Introduction

The market for accommodation services involves consumers and prospects who rent accommodations from traditional economy service providers, such as hotels. Sharing economy platforms like Airbnb have disrupted the accommodation market by hosting digital platforms that enable the large-scale rental of private/shared rooms and entire houses from individuals to other individuals (Bardhi and Eckhardt 2012; Zervas et al. 2014). While other platforms such as Vrbo, Wimdu, 9flats, and Roomorama offer similar services, the Airbnb’s global presence expanded rapidly, boasting more than 7 million listings across 220 countries as of 2022 (Airbnb 2023). Due to Airbnb’s affordable and unique experiences with a range of accommodations worldwide, the number of guestsFootnote 1 staying at Airbnb hostsFootnote 2 has been rapidly increasing. Since its founding in 2007, Airbnb has hosted more than 1.5 billion guests worldwide (Airbnb 2023). Globally, over 4 million Airbnb hosts have generated an estimated $180 billion in revenue. Perhaps reflecting this success, Airbnb’s total market capitalization stands at $90 billion, positioning it alongside major hotel brands such as Hilton Group, valued at $44 billion, and Marriott, valued at $61 billion (YahooFinance 2023).

Given its recent emergence and growth, insights into how consumers make their purchase decisions on Airbnb are limited. In contrast to the choice of goods, the selection of services, especially accommodations, is perceived by consumers as involving greater purchase risk (Jain and Mishra 2020; Mao et al. 2020). This perception is particularly pronounced in the case of first purchases (Conchar et al. 2004; Herzenstein et al. 2007; Savas-Hall et al. 2022). Consequently, consumer prospects seek information about accommodation features before making a choice, often delaying their purchase. In the context of branded hotels, standardization primarily provides this information. However, sharing economy platforms like Airbnb cannot rely on standardization, as the accommodation services are offered by various hosts, leading to variations in the accommodations features and service quality. Therefore, consumer prospects must depend on alternative sources of information, such as user-generated ratings and reviews. Extant research focused on sharing economy has examined the impact of different risk reduction strategies such as brand credibility, past experience, word-of-mouth, higher price, trust in the platform, trust in the service providers, online review contents, money-back guarantee, reputation, special offers, endorsement, familiarity, payment security, etc. on consumer purchase intentions (Cheng et al. 2019; Hong et al. 2019; Hsieh et al. 2022; Jun 2020; Mittendorf 2018; Tian et al. 2022; Zhang et al. 2021).

Additionally, sharing economy platforms implement various features to mitigate perceived risk and enhance the user experience. For instance, transportation-focused platforms like Uber and DoorDash incorporate real-time GPS tracking to ensure safety and convenience for users. Similarly, platforms like Home Exchange encourage the development of personal connections between members through messaging and detailed profile information. Airbnb offers information on whether the host or any previous guests of an accommodation is an acquaintance of a prospective guest. Specifically, Airbnb offers this through a search filter called social connections, which allows consumer prospects to see only those accommodations reviewed by their friends or friends of friends on Facebook or offered by one of their social connections on the Facebook network.

Arndt (1967) emphasizes the role and importance of social relationships in customers’ purchase decision-making processes. Extant research has examined the effect of social ties strength on user behavior (Dong and Wang 2018) and purchase intention (Sun et al. 2021). However, there is still limited understanding of the role of social relationships in consumers' first purchase decisions in the accommodation sector of the sharing economy. Existing theories of perceived risk and risk reduction strategies which were developed for purchase decisions in the traditional economy such as hotels or brands in general, cannot be automatically extended to sharing economy because Airbnb hosts are individuals with whom consumers prospects can have social relationships. The present research, thus, aims to fill this void in the literature.

Our empirical analysis utilizes data on the search activity and time to the first purchase of a sharing accommodation by individuals who registered on the Airbnb site. We employ a censored proportional hazards model to examine the relationship between the time to the first purchase and our primary variable of interest – social connections. We operationalized social connections as the number of times that a registered consumer prospect uses the social connections feature. Additionally, we control for the effects of demographics (gender and age), the method by which a registered user initially arrived at the Airbnb site (e.g., via a link on Facebook or a search engine), and the number devices used to access the Airbnb site. Our findings indicate a significant effect of social connections in reducing the time to the first purchase.

The variable representing social connections could be endogenous with search time. Individuals with friends on Facebook might be more experienced online users, thereby searching faster and being more inclined to make online purchases. Additionally, they may utilize the social connections feature primarily to see which of their friends might be hosts or have previously used accommodations they are considering. To explore whether alternative explanations exist for our findings, we adopt a quasi-experimental (causal) approach. Specifically, we employ propensity score matching (PSM) to pair consumer prospects who have used social connection feature at least once (treatment group) with those who have not used this feature (control group), and we estimate a model on this combined sample. We consider the method of signup, indicating whether individuals used Facebook/Google to create an Airbnb account before searching for accommodations, as well as age as a matching variable to serve as a proxy for experience with and interest in using social media and learning about friends' activities. The results from this re-estimation align with our initial findings and indicate that the using social connection feature indeed shortens the search time for the first purchase on Airbnb.

We also explore potential geographic differences in purchase decision-making to validate our findings on social connections. Specifically, we posit that the effect of social connections should be larger when someone is searching internationally rather than domestically within the US, as the perceived risk likely higher for the international travel. Therefore, we re-estimate our model incorporating geographic-specific estimates of the effect of social connections. Our analysis confirms that the effect is larger on the time to make the first purchase for international than domestic ones.

The present research contributes to the literature in two significant ways. First, this study extends beyond previous research on social relationships and enriches the literature on perceived risk, consumer behavior, and marketing strategy by examining the role of social relationships in consumer purchase decision-making, specifically in the first purchase situations. Second, the empirical setting of Airbnb allowed us to gain a deeper understanding of consumer behavior in the rapidly growing emerging industry, sharing economy. This research is managerially important as well. It offers insights into how sharing economy platforms can leverage social relationships as a strategic variable to maximize their business potential.

We next provide the theoretical and contextual background for our research, followed by a description of our data and the modeling approach for the empirical analysis. This is succeeded by a section discussing our results and substantive findings. We conclude with the managerial implications of our findings and suggestions for future research in this area.

Theoretical background

The concept of perceived risk has a long-standing history and has established a tradition of research unparalleled in the field of consumer behavior research (Mitchell 1999). We organize this extensive literature by dividing it into three sections: (1) Reasons for Delay, (2) Perceived Risk and (3) Risk Reduction Strategies. Then, based on the literature review, we develop our research framework in this section.

Reasons for delay

Several research studies have investigated why individuals delay their decisions or tasks across different contexts: daily tasks (Milgram et al. 1988), personal projects (Lay 1986), and term-paper writing by students (Solomon and Rothblum 1984). In the context of consumers’ purchase decisions, Greenleaf and Lehman (1995) developed comprehensive typologies of reasons why consumers delay purchase decisions. They suggest that the delay time considerably exceeds the active decision time—time used for gathering additional information, evaluating different alternatives, and making the actual purchase decision. Studying total delay time could explain why some purchases are made quickly while others are delayed for months.

The authors developed six propositions: (1) perceived lack of time to devote to the decision, (2) shopping for the product is unpleasant, (3) perceived risk, (4) seeking advice from others, (5) procedural uncertainty, and (6) gathering more information on alternatives. Although many reasons contribute to consumers’ decision to delay a purchase in the general shopping context, one of the most critical reasons for the delay is to reduce “perceived risk” or “uncertainty” (Corbin 1980; Darpy 2000; Greenleaf and Lehman 1995).

Perceived risk

The concept of perceived risk, introduced by Raymond A. Bauer in 1960, has been a fundamental element in understanding consumer behavior for over six decades. Perceived risk represents the uncertainty and potential for negative outcomes that consumers associate with their purchasing decisions. Bauer's seminal work laid the groundwork for the exploration of how risk perception affects consumer choices, highlighting that every consumer action carries potential consequences that cannot be fully anticipated, some of which may be unfavorable (Bauer 1960).

This notion of risk in consumer behavior has been further developed by scholars such as Cox (1967) and Cunningham (1967). Cox emphasized the role of uncertainty in decision-making, suggesting that a situation is perceived as risky when the consumer cannot confidently predict the outcome. Cunningham expanded on this by defining perceived risk in terms of two components: the potential loss if the outcome of an action is unfavorable, and the individual’s subjective feeling of uncertainty regarding the outcome.

Perceived risk is a multi-faceted concept, encompassing various dimensions that influence consumer decision-making. These dimensions include financial risk, performance risk, physical risk, social risk, psychological risk, and time or convenience risk. Financial risk involves concerns about the monetary cost of a purchase and the fear of financial loss. Performance risk relates to doubts about whether a product or service will function as expected. Physical risk pertains to the potential harm a purchase might cause to the consumer's well-being. Social risk revolves around how a purchase may affect one's social standing and relationships. Psychological risk involves the impact on one's self-esteem and self-image. Finally, time risk concerns the potential loss of time and effort involved in a purchase (Jacoby and Kaplan 1972; Roselius 1971; Stone and Gronhaug 1993). Extant research highlighted how perceived risk influence consumer purchase decisions and behavior (e.g., Godovykh et al. 2021; Li et al. 2020).

In the realm of services, particularly those that are intangible, consumer perceive greater risk (Bebko 2000; Jain and Mishra 2020; Mao et al. 2020). Services, by their very nature, lack the physical tangibility of products, making their quality and outcome more difficult to assess prior to consumption. This intangibility amplifies the perceived risk, as consumers often rely on personal experiences or the experiences of others to gauge service quality. Consequently, social risk – the potential impact of a service on one's social standing or acceptance within a group – becomes a critical consideration (Mitchell and Greatorex 1993; Murray and Schlacter 1990). Consumers may worry about how their use of a particular service will be perceived by their peers, fearing social disapproval or embarrassment. Similarly, psychological risk – the impact of a service on one's self-esteem or self-concept – is heightened in service contexts. Therefore, understanding and addressing these social and psychological aspects of perceived risk is crucial in service marketing and consumer decision-making processes (Mitchell and Greatorex 1993; Murray and Schlacter 1990).

Perceived risk in the sharing economy

In the sharing economy, exemplified by platforms like Airbnb, perceived risk takes on additional complexity (Jain and Mishra 2020; Mao et al. 2020). The sharing economy's nature, characterized by transactions between strangers and the provision of services by diverse and less established providers, heightens certain risk dimensions, particularly social and psychological risks. Consumers in this environment face uncertainties regarding the trustworthiness and reputation of service providers, the quality and safety of the services offered, and the potential for social disapproval or psychological discomfort (Anaya and Vega 2022; Jain and Mishra 2020; Jiang and Lau 2021; Xu 2020).

If consumer is new to the service, perceived risks are further elevated. Consumers must navigate unfamiliar features and service quality, which can significantly increase the perceived risk, particularly in terms of performance and psychological risks (Hall et al. 2022). Extant research suggests that perceived risk negatively influence the adoption decisions of first time purchase decisions such new products (Conchar et al. 2004; Herzenstein et al. 2007).

Strategies to mitigate perceived risk

Recognizing the impact of perceived risk on their choices, consumers adopt various strategies to mitigate these uncertainties. One primary method is information seeking, where consumers engage in extensive research to gather data that can alleviate their concerns. This search for information often involves consulting different types of sources, classified as marketer-dominated, consumer-dominated, and neutral. Marketer-dominated sources include advertising and promotional materials, consumer-dominated sources encompass word-of-mouth and peer reviews, and neutral sources involve independent reports and expert reviews (Cox 1967; Dowling and Staelin 1994). As shown in Table 1, extant research focused on sharing economy has examined the impact of different risk reduction strategies on consumer purchase intentions (Fig. 1).

Table 1 List of selected studies
Full size table
Fig. 1
figure 1

Framework

Full size image

Another key strategy in reducing perceived risk is leveraging social connections. Extant research has examined the effect of social ties strength on user behavior (Dong and Wang 2018) and purchase intention (Sun et al. 2021). These social relationships play an increasingly important role in the sharing economy. They provide platforms where individuals can share experiences, opinions, and insights about products and services. This sharing of information fosters a sense of community and trust, which can significantly minimize perceived risks, especially those related to social and psychological aspects. Extant research has explored how risk perceptions influence behavioral intentions and the role of trust in mitigating risks in peer-to-peer accommodation (Cheng 2016; Tian et al. 2022). Therefore, trust, built through social interactions and networks, becomes a crucial factor in alleviating consumer apprehensions and encouraging purchase decisions (Chen et al. 2011; Gefen 2002; Hajli 2015).

Social connections in Airbnb as a risk reduction tool

Airbnb’s introduction of the social connections feature in 2011 serves as an exemplar of how social networks can be harnessed to mitigate perceived risk in the sharing economy. This feature allows users to identify mutual connections with hosts or previous guests, fostering a sense of familiarity and trustworthiness. Such connections provide social proof and validation, which are particularly effective in reducing social and psychological risks. The feature enhances the transparency of transactions and the credibility of hosts, making prospective renters more comfortable and confident in their booking decisions. This use of social connections illustrates how modern platforms are integrating social elements into their services to address the inherent uncertainties of the sharing economy (Airbnb 2011). Given the discussion above, we propose the following hypothesis:

H1: The number of times a consumer prospect used the social connection feature to find a host is positively associated with the hazard rate of making the first purchase at Airbnb.

We further expect that the consumer characteristics directly affect the consumer purchase decision. The consumer-related variables include gender, age, signup method, number devices used for browsing, and how the consumer is acquired. Extant literature on consumer choice (e.g., Guadagni and Little 1983) suggests that individual differences significantly influence consumer purchase behavior. Thus, we examine the effect of the aforementioned customer characteristics on the purchase decision.

Data and methodology

Data

Airbnb provides a digital marketplace for individuals to list, discover, and book unique accommodations around the world. Consumer prospects can use the web application or the android/iOS application. The dataset used in this research is available as part of the Kaggle challenge (www.kaggle.com). The dataset includes demographics, web session records, and some summary statistics for users from the USA. Specifically, it contains information for each consumer prospect such as user id, language, age, gender, date of creating the account, date of first booking, signup method—Facebook, Basic, first device type—Mac, Windows, iPhone, etc., and affiliate channel. Another dataset contains information about the users’ sessions, such as action, action type, action detail, device type, and seconds elapsed. The data set consists of 35,741 consumer prospects with data on all variables considered. We discuss our variables in the following section.

Variables

\({Time}_{i}\) = a continuous variable that represents the time until consumer prospect i made the reservation if the purchase occurs. If the purchase does not occur, the time is the observation period, which 365 days from the registration date.

\({Censored}_{i}\) = an indicator variable set to 1 if the consumer prospect i did not make a reservation and 0 otherwise.

\({Female}_{i}\) = an indicator variable set to 1 if the consumer prospect i is a female and 0 if the consumer prospect is male.

\({Age}_{i}\) = a log-transformed continuous variable to represent the age of the consumer prospect i.

\({FB\_Sign\_up}_{i}\) = an indicator variable set to 1 if the consumer prospect i uses Facebook account to sign up on Airbnb and 0 otherwise.

\({Go\_Sign\_up}_{i}\) = an indicator variable set to 1 if the consumer prospect i uses Google account to sign up on Airbnb and 0 otherwise.

\({Two\_Device}_{i}\) = an indicator variable set to 1 if the consumer prospect i uses Two different devises to access Airbnb website and 0 otherwise.

\({Three\_Device}_{i}\) = an indicator variable set to 1 if the consumer prospect i uses three or more different devises to access Airbnb website and 0 otherwise.

\({Affiliate}_{i}\) = an indicator variable set to 1 if the consumer prospect i was acquired through an affiliate website and 0 otherwise.

\({API}_{i}\) = an indicator variable set to 1 if the consumer prospect i was acquired through Airbnb API and 0 otherwise.

\({Social}_{i}\) = a continuous variable to represent the number of times the consumer prospect i used social connection feature.

Model specification

Consumer prospects register on the Airbnb website and make their purchases at different points in time. Researchers investigating the time to event (first purchase in our study) usually focus on data within a specific period. Thus, each consumer prospects in our sample has her own registration time and purchase time. Some consumer prospects may make their first purchase during the observation period, while other may not make their first purchase before the end of the study period. Moreover, we do not have information about whether these consumer prospects will make their first purchase after the observation period. To accommodate these censored observations in the modeling, we model the time to first reservation using a censored proportional hazard model (Cox 1967; Heckman and Singer 1984), widely used in marketing to analyze time to event and interpurchase times (Helsen and Schmittlein 1993; Jain and Vilcassim 1991).

To accommodate censoring, hazard models focus on survival probability \({S}_{i}(t)\) and the hazard probability \({h}_{i}(t)\). The survival probability represents the probability that consumer prospect \(i\) does not make her first purchase by time \(t\), and the hazard probability is the probability that consumer prospect \(i\) makes her first purchase at time \(t\). Therefore, the survival function is defined as follows:

$${S}_{i}\left(t\right)=P ({Time}_{i}>t)$$
(1)

The hazard function of consumer prospect \(i\) at time \(t\) is:

$${h}_{i}\left(t\right)=\underset{\Delta t \to 0}{{\text{lim}}} \frac{P \left(t<{Time}_{i}\le t+ \Delta t \right|{Time}_{i}>t)}{\Delta t} =\frac{{f}_{i}(t)}{{S}_{i}(t)}$$
(2)

Most hazard models are formulated with a specific functional form. In our case, we utilize the Weibull proportional hazard model. Then the hazard rate is specified as follows:

$${h}_{i}\left(t\right)= {\lambda }_{i}\gamma {Time}_{i}^{\gamma -1} ; {Time}_{i}, {\lambda }_{i},\gamma >0$$
(3)

where \(\gamma\) is a shape parameter, and \({\lambda }_{i}\) is a scale parameter, which is reparametrized as shown below to capture the impact of explanatory variables on the hazard rate \({h}_{i}\left(t\right)\).

$${\lambda }_{i}={\text{exp}}({\beta }_{0}+{\beta }_{1}{Female}_{i}+{\beta }_{2}{\text{log}}({Age}_{i})+{\beta }_{3}{FB\_Sign\_up}_{i}+{\beta }_{4}{Go\_Sign\_up}_{i}+{\beta }_{5}{Two\_Device}_{i}+{\beta }_{6}{Three\_Device}_{i}+{\beta }_{7}{Affiliate}_{i}+{\beta }_{8}{API}_{i}+{\beta }_{9}{\text{log}}({Social}_{i}))$$
(4)

Then the log-likelihood function is given by:

$$LL\left(\beta ,\gamma \right)=\sum_{i=1}^{N}{\text{log}}({\left[\gamma {{\lambda }_{i}\left({Time}_{i}\right)}^{\gamma -1}{\text{exp}}\left(-{\lambda }_{i}{\left({Time}_{i}\right)}^{\gamma }\right)\right]}^{1-{Censored}_{i}} {\left[{\text{exp}}\left(-{\lambda }_{i}{\left({Time}_{i}\right)}^{\gamma }\right)\right]}^{{Censored}_{i}})$$
(5)

Model estimation

We obtain the posterior distribution of our parameters within a Bayesian framework using JAGS (Plummer 2003). The prior specifications for the coefficients of the covariates (all β’s) are normal with zero mean and large variance. The prior for shape parameter is Gamma with a unit mean and variance \({\sigma }^{2}\). We draw a chain of 50,000 samples with random starting values for the parameters in the Markov chain. We discard the first 25,000 samples in each chain as burn-in, and in the remaining samples, we select every 5th sample and retain 5,000 samples for posterior inference. We monitor the convergence of parameters graphically.

Results

To investigate the role of social relationships in consumer purchase decisions, we selected data for 4,316 consumer prospects who have used the social connection feature at least once. Table 2 displays a summary of descriptive statistics for both indicator and continuous variables for the selected sample. The results for the proportional hazard model, presented in Table 3, are discussed next.

Table 2 Descriptive statistics of sample for main model
Full size table
Table 3 Results for main model
Full size table

The estimate for the shape parameter (0.324) of the Weibull distribution is less than 1, indicating a decreasing hazard rate over time. This reflects time dependence, implying an “inertial” effect, where the longer the time since registration, the less likely a consumer prospect is to make a purchase soon. The coefficients of the covariates can be interpreted in similarly to those in a regression model (Jain and Vilcassim 1991). A positive (negative) coefficient indicates that the rate of purchase increases (decreases) as the covariate increase, i.e., time to purchase decreases. The parameter related to social connection variable (0.059) is significant and has the expected sign, indicating that the number of times the consumer prospect used the social connection feature is significantly positively associated with the hazard rate. For better interpretation, in the final column of Table 3, we report the hazard ratio for each covariate in the model. The hazard ratio for the social connection variable is 1.06, indicating that the likelihood of purchase for a given time increases by 6 percent for a unit increase in the use of social connection feature. This finding is consistent with our theory that consumer prospects use social relationships in searching for accommodations, thereby reducing perceived risk.

Other estimated parameters The parameter estimates for customer demographic variables – Female and Age are not significant. However, these estimates offer some intriguing insights for service providers, particularly for Age (− 0.021). The hazard ratio indicates that as customers age by one year, the chance of making a purchase at a given time decreases by approximately 2 percent. This indicates that the younger customers are more inclined towards Airbnb accommodations than older ones. However, we do not observe any significant gender difference (− 0.005) in the likelihood of making a purchase at a given time.

Parameter estimates related to the signup method via Facebook, and the number of devices (two and three) used are significant. Specifically, the parameter related to the signup method—\({FB\_Sign\_up}_{i}\) (− 0.782) is significantly negatively associated with the hazard rate. This finding indicates that, compared to consumer prospects who used a direct signup (i.e., an Airbnb account), those who signed up using their Facebook account are expected to take more time to book an accommodation on Airbnb. The parameter related to another signup method, Google, is not significant.

Both the parameters related to the number of devices used by consumer prospects—\({Two\_Device}_{i}\) (0.663) and \({Three\_Device}_{i}\) (0.805)—are significantly, positively associated with the hazard rate of purchase. These findings are particularly interesting as consumer prospects who use more devices to browse Airbnb website take less time to make a purchase compared to consumers who use a single device. This seems intuitively reasonable, as consumers more likely to make reservations may search more for accommodation and on multiple devices, such as office computer, home computer, cell phones etc. Parameters related to both acquisition channels, Affiliate and API, are not significant.

Robustness checks

Endogeneity of social connections

For the main model previously discussed, we used a sample consisting of customers who have used the social connections feature at least once. However, this social connections variable could be endogenous with search time. Individuals with friends on Facebook may be more experienced online users, thus faster in searching and more inclined to make online purchases. Additionally, they may utilize the social connections feature solely because it allows them to identify which of their friends may be hosts or have previously used accommodations they are considering. While field experiments (randomizing the sample) can mitigate such endogeneity, propensity score matching (PSM) is increasingly employed in marketing literature to address this issue (e.g., Kannan et al. 2016).

We created an indicator variable \({Social\_Used}_{i}\) to indicate whether customer prospects used the social connection feature at least once while searching for accommodations. We selected two customer characteristic variables as matching variables. First, we considered the signup method, which indicates whether individuals used Facebook/Google to set up an account on Airbnb. Second, we used age as a matching variable, serving as a proxy for experience with and interest in using social media and learning about friends’ activities. We employed the R-package MatchIt (Stuart et al. 2011), which estimates the propensity score in the background and then matches observations based on the chosen method (“nearest” in our case). Thus, the selected matched sample includes a total of 8,632 customers (4,316 consumer prospects who have used social connection feature at least once and 4,316 consumer prospects who have never used it). Table 4 displays the summary of descriptive statistics for indicator and continuous variables, respectively. Results from this re-estimation (Eqs. 1–5) are displayed in Table 5. The parameter estimate related to the social connections variable is significantly positively associated with the hazard rate of purchase. Findings from this causal approach are consistent with the main findings, indicating that social relationships are indeed reduce search time.

Table 4 Descriptive statistics of matched sample for endogeneity model
Full size table
Table 5 Results for endogeneity model
Full size table

Geographic differences

We further explore potential geographic differences in purchase decision-making to validate our findings from the main model. Specifically, we argue that perceived risk will be higher when searching accommodations internationally rather than domestically within the USA; hence, the impact of social relationships on the hazard rate of purchase should be greater for international searches. To investigate these geographic differences, we select 1983 consumer prospects who meet the two conditions: (1) they have used the social connection feature at least once and (2) they have made a purchase during the observation period. This selection enables us to estimate the destination-specific effect of social connection variable, one for domestic and another one for international purchases.

We operationalize the geographic variable, \({Dest}_{i}\), as a categorical variable, and we set it to 1 if the consumer prospect i booked an accommodation in the domestic market and to 2 if the booking was in a foreign market. Table 6 displays the summary of descriptive statistics for this sample. Let \({T}_{i}\) be a random variable representing the time it takes for consumer i to make a purchase. We can then express this as:

$${T}_{i}\sim Weibull( {\lambda }_{i}\gamma )$$
(6)

where \(\gamma\) is a shape parameter, and \({\lambda }_{i}\) is a scale parameter, which is reparametrized as shown below to captures the impact of explanatory variables on the hazard rate:

$${\lambda }_{i}={\text{exp}}({\beta }_{0}+{\beta }_{1}{Female}_{i}+{\beta }_{2}{\text{log}}({Age}_{i})+{\beta }_{3}{FB\_Sign\_up}_{i}+{\beta }_{4}{Go\_Sign\_up}_{i}+{\beta }_{5}{Two\_Device}_{i}+{\beta }_{6}{Three\_Device}_{i}+{\beta }_{7}{Affiliate}_{i}+{\beta }_{8}{API}_{i}+{\beta }_{9k=\mathrm{1,2}}{\text{log}}({Social}_{i}))$$
(7)

where \({\beta }_{9k=\mathrm{1,2}}\) capture the geographic specific-effects of the social connections variable.

Table 6 Descriptive statistics of sample for geographic specific effects model
Full size table

Parameter estimates for the above model are provided in Table 7. Results indicate that the effect of social connection variable on the hazard rate of the first purchase is indeed larger for international listings than domestic ones. This finding is consistent with our argument that consumers perceive more risk when searching for accommodation internationally than domestically and rely more on their social relationships to mitigate the perceived risk associated with the purchase.

Table 7 Results for geographic specific effects model
Full size table

Conclusions

Consumers are increasingly purchasing accommodation services through sharing economy platforms such as Airbnb, which are growing in popularity. It is critical, therefore, for providers to understand how to encourage consumers who have never used accommodation services to make their first purchase. Consumers perceive more risk while making a purchase decision of services compared to goods, and this perception of risk is expected to higher for first-time consumers and international travel. One mitigating source that Airbnb has offered is information on whether the host or any previous guests of a shared accommodation are acquaintances of a prospective consumer. Airbnb offers this through a feature called social connections that allows consumer prospects to see only those accommodations reviewed by their friends or friends of friends on Facebook.

In this research, we investigated the time taken to make the first purchase from when the Airbnb service is first considered by the consumer prospect, including the extreme case of infinite time, meaning that the choice never occurs. Our findings indicate a significant effect of social connections on the hazard rate of purchase, highlighting the role of social relationships in consumer purchase decisions.

Moreover, our analysis revealed significant findings regarding the role of the sign-up method and the number of devices used by consumer prospects. Specifically, consumers who signed up using their Facebook account are expected to take longer to book an accommodation on Airbnb compared to those who used a direct sign-up method, indicating a negative association with the hazard rate of purchase. In contrast, the use of multiple devices for browsing Airbnb's website is significantly positively associated with the hazard rate, suggesting that consumers using more devices are likely to make a purchase sooner. This behavior reflects a higher engagement level with the platform, as these consumers may actively search for accommodation across various devices.

While the effects of age and gender were not significant, indicating no marked influence on the initial purchase decision from demographic factors, these insights still contribute to a nuanced understanding of consumer behavior in the sharing economy. Additionally, the lack of significant effects for other sign-up methods, like Google, and acquisition channels, such as Affiliate and API, points to the varied impact of different consumer engagement strategies on the purchasing process. In sum, this research provides a comprehensive overview of factors influencing consumer behavior in the sharing economy, highlighting the crucial role of social connections and the interesting dynamics of sign-up methods and device usage in shaping consumer purchase decisions.

Theoretical contributions

The theoretical contributions of this research are manifold, offering valuable insights into the interplay between social relationships and consumer behavior within the sharing economy, specifically through platforms like Airbnb. This study has advanced the understanding of how social connections influence consumer purchase decision, particularly in the context of first purchases and international travel. By integrating theories perceived risk and social influence with empirical findings from a sharing economy setting, we have provided a richer, more nuanced understanding of the mechanisms through which social relationships mitigate perceived risks.

Furthermore, our research has expanded the scope of marketing analytics by applying advanced analytical techniques to dissect the complex dynamics of consumer behavior in the sharing economy. Through the use of propensity score matching and censored proportional hazard modeling, we have quantified the impact of social connections on time to purchase, offering a novel perspective on the strategic importance of social relationships in marketing strategies.

Managerial implications

From a managerial perspective, the findings from current research can yield significant insights for marketing strategy in the context of sharing economy platforms. First, findings provide insights into how sharing economy platforms by using social connection feature can shorten the time taken by prospective customers to make their first purchases through these platforms. For example, consumers prospects now have a more personal way to search for unique accommodations around the globe on Airbnb. With millions of guests stayed at Airbnb hosts to date, chances are someone a consumer prospect know has already used one of these hosts. On top of being able to search for accommodation that friends or someone in the network have reviewed, social connection feature also allows consumers and prospects to find unique places to rent from hosts who are direct friends, friends of friends, or share similar affiliations. Sharing economy platforms, thus, have to encourage hosts and guests to develop social relationships, add social profiles etc.

Second, sharing economy platform must take social relationships into account and develop personalized marketing communications strategies such as recommender system to fulfill consumer’ needs (e.g., reducing perceived risk, reduce search efforts). For example, when targeting consumers who are susceptible to delay their purchase decision, sharing economy platforms can use recommender system to suggest the consumer prospects about the offerings that have been reviewed or offered by someone in their network. Sharing economy providers, thus, can use social relationships as a strategic variable and adapt their marketing strategies to maximize their business.

Limitations and future research

Though we provide insights into how consumers of Airbnb or other service providers can benefit by providing information through social connection feature to reduce consumers’ perceived risk, we acknowledge several limitations of this study. First, we focused mainly on consumer prospects. However, findings from existing customers purchase behavior or usage of social connection feature is also important to generalize the findings. Therefore, future studies can incorporate data from existing customers into analyses.

Though we provide some insights on how consumer prospects reduce perceived risk and make their first purchase on Airbnb, there are several avenues for additional research in this area. For instance, the role of perceived risk in consumer choice between traditional service providers such as Hotels and sharing economy platforms such as Airbnb. How providers’ content like photos or videos of the services can reduce perceived risk is another such avenue.

A comparison of confirmed purchases on Airbnb before and after the introduction of the social connection feature would indeed offer valuable insights. However, due to data limitation issue, such an analysis could not be conducted for this study. Future research could benefit from analyzing the impact of the social connection feature on Airbnb's confirmed purchases by accessing pre- and post-introduction data, thereby providing a more comprehensive understanding of its effects. While this study focuses on social connections as a risk reduction strategy, future research could benefit from exploring additional factors such as user reviews, ratings, and brand reputation, which were not covered in this research due to data limitations. Another limitation of this study is the lack of data on the frequency of platform visits by users, which could potentially influence the use of the social connection feature. Future research could benefit from exploring this aspect to gain a deeper understanding of how visit frequency and feature usage interact in affecting purchase decisions in the sharing economy.