Does time dull the pain? The impact of temporal contiguity on review extremity in the hotel context

Abstract

This study aims to investigate how the timing of review posting influences the associated hotel rating. Utilizing data collected from a major travel review website, the authors estimate a hierarchical linear regression that reveals a positive relationship between temporal contiguity (i.e., the closeness between check-in time and the time when a review is posted) and review extremity, as measured by deviation from the hotel’s average rating. Moreover, two moderating factors in this relationship are highlighted: experience valence and reviewer expertise. More specifically, the positive effect of temporal contiguity on review extremity is significant only for negative traveler experiences, and this effect decreases as reviewer expertise increases. The major empirical results are further confirmed through robustness checks that apply a different range of temporal contiguity, alternative rules defining positive/negative valence, different estimation methods, and correction for endogeneity bias, respectively. Lastly, theoretical and practical implications are provided based on the empirical findings.

Full text

DOES TIME DULL THE PAIN? THE IMPACT OF TEMPORAL CONTIGUITY ON
REVIEW EXTREMITY IN THE HOTEL CONTEXT
Yang Yang, Ph.D.
Assistant Professor
Department of Tourism and Hospitality Management
Temple University, Philadelphia, Pennsylvania, USA
Email: yangy@temple.edu
Phone: (215)204-8701
Laurie Wu, Ph.D.*
Assistant Professor
Department of Tourism and Hospitality Management
Temple University, Philadelphia, Pennsylvania, USA
Email: laurie.wu@temple.edu
Phone: (215)204-8701
Fax: (215)204-8705
Wan Yang, Ph.D.
Assistant Professor
The Collins College of Hospitality Management
California State Polytechnic University, Pomona, California, USA
Email: wanyang@cpp.edu
Phone: (909) 869-3128
* Dr. Laurie Wu is the corresponding author of this paper
Please cite as
Yang, Y, Wu, L., and Yang, W. (2018). Does time dull the pain? The impact of temporal contiguity on
hotel review extremity. International Journal of Hospitality Management. 75, 119-130.

We reveal a positive relationship between temporal contiguity and review extremity.
Experience valence and reviewer expertise moderate the relationship.
Positive effect of temporal contiguity is significant only for negative experiences.
Positive effect of temporal contiguity decreases as reviewer expertise increases.

Does time dull the pain? The impact of temporal contiguity on review extremity in the hotel
context
Abstract
This study aims to investigate how the time at which a hotel review is posted influences the
associated hotel rating. Utilizing data collected from a major travel review website, the authors
estimate a hierarchical linear regression that reveals a positive relationship between temporal
contiguity (i.e., the closeness between the time of hotel stay and the time when a review is posted)
and review extremity, as measured by deviation from the hotel’s average rating. Moreover, two
moderating factors in this relationship are highlighted: experience valence and reviewer expertise.
More specifically, the positive effect of temporal contiguity on review extremity is significant
only for negative traveler experiences, and this effect decreases as reviewer expertise increases.
The major empirical results are further confirmed through robustness checks that apply a
different range of temporal contiguity, alternative rules defining positive/negative valence,
different estimation methods, and correction for endogeneity bias, respectively. Lastly,
theoretical and practical implications are provided based on the empirical findings.
Keywords: temporal contiguity; review extremity; experience valence; reviewer expertise;
construal level theory

1. Introduction
In line with the constant proliferation of information and communications technology, various
types of online reviews have become major players in consumers’ travel-related decision making
(Pan and Yang, 2016). Reviewers’ opinions and explanations can inform travelers’ perceived
product quality (Koh, Hu, and Clemons, 2010), trust and attitudes toward travel service providers
(Ladhari and Michaud, 2015), and hotel booking intentions (Sparks and Browning, 2011). In
particular, extreme reviews are more likely to sway the overall consensus about an establishment
(Zhang, Zhang, and Yang, 2016) and may potentially prove more helpful and influential for
customers compared to less-extreme reviews (Fang, Ye, Kucukusta, and Law, 2016). As such, it
is crucial for managers and tourism scholars to be aware of and understand the factors that can
influence the contents and extremity of online reviews.
Existing research shows that review contents, and notably review extremity, may be
influenced by hotel-specific factors including environmental certification (Peiró-Signes, Segarra-
Oña, Verma, Mondéjar-Jiménez, and Vargas-Vargas, 2014) and management response
frequencies (Liang, Schuckert, and Law, 2017). Reviews can also be colored by reviewer-
specific characteristics including socio-demographics (Del Chiappa and Dall’Aglio, 2012), travel
patterns (Yang, Mao, and Tang, 2017), and online review experiences (Liang, et al., 2017; Liu,
Schuckert, and Law, 2016). Yet one intriguing factor related to online reviews has yet to capture
scholarly attention: temporal contiguity. In today’s technology era, consumers may choose to
post reviews that reflect their personal travel experiences wherever and whenever they want. For
example, a traveler might post his/her review onsite, immediately after the experience, or from
another location weeks or months after the trip. These diverse possibilities raise questions: does
time matter in review rating and extremity? Does time alleviate the pain associated with a

negative travel experience? And what are the factors that moderate the effect of temporal
contiguity? The current research addresses these inquiries.
To answer these research questions, we used online review data collected for Manhattan
hotels in 2015 to investigate the relationship between temporal contiguity and review extremity
along with the moderating roles of experience valence and reviewer expertise. By doing so, this
study represents the very first research effort that highlights the importance of relative posting
time on hotel reviews to unveil the effect of temporal contiguity. Moreover, our research findings
are expected to provide important implications for hotels in terms of online reputation
management and how to decide when to solicit hotel reviews from customers.
2. Theoretical Background
2.1. Online Reviews
Over the past decade, online reviews have become a powerful tool and received increased
attention from both scholars and industry practitioners. Studies have demonstrated that online
WOM can heavily influence consumers’ attitudes, preferences, and purchase behaviors (e.g.:
Ludwig, et al., 2013; Tsao, Hsieh, Shih, and Lin, 2015; Viglia, et al., 2016; Ye, Law, and Gu,
2009). Besides the impact of eWOM, scholars also started investigating why and how people
post online reviews. For example, Chu and Kim (2011) demonstrated that tie strength, trust,
normative and informational interpersonal can trigger online posting behaviors. Yoo and Gretzel
(2008) discovered some important eWOM motivators including the need to help a service
provider, concerns for other consumers, and needs for positive self-enhancement. A group of
scholar studied the impact of status seeking on online review posting. Lampel and Bhalla (2007)

revealed that one of the most important motivators of online posting is an individual’s status
seeking intention. They argue that an individual can build a positive reputation and achieve
desired status by posting quality information and conveying his/her consumption experiences
through textual communication, and ultimately creating an ideal virtual self in the online
community (Goffman, 1955; Jensen Schau and Gilly, 2003). Burtch and Hong (2014) compared
mobile and non-mobile eWOM, and they found that consumers tend to give lower ratings and
provide more concrete and emotional text when posting reviews via a mobile device.
Unfortunately, till today (to the best of the authors’ knowledge), it is not clear whether timing
matters on eWOM behaviors, especially in the hotel eWOM context. Therefore, in the current
study, we introduce a new concept called temporal contiguity and aim to explore the impact of
posting time on review rating extremity.
2.2. Temporal contiguity in online reviews
Temporal contiguity can be defined as the temporal closeness between two stimuli or events
(Chen and Lurie, 2013). In the context of travel-related online reviews, the construct of temporal
contiguity refers to the temporal closeness between travel consumption and the time at which a
review is posted (Chen and Lurie, 2013). Temporal contiguity has been underscored as an
important factor that shapes individuals’ causal judgment and decision making (Buehner and
May, 2003; Chen and Lurie, 2013; Reed, 1992, 1999). Previous social psychology research has
revealed that the more closely two events occur in sequence, the more likely individuals are to
judge the subsequent event as having been caused by the earlier event (Buehner and May, 2003;
Einhorn and Hogarth, 1986; Kummer, 1995). Due to this psychological attribution, peripheral
informational cues that indicate temporal contiguity (also called temporal contiguity cues) can

strongly influence individuals’ causal judgment (e.g.: Buehner and May, 2003; Burtch and Hong,
2014; Chen and Lurie, 2013; Kelley, 1973; Wu, Shen, Li, and Deng, 2017). For example, in an
experiment where respondents were asked to judge the extent to which an action caused an
outcome, Shanks, Pearson, and Dickinson (1989) revealed that a low temporary contiguity (a
delay) between the action and outcome reduced causality judgments. Similarly, Topolinski and
Reber (2010) conducted three experiments and demonstrated that respondents were more likely
to judge a solution as the correct one for a problem when it appeared in higher temporal
contiguity to that problem.
In the context of the present study, the disclosure of temporal contiguity cues could also
influence travelers’ causal judgment of online reviews. Some consumers may directly indicate
temporal contiguity in their travel reviews by using phrases such as “We just came back from the
trip” (Chen and Lurie, 2013). As indicated in prior research, the disclosure of such temporal
contiguity cues can indeed influence other customers’ judgment of reviewers’ posting intentions
(Chen and Lurie, 2013; Wu, et al., 2017). The closer the temporal distance between travel
consumption and time of review posting, the more likely it is that readers will consider a positive
review a true reflection of a positive travel experience (rather than attributing the review to other
posting intentions, such as to receive a financial reward from the business) (Chen and Lurie,
2013; Wu, et al., 2017). Such attributions can lead to increased purchase intention toward the
reviewed business (Wu et al., 2017).
2.3. The impact of temporal contiguity on review extremity: the perspective of construal
level theory

While the majority of existing research on temporal contiguity examines the effect on customers’
perceptions of posted reviews and the reviewed businesses, the current study takes a unique
angle when assessing how temporal contiguity may influence reviewers’ rating behavior. In
particular, this research aims to understand if temporal contiguity might influence review
extremity. Review extremity can be conceptually defined as the extent to which an individual’s
review deviates from the general consumer consensus about the reviewed business (Schlosser,
2011); thus, review extremity can be measured as the deviation between an individual review
rating and the forum consensus rating for the corresponding business (Purnawirawan, De
Pelsmacker, and Dens, 2012). Previous marketing research has shown that extreme reviews
significantly shape consumer attitudes towards a reviewed business (Schlosser, 2011).
Drawing on construal level theory (CLT), the current research proposes that the temporal
contiguity between travel consumption and the time when a review is posted could influence the
extent of review extremity. CLT posits that individuals construe an object or event based on their
perceived psychological distance from that object or event (Dhar and Kim, 2007; Fiedler, 2007;
Trope and Liberman, 2010). While psychologically distant events tend to be construed in an
abstract manner (i.e., the desirability of an event/object), psychologically close events are more
likely to be reflected in a more detailed and tangible way (i.e., the feasibility of an event/object)
(Liberman, Trope, and Wakslak, 2007; Trope and Liberman, 2010). Further, psychological
distance can take various forms depending on its driving factors: temporal distance (i.e., the time
distance from when the event occurred), spatial distance (i.e., the space distance from where the
event occurred), social distance (i.e., the interpersonal distance between individuals who are
involved in the event), and hypothetical distance (i.e., imagining the event as likely or unlikely)
(Trope and Liberman, 2010). According to CLT, any of these four types of psychological

distance could influence the degree of construal along the “concrete-abstract” continuum
(Liberman, et al., 2007; Trope and Liberman, 2010).
In the context of travel-related online reviews, the temporal contiguity between travel
consumption and the time at which a review is posted may also affect how consumers construe
their travel experiences. After all, time is a key dimension behind construal-level effects
(Liberman and Trope, 2003; Trope and Liberman, 2003). Immediately following a travel
experience, the event is psychologically close, and so it can be reflected in a detailed, concrete
manner. As time passes, however, the event will be construed more abstractly, and detailed
aspects of the experience will gradually fade (Kim, Zhang, and Li, 2008). In such circumstances,
individuals are more likely to rely on external information to help them recall their travel
experiences (Trope and Liberman, 2003, 2010). One such source of information could be the
consensus of other reviewers’ attitudes towards the business. Given that a psychologically distant
experience is vague (Hong and Lee, 2010), reviewers are more likely to rely on others’ posted
reviews as a social anchor when writing their own reviews. Therefore, the longer the temporal
distance between travel consumption and the time when a review is posted, the less a review will
deviate from the social consensus about the reviewed business. Thus, we propose that:
H1: Temporal contiguity between travel consumption and the time when a review is posted
is positively related to review extremity.
2.4. The moderating role of experience valence
Further, the effect of temporal contiguity on review extremity is conditioned by experience
valence. For our purposes, experience valence refers to the general nature of a travel experience

(i.e., whether it is positive and satisfactory or negative and disappointing) (Vermeulen and
Seegers, 2009). Experience valence is an influential factor on consumers’ review posting
behaviors (Vermeulen and Seegers, 2009).
The proposition of this valence-conditioned effect of temporal contiguity is also
grounded in CLT, which suggests that psychological distance shapes one’s focus on positive vs.
negative attributes when perceiving social events and objects (Eyal, Liberman, and Trope, 2008;
Eyal, Liberman, Trope, and Walther, 2004; Hong and Lee, 2010). This relationship is derived
from the theoretical connection between abstractness/concreteness and desirability/feasibility. As
suggested by CLT, abstract construal of a social event is manifested in answers to the “why”
question: why should something be done (i.e., the desirability of an event) (Trope, Liberman, and
Wakslak, 2007)? Such connections are similar to the rosy memory effect, implying that
perceptions about distant events or objects tend to highlight positive features because construal
of such events or objects focuses on their desirability and benefits (Eyal, et al., 2008; Eyal, et al.,
2004). Contrarily, concrete construal of a psychologically close social event is manifested in
answers to the “how” question: how should something be achieved (i.e., the feasibility of an
event) (Trope, et al., 2007)? As feasibility is typically associated with costs and investment,
which tend to require unpleasant personal sacrifices, perceptions underpinned by feasibility
concerns tend to focus on negative attributes (Eyal, et al., 2008; Eyal, et al., 2004; Hong and Lee,
2010). Therefore, individuals are more likely to focus on negative features when reflecting on
psychologically close events or objects (Eyal, et al., 2008; Eyal, et al., 2004; Hong and Lee,
2010).
In the context of the current study, perceptions of a psychologically distant consumption
experience are likely to highlight positive aspects while perceptions of a psychologically close

consumption experience will emphasize negative attributes (Eyal, et al., 2008; Eyal, et al., 2004).
Given the theoretical association between temporal distance/contiguity and positivity vs.
negativity, it is important to consider how a consumer’s actual experience valence may moderate
the effect of temporal contiguity on review extremity. A negative service experience, by its very
nature, is consistent with the directional association between temporal contiguity and negativity
focus. In negative service experiences, the temporal closeness between the actual experience and
the time when a corresponding review is posted will further amplify the magnitude of the
experience’s negative valence. In other words, for negative travel experiences, temporal
contiguity should increase review extremity.
With positive experiences, however, such an effect is unlikely. As discussed above,
temporal contiguity enhances one’s focus on negative attributes (Eyal, et al., 2008; Eyal, et al.,
2004; Hong and Lee, 2010). The fact that temporal contiguity guides perceptions of negative
attributes shall offset individuals’ overall favorable perceptions of a positive travel experience. It
can therefore be concluded that, although temporal contiguity is generally expected to lead to
review extremity, positive travel experiences may be a different case. For positive travel
experiences, the two conflicting effects of temporal contiguity, namely (1) the association of
temporal contiguity with a negativity focus and (2) the association of temporal contiguity with
positive perception extremity, may offset each other. Thus, for positive travel experiences, the
positive effect of temporal contiguity on review extremity should be diminished. Hence, we
propose that:
H2: Experience valence moderates the effect of temporal contiguity on review extremity.
Specifically:

H2a: For negative experiences, temporal contiguity from a travel experience increases
review extremity.
H2b: For positive experiences, temporal contiguity from a travel experience does not
influence review extremity.
2.5. The moderating role of reviewer expertise
Furthermore, the effect of temporal contiguity on review extremity is conditioned by reviewer
expertise. Previous research suggests there are varied levels of expertise among travel reviewers,
and reviewer expertise significantly influences consumers’ perceptions of a business (Vermeulen
and Seegers, 2009; Zhang, Ye, Law, and Li, 2010; Zhang, Zhang, et al., 2016).
This study proposes that reviewer expertise shapes individuals’ perceptions of travel
experiences, thereby affecting travelers’ ratings in online reviews, especially in terms of review
extremity. Previous research in managerial decision making suggests that experts tend to follow
a more analytical path in their judgment and decision-making processes (Dew, Read, Sarasvathy,
and Wiltbank, 2009; Klein, 1999; Sonnentag, Niessen, and Volmer, 2006). Compared to novices,
experts are more familiar with the processed task at hand. Expertise and accumulated experience
lead in turn to a more systematic and integrative processing style, which tends to be inherently
cautious (Dew, et al., 2009; Klein, 1999; Sonnentag, et al., 2006). Such methodical processing
enhances judgmental and decision-making accuracy, as it can overcome any peripheral influence
of irrelevant contextual factors (Dew, et al., 2009; Klein, 1999; Sonnentag, et al., 2006).
In the context of the current research, compared with the influence of actual travel
experience, the impact of temporal contiguity on review contents should be somewhat peripheral.

Expertise should help experienced reviewers overcome peripheral influences and rely on their
own experiences when posting reviews. Therefore, we propose that:
H3: Reviewer expertise moderates the effect of temporal contiguity on review extremity so
that the effect is smaller for more experienced reviewers.
Last but not least, the expertise-conditioned effect of temporal contiguity on review
extremity (Hypothesis 3) may not apply to everyone. As discussed in Hypothesis 2, the effect of
temporal contiguity is evident only for negative reviews. Therefore, the expertise-conditioned
effect of temporal contiguity should apply to reviewers with negative experiences but not to
reviewers with positive experiences. Accordingly, we propose that the expertise-conditioned
effect of temporal contiguity on review extremity depends on experience valence, such that:
H4: The interaction effect of reviewer expertise and temporal contiguity on review
extremity is moderated by experience valence, such that:
H4a: For reviewers with negative experiences, reviewer expertise lowers the effect of
temporal contiguity on review extremity.
H4b: For reviewers with positive experiences, reviewer expertise does not lower the
effect of temporal contiguity on review extremity.
In sum, Figure 1 depicts the conceptual research framework.
(Please insert Figure 1 about here)
3. Models and Data

3.1. Research area and data source
We chose hotels in Manhattan, New York (NYC) for our sample. NYC is the most populous U.S.
city (United States Census Bureau, 2016), and Manhattan is one of NYC’s five boroughs with a
total population of 1,636,268 as of 2014. The city is known as the cultural and financial capital
of the world; many renowned attractions reside in Manhattan, such as the United Nations
Headquarters and Wall Street in the Financial District. With its unique global reputation for an
array of world-class cultural, historical, and business attractions, NYC is one of the most popular
cities for domestic and international tourists alike. In 2015, NYC attracted 43.2 million domestic
visitors and another 12.7 million international travelers who brought a total of $70.0 billion USD
economic revenue to the city (Tourism Economics and NYC & Company, 2016). As the third-
largest hotel market in the nation, the NYC hotel industry offered over 107,000 hotel rooms in
2015, with two-thirds of hotels located in Manhattan (Office of the New York State Comptroller,
2016).
To test the proposed research hypotheses, we crawled online review data for Manhattan
hotels on TripAdvisor using an automated program that retrieved data about hotel information,
review ratings, and reviewers’ individual profiles. TripAdvisor is the largest online travel
community and allows users to write, share, and search for travel reviews on various types of
businesses including hotels, attractions, and restaurants (Yoo, Sigala, and Gretzel, 2016).
Customers nowadays rely heavily on online reputations when booking hotels, and online ratings
have been shown to be positively related to hotel booking intentions (Sparks and Browning,
2011). The accuracy and consistency of our crawling program was manually evaluated by the
authors to ensure all information was adequately processed and stored. Using the program, we
obtained a list of Manhattan hotels from TripAdvisor and retained 364 hotels that had received at

least 20 reviews throughout the year 2015. After cleaning the dataset and deleting reviews with
incomplete information, we obtained a total of 79,107 reviews in 2015. Figure 2 pinpoints the
locations of hotels sampled in this study.
(Please insert Figure 2 about here)
3.2. Model specification
We used the absolute deviation of a five-point TripAdvisor rating scale, deviation, as the
dependent variable to measure review extremity. Deviation was calculated as the absolute value
of the difference between individual review ratings and the average rating of the corresponding
hotel in 2015. We chose to mean-center review ratings to reduce between-hotel rating variation
because many factors that contribute to between-hotel variation are challenging to measure. Our
sample encompassed a list of highly heterogeneous hotels in terms of service quality, hotel class,
and room rate. For example, out of 364 hotels, 35 were five-star hotels that specialized in
offering luxury products/services to high-end markets, whereas 14 were one-star hotels targeting
low-end segments. Therefore, after mean-centering the ratings, we could focus on leveraging
within-hotel variation to investigate the effect of temporal contiguity on hotel review ratings. We
calculated temporal contiguity as the number of months the reviewer took to post a review after
his/her hotel stay. Note that TripAdvisor records hotel stay times by month but records review
posting times by date. Therefore, we could not code in days the time from a hotel stay to the date
on which a review was posted. Figure 3 presents the frequency graph of temporal contiguity in
the sample. The graph shows that more than 60% of reviews were posted in the same month of a
hotel stay, whereas more than 20% of reviews were posted during the following month. Because
the distribution of this variable was severely left-skewed, we used a natural logarithm in our
empirical analysis.

(Please insert Figure 3 about here)
A hierarchical structure was naturally embedded in the hotel review dataset. Two levels
existed in our data, with each review representing a low-level unit and each hotel that the review
matched representing a high-level unit. Basically, the low-level unit (review) was nested in the
high-level unit (hotel). Hence, we decided to apply a hierarchical linear model (HLM) because it
presented two major advantages. First, the HLM was able to take full advantage of the
hierarchical structure of our dataset and accounted for unobserved hotel-specific factors that have
been found to influence review extremity but were not incorporated in the model, such as a
hotel’s relative location (Xiang and Krawczyk, 2016), brand effect (Nam, Ekinci, and Whyatt,
2011), and operating team performance (Salanova, Agut, and Peiró, 2005). Second, the
traditional simple linear regression imposes the independence assumption on error terms, but in
the hospitality industry, it is normal to observe spatial spillovers or economic externalities across
nearby hotels (Lee and Jang, 2015; Yang and Mao, 2017). The HLM relaxes this strong
assumption by structurally incorporating the inter-dependence of error terms (Hox, Moerbeek,
and van de Schoot, 2010). The benchmark model was specified as follows:
( )
( )
5
12 3 4
1
12
5
1
ln _ _
_
ij ij ij ij m ij
m
n ij i ij
n
deviation month neg valence expertise traveler type m
month stay n
αβ β β β
β µε
=
=
=++ ++ =
+ =++
∑
∑
(1)
where i indexes each hotel, and j indexes each review on hotel i. As discussed previously, we
used the absolute mean-centered rating, deviation, as the dependent variable. To measure
temporal contiguity, lnmonth is the log of one plus the number of months the reviewer took to
post the review after his/her hotel stay. Regarding other independent variables, a dummy variable

neg_valence indicates if the review rating is lower than the hotel’s average rating; neg_valence =
1 indicates a negative review, and neg_valence = 0 indicates a positive review. Furthermore,
expertise denotes the reviewer’s contribution level on TripAdvisor, from level 1 to level 6, based
on his/her participation in a wide variety of activities on the website. We assigned level 0 to any
reviewers whose contribution level did not reach level 1. Another independent variable,
traveler_type, refers to the reviewer’s traveler type indicated in the review. TripAdvisor lists five
traveler types: (1) couple travelers (the reference group), (2) business travelers, (3) solo travelers,
(4) family travelers, and (5) travelers traveling with friends. Lastly, month_stay is the month
when the reviewer stayed in the hotel, which included a set of 11 monthly dummies. In the
model,
ij
ε
is the normal error term like in the conventional linear regression model, and
i
µ
denotes the hotel-specific effect that captures unobserved hotel characteristics;
ij
ε
and
i
µ
follow
an independent normal distribution with a mean of zero and a variance of
2
ε
σ
and
2
µ
σ
,
respectively, to be estimated. We estimated the model using the maximum likelihood estimation
with the expectation-maximisation algorithm, which is expected to generate asymptotically
efficient and consistent estimates (Hox, et al., 2010).
In Equation 1, we tested Hypothesis 1 based on the estimate of β1. Hypothesis 1 predicts
a negative and significant coefficient of β1. To test Hypothesis 2 on the moderating role of
negativity on the effect of temporal contiguity, we introduced an interaction term between
lnmonth and neg_valence in the following model. A negative and significant coefficient of β6
would lend support to Hypothesis 2:

( )
( )
5
12 3 4
1
12
56
1
ln _ _
_ ln _ +
ij ij ij ij m ij
m
n ij ij ij i ij
n
deviation month neg valence expertise traveler type m
month stay n month neg valence
αβ β β β
β β µε
=
=
=++ ++ =
+ =+⋅ +
∑
∑
(2)
To test Hypothesis 3 on the moderating role of reviewer’s expertise on the effect of temporal
contiguity, we included an interaction term between lnmonth and expertise in the following
model. A positive and significant coefficient of β7 would lend support to Hypothesis 3:
( )
( )
5
12 3 4
1
12
57
1
ln _ _
_ ln +
ij ij ij ij m ij
m
n ij ij ij i ij
n
deviation month neg valence expertise traveler type m
month stay n month expertise
αβ β β β
β β µε
=
=
=++ ++ =
+ =+⋅ +
∑
∑
(3)
To test Hypothesis 4, we added a three-way interaction of lnmonth, expertise, and
positive. A positive and significant coefficient of β9 in the following model would lend support
to Hypothesis 4:
( )
( )
5
12 3 4
1
12
56 7
1
8
ln _ _
_ ln _ + ln
+
ij ij ij ij m ij
m
n ij ij ij ij ij
n
deviation month neg valence expertise traveler type m
month stay n month neg valence month expertise
αβ β β β
ββ β
β
=
=
=++ ++ =
+ = ++ ⋅ ⋅
∑
∑
9
_+ln _+
ij ij ij ij ij i ij
expertise neg valence month expertise neg valence
β µε
⋅ ⋅⋅ +
(4)
As mentioned previously, we cannot obtain a precise day-level temporal contiguity measure
due to data limitation. However, we used the day of month as a rough proxy for those posting
the reviews in the same month of hotel stay. For these guests, those posting reviews at the
beginning of the month are more likely to be characterized by a higher level of temporal
contiguity than those posting at the end of the month. Although we cannot gauge each review's
day-level temporal contiguity accurately, the statistical inference based on a large sample of
reviews may still present meaningful results. Therefore, in the last step of empirical analysis,

we will replace lnmonth in Equations 1-4 by lnday, the log of the day of month for reviews
posted in the same month of hotel stay.
3.3. Descriptive data analysis
Table 1 presents the descriptive statistics for the continuous variables incorporated in our model.
The dependent variable, deviation, had a mean value of 0.724 with a standard deviation of 0.604,
and its distribution was slightly left-skewed as indicated by a median value larger than the mean.
For independent variables, the mean and median values of expertise were 2.356 and 2,
respectively. The mean value of lnmonth was 0.341, which corresponds to 0.41 months after the
hotel stay, indicating that most reviewers posted their reviews shortly after their stay. Moreover,
the median value of lnmonth was 0, indicating that at least half of reviewers left their reviews in
the same month as their stay. The mean value of lnday was 2.798, and its median value was
2.944, corresponding to the 19th of each month.
(Please insert Table 1 about here)
Table 2 presents the descriptive statistics of our categorical variables. The statistics
suggest that nearly 40% of reviews were negative and gave a lower-than-average hotel rating.
Furthermore, out of the five types of travelers, couples dominated our sample (34.93%), although
family travelers and business travelers accounted for a large share of the Manhattan lodging
market overall (27.54% and 19.63%, respectively). Relatively few hotel guests traveled alone or
with friends. Lastly, according to the distribution of month_stay, we found that the summer
months (May-July) and Christmas holidays (December) were high seasons for Manhattan hotels,
whereas the winter (January and February) was the shoulder season.

(Please insert Table 2 about here)
To detect any potential multi-collinearity problem, we calculated the correlation coefficient
matrix and variance inflation factors (VIFs) based on the benchmark model specifications
(Equation 1). All correlation coefficients had a magnitude of less than 0.2, except for the
coefficient between traveler_type = 2 and 4 (-0.305) and between traveler_type = 4 and 5 (-
0.224). Independent variables had the highest VIF of 2.29 and a mean VIF of 1.75. VIF measures
for all variables were far below the suggested cutoff value of 10 (Dormann, et al., 2013). Another
diagnostic metric of multicollinearity, the condition number, was calculated to be 11.744, which
is lower than the threshold value of 30 (Dormann, et al., 2013). All of these results confirmed the
absence of a multicollinearity problem in the model (Leeflang, Wittink, Wedel, and Naert, 2000).
The detailed results of our multi-collinearity check are available upon request.
4. Empirical Results
4.1. Main results
We started our econometric modeling efforts by estimating the benchmark model without any
interaction terms (Equation 1). Model 1 in Table 3 presents this result. The variable of major
interest, lnmonth, was estimated to be negative and significant, suggesting that the less time there
is between a hotel stay and the time when a corresponding review is posted, the higher the level
of review extremity. This result supports Hypothesis 1. Moreover, neg_valence was estimated to
be positive and statistically significant, illustrating that negative reviews deviated more than
positive reviews from the average rating. The other variable, expertise, was found to be negative

and statistically significant, indicating that more experienced and expert reviewers posted ratings
closer to hotels’ average ratings (i.e., the general consensus) compared to other reviewers. For
reviewers of different travel types, the results indicated that compared to couple travelers
(traveler_type = 1), business travelers (traveler_type = 2) and family travelers (traveler_type = 4)
tended to leave more extreme ratings of their hotel experiences.
(Please insert Table 3 about here)
Model 2 in Table 3 presents the estimates from testing Hypothesis 2 using Equation 2.
The interaction term between lnmonth and neg_valence was estimated to be negative and
statistically significant, whereas lnmonth was found to be insignificant. Therefore, Hypothesis 2
is supported; the significant effect of temporal contiguity exists only for reviewers posting
negative reviews. The estimates of other independent variables were similar to those of Model 1.
To better explain the estimated coefficient of this interaction term, we plotted predicted
deviation-time curves for negative and positive reviews to demonstrate how predicted absolute
deviations (i.e., the degree of review extremity) changed at different levels of temporal
contiguity (i.e., number of months after hotel stay) (see Figure 4). We predicted the deviation
score based on the fixed portion of the model after setting other control variables at their mean
values. As shown in the graph, the curve for negative reviews illustrated a positive relationship
between predicted deviation and temporal contiguity, suggesting that the less time that has
passed between a hotel stay and the time when a review is posted, the greater extremity a review
indicates. As incorporated in the logarithm form of the time variable (month), the slope of this
curve was not constant and became progressively less steep as time went by, indicating that the
deviation of negative reviews shrinks more quickly over time during the first couple months after

a hotel stay. In contrast, the curve for positive reviews remained nearly flat over time, suggesting
that the deviations of positive reviews do not necessarily change over time. The curve for
positive reviews was consistently lower than for negative reviews; thus, positive reviews appear
to be characterized by a lesser degree of review extremity.
(Please insert Figure 4 about here)
We further split the sample into the positive review sample (neg_valence = 0) and the
negative review sample (neg_valence = 1) and re-estimated the model using Equation 1’s
specification. Models 3 and 4 in Table 3 present these results; lnmonth was estimated to be
insignificant in the positive review sample (Model 3) but negative and significant in the negative
review sample (Model 4). Therefore, Hypotheses 2a and 2b are supported. Regarding other
variables, although expertise was found to be negative and significant in both samples, the
estimates of traveler_type were different. In particular, we found that the coefficient of
traveler_type = 2 (business travelers) was negative and significant in Model 3 but positive and
significant in Model 4. This result suggests that compared to couple travelers, business travelers
leave less positive and more negative ratings; in other words, business travelers are more
demanding and tend to post lower ratings when reviewing their hotel stays.
We tested Hypothesis 3 using Equation 3’s specification. Model 5 in Table 3 presents
these estimation results. In the model, lnmonth was estimated to be negative and significant, and
the interaction term between lnmonth and expertise was estimated to be positive and significant.
The results show that temporal contiguity, in general, significantly increases the level of review
extremity, and with an increase in expertise level, the effect of temporal contiguity on review

extremity significantly decreases. More specifically, the marginal effect of lnmonth was
estimated to be -0.0376 (95% CI of -0.0505 and -0.0248) for the least experienced reviewers
(expertise = 0) and -0.0130 (95% CI of -0.0258 and -0.0002) for the most experienced reviewers
(expertise = 6). Therefore, Hypothesis 3 is supported. Figure 5 demonstrates the deviation-time
curves for reviewers at different expertise levels (expertise) based on Model 5 estimates. To
better visualize the curves, only four out of seven levels of expertise were selected. As shown in
the graph, a negative slope characterizes these curves, and the predicted deviation of review
rating (i.e., review extremity) decreased more sharply over time for less experienced reviewers,
suggesting a larger impact of temporal contiguity on their reviews. Also, the curves do not
intersect, and reviews from more experienced reviewers had a consistently lower level of
extremity irrespective of how long the reviewer waited to post a review after his/her hotel stay.
(Please insert Figure 5 about here)
We applied Equation 4’s specification to test Hypothesis 4 in Model 6 (see Table 3).
Among various interaction terms, lnmonth*neg_valence and lnmonth*expertise*neg_valence
were positive. This result suggests that the role of expertise level in moderating the relationship
between temporal contiguity and review extremity is only significant for reviewers who post
negative reviews. Therefore, Hypothesis 4 is supported. Figure 6 presents the deviation-time
curves for negative and positive reviews from reviewers at different expertise levels (expertise)
to better interpret the three-way interaction. The left pane presents the curves for negative
reviews, whereas the right pane presents those for positive reviews. In general, these curves
confirm the major findings from Figures 3 and 4. More specifically, the deviation-time curves for
positive reviews were consistently flat and therefore suggest that a positive relationship

characterizes only negative reviews between temporal contiguity and review extremity. Also, the
left pane of the graph shows that the slope of deviation-time curves for negative reviews was
steeper for reviewers with a lower level of expertise. We also split the sample into positive
reviews and negative reviews and re-estimated the model using Equation 3. Models 7 and 8 in
Table 3 present these results, specifically that lnmonth and lnmonth*expertise were statistically
significant for negative reviews only. These results support Hypotheses 4a and 4b.
(Please insert Figure 6 about here)
4.2. Robustness check of results
We conducted a series of robustness checks to evaluate whether our major results would change
when applying different sample coverage, variable specifications, estimation methods, and
endogeneity specifications to replicate results from Models 1 to 8 (see Table 3). First, because
96.86% of reviewers left reviews within five months of their stay, we limited our sample to these
reviewers and re-ran the model. Second, we constructed an alternative median-centered deviation
measure, deviation2, as the absolute difference between the review rating and the median score
of all ratings for a hotel. We used neg_valence2 to indicate if the review rating was lower than
the median score. Third, we used ordinary least squares regression by using a set of dummies to
capture hotel-specific effects; then, we estimated the variances using the clustered sandwich
estimator to allow for intra-hotel correlation. Due to space limitations, the results of these
robustness checks are presented in the supplementary materials. In general, we found that our
major results still held, and all hypotheses we proposed could not be rejected.

We also conducted a robustness check after correcting for endogeneity bias of temporal
contiguity, lnmonth. We suspected a simultaneous relationship between temporal contiguity and
review extremity, which could potentially lead to an endogeneity problem with lnmonth as an
independent variable. An ideal instrumental variable (IV) to solve this problem should be highly
correlated with the endogenous variable but not associated with the dependent variable (Greene,
2007). We chose variable mobile as the IV for lnmonth, which indicated if the review was posted
via a mobile device. This IV is closely related to temporal contiguity because it is easier and
more convenient for customers to post online reviews instantly from mobile devices such as
smartphones and iPads (Wang, Shen, and Sun, 2013). On the other hand, this IV was not
expected to determine the dependent variable because mobile devices are used to post reviews
after an individual has already judged his/her experience; hence, it is unlikely to influence a
reviewer’s evaluation of the experience. We found a statistically significant difference in
temporal contiguity between mobile-posting and non-mobile-posting groups. In general,
travelers who posted reviews via mobile devices took significantly less time to post after their
actual hotel stay. Therefore, we used mobile as an IV to re-estimate the model using the
conventional two-step method (Wooldridge, 2010). In the first step, we regressed lnmonth on
mobile and all other exogenous variables and obtained the fitted value of lnmonth from the
regression. In the second step, we used the fitted value of lnmonth from the first step to estimate
the HLM. The results of these models are presented in the supplementary materials. All proposed
research hypotheses could not be rejected at the 0.01 significance level, and we found that the
magnitudes of lnmonth-related estimates were larger than their counterparts reported in Table 3.
4.3. Results for sub-ratings

TripAdvisor also allows users to post sub-ratings about particular aspects of their hotel stay. We
used Equations 1-4 to estimate the models for four experience-based sub-ratings: room rating,
sleep quality rating, cleanliness rating, and service rating. Note that not all reviewers posted all
sub-ratings. In our sample, we retained hotels that received at least 20 sub-ratings in 2015. Table
4 presents brief results on the estimates of variables directly related to the temporal contiguity
measure (lnmonth). Hypotheses 1 and 2 were supported in all four sub-ratings, with the positive
relationship between temporal continuity and review extremity present exclusively in negative
reviews. Judging by the magnitudes of estimated coefficients, we found the effect of temporal
contiguity to be largest for service rating and smallest for room rating. However, Hypothesis 3
was supported in room rating only, and Hypothesis 4 was not supported in any sub-rating results.
(Please insert Table 4 about here)
4.4. Results with day-level temporal contiguity proxy
In the last step of empirical analysis, we tested all hypotheses using the day-level temporal
contiguity proxy, lnday, for reviews posted in the same month of hotel stay. Table 5 presents the
brief estimation results. The variable of interests, lnday, was estimated to be statistically
significant and negative in Model 9, and Hypothesis 1 was supported. Also, results in Models 10-
12 indicate that the estimated coefficient of lnday was significant and negative for negative
reviews only, lending supports to Hypothesis 2. However, the estimated coefficients of
lnday*expertise and lnday*expertise*neg_valence were not statistically significant in Models 13-
16. Hence, Hypotheses 3 and 4 were rejected based on day-level temporal contiguity proxy.
(Please insert Table 5 about here)

5. Conclusion
Using 79,107 TripAdvisor hotel reviews for 364 Manhattan hotels in 2015, this study applied an
HLM to explore the relationship between temporal contiguity and review extremity. The
estimation results suggested that the less time a traveler took to post a review after his/her hotel
stay, the more extreme the review rating was compared to the hotel’s average rating. Moreover,
we proposed and empirically confirmed two moderating factors: experience valence and
reviewer expertise. More specifically, the effect of temporal contiguity on review extremity was
statistically significant for reviewers with negative experiences but insignificant for those with
positive experiences. As another significant moderating factor, reviewer expertise demonstrated
a marginal effect of temporal contiguity on review extremity, an effect that was found to
decrease as reviewers’ level of expertise increased. Lastly, the results demonstrated that the
moderating effect of reviewer expertise was evident only for reviewers with negative experiences.
Our major results were generally confirmed after several types of robustness checks, including
choosing a different contiguity time span, using alternative rules to define positive/negative
valence, applying different estimation methods, and correcting for endogeneity bias.
6. Theoretical Implications
The current research contributes to the extant hospitality management research in three unique
ways. Despite the growing attention towards online review management in hospitality literature
(Sparks, So, and Bradley, 2016; Xie, Zhang, Zhang, Singh, and Lee, 2016; Ye, et al., 2009), no
known studies hitherto have examined the effect of temporal contiguity on hotel ratings. To fill
this research gap, the current study contributes to the hospitality literature on online review

management by exploring the question of how temporal contiguity influences review extremity.
Previous marketing research has found that extreme reviews play a significant role in shaping
other customers’ attitudes towards a reviewed business (Schlosser, 2011). However, the existing
body of research does not explain when and why such extreme reviews are posted in the first
place. In addressing this question, our research indicated that the timing of review postings could
be a driving factor behind review extremity. Based on CLT, our research found that temporal
contiguity of review postings can actually amplify the extremity of their contents.
Moreover, our research revealed two moderating factors that conditioned the effect of
temporal contiguity on review extremity: experience valence and reviewer expertise. Based on
CLT, previous research has found an association between temporal contiguity and a cognitive
focus on negative attributes (Eyal, et al., 2008; Eyal, et al., 2004; Hong and Lee, 2010).
Consistent with the literature (Eyal, et al., 2008; Eyal, et al., 2004; Zhang, Wu, and Mattila,
2016), we found that temporal contiguity amplified review extremity only in negative service
experiences. For positive service experiences, however, the association between temporal
contiguity and a negativity focus was found to attenuate consumers’ favorable perceptions
toward the service experience, resulting in a significant effect of temporal contiguity on review
extremity.
Another important moderating factor for the temporal contiguity cue effect is reviewer
expertise (Vermeulen and Seegers, 2009; Zhang, Wu, et al., 2016; Zhang, et al., 2010). Previous
studies in managerial decision making have suggested that experts and novices do not employ
the same cognitive approach (Dew et al., 2009). This line of research indicates that experts tend
to adhere to a more systematic and analytical processing style, which can help them overcome
the peripheral influence of irrelevant contextual factors such as temporal contiguity (Dew, et al.,

2009; Klein, 1999; Sonnentag, et al., 2006). Our results confirmed this theory. In fact, as
revealed by our findings, reviews posted by expert reviewers tended to be more stable and
consistent along the temporal dimension; that is, for reviewers with high levels of expertise, the
effect of temporal contiguity on review extremity was smaller than for novice reviewers.
7. Managerial Implications
Findings from our research provide direct and practical insights for social media
managers and marketers in the hotel industry. In today’s business, hotels’ online reputations are
critical; they play a dominant role in consumers’ decision-making process. Management of a
hotel’s online reputation is thus becoming a central focus for many hotel brands. Reflecting such
a practical urgency, the current research has examined a managerial question that has not been
adequately answered by the existing hospitality literature: does the timing of an online matter? If
yes, how should hotels cope with the timing issue in online reviews?
First of all, our research revealed that timing is indeed an important factor on which hotel
managers should focus. As demonstrated by our study, the time at which a review is posted can
significantly influence the review contents. More specifically, for negative service experiences,
consumers are more likely to post an extremely unfavorable review if they do so immediately
after their unsatisfactory experience. In fact, many unhappy customers who had bad experiences
would actually immediately (and voluntarily) post their negative reviews right after their
consumption experiences to vent or seek compensations. Our study results suggest that such
reviews tend to be extremely negative and can cause serious damages to the business. Therefore,
in order to avoid those extremely low ratings, it is imperative for hotels to implement a
quality/satisfaction check before guests leave the property. For example, front desk employees

should be trained to inquire guest satisfaction during the checkout process with the intention to
correct any problems on the spot. They need to provide a venue for unhappy customers to vent
and/or offer a solution to correct problems. Moreover, express checkout is gaining popularity in
the lodging industry. Although it provides convenience to hotel guests, it removes the human
interaction and consequently eliminates an opportunity for frontline employees to turn an angry
customer into a happy one. Therefore, hotels must closely monitor customer satisfaction during
the entire consumption experiences and put best efforts to recover any service failures. They
need to train all employees who may interact with guest (e.g. front desk agents, housekeepers,
servers, etc.), and clearly communicate the service complaint procedure and satisfaction
guarantee to the guests. For example, front desk agents may communicate about the service
satisfaction guarantee with hotel guests during the check-in process. On the express check-out
folio, hotels can print a statement reminding customers to voice any complaint before they leave
the property. Hotels can also implement a quality monitor system by sending guests text
messages during their stays. In the text message, hotels may check guest satisfaction, provide a
venue for customers to file a complaint, and initiate a service recovery as needed.
Secondly, our study suggests that when soliciting customer feedback, timing is important.
While in practice most hotels and review sites tend to invite customers to post online reviews
right after a trip (Shu, 2013), our research would suggest otherwise. Our study discovers that for
positive service experiences, posting time does not have a major impact on review extremity.
However, for negative service experiences, guests tend to give extremely low ratings if they do
so immediately after their unhappy experiences. But as time goes by, the review contents of a
negative experience tend to be more neutral. Based on such findings, we would suggest that an

acceptable level of delay of online review solicitation might actually help protect hotels from the
reputational damage caused by extremely negative reviews.
In addition to recommending that hotel managers impose a moderate delay after
consumers’ hotel stays when soliciting online reviews, we would also suggest that practitioners
seamlessly integrate offline and online strategies in managing consumers’ post-experience
perceptions and behaviors. As suggested by the “tip of the iceberg” rule, very few consumers
voice their dissatisfaction after a negative service experience; instead, customers will choose to
avoid the business altogether in the future (Zeithaml, Bitner, and Gremler, 2012). To that end,
our recommended approach in delaying review solicitation may cause losses in a hotel’s
customer base due to delays in social listening. In a preemptive effort to mitigate potential
negative consequences of our previous managerial suggestion, we recommend that practitioners
first gauge customer satisfaction through an email survey before sending out a review solicitation
email. In the survey email, managers should clearly communicate the satisfaction guarantee
promise and encourage customers to contact the business to solve any problem offline. We
would encourage hotel managers to be as proactive and thorough as possible in helping
consumers recover from poor hotel experiences. This way, when delayed review solicitations are
sent, dissatisfied consumers will have hopefully recovered already from their negative service
experiences and may be less likely to post extremely negative reviews. If recovered customers do
share their experiences via online reviews, they should be encouraged to tell the whole story,
including service failure and the recovery process. This way, firms may be able to turn a service
failure into a marketing opportunity to gain or maintain a positive online reputation and
consumer favorability.

Lastly, the implications of our research also extend to brand-level management as well as the
rating/ranking design of travel review sites. As suggested by our findings, the contents of online
reviews can be influenced by many subjective and contextual factors including the time when
reviews are posted. Thus, a single number of online ratings may not be an objective and unbiased
indication of service quality. Caution is therefore warranted when using online review data to
monitor property-level service consistency. Specifically, brand managers should be aware of the
possible effect of temporal contiguity in shifting review ratings, and some adjustments can be
made to correct this bias.
8. Limitations and Future Research
There are some limitations to our research. First, due to limited data, we were unable to
incorporate some important control variables in the model, such as social networking (Luo and
Zhong, 2015) and past travel experiences (Weaver, Weber, and McCleary, 2007). Second, our
online review data were cross-sectional, and we were unable to track how individual reviewers’
evaluations and attitudes changed at multiple points over time. Third, the data allowed us to
examine temporal contiguity over a maximum time span of 11 months. Hence, we could not
uncover the impact of temporal contiguity over a longer time period, such as multiple years or
decades. We are therefore calling for future research efforts to evaluate the impact of temporal
contiguity using a longitudinal dataset and covering a longer period of time.
References

Buehner, M. J., May, J., 2003. Rethinking temporal contiguity and the judgement of causality:
Effects of prior knowledge, experience, and reinforcement procedure. The Quarterly
Journal of Experimental Psychology Section A 56 (5), 865-890.
Burtch, G., Hong, Y., 2014. What happens when word of mouth goes mobile? In Proceedings of
the International Conference on Information Systems (ICIS). Auckland, New Zealand.
Chen, Z., Lurie, N. H., 2013. Temporal contiguity and negativity bias in the impact of online
word of mouth. Journal of Marketing Research 50 (4), 463-476.
Chu, S.-C., Kim, Y., 2011. Determinants of consumer engagement in electronic word-of-mouth
(eWOM) in social networking sites. International journal of Advertising 30 (1), 47-75.
Del Chiappa, G., Dall’Aglio, S., 2012. Factors influencing Travellers’ e-Ratings and e-
Complaints about Hotel Services: Insights from an Italian Tourism Destination. In M.
Fuchs, F. Ricci & L. Cantoni (Eds.), Information and Communication Technologies in
Tourism 2012: Proceedings of the International Conference in Helsingborg, Sweden,
January 25–27, 2012 (pp. 448-459). Vienna: Springer Vienna.
Dew, N., Read, S., Sarasvathy, S. D., Wiltbank, R., 2009. Effectual versus predictive logics in
entrepreneurial decision-making: Differences between experts and novices. Journal of
business venturing 24 (4), 287-309.
Dhar, R., Kim, E. Y., 2007. Seeing the forest or the trees: Implications of construal level theory
for consumer choice. Journal of Consumer Psychology 17 (2), 96-100.
Dormann, C. F., Elith, J., Bacher, S., Buchmann, C., Carl, G., Carré, G., Marquéz, J. R. G.,
Gruber, B., Lafourcade, B., Leitão, P. J., 2013. Collinearity: A review of methods to deal
with it and a simulation study evaluating their performance. Ecography 36 (1), 27-46.

Einhorn, H. J., Hogarth, R. M., 1986. Judging probable cause. Psychological Bulletin 99 (1), 3-
19.
Eyal, T., Liberman, N., Trope, Y., 2008. Judging near and distant virtue and vice. Journal of
experimental social psychology 44 (4), 1204-1209.
Eyal, T., Liberman, N., Trope, Y., Walther, E., 2004. The pros and cons of temporally near and
distant action. Journal of personality and social psychology 86 (6), 781-795.
Fang, B., Ye, Q., Kucukusta, D., Law, R., 2016. Analysis of the perceived value of online
tourism reviews: Influence of readability and reviewer characteristics. Tourism
Management 52 (498-506.
Fiedler, K., 2007. Construal level theory as an integrative framework for behavioral decision-
making research and consumer psychology. Journal of Consumer Psychology 17 (2),
101-106.
Goffman, E., 1955. On face-work: An analysis of ritual elements in social interaction. Psychiatry
18 (3), 213-231.
Greene, W. H., 2007. Econometric Analysis (6th ed.). Upper Saddle River, NJ: Prentice Hall.
Hong, J., Lee, A. Y., 2010. Feeling mixed but not torn: The moderating role of construal level in
mixed emotions appeals. Journal of Consumer Research 37 (3), 456-472.
Hox, J. J., Moerbeek, M., van de Schoot, R., 2010. Multilevel Analysis: Techniques and
Applications (2nd ed.). New York: Routledge.
Jensen Schau, H., Gilly, M. C., 2003. We are what we post? Self-presentation in personal web
space. Journal of Consumer Research 30 (3), 385-404.
Kelley, H. H., 1973. The processes of causal attribution. American psychologist 28 (2), 107-128.

Kim, K., Zhang, M., Li, X., 2008. Effects of temporal and social distance on consumer
evaluations. Journal of Consumer Research 35 (4), 706-713.
Klein, G. A., 1999. Sources of Power: How People Make Decisions. Cambridge, MA: MIT press.
Koh, N. S., Hu, N., Clemons, E. K., 2010. Do online reviews reflect a product’s true perceived
quality? An investigation of online movie reviews across cultures. Electronic Commerce
Research and Applications 9 (5), 374-385.
Kummer, H., 1995. Causal Knowledge in Animals. In D. Sperber, D. Premack & A. Premack
(Eds.), Causal Cognition: A Multidisciplinary Debate, Symposia of the Fyssen
Foundation (pp. 26-39). Oxford: Clarendon Press.
Ladhari, R., Michaud, M., 2015. eWOM effects on hotel booking intentions, attitudes, trust, and
website perceptions. International Journal of Hospitality Management 46 (36-45.
Lampel, J., Bhalla, A., 2007. The role of status seeking in online communities: Giving the gift of
experience. Journal of Computer‐Mediated Communication 12 (2), 434-455.
Lee, S. K., Jang, S., 2015. Conditional agglomeration externalities in lodging markets. Journal of
Hospitality & Tourism Research 39 (4), 540-559.
Leeflang, P. S. H., Wittink, D. R., Wedel, M., Naert, P. A., 2000. Building Models for Marketing
Decisions. Boston: Kluwer Academic.
Liang, S., Schuckert, M., Law, R., 2017. Multilevel analysis of the relationship between type of
travel, online ratings, and management response: Empirical evidence from international
upscale hotels. Journal of Travel & Tourism Marketing 34 (2), 239-256.
Liberman, N., Trope, Y., 2003. Construal level theory of intertemporal judgment and decision.
Time and decision 245-276.

Liberman, N., Trope, Y., Wakslak, C., 2007. Construal level theory and consumer behavior.
Journal of Consumer Psychology 17 (2), 113-117.
Liu, X., Schuckert, M., Law, R., 2016. Online incentive hierarchies, review extremity, and
review quality: Empirical evidence from the hotel sector. Journal of Travel & Tourism
Marketing 33 (3), 279-292.
Ludwig, S., De Ruyter, K., Friedman, M., Brüggen, E. C., Wetzels, M., Pfann, G., 2013. More
than words: The influence of affective content and linguistic style matches in online
reviews on conversion rates. Journal of marketing 77 (1), 87-103.
Luo, Q., Zhong, D., 2015. Using social network analysis to explain communication
characteristics of travel-related electronic word-of-mouth on social networking sites.
Tourism Management 46 (274-282.
Nam, J., Ekinci, Y., Whyatt, G., 2011. Brand equity, brand loyalty and consumer satisfaction.
Annals of Tourism Research 38 (3), 1009-1030.
Office of the New York State Comptroller. 2016. The Hotel Industry in New York City. In.
Pan, B., Yang, Y., 2016. Monitoring and forecasting tourist activities with big data. In M. Uysal;,
Z. Schwartz; & E. Sirakaya-Turk (Eds.), Management science in hospitality and tourism:
Theory, practice, and applications (pp. 43-62). Watertown, NJ: Apple Academic Press.
Peiró-Signes, A., Segarra-Oña, M.-d.-V., Verma, R., Mondéjar-Jiménez, J., Vargas-Vargas, M.,
2014. The impact of environmental certification on hotel guest ratings. Cornell
Hospitality Quarterly 55 (1), 40-51.
Purnawirawan, N., De Pelsmacker, P., Dens, N., 2012. Balance and sequence in online reviews:
How perceived usefulness affects attitudes and intentions. Journal of Interactive
Marketing 26 (4), 244-255.

Reed, P., 1992. Effect of a signalled delay between an action and outcome on human judgement
of causality. Quarterly Journal of Experimental Psychology: Section B 44 (2), 81-100.
Reed, P., 1999. Role of a stimulus filling an action-outcome delay in human judgments of causal
effectiveness. Journal of Experimental Psychology: Animal Behavior Processes 25 (1),
92-102.
Salanova, M., Agut, S., Peiró, J. M., 2005. Linking organizational resources and work
engagement to employee performance and customer loyalty: the mediation of service
climate. Journal of applied Psychology 90 (6), 1217-1227.
Schlosser, A. E., 2011. Can including pros and cons increase the helpfulness and persuasiveness
of online reviews? The interactive effects of ratings and arguments. Journal of Consumer
Psychology 21 (3), 226-239.
Shanks, D. R., Pearson, S. M., Dickinson, A., 1989. Temporal contiguity and the judgement of
causality by human subjects. The Quarterly Journal of Experimental Psychology 41 (2),
139-159.
Shu, C., 2013. TripAdvisor review express service lets businesses bulk email visitors to ask for
reviews. In.
Sonnentag, S., Niessen, C., Volmer, J., 2006. Expertise in software design. In K. A. Ericsson, N.
Charness, P. J. Feltovich & R. R. Hoffman (Eds.), The Cambridge Handbook of
Expertise and Expert Performance (pp. 373-388). Cambridge, U.K.: Cambridge
University Press.
Sparks, B. A., Browning, V., 2011. The impact of online reviews on hotel booking intentions and
perception of trust. Tourism Management 32 (6), 1310-1323.

Sparks, B. A., So, K. K. F., Bradley, G. L., 2016. Responding to negative online reviews: The
effects of hotel responses on customer inferences of trust and concern. Tourism
Management 53 (74-85.
Topolinski, S., Reber, R., 2010. Immediate truth–Temporal contiguity between a cognitive
problem and its solution determines experienced veracity of the solution. Cognition 114
(1), 117-122.
Tourism Economics and NYC & Company. 2016. Economic Impact of New York City Tourism.
In.
Trope, Y., Liberman, N., 2003. Temporal construal. Psychological review 110 (3), 403-421.
Trope, Y., Liberman, N., 2010. Construal-level theory of psychological distance. Psychological
review 117 (2), 440-463.
Trope, Y., Liberman, N., Wakslak, C., 2007. Construal levels and psychological distance: Effects
on representation, prediction, evaluation, and behavior. Journal of Consumer Psychology
17 (2), 83-95.
Tsao, W.-C., Hsieh, M.-T., Shih, L.-W., Lin, T. M. Y., 2015. Compliance with eWOM: The
influence of hotel reviews on booking intention from the perspective of consumer
conformity. International Journal of Hospitality Management 46 (99-111.
United States Census Bureau. 2016. American FactFinder – Results. In.
Vermeulen, I. E., Seegers, D., 2009. Tried and tested: The impact of online hotel reviews on
consumer consideration. Tourism Management 30 (1), 123-127.
Viglia, G., Viglia, G., Minazzi, R., Minazzi, R., Buhalis, D., Buhalis, D., 2016. The influence of
e-word-of-mouth on hotel occupancy rate. International Journal of Contemporary
Hospitality Management 28 (9), 2035-2051.

Wang, N., Shen, X.-L., Sun, Y., 2013. Transition of electronic word-of-mouth services from web
to mobile context: A trust transfer perspective. Decision Support Systems 54 (3), 1394-
1403.
Weaver, P. A., Weber, K., McCleary, K. W., 2007. Destination evaluation: The role of previous
travel experience and trip characteristics. Journal of Travel Research 45 (3), 333-344.
Wooldridge, J. M., 2010. Econometric analysis of cross section and panel data. Massachusetts:
MIT press.
Wu, L., Shen, H., Li, M., Deng, Q., 2017. Sharing information now vs later: The effect of
temporal contiguity cue and power on consumer response toward online reviews.
International Journal of Contemporary Hospitality Management 29 (2), 648-668.
Xiang, Z., Krawczyk, M., 2016. What does hotel location mean for the online consumer? Text
analytics using online reviews. In Information and Communication Technologies in
Tourism 2016 (pp. 383-395): Springer.
Xie, K. L., Zhang, Z., Zhang, Z., Singh, A., Lee, S. K., 2016. Effects of managerial response on
consumer eWOM and hotel performance: Evidence from TripAdvisor. International
Journal of Contemporary Hospitality Management 28 (9), 2013-2034.
Yang, Y., Mao, Z., 2017. Do independent hotels benefit from the presence of branded ones?
Journal of Business Research 76 (108-117.
Yang, Y., Mao, Z., Tang, J., 2017. Understanding guest satisfaction with urban hotel location.
Journal of Travel Research 0047287517691153.
Ye, Q., Law, R., Gu, B., 2009. The impact of online user reviews on hotel room sales.
International Journal of Hospitality Management 28 (1), 180-182.

Yoo, K.-H., Sigala, M., Gretzel, U., 2016. Exploring TripAdvisor. In R. Egger, I. Gula & D.
Walcher (Eds.), Open Tourism: Open Innovation, Crowdsourcing and Co-Creation
Challenging the Tourism Industry (pp. 239-255). Berlin, Heidelberg: Springer Berlin
Heidelberg.
Yoo, K. H., Gretzel, U., 2008. What motivates consumers to write online travel reviews?
Information Technology & Tourism 10 (4), 283-295.
Zeithaml, V. A., Bitner, M. J., Gremler, D. D., 2012. Services Marketing: Integrating Customer
Focus across The Firm (6th ed.): McGraw Hill.
Zhang, L., Wu, L., Mattila, A. S., 2016. Online reviews: The role of information load and
peripheral factors. Journal of Travel Research 55 (3), 299-310.
Zhang, Z., Ye, Q., Law, R., Li, Y., 2010. The impact of e-word-of-mouth on the online
popularity of restaurants: A comparison of consumer reviews and editor reviews.
International Journal of Hospitality Management 29 (4), 694-700.
Zhang, Z., Zhang, Z., Yang, Y., 2016. The power of expert identity: How website-recognized
expert reviews influence travelers' online rating behavior. Tourism Management 55 (15-
24.

Figure 1. Conceptual research framework.
Experience Valence
Reviewer Expertise
Temporal Contiguity
Review Extremity
H1
H4
H3
H2

Figure 2. Map of Manhattan hotels in the sample.

Figure 3. Distribution of temporal contiguity in the sample.

Figure 4. Predicted deviation-time curves for negative and positive reviews.

Figure 5. Predicted deviation-time curves for reviewers at different expertise levels.

Figure 6. Predicted deviation-time curves for negative and positive reviews from reviewers at different
expertise levels.

Table 1. Descriptive statistics of continuous variables in the empirical models.
Variable
Observation
s
Mean Median Std. Dev. Min Max
deviation 79,107 0.724 0.581 0.604 0 3.757
expertise 79,107 2.356 2 2.033 0 6
lnmonth 79,107 0.341 0 0.547 0 2.485
lnday 53,866 2.798 2.944 0.548 0 3.434

Table 2. Descriptive statistics of categorical variables in the empirical models.
Variable
Freq.
Percent
Cumulative
Percent
neg_valence
0=Positive
47,513
60.06
39.94
1=Negative
31,594
39.94
100.00
traveler_type
1=Couples
27,632
34.93
34.93
2=Business
15,525
19.63
54.56
3=Solo
4,945
6.25
60.81
4=Family
21,790
27.54
88.35
5=With friends
9,215
11.65
100.00
month_stay
Jan
5,041
6.37
6.37
Feb
4,862
6.15
12.52
Mar
6,049
7.65
20.17
Apr
6,634
8.39
28.55
May
7,720
9.76
38.31
Jun
7,135
9.02
47.33
Jul
7,412
9.37
56.70
Aug
6,968
8.81
65.51
Sep
6,211
7.85
73.36
Oct
7,139
9.02
82.38
Nov
6,340
8.01
90.40
Dec
7,596
9.60
100.00

Table 3. Estimation results of all empirical models.
Model 1
Model 2
Model 3
Model 4
Model 5
Model 6
Model 7
Model 8
All
reviews
All
reviews
Positive
reviews
Negative
reviews
All
reviews
All
reviews
Positive
reviews
Negative
reviews
lnmonth
-0.0261***
0.00155
-0.00197
-0.0573***
-0.0376***
-0.00747
-0.00526
-0.0899***
(0.004)
(0.005)
(0.003)
(0.009)
(0.007)
(0.006)
(0.004)
(0.019)
lnmonth*
neg_valence
-0.0688***
-0.0917***
(0.011)
(0.024)
lnmonth*expertise
0.00410**
0.00123
0.00125
0.0105**
(0.002)
(0.002)
(0.001)
(0.005)
lnmonth*expertise*
neg_valence
0.0111**
(0.005)
expertise*neg_vale
nce
-0.0354***
(0.007)
neg_valence
0.312***
0.336***
0.312***
0.0662***
(0.030)
(0.031)
(0.030)
(0.007)
expertise
-0.0343***
-0.0343***
-0.0173***
-0.0538***
-0.0356***
0.0107
-0.0177***
-0.0569***
(0.002)
(0.002)
(0.002)
(0.003)
(0.002)
(0.009)
(0.002)
(0.003)
traveler_type=2
0.0653***
0.0649***
-0.0177***
0.153***
0.0650***
0.0183***
-0.0178***
0.152***
(0.007)
(0.007)
(0.006)
(0.013)
(0.007)
(0.005)
(0.006)
(0.013)
traveler_type=3
0.00890
0.00918
0.0115***
0.00330
0.00903
-0.00398
0.0115***
0.00352
(0.009)
(0.009)
(0.004)
(0.018)
(0.009)
(0.006)
(0.004)
(0.018)
traveler_type=4
0.0180***
0.0181***
0.0133***
0.0200*
0.0179***
0.419***
0.0133***
0.0198*
(0.005)
(0.005)
(0.004)
(0.012)
(0.005)
(0.040)
(0.004)
(0.012)
traveler_type=5
-0.00456
-0.00426
0.00907**
-0.0328**
-0.00464
-0.0212***
0.00903**
-0.0328**
(0.006)
(0.006)
(0.004)
(0.016)
(0.006)
(0.003)
(0.004)
(0.016)
constant
0.956***
0.970***
0.679***
1.033***
0.960***
0.608***
0.680***
1.042***
(0.027)
(0.027)
(0.015)
(0.027)
(0.027)
(0.017)
(0.015)
(0.027)
month_stay
Controlled
Controlled
Controlled
Controlled
Controlled
Controlled
Controlled
Controlled
lnσ
μ
-1.903***
-1.906***
-1.480***
-1.318***
-1.903***
-1.918***
-1.480***
-1.318***
(0.050)
(0.050)
(0.041)
(0.035)
(0.050)
(0.051)
(0.041)
(0.035)
lnσ
ɛ
-0.582***
-0.582***
-1.264***
-0.276***
-0.582***
-0.584***
-1.264***
-0.276***
(0.014)
(0.014)
(0.068)
(0.009)
(0.014)
(0.014)
(0.068)
(0.009)
Number of reviews
79107
79107
47513
31594
79107
79107
47513
31594
Number of hotels
364
364
364
364
364
364
364
364
AIC
133378.0
133291.4
16215.1
73023.0
133375.6
133028.7
16216.1
73018.9
BIC
133572.8
133495.6
16390.5
73190.2
133579.7
133260.7
16400.3
73194.5
Log likelihood
-66668.0
-66623.7
-8087.5
-36491.5
-66665.8
-66489.4
-8087.1
-36488.4
(Notes: *** indicates significance at the 0.01 level; ** indicates significance at the 0.05 level; * indicates
significance at the 0.10 level. Robust standard errors are presented in parentheses. Estimates of month_stay
dummies are not presented for brevity purposes.)

Table 4. Brief results of key estimates based on sub-ratings.
Room
Sleep
Quality
Cleanliness
Service
Equation 1
lnmonth
-0.0168***
-0.0209***
-0.0130**
-0.0286***
(0.005)
(0.005)
(0.005)
(0.004)
AIC
68930.3
68726.9
59266.1
138343.4
BIC
69111.4
68907.9
59447.2
138538.2
Equation 2
lnmonth
-0.00705
-0.000240
0.00698
0.00310
(0.006)
(0.005)
(0.005)
(0.004)
lnmonth*neg_valence
-0.0308***
-0.0553***
-0.0520***
-0.0827***
(0.012)
(0.015)
(0.017)
(0.012)
AIC
68927.7
68705.2
59235.2
138226.1
BIC
69117.4
68894.9
59425.0
138430.1
Equation 3
lnmonth
-0.0341***
-0.0129
-0.0152*
-0.0255***
(0.009)
(0.009)
(0.008)
(0.008)
lnmonth*expertise
0.00660**
-0.00305
0.000850
-0.00109
(0.003)
(0.003)
(0.002)
(0.002)
AIC
68926.9
68727.8
59268.0
138345.1
BIC
69116.6
68917.4
59457.8
138549.1
Equation 4
lnmonth
-0.0124
-0.000227
0.00532
0.00218
(0.009)
(0.002)
(0.007)
(0.005)
lnmonth*neg_valence
-0.0568**
-0.0329
-0.0703**
-0.0956***
(0.026)
(0.033)
(0.033)
(0.027)
lnmonth*expertise
-0.00144
-0.0153*
-0.0278***
-0.0442***
(0.007)
(0.008)
(0.010)
(0.007)
lnmonth*expertise* neg_valence
0.0112
-0.00422
0.00972
0.00905
(0.007)
(0.009)
(0.008)
(0.006)
AIC
68923.2
68676.5
59142.7
137856.9
BIC
69138.8
68892.0
59358.3
138088.7
Number of reviews
41062
40968
41121
78661
Number of hotels
325
323
317
364
(Notes: *** indicates significance at the 0.01 level; ** indicates significance at the 0.05 level; * indicates
significance at the 0.10 level. Robust standard errors are presented in parentheses. Only estimates related to
lnmonth are presented for brevity purposes.)

Table 5. Brief estimation results of empirical models with day-level temporal contiguity proxies
Model 9
Model 10
Model 11
Model 12
Model 13
Model 14
Model 15
Model 16
All
reviews
All
reviews
Positive
reviews
Negative
reviews
All
reviews
All
reviews
Positive
reviews
Negative
reviews
lnday
-0.0115**
0.000045
2
0.000874
-0.0232**
-0.00974
0.00418
0.00270
-0.0236
(0.005)
(0.004)
(0.003)
(0.011)
(0.008)
(0.006)
(0.005)
(0.020)
lnday* neg_valence
-0.0287***
-0.0396*
(0.011)
(0.022)
lnday*expertise
-0.000733
-0.00149
-0.000856
0.000151
(0.002)
(0.002)
(0.001)
(0.005)
lnday*expertise*ne
g_valence
0.00361
(0.006)
Number of reviews
51311
51311
30762
20549
51311
51311
30762
20549
Number of hotels
363
363
363
361
363
363
363
361
AIC
88005.9
87998.6
9763.4
48519.1
88007.8
87796.9
9765.0
48521.1
BIC
88191.7
88193.2
9930.1
48677.7
88202.4
88018.0
9940.0
48687.7
Log likelihood
-43982.0
-43977.3
-4861.7
-24239.6
-43981.9
-43873.4
-4861.5
-24239.6
(Notes: *** indicates significance at the 0.01 level; ** indicates significance at the 0.05 level; * indicates
significance at the 0.10 level. Robust standard errors are presented in parentheses. Estimates of month_stay
dummies are not presented for brevity purposes. Only estimates related to lnday are presented for brevity
purposes.)

Supplementary Material

Table S1. Estimation results of empirical models with low temporal contiguity
(Notes: *** indicates significance at 0.01, ** indicates significance at 0.05, * indicates significance at 0.1. Robust
standard errors are presented in parentheses. Estimates of month_stay dummies are not presented for brevity
purposes)
Model S1
Model S2
Model S3
Model S4
Model S5
Model S6
Model S7
Model S8
All
reviews
All
reviews
Positive
reviews
Negative
reviews
All
reviews
All
reviews
Positive
reviews
Negative
reviews
lnmonth
-0.0295***
-0.00133
-0.00362
-0.0616***
-0.0416***
-0.00835
-0.00411
-0.103***
(0.005)
(0.005)
(0.003)
(0.011)
(0.008)
(0.008)
(0.005)
(0.023)
lnmonth*neg_valen
ce
-0.0727***
-0.0994***
(0.013)
(0.030)
lnmonth*expertise
0.00450**
0.000354
0.000197
0.0140**
(0.002)
(0.002)
(0.002)
(0.006)
lnmonth*expertise*
neg_valence
0.0142**
(0.007)
expertise*neg_vale
nce
-0.0360***
(0.007)
neg_valence
0.317***
0.337***
0.317***
0.421***
(0.030)
(0.031)
(0.030)
(0.040)
expertise
-0.0346***
-0.0346***
-0.0175***
-0.0542***
-0.0358***
-0.0210***
-0.0175***
-0.0577***
(0.002)
(0.002)
(0.002)
(0.003)
(0.002)
(0.003)
(0.002)
(0.003)
traveler_type =2
0.0662***
0.0658***
-0.0175***
0.154***
0.0661***
0.0673***
-0.0175***
0.154***
(0.007)
(0.007)
(0.006)
(0.013)
(0.007)
(0.007)
(0.006)
(0.013)
traveler_type =3
0.00988
0.00997
0.0112**
0.00682
0.0100
0.0115
0.0112**
0.00723
(0.009)
(0.009)
(0.004)
(0.019)
(0.009)
(0.009)
(0.004)
(0.019)
traveler_type =4
0.0174***
0.0175***
0.0124***
0.0201*
0.0173***
0.0176***
0.0124***
0.0198*
(0.005)
(0.005)
(0.004)
(0.012)
(0.005)
(0.005)
(0.004)
(0.012)
traveler_type =5
-0.00632
-0.00626
0.00707
-0.0361**
-0.00638
-0.00591
0.00706
-0.0360**
(0.006)
(0.006)
(0.004)
(0.016)
(0.006)
(0.006)
(0.004)
(0.016)
constant
0.960***
0.972***
0.681***
1.030***
0.963***
0.608***
0.681***
1.040***
(0.027)
(0.028)
(0.015)
(0.027)
(0.027)
(0.017)
(0.016)
(0.027)
month_stay
Controlled
Controlled
Controlled
Controlled
Controlled
Controlled
Controlled
Controlled
lnσ
μ
-1.896***
-1.898***
-1.474***
-1.317***
-1.896***
-1.910***
-1.474***
-1.317***
(0.051)
(0.051)
(0.043)
(0.035)
(0.051)
(0.051)
(0.043)
(0.035)
lnσ
ɛ
-0.580***
-0.581***
-1.268***
-0.273***
-0.580***
-0.582***
-1.268***
-0.274***
(0.014)
(0.014)
(0.069)
(0.009)
(0.014)
(0.014)
(0.069)
(0.009)
Number of reviews
76620
76620
45998
30622
76620
76620
45998
30622
Number of hotels
364
364
364
364
364
364
364
364
AIC
129442.5
129388.5
15393.4
70952.9
129441.0
129129.9
15395.4
70947.7
BIC
129636.7
129591.9
15568.2
71119.5
129644.4
129361.0
15578.9
71122.6
Log likelihood
-64700.2
-64672.2
-7676.7
-35456.4
-64698.5
-64539.9
-7676.7
-35452.9

Table S2. Estimation results of empirical models with median-center rating as dependent variable.
Model S9
Model
S10
Model
S11
Model
S12
Model
S13
Model
S14
Model
S15
Model
S16
All
reviews
All
reviews
Positive
reviews
Negative
reviews
All
reviews
All
reviews
Positive
reviews
Negative
reviews
lnmonth
-0.0135***
0.000725
-0.00648*
-0.0435***
-0.0230***
-0.0101
-0.0154***
-0.0617***
(0.004)
(0.004)
(0.004)
(0.009)
(0.006)
(0.007)
(0.005)
(0.020)
lnmonth*neg_valen
ce2
-0.0471***
-0.0476**
(0.009)
(0.022)
lnmonth*expertise
0.00336*
0.00258
0.00329
0.00593**
(0.002)
(0.002)
(0.005)
(0.002)
lnmonth*expertise*
neg_valence2
0.00361
(0.005)
expertise*neg_vale
nce2
-0.0300***
(0.005)
neg_valence2
-1.249***
-1.265***
-1.249***
1.336***
(0.017)
(0.018)
(0.017)
(0.027)
expertise
-0.0355***
-0.0355***
-0.0289***
-0.0613***
-0.0366***
-0.0268***
-0.0299***
-0.0630***
(0.002)
(0.002)
(0.002)
(0.003)
(0.002)
(0.002)
(0.002)
(0.003)
traveler_type =2
0.0200***
0.0198***
-0.0312***
0.127***
0.0198***
0.0206***
-0.0313***
0.126***
(0.007)
(0.007)
(0.006)
(0.014)
(0.007)
(0.007)
(0.006)
(0.014)
traveler_type =3
0.00311
0.00307
0.0128*
-0.0163
0.00321
0.00452
0.0129*
-0.0163
(0.008)
(0.008)
(0.007)
(0.021)
(0.008)
(0.008)
(0.007)
(0.021)
traveler_type =4
0.0193***
0.0195***
0.0177***
0.0206*
0.0193***
0.0198***
0.0176***
0.0204
(0.005)
(0.005)
(0.005)
(0.012)
(0.005)
(0.005)
(0.005)
(0.012)
traveler_type =5
0.0115*
0.0118*
0.0145***
-0.0136
0.0115*
0.0122**
0.0145***
-0.0137
(0.006)
(0.006)
(0.005)
(0.018)
(0.006)
(0.006)
(0.005)
(0.018)
constant
1.590***
1.600***
0.347***
1.583***
1.592***
0.318***
0.350***
1.588***
(0.015)
(0.015)
(0.019)
(0.025)
(0.015)
(0.020)
(0.020)
(0.026)
month_stay
Controlled
Controlled
Controlled
Controlled
Controlled
Controlled
Controlled
Controlled
lnσ
μ
-1.623***
-1.626***
-1.342***
-2.055***
-1.623***
-1.639***
-1.341***
-2.055***
(0.021)
(0.021)
(0.025)
(0.084)
(0.021)
(0.021)
(0.025)
(0.083)
lnσ
ɛ
-0.640***
-0.640***
-0.992***
-0.303***
-0.640***
-0.642***
-0.992***
-0.303***
(0.013)
(0.013)
(0.032)
(0.010)
(0.013)
(0.013)
(0.032)
(0.010)
Number of reviews
79107
79107
54720
24387
79107
79107
54720
24387
Number of hotels
364
364
364
364
364
364
364
364
AIC
124383.0
124344.8
48229.6
54803.8
124381.7
124138.2
48227.1
54804.2
BIC
124577.9
124548.9
48407.8
54965.8
124585.8
124370.2
48414.2
54974.3
Log likelihood
-62170.5
-62150.4
-24094.8
-27381.9
-62168.9
-62044.1
-24092.6
-27381.1
(Notes: *** indicates significance at 0.01, ** indicates significance at 0.05, * indicates significance at 0.1. Robust
standard errors are presented in parentheses. Estimates of month_stay dummies are not presented for brevity
purposes)

Table S3. Estimation results of empirical models with fixed-effect OLS regression
(Notes: *** indicates significance at 0.01, ** indicates significance at 0.05, * indicates significance at 0.1. Clustered
standard errors are presented in parentheses. Estimates of month_stay dummies are not presented for brevity
purposes)
Model
S17
Model
S18
Model
S19
Model
S20
Model
S21
Model
S22
Model
S23
Model
S24
All
reviews
All
reviews
Positive
reviews
Negative
reviews
All
reviews
All
reviews
Positive
reviews
Negative
reviews
lnmonth
-0.0267***
0.000606
-0.00211
-0.0580***
-0.0383***
0.00804
-0.00530
-0.0902***
(0.004)
(0.005)
(0.003)
(0.009)
(0.007)
(0.006)
(0.004)
(0.019)
lnmonth*neg_valen
ce
-0.0691***
-0.0919***
(0.011)
(0.024)
lnmonth*expertise
0.00410**
0.00113
0.00121
0.0104**
(0.002)
(0.002)
(0.001)
(0.005)
lnmonth*expertise*
neg_valence
0.0113**
(0.005)
expertise*neg_vale
nce
-0.0350***
(0.007)
neg_valence
0.311***
0.335***
0.312***
0.417***
(0.030)
(0.031)
(0.030)
(0.040)
expertise
-0.0344***
-0.0344***
-0.0173***
-0.0535***
-0.0356***
-0.0214***
-0.0177***
-0.0566***
(0.002)
(0.002)
(0.002)
(0.003)
(0.002)
(0.003)
(0.002)
(0.003)
traveler_type =2
0.0642***
0.0639***
-0.0179***
0.152***
0.0640***
0.0651***
-0.0180***
0.151***
(0.007)
(0.007)
(0.006)
(0.013)
(0.007)
(0.007)
(0.006)
(0.013)
traveler_type =3
0.00732
0.00758
0.0107**
0.00376
0.00744
0.00911
0.0108**
0.00397
(0.009)
(0.009)
(0.004)
(0.018)
(0.009)
(0.009)
(0.004)
(0.018)
traveler_type =4
0.0176***
0.0177***
0.0130***
0.0223*
0.0175***
0.0179***
0.0130***
0.0220*
(0.005)
(0.005)
(0.004)
(0.012)
(0.005)
(0.005)
(0.004)
(0.012)
traveler_type =5
-0.00603
-0.00573
0.00854*
-0.0317**
-0.00611
-0.00549
0.00851*
-0.0318**
(0.006)
(0.006)
(0.005)
(0.016)
(0.006)
(0.006)
(0.005)
(0.016)
constant
1.034***
1.052***
0.926***
0.843***
1.037***
0.687***
0.927***
0.850***
(0.020)
(0.021)
(0.007)
(0.023)
(0.020)
(0.019)
(0.007)
(0.023)
month_stay
Controlled
Controlled
Controlled
Controlled
Controlled
Controlled
Controlled
Controlled
Number of reviews
79107
79107
47513
31594
79107
79107
47513
31594
Number of hotels
364
364
364
364
364
364
364
364
R-square
159103.1
159038.7
16681.0
73372.0
159100.8
0.152
16682.1
73367.8

Table S4. Estimation results of empirical models after endogeneity bias correction
(Notes: *** indicates significance at 0.01, ** indicates significance at 0.05, * indicates significance at 0.1. Clustered
standard errors are presented in parentheses. Estimates of month_stay dummies are not presented for brevity
purposes)
Model
S25
Model
S26
Model
S27
Model
S28
Model
S29
Model
S30
Model
S31
Model
S32
All
reviews
All
reviews
Positive
reviews
Negative
reviews
All
reviews
All
reviews
Positive
reviews
Negative
reviews
lnmonth
-1.365***
0.109
0.133
-1.934***
-1.503***
1.320***
0.122
-2.220***
(0.059) (0.208) (0.186) (0.074) (0.062) (0.222) (0.192) (0.078)
lnmonth*neg_valen
ce
-1.275*** -2.770***
(0.157) (0.224)
lnmonth*expertise
0.127*** 0.00379 0.00542 0.278***
(0.020) (0.017) (0.013) (0.039)
lnmonth*expertise*
neg_valence
0.274***
(0.049)
expertise*neg_vale
nce
-0.0610***
(0.022)
neg_valence
0.288*** 0.738*** 0.291*** 1.147***
(0.031) (0.071) (0.031) (0.079)
expertise
0.00894** -0.0213*** -0.0217*** 0.00711 -0.0419*** -0.0654*** -0.0237*** -0.103***
(0.004) (0.006) (0.006) (0.004) (0.009) (0.010) (0.008) (0.017)
traveler_type =2
-
0.000082
5
0.0444*** -0.0111 0.0611*** 0.00609 0.0841*** -0.0112 0.0801***
(0.008) (0.009) (0.011) (0.015) (0.008) (0.011) (0.011) (0.016)
traveler_type =3
0.0311*** 0.0175* 0.00927* 0.0342* 0.0281*** 0.00184 0.00925* 0.0258
(0.009) (0.010) (0.005) (0.018) (0.009) (0.009) (0.005) (0.018)
traveler_type =4
0.0303*** 0.0216*** 0.0121*** 0.0370*** 0.0296*** 0.0141*** 0.0121*** 0.0357***
(0.005) (0.005) (0.004) (0.012) (0.005) (0.005) (0.004) (0.011)
traveler_type =5
0.0101 -0.000540 0.00760 -0.0127 0.00893 -0.0102 0.00761 -0.0158
(0.006) (0.006) (0.005) (0.015) (0.006) (0.006) (0.005) (0.016)
constant
1.044*** 0.579*** 0.638*** 1.560*** 1.087*** 0.231*** 0.642*** 1.647***
(0.028) (0.069) (0.053) (0.031) (0.029) (0.070) (0.055) (0.032)
month_stay
Controlled
Controlled
Controlled
Controlled
Controlled
Controlled
Controlled
Controlled
lnσ
μ
-1.907*** -1.924*** -1.480*** -1.320*** -1.907*** -1.922*** -1.480*** -1.319***
(0.051) (0.051) (0.041) (0.035) (0.051) (0.051) (0.041) (0.035)
lnσ
ɛ
-0.582*** -0.585*** -1.264*** -0.275*** -0.582*** -0.586*** -1.264*** -0.277***
(0.014) (0.014) (0.068) (0.009) (0.014) (0.014) (0.068) (0.009)
Number of reviews
79107 79107 47513 31594 79107 79107 47513 31594
Number of hotels
364 364 364 364 364 364 364 364
AIC
133426.9 132894.6 16213.8 73073.0 133359.0 132693.7 16215.5 72997.9
BIC
133621.8 133098.7 16380.4 73240.2 133563.2 132925.6 16390.9 73173.5
Log likelihood
-66692.5 -66425.3 -8087.9 -36516.5 -66657.5 -66321.8 -8087.7 -36477.9