. Average number of ratings per unit by type of host and by year
Full size table

The results of this analysis show statistically-significant differences. However, these differences occur between hosts that rent less than 10 units (type of hosts 1, 2, and 3) and hosts that rent 10 or more units (type of host 4). The latter systematically receive fewer reviews than those managed by hosts that rent out less than 10 properties.

4.3 Content Analysis of Reviews

1,000 comments in which the global rating given by the guest was 3, 2 or 1 stars were downloaded. As noted in Sect. 3, comments with this sort of low rating are not frequent within the Airbnb platform. For this analysis, information for more than 5,000 properties had to be retrieved in order to find these 1,000 negative comments. This pool came from the total reviews in Spanish and English, which exceeded 100,000. In other words, negative comments rated 3, 2, or 1 stars accounted for less than 1% of the reviews.

Out of these 1,000 comments, 16.3% (163) did not mention any specific negative aspects. Most of these reviews were 3-star ratings. Furthermore, in some cases, the user made an error by giving 1-star rating, mistakenly understanding that that was very good evaluation. The remaining ratings account for 83.7% (837) of the sample. These are comments that provided a low rating and negative accompanying comments.

27.4% (229) of these 837 comments mentioned factors that were not under the control of the host, including: problems finding parking, bad weather at the destination, and problems in the building were the unit was located, among other things.

The rest of the cases (72.6% of the 837 comments, that is 608 comments) involved issues that could be solved by hosts, with two main trends: the accuracy of the description of the unit, and the need for the host to take a bit more care of the property, in what many guests specifically named as TLC (Tender Loving Care).

Problems with the description of the unit were mentioned in 46.6% of the 837 cases (390 comments). These included issues such as missing equipment, extra charges that were not made clear, and issues related to the specific location of the unit, among other things.

TLC or similar issues were mentioned in 26.4% (221) of the comments. Comments in this category included issues such as equipment and areas of the unit that were not well maintained, missing curtains or blinds, and units that were clean in the better-seen areas but not so much in other areas (cupboards, under the beds, etc.).

The research also wanted to establish whether this type of complaints could be different regarding the type of host. It was found that although 58.9% of units are rented by hosts that only rent out one unit (defined as amateurs), the total number of complaints for these properties was only 47.0%. In the specific case of accuracy problems, this number goes down to 42.3%, while in the case of TLC, it is to 49.3%.

From the analysis of the comments, in 8.7% (73) of the cases a specific negative comment was made regarding the outsourcing of the management of the property: delays in the check-in process due to simultaneous arrivals at different properties, little information provided, and inflexibility of checkout times, among other things. Other comments that appear frequently were related to communication problems (9.9%, 83 comments) and not having met personally the host (8.6%, 72). In all these three types of problems, non-amateur hosts accounted for a higher percentage than the amount of units these hosts were renting out.

5 Conclusions

The main objective of this research was to determine whether the sharing economy, and its associated benefits, is really something that everyone can access easily. To this end, data downloaded from Airbnb was used, studying all of the units currently available for rental in the whole of Spain.

The results yielded important findings. First, based on the comparison of average prices, user satisfaction, and number of reviews between neighbourhoods, it is possible to conclude that although anyone can offer products and services in the sharing economy, guests usually prefer certain characteristics. In the case of accommodation units, listings placed in areas with higher income are preferred to those in areas with lower income. These properties in areas with high income have a greater demand, guests are more satisfied with the rentals, and are willing to pay higher prices. Areas with higher income are usually nicer, with better services, and close to the main points of interest of a city or a destination, among other things. Previous research [18, 20] had already found that users prefer listings close to public transport, tourist attractions, and to the city centre. The findings of this research coincide and add to this knowledge, since it is not only those attributes that matter, but also the average income of the neighbourhood.

Second, the analysis showed that users prefer to be housed by amateur hosts than by professional ones. These results exposed that user satisfaction declines as hosts rent out more units. This result coincides with previous research that had already highlighted the key role of the host in peer-to-peer accommodations [22] and how multi-listing hosts generate less user satisfaction [23]. Especially interesting is the finding that professional hosts tend to charge more than amateurs do, when all other factors are equal. This finding coincides with previous research in the area [25]. Thus, it seems as if professional hosts use some type of revenue management strategy that amateurs do not. This constant monitoring and adjusting of prices yield as a result a higher average price. However, this finding will require additional research in the future in order to be confirmed.

Finally, results show that amateurs receive fewer complaints, especially in the case of complaints involving the accuracy of the description of the space (i.e., differences between what was announced and what was received by the guests). These results allow advising property owners that if they are considering outsourcing the management of their properties to a professional company, they should be especially careful with how the property is described and how the unit is maintained. Of course, this point is also valid for owners taking care of their own properties; however, it seems more relevant in cases where a professional is taking care of the property or when owners run a high amount of listings. Anyhow, these results show that in peer-to-peer accommodations guests prefer authentic experiences to professional and standard services.

Based on these findings, the primary conclusion is that sharing accommodation is becoming one more option for users, competing directly with traditional accommodation options. For some time, there was a novelty effect by which some of the problems of these accommodations may have been masked, but this effect has now passed. Thus, guests are now demanding a similar quality from sharing economy services as they have in the past from other traditional services (e.g., hotels, taxis, restaurants, etc.). In this same vein, it seems as if the sharing economy is not as egalitarian as it was initially believed. Users look for rentals in specific areas of a city, specific amenities, and good and reliable service, even when this service is delivered by an amateur. Thus, although anyone can upload their listing to a platform such as Airbnb, some units will be more successful than others depending on their location. However, a real effort from the host in trying to make guests have a great experience could help overcome a slightly worse location.

6 Limitations

This article has certain limitations. First, when the hosts were classified based on the number of properties that they rent, the Airbnb host code was used. However, it is possible that one host rents properties using more than one host code. For example, if an individual created his/her profile in Airbnb and then decided to outsource the management to a professional company, they would appear as an amateur host even when the property is being run by a professional service. Additionally, the data set only included units from Spain. It is possible that a host offers properties in other countries as well.

Second, the number of ratings was used as a proxy for the interest in the property, since data on occupancy rates was not available. However, as is always the case with proxies, this strategy can yield certain mistakes. If some properties are rented for longer periods than others are, that can lead to incorrect interpretations of the data. If, for some reason, there is a greater trend to rate some properties than others, that can also lead to incorrect interpretations. However, the fact that such a large sample was used (more than 40,000 properties in all in the analysis) can lead to expect that most of these possible negative effects will be diluted by the sample size.

Third, since the data is quantitative and based on the information of the listings that Airbnb provides, other aspects such as the relationship between Airbnb and the hosts, specifically the guidance that Airbnb provides the hosts in relation to pricing, descriptions, pictures, and so on were not analysed.

Finally—and probably most importantly—the analysis was focused on the global issue of the sharing economy. However, only data from one platform was used, and thus, from one type of activity. Although some of the results are likely generalizable, further tests should be conducted on other types of services and in other sectors within the sharing economy.