ELECTRONIC WORD OF MOUTH AND HOTEL PERFORMANCE:
Yang Yang, Ph.D
1810 N.13th Street, Speakman Hall 304,
Philadelphia, PA 19122, USA
Sangwon Park, Ph.D*
School of Hotel and Tourism Management
The Hong Kong Polytechnic Univeristy
17 Science Museum Road, TST East, Kowloon, Hong Kong
Tel: +852 3400-2262
Fax: +852 2362-9362
1810 N.13th Street, Speakman Hall 304,
Philadelphia, PA 19122, USA
* = corresponding author
Please cite as:
Yang, Y., Park, S. and Hu, X. (2018). Electronic word of mouth and hotel
performance: A meta-analysis. Tourism Management. 67, 248-260
ELECTRONIC WORD OF MOUTH AND HOTEL PERFORMANCE:
Abstract: This study synthesizes existing empirical results about the relationship between
electronic word of mouth (eWOM) and hotel performance via meta-analysis. Based on estimates
from 25 articles, the average eWOM valence-based elasticity is estimated to be 0.888, whereas
the average volume-based elasticity is 0.055. A hierarchical linear model is applied to uncover
five aspects that explain variations in eWOM elasticities: research setting, data structure, variable
measurement, model specification, and research outlet. The estimation results highlight several
significant aspects affecting elasticity, such as year of study, geographic setting, panel data
structure, data frequency, performance measurement, control of price variable, and function form.
Finally, implications are provided for researchers and hoteliers.
Keywords: meta-analysis; eWOM effect; hotel performance; hierarchical linear model (HLM)
It is well known that the intangible, perishable nature of tourism and hospitality products makes
it difficult for consumers to gauge their quality prior to purchase. This phenomenon underscores
the perceived uncertainty in consumers’ decision-making process as it relates to travel, which
fosters a fundamental need to obtain reliable, useful information when considering travel options
(Liu & Park, 2015). With the recent proliferation of social media websites that facilitate the
sharing of travel experiences with others, the role of online consumer reviews has become
increasingly pertinent for the tourism and hospitality industry. A recent report by Mintel (2016)
revealed that consumer review websites have been identified as the second most frequently used
information source apart from search engines (e.g., Google) when travelers are researching a trip.
Service providers have therefore begun to leverage online consumer reviews, also known as
electronic word of mouth (eWOM), as marketing tools by inviting consumers to post their
personal experiences for others (Litvin & Dowling, 2016).
In the same vein, the relationship between online consumer reviews and hotel
performance has gained considerable attention from tourism and hospitality scholars (Schuckert,
Liu, & Law, 2015). Importantly, however, existing research remains largely inconclusive. For
example, while some studies suggest positive effects of eWOM on predicting hotel performance
(e.g., Ogut & Tas, 2012; Ye, Law, Gu, & Chen, 2011), others have found that the influence is
negligible or context dependent (Lu, Ye, & Law, 2014). Although prior research has provided
essential insight into the role of eWOM, a consensus about its utility vis-à-vis the tourism and
hospitality industry is elusive. Thus, the present study argues that the range of research
approaches, settings, designs, data sources, and estimation methods used in earlier analyses may
hinder generalizability (Floyd, Freling, Alhoqail, Cho, & Freling, 2014). For instance, eWOM’s
influence on hotel performance has been found to vary across different hotel classes (Blal &
Sturman, 2014). Using panel data to facilitate sound analyses, Duverger (2013) demonstrated a
non-linear (i.e., inverted U-shaped) relationship between eWOM and hotel performance.
Therefore, the aim of this research is to assess eWOM’s elasticity relevant to hotel performance
by considering five aspects as contextual variables: research setting, data structure, variable
measurement, model specification, and research output.
To fill this theoretical gap in the current literature, this study employs a meta-analysis
method, reviewing 25 research articles related to the tourism and hospitality industry. Meta-
analysis enables the authors to synthesize the literature stream quantitatively and therefore assess
eWOM elasticity as it corresponds to hotel performance (You, Vadakkepatt, & Joshi, 2015).
More specifically, this paper takes into account two separate metrics in measuring eWOM
elasticity—valence- and volume-based approaches—to identify contributing factors to the
sizable variation among estimated eWOM elasticities. In this vein, eWOM elasticity refers to
percentage changes of firm performance against percent changes of eWOM consisting of
valances and volumes of review ratings (Floyd et al., 2014; You et al., 2015). This research
considers eWOM as numerical review ratings provided by consumers on travel review/booking
websites (Park & Nicolau, 2015). By doing so, this research represents a pioneering effort to
synchronize the eWOM–performance relationship in the tourism and hospitality literature. Use
of regression-based meta-analysis makes it possible to accommodate and correct potential biases
in previous econometric results objectively (Stanley & Doucouliagos, 2012).
2. Literature Review and Conceptual Framework
2.1 Theoretical background
Online consumer reviews have long been considered important sources of information; they
allow potential customers to assess the quality of a product or service and develop an associated
image (Filieri, 2016). Online reviews play a particularly central role in the hospitality and
tourism industry due to the services’ inherent intangibility and perishability. Indeed, consumers
find it difficult to evaluate the quality of services before actually consuming them (O’Connor,
2010; Yang, Mueller, & Croes, 2016). This unique characteristic implies that people experience
significant uncertainty over their choices and, thus, require substantial information in order to
reduce perceived risks and make informed decisions. Dickinger (2011) found that online reviews
posted by other travelers are often thought of as more up-to-date, informative, enjoyable, and
reliable than information from travel service providers.
The prevalence of social media websites has created an environment where people face
information overload when confronted with numerous online consumer reviews. As such, in an
effort to reduce decision-making costs, they tend to rely more on review ratings than textual
comments. In other words, consumers who peruse multiple reviews are likely to focus on the
reviews’ valence and volume, which serve as proxies for underlying product quality (Chaiken &
Maheswaran, 1994) and hotel reputation (Anderson & Lawrence, 2014). Previous studies have
indicated that the valence and volume of online reviews indeed influence tourists’ decision-
making process (Liu & Park, 2015), along with organizations’ pricing strategies and performance
(Xie, Chen, & Wu, 2016). Signaling theory posits that people rely on signals when tasked with
making a judgment that requires balancing uncertainties. Given that most travelers today use
online channels to purchase tourism products, information asymmetry about product quality has
become widespread (Lu et al., 2014). Abrate, Capriello, and Fraquelli (2011) discussed the
importance of quality signals associated with hotel reputation as indicated by star rating and
brand affiliation. Online consumer reviews that reflect and justify users’ attitudes toward
products help readers reduce information asymmetry about product quality and, as a result,
increase their likelihood of purchasing relevant products (Park & Nicolau, 2015). In addition,
online reviews’ perceived effectiveness and reliability reduce search costs for consumers and
enhance sellers’ trustworthiness, which persuades people to pay more for products and ultimately
increases sales (Pavlou & Dimoka, 2006).
The valence of online reviews—that is, the evaluative direction (positive or negative) of
the review in terms of service experience—is more effective than the midpoint of an individual’s
attitude toward the product (Mudambi & Schuff, 2010). Put another way, one-sided reviews that
clearly indicate the direction of a consumer’s opinion generate more diagnosticity and greater
salience than moderate reviews. The accessibility-diagnosticity model suggests that a piece of
information is perceived as diagnostic when it assists consumers in deeming an alternative
worthy of further consideration (Feldman & Lynch, 1988). In contrast, online comments that do
not help consumers determine whether a product warrants deliberation are not considered
diagnostic (Herr, Kardes, & Kim, 1991). Park and Nicolau (2015) demonstrated that consumers
consistently consider extreme ratings (whether positive or negative) to be more useful and
enjoyable than moderate ratings. Furthermore, people find negative reviews more helpful than
positive comments, as the guidance they receive is likely to reduce loss rather than increase gain
(c.f. Kahneman & Tversky, 1979).
The volume of online consumer reviews underpins the bandwagon effect (see Van den
Bulte & Lilien, 2001). A greater volume of opinions provided by other consumers positively
affects customers’ judgment, regardless of whether the opinions are positive or negative (Babić
Rosario, Sotgiu, De Valck, & Bijmolt, 2016). This phenomenon is tied to herding behavior,
otherwise known as social contagion, where people are likely to mimic others’ situational
behavior in order to reduce their own risk (Banerjee, 1992; Iyengar, Van den Bulte, & Choi,
2011). For purposes of this paper, the greater the number of opinions shared online by existing
consumers, the greater the chance that other customers will become aware of the product;
message repetition attracts consumers’ attention (Duan, Gu, & Whinston, 2008).
2.2 The effects of online consumer reviews on hotel performance in the hospitality and tourism
Numerous studies have empirically examined the relationship between online consumer reviews
and hotel performance. Ogut and Tas (2012) discovered that customer ratings boost hotel
performance and affect hotel room prices (Nieto et al., 2014). Several studies measured hotel
performance by the proxy variable of number of reviews for a property (see Ogut & Tas, 2012;
Ye, Law, & Gu, 2009). Ye et al. (2011) found that a 10% increase in review ratings posted on a
major Chinese online travel agency (OTA) increased online hotel bookings (measured by
number of consumer reviews on hotels) by more than 5%. In a more comprehensive study
investigating 10 major cities (five in Europe and five in the United States), Anderson and
Lawrence (2014) assessed hotel-level word of mouth (WOM) by using ReviewPro’s Global
Review Index. The index is an aggregate of millions of social media reviews in more than 35
languages, from OTAs, review websites, and social media platforms. Results revealed that the
review index positively influenced not only the elasticity of hotel performance but also room
Consistent with the above literature, two elements comprise consumers’ online reviews of
hotels—valence (i.e., review ratings) and volume (i.e., review number)—each of which plays a
distinct role in shaping business performance. For example, a positive valence affects revenue
per available room (RevPAR) in luxury hotels, whereas review volume influences low-tier
hotels’ performance (Blal & Sturman, 2014). However, these results are inconsistent across
studies: Kim, Lim, and Brymer (2015) determined that only valence, not volume, affects hotel
Different from other research, Duverger (2013) focused on a longitudinal panel data
sample (138 hotels with monthly performance) in evaluating the temporal dynamic nature of
market share via online reviews collected from three major OTAs (TripAdvisor, Orbitz, and
Expedia). A positive relationship was identified between average hotel property ratings and the
ratio of a property’s RevPAR to that of the market. Interestingly, the study revealed a quadratic
(i.e., inverted U-shaped) relationship, implying that the effect of review ratings has diminishing
From a methodological perspective, caution must be exercised around the endogeneity of
eWOM. In the hotel industry, managerial expertise contributes to performance in several ways:
hotels with more effective operational management tend to receive high consumer ratings, and
customers will likely prefer to stay in hotels that are operating well. Taken together, these factors
encourage increased bookings compared to hotels with poorer management. To account for
endogeneity in the statistical model, Lu et al. (2014) applied a difference-in-difference approach
by using large panel data (more than 40,000 observations) to eliminate unobservable factors
along with the endogenous effect. They also discussed the importance of considering a
moderation effect associated with hotel scales, which interacts with the relationship between
consumer ratings and hotel performance (see Blal & Sturman, 2014). Several scholars have also
considered other contextual variables involved in review patterns, such as rating variations (Xie,
Zhang, & Zhang, 2014) and hotel features, including operational strategy (e.g., management
response to consumer reviews) and dates of demand between weekends and weekdays (Schamel,
2012). Phillips, Barnes, Zigan, and Schegg (2017) investigated textual review and rating datasets
from 68 online platforms to assess their effects on performance for 442 hotels. Based on the
results of sentiment analysis, they concluded that positive hotel experiences in general have the
greatest impact on hotel demand and revenue.
These research efforts provide valuable information. To develop a better understanding of
the effects of online consumer reviews, however, it is necessary to address the systematic
variation resulting from diverse contextual factors (Blal & Sturman, 2014). It is also plausible
that the multiple research approaches, settings, and data sources utilized in analyses of online
consumer reviews may render generalizable conclusions nearly impossible (Floyd et al., 2014).
Even when studies focus on similar empirical contexts, there are conflicting findings about
eWOM volume and valence. Given the importance of these metrics to the topic at hand, this
study attempts to synthesize extant research on eWOM and compare the elasticities of eWOM
volume and valence to uncover their effects on hotel performance.
2.3 Conceptual framework of meta-analysis
The conceptual model of meta-analysis addresses the following factors related to the varied
effects of online consumer reviews (i.e., eWOM elasticity) on hotel performance: research
setting (i.e., research time, geographic setting, and hotel class); data structure (i.e., panel data,
sample size, and data frequency); variable measurement (i.e., eWOM platform, performance
measures, and elasticity type); model specifications (e.g., omitted variables such as price and
other eWOM, dynamic specifications, functional form, and model complexity); and research
outlet (i.e., journal article; Babić Rosario et al., 2016; De Maeyer, 2012; Floyd et al., 2014;
Purnawirawan, Eisend, De Pelsmacker, & Dens, 2015; You et al., 2015; see Figure 1). The
following section discusses these factors in greater detail.
(Please insert Figure 1 about here)
2.3.1 Research setting
In line with the expansion of IT technology, more and more travelers have come to embrace
eWOM platforms to familiarize themselves with hotels prior to booking. In previous years, when
eWOM platforms were less popular, eWOM elasticity was small and even negligible. Lewis and
Zervas (2016) found that ratings posted on review platforms became more influential over time
in determining hotel demand, a fact that is explained by these platforms’ growing popularity.
Therefore, the eWOM elasticity of hotel performance is expected to increase over time. Formally,
this research proposes that
H1: eWOM elasticities increase over time.
Consumers’ cultural background is generally considered an important factor in dictating
information-seeking and purchasing behaviors. Hofstede’s cultural dimensions theory (2010)
suggests that national cultures vary based on the six dimensions of power distance,
individualism/collectivism, uncertainty avoidance, masculinity/femininity, long-term orientation,
and indulgence/self-restraint. This theory has been well applied in the tourism industry, unveiling
that Chinese travelers’ behavior is largely shaped by relatively higher levels of power distance,
collectivism, femininity, and long-term orientation compared to other cultures (Reisinger &
Crotts, 2010). In exploring differences in web communication behaviors between American and
Chinese travelers, Park and Reisinger (2012) discovered heterogeneous behaviors across
information seeking, communication, and transactions. More specifically, Chinese travelers from
highly collectivistic societies tend to use the Internet for social purposes and entertainment, and
they generally perceive higher risks when shopping online. Thus, it is important to assess
geographical context—in this case, whether the hotels investigated are located in China or
elsewhere. These arguments lead to the following hypothesis:
H2: eWOM elasticities are lower with hotels in China than elsewhere.
Consumers’ online review habits appear to vary by hotel class. Indeed, consumers are more
likely to leave comments when they stay in higher-end hotels versus lower-end accommodations
(Miguens, Baggio, & Costa, 2008). Blal and Sturman (2014) found that eWOM volume appears
to exert positive effects on economy, midscale, and upper-midscale hotels, whereas effects are
negative for upscale and luxury hotels. Furthermore, the star ratings reflecting hotel classes
serve as another signaling factor (Lu et al., 2014). A five-star hotel is thought to provide guests
with high-quality facilities and services, which affects customers’ expectations for service
consumption. Indeed, although a five-star hotel may have negative reviews, consumers are still
confident in anticipating better quality compared to lower hotel classes with higher consumer
ratings. Thus, the following is hypothesized:
H3: eWOM elasticities are higher with low-class hotels than mid- and high-class hotels.
2.3.2 Data structure
Longitudinal panel data provide researchers clearer insights into causality while controlling for
potential confounds (Rindfleisch, Malter, Ganesan, & Moorman, 2008). For example, because
panel data generally contain a greater degree of freedom and within-in-unit variation than cross-
sectional data, more meaningful model parameters are likely to be obtained (Baltagi, 2008). As a
result, findings derived from panel data are often considered more informative when inferring a
factual relationship between the variables being analyzed (Floyd et al., 2014). Therefore, the
following is proposed:
H4: eWOM elasticities are different between those studies analyzing panel and cross-sectional
There is substantial evidence that a sufficient sample size is essential to achieving acceptable
error in model estimates and to determining error in the parameters being estimated, which is
associated with coefficient variation (Beaman, Huan, & Beaman, 2004). Sample size ultimately
reflects the accuracy and reliability of findings. Therefore, this study argues that an appropriate
sample size is necessary to ensure the accuracy of empirical findings delineating the eWOM–
performance relationship. Additionally, sample size can influence the magnitude of estimated
eWOM elasticities. Thus, the following is hypothesized:
H5: eWOM elasticities vary across models using different sample sizes.
A finer level of temporal aggregation of the variable (e.g., a daily rather than monthly, quarterly,
or yearly level of data) used in the empirical model influences eWOM volume and valence
elasticities. Essentially, data aggregation at a coarser level is more likely to dilute variations of
performance as an outcome variable (You et al., 2015). Thus, the following hypothesis is
H6: eWOM elasticities are different across levels of data aggregations across weekly, quarterly,
and yearly formats.
2.3.3 Variable measurement
It is important to address whether all eWOM communication channels generate equal or
disparate effects (Ya et al., 2015). A large number of studies in tourism and hospitality have
focused on specific eWOM platforms, such as TripAdvisor, booking.com, ctrip.com, or Yelp
(e.g., Park & Nicolau, 2015; Xie et al., 2014). Research has also found that disclosure of a
reviewer’s identity increases the perceived usefulness and trustworthiness of online reviews (Liu
& Park, 2015). Comments posted on platforms that offer structured fields for user information
and encourage reviewers to reveal their true identities (e.g., TripAdvisor) are more valuable to
other consumers than information provided on platforms that allow reviewers to remain
anonymous. In this respect, Ya et al. (2015) found that platforms such as community-based sites,
blogs, and online review websites play an important role in eWOM valence elasticities. In
particular, because TripAdvisor is the largest online community for travelers and has a well-
developed eWOM metric system (Levy, Duan, & Boo, 2013), and because many OTA websites
share TripAdvisor’s hotel eWOM metrics, TripAdvisor as a platform can be expected to provide
the largest eWOM effect. These arguments lead to the following hypothesis:
H7: eWOM elasticities are different depending on types of review platforms.
There are two main approaches to measuring hotel performance in studies investigating eWOM
effects in tourism: direct performance measures and proxy measures of performance. The former
(e.g., room nights sold, room revenue, occupancy rate, and RevPAR) is relatively straightforward
when interpreting performance elasticities (see Anderson & Lawrence, 2014). However, due to
restricted access to actual performance data, numerous studies have utilized proxy measures of
performance, such as the number of reviews written on review websites, based on the assumption
that review volume reflects the number of consumers who have experienced a specific product or
service (see Ogut & Tas, 2012; Ye et al., 2009, 2011). Thus, it is important to understand the
influence of different performance measures on eWOM elasticity.
H8: eWOM elasticities are different depending on the type of hotel performance measurement.
As discussed in Section 2.1, online consumer reviews vary in volume (i.e., the total number of
eWOM units regarding a specific product or service) and valence (i.e., the idea that eWOM can
be positive, negative, or neutral; Liu, 2006). Volume refers to the number of consumers who
have experienced a level of popularity about a product in the market. Hence, it can be expected
that eWOM volume enhances consumers’ awareness of a given product and reduces their
uncertainty about it, which ultimately encourages sales (Chintagunta, Gopinath, &
eWOM valence captures the favorability, sentiment, or polarity of consumer reviews
related to evaluations and reputation of a product or firm. Consumers’ attitudes toward and/or
preferences for a product can be formed, reinforced, or changed by directional consumer reviews
(Kim & Gupta, 2012). Thus, discerning two metrics of eWOM— valence and volume—makes it
possible to test the heterogeneity of sales elasticities. Therefore, the following is hypothesized:
H9: eWOM elasticities are different according to the types of elasticities estimated.
2.3.4 Model specification
Tourism and hospitality services have long been regarded as high-involvement products that
present significant risk, complexity, and price (McKercher, 2016). In addition, people do not
purchase these types of service products as often as they do routine products (i.e., low-
involvement products). As a result, consumers who buy high-involvement products engage in an
extensive problem-solving process that requires significant time and effort as they search for and
review relevant information (Mathwick & Rigdon, 2004). This implies that when consumers
purchase expensive (i.e., high-involvement) products, they are likely to look for reliable and
useful information by depending more on eWOM compared to those who buy inexpensive
products. Moreover, because price is closely associated with other unobservable characteristics
that may influence hotel performance (e.g., location), excluding a price variable can lead to an
omitted variable bias of estimates. Therefore, estimated eWOM elasticities can be expected to
depend on an empirical model’s specifications, including the inclusion or lack of a price measure.
These arguments lead to the following hypothesis:
H10: eWOM elasticities are different between those studies controlling and overlooking room
Because different eWOM metrics (i.e., valence, volume, and valence) capture different aspects
of eWOM effect, disregarding other eWOM metrics in the model may lead to some omitted
variable bias as well. Therefore, the following is hypothesized:
H11: eWOM elasticities are different between those studies controlling and overlooking other
It may take time for hotels to internalize the benefits resulting from a positive online reputation;
therefore, some studies specify lagged independent variables in the empirical model (Blal &
Sturman, 2014; Lu, Xiao, & Ye, 2012; Xie, Zhang, Zhang, Singh, & Lee, 2016), especially for
weekly or monthly data. In general, specifying lagged eWOM independent variables can be
expected to result in greater eWOM elasticity of hotel performance.
H12: eWOM elasticities are different between those studies stipulating and overlooking lag of
independent variables and others.
There are four typical functional forms in econometric models that estimate eWOM effects on
hotel performance. A log-log form includes the dependent variable (i.e., performance) and the
independent variable (i.e., eWOM valence or volume) in the estimated model as the natural log
of the original variable. A log-lev form refers to the specification that log-transformed only the
dependent variable while leaving the independent variable untransformed, whereas a lev-log
form suggests the opposite. Lastly, a lev-lev form includes untransformed dependent and
independent variables of interest (Floyd et al., 2014). As suggested by previous literature, the
functional form can influence eWOM elasticity (Floyd et al., 2014; You et al., 2015). Formally,
the following hypothesis is proposed:
H13: eWOM elasticities are different according to different functional forms in estimating
The number of independent variables indicates an empirical model’s complexity. Because
eWOM valence and volume also reflect some hotel characteristics that shape hotel performance,
omitting these key factors may lead to an “omitted variable bias” (Greene, 2007) associated with
the eWOM elasticity estimate. Therefore, eWOM elasticity can be expected to depend on model
complexity as measured by the number of independent variables in the empirical model. Thus,
the following is hypothesized:
H14: eWOM elasticities are influenced by the number of independent variables being estimated
2.3.5. Research outlet
This research analyzed whether articles were published in research journals and how the research
outlet influences the estimated eWOM elasticities. If such influences exist, they can be explained
by the issue of publication bias, which refers to “the selective publication of studies with a
particular outcome, generally those which are statistically significant, at the expense of null
studies” (Ferguson & Brannick, 2012, p. 120). Publication bias may result in reduced inter-study
variability and inflation of biased mean estimates (Dickersin, 2005), which may potentially affect
estimated eWOM elasticities (Floyd et al., 2014). More specifically for journal publication bias,
this bias indicates that researchers are less motivated to submit manuscripts with non-significant
eWOM effect to journals because editors and reviewers are less likely to accept such manuscripts
(Card, 2015, p. 257). As a result, less significant and non-significant eWOM effects are more
likely to be seen in research from other outlets, such as conference papers and working papers,
which receive less pressure from editors and reviewers. These arguments lead to the following
H15: eWOM elasticities are less in journal articles than other resources.
3. Coding and Empirical Models
3.1 Data collection and coding
A search was conducted for relevant empirical literature investigating the relationship between
hotel performance and eWOM in online reviews by using the Google Scholar search engine,
EBSCO Hospitality & Tourism Complete database, SSRN’s eLibrary, and the ProQuest
Dissertations & Theses database. The final search was completed on Dec. 11, 2016. The
keywords used in the literature search included “online rating”, “hotel sales”, “hotel
performance”, “online reviews”, “hotel demand”, and “TripAdvisor”. The references in
identified papers were consulted in a search for further studies. The search was restricted to
English-language materials but was not limited to journal articles.
After the papers were collected, they were screened according to the following criteria:
(1) valence and/or volume metrics of online reviews were used as independent variables, and
hotel performance measures (e.g., revenue, RevPAR, occupancy rate, room nights sold, review
volume as sales proxy) as dependent variables; (2) the studies’ empirical models had no latent
variable with which to measure hotel performance or eWOM metrics; and (3) elasticities were
either provided in each paper or could be clearly derived from the estimated coefficients and
other information. Room rate was not considered as a performance measure; therefore, estimates
using room rate as the dependent variable were excluded. Elasticity was chosen as the effect size
in the meta-analysis for several reasons. First, elasticity has an inherent economic meaning and
has been widely used in meta-analyses of marketing literature (Floyd et al., 2014; You et al.,
2015). Second, many studies in the sample reported elasticity estimates directly without
providing other necessary information to calculate alternative effect sizes, such as t statistics and
partial correlations (Stanley & Doucouliagos, 2012). Third, some elasticity estimates were
obtained from the interactions between eWOM metrics and dummy/categorical variables, which
compounded the challenges of obtaining standard errors and t statistics.
The literature search resulted in a total of 25 papers comprising the meta-sample. This
sample covered a variety of article types, including journal articles (19), dissertations and theses
(2), an academic report (1), a book chapter (1), a conference paper (1), and a working paper (1).
eWOM elasticities that used metrics of variance and ranking were excluded; very few studies
reported them, and the ranking metric is a function of review valence and volume (Rianthong,
Dumrongsiri, & Kohda, 2016). Because studies in the sample used different functional forms in
their empirical models, regression coefficients were transformed into elasticities by using
formulas suggested by Floyd et al. (2014) and You et al. (2015). For the log-log function, the
elasticity is the regression coefficient β itself; for the lev-lev function, the elasticity is β (
and are the mean values of the dependent and independent variables, respectively); for the log-
lev function, the elasticity is β; and for the lev-log function, the elasticity is β. If some
studies failed to report information necessary for the meta-analysis, every effort was made to
contact the authors.
Most studies in the sample presented multiple estimations using different model
specifications and/or different sub-samples. Moreover, they sometimes presented multiple
elasticity estimates based on different eWOM metrics in a single estimation. Therefore, multiple
elasticity estimates were coded from different estimations in a single study. By doing so,
efficiency was improved from the expanded sample size and additional within-study variation
across elasticity estimates (Melo, Graham, & Noland, 2009). The final sample was composed of
a total of 161 valence- and volume-based estimates taken from 25 studies. Apart from elasticities,
15 independent variables discussed previously were also coded to explain variations in eWOM
elasticities. Table 1 presents the coding schemes and descriptions of these variables. In the
research team, two coders recorded and calculated elasticities and the abovementioned
independent variables. After independent coding, the coders verified the consistency of results
and resolved any disagreements by discussion.
(Please insert Table 1 about here)
3.2 Empirical models
Hierarchical linear modelling (HLM) was used to uncover the impacts of various factors on
eWOM elasticities. HLM is particularly suited to modelling data with a nested structure, as in the
case of this dataset: an elasticity estimate is nested in a study. The traditional ordinary least
squares (OLS) regression assumes each observation to be independent from the others; however,
in this data set, different elasticities from the same paper can be correlated with one another. In
this research context, the HLM model can effectively account for within-study error correlations
stemming from unobserved study-specific factors (Edeling & Fischer, 2016), providing more
reliable estimation results. An HLM model is specified as follows:
ij ij i ij
where i indexes each study, and j indexes each elasticity estimate in study i. In the model,
denotes the jth elasticity estimate in study i, and
represents a set of independent variables.
is the normal error term (as in OLS regression), and
denotes the study-specific
effect that captures unobserved characteristics of the study that remained unchanged across
estimates within the study;
follow an independent normal distribution with a mean of
zero and a variance of
, respectively. The HLM model was estimated by using the
maximum likelihood estimation with the expectation-maximization algorithm, which is expected
to generate asymptotically efficient and consistent estimates (Hox, Moerbeek, & van de Schoot,
4.1 Coding results
First, all 161 eWOM elasticities from 25 studies were coded. Most of these studies reported a
positive and significant effect of eWOM on hotel performance. However, some studies found
insignificant effects based on model estimation results. For example, Blal and Sturman (2014)
obtained an insignificant effect of eWOM volume on hotel performance based on 319 hotels in
London, and similar results on this insignificant effect were identified by Kim et al. (2015) from
128 U.S. hotels in a hotel chain. Furthermore, some studies investigated the eWOM effect for
hotels in different classes; therefore, different eWOM elasticities were extracted for them. More
specifically, Anderson (2012) and Anderson and Lawrence (2014) found larger eWOM
elasticities for mid-scale hotels compared to high-end ones; however, Duverger (2013) and Blal
and Sturman (2014) showed that eWOM elasticities are larger for high-end hotels.
Figure 2 shows the distributions of reported elasticity estimates in the data. For all
elasticities (in the upper panel), the mean value is 0.722 with a standard deviation of 0.965. As
shown in the histogram, the majority (92.55%) of elasticity estimates are positive. For valence-
based elasticities (N = 129), the mean value is 0.888 and the standard deviation is 0.999; both are
larger than their volume-based elasticity counterparts (N = 32), which are 0.055 and 0.329,
respectively. These results suggest that a 1% increase in eWOM valence would lead to a 0.722%
increase in hotel performance, whereas a 1% increase in eWOM volume would lead to a 0.055%
increase in hotel performance. As shown in the middle- and lower-panel histograms, compared
to valence-based elasticity, volume-based elasticity covers a narrower range of positive values.
Note that there is an apparent outlier, as indicated in the left tail of the distribution of volume-
based estimates. Further empirical investigation showed that the results would not change if this
estimate was removed; these results are available upon request.
(Please insert Figure 2 about here)
Table 2 presents the descriptive statistics of continuous independent variables in the specified
HLM model. The average data year of the sample was around 2011, and the year ranged from
2007 to 2014. Out of all elasticity estimates, 12.4% were based on Chinese hotel data, and 59.0%
used panel data covering hotel properties over time. A large proportion (58.4%) of the estimated
elasticity stemmed from TripAdvisor-based metrics. Notably, 16.8% of estimates were estimated
by using review volume as a proxy of hotel performance, which may introduce significant
measurement errors. As for variables related to the empirical model specification, 17.4% of
estimates were obtained from a model incorporating a price variable, 23.6% from models using
the lagged value of eWOM metrics as independent variables, 53.4% from models covering other
eWOM metrics, and 39.1% from models in a linear functional form (lev-lev) without log-
transformation of variables. Moreover, the average number of independent variables specified in
the empirical model was around nine. Lastly, 68.9% of elasticity estimates were collected from
journal articles. Table 3 presents the descriptive statistics of categorical independent variables.
Across the sample, 71.43% of estimates were based on a sample of hotels across different classes.
In this research, the sample included a variety of observation frequencies, with month-based
estimates outnumbering quarter-, year-, and week-based ones. As stated earlier, the sample was
dominated by valence-based elasticity estimates, which accounted for 80.12% of all estimates.
The correlation matrix of independent variables was also calculated. Most coefficients were
below 0.5, suggesting the absence of multi-collinearity (Gujarati & Porter, 2010).
(Please insert Table 2 about here)
(Please insert Table 3 about here)
4.2 HLM results
Table 4 presents the estimation results of the HLM models. A series of models was estimated
depending on the type of eWOM elasticities (i.e., the general eWOM elasticity, the eWOM
valence elasticity, and the eWOM volume elasticity). Moreover, because some studies used the
review number to proxy hotel sales, which may yield problematic results, different HLM
specifications were run with and without elasticities from these studies. In Model 1, only a few
estimated coefficients were statistically different from zero. First, proxy_perf had the largest
significant estimated coefficient across independent variables, and using review volume as a
performance proxy substantially inflated the estimated elasticity by 1.343 points. H8 is therefore
supported. Second, this result suggests that volume-based elasticity (elasticity_type = volume) is
0.751 point less than valence-based after controlling for other factors. H9 is therefore supported.
Third, elasticity estimates from the linear function without log-transformation of any variables
were 0.524 point less than estimates from other function forms. H13 is therefore supported.
Fourth, sample size appeared to be negatively associated with estimated eWOM elasticity, and a
larger sample size tended to generate a smaller elasticity estimate. H14 is therefore supported.
(Please insert Table 4 about here)
As shown in Model 1, using review volume as a performance proxy would substantially inflate
the elasticity estimate; therefore, Model 2 was run using the same specification after excluding
27 estimates from relevant studies. In Model 2, the insignificant estimate of year provides little
support to H1. The coefficient of China was negative and significant, suggesting that hotels in
China derive considerably fewer benefits from eWOM valence and volume compared to their
American and European counterparts. This result lends support to H2. For categorical variables,
a positive coefficient indicates a more elastic effect of eWOM in this category compared to the
reference category, and a negative coefficient indicates the opposite. Therefore, a positive
coefficient of class = low indicates that the eWOM elasticity of low-class hotels is significantly
lower than that obtained for the reference category: all types of hotels. H3 is therefore supported.
Moreover, panel_data was estimated to be negative and significant, and the use of panel data
substantially reduced the elasticity estimate by 0.908 point. H4 is therefore supported. Similarly,
as suggested by the results, a larger sample size used in the empirical model was associated with
a smaller calibrated elasticity estimate. H5 is therefore supported. As for data observation
frequency, compared to monthly data (the reference category freq = monthly), weekly data
generated a larger elasticity estimate by 0.306 point. However, the insignificant coefficients of
freq = quarterly and freq = yearly indicated that quarterly and yearly data produced no statistical
difference in the size of eWOM elasticities compared to monthly data. H6 is therefore partly
confirmed. In addition, no empirical evidence was found to support a larger elasticity from
TripAdvisor-based reviews. H7 is not supported by the result. Like Model 1, Model 2 highlights
a smaller size for volume-based elasticity and is 0.776 point less than a valence-based elasticity,
Regarding variables related to model specification, the Model 2 results highlight the
important role of price measures as a control variable in estimating eWOM elasticity. More
specifically, controlling the price measure increased the size of eWOM elasticity by 0.130 point,
suggesting that omitting the price measure leads to a significant downward bias. This result
corroborates H10. However, the use of a lagged eWOM variable and inclusion of other eWOM
variables did not explain the variation of eWOM elasticity in the sample, leaving H12
unsupported. The negative and significant coefficients of linear_function and indep_vars
indicate that the use of a linear functional form (lev-lev) and more independent variables leads to
a smaller elasticity estimate, lending support to H13 and H14. Lastly, the coefficient of
journal_article was insignificant but positive, demonstrating that the magnitude of eWOM
elasticity from journal articles is not statistically different from that of other types of studies.
Therefore, H15 is rejected.
Because valence-based elasticity estimates dominated the sample, Model 3 presents the
HLM results for valence-based elasticities only. Similar to the results of Model 1, online review
volume was found to be a very poor proxy of hotel performance because of the significant and
sizable estimated coefficient of proxy_perf. Therefore, in Model 4, those elasticity estimates that
used review volume to proxy performance were excluded. Model 4 provided similar results to
Model 2, such as negative and significant estimated coefficients of year, China, class = low,
panel_data, lnsample_size, and linear_function. However, some differences were noted. For
example, using weekly data (freq = weekly), omitting price controls (price_control), and
including more independent variables (indep_vars) did not necessarily induce a statistically
different valence-based elasticity estimate.
Model 5 presents the estimation results of volume-based elasticities exclusively. Because
only 32 observations were included, some categorical independent variables had to be omitted,
such as freq and class, to provide more reliable statistical inference of the results. Note that no
studies adopted review volume as a performance proxy in any of the 32 observations. As
suggested by the significant coefficients of year and TripAdvisor, volume-based elasticity was
also found to decrease over time, and volume-based elasticities from TripAdvisor reviews were
significantly larger than others, revealing the central role of hotels’ visibility on TripAdvisor in
spurring performance. Therefore, H1 and H7 are supported for the valence-based eWOM
elasticities. Furthermore, the model specification largely explained variation in volume-based
elasticities across studies. More specifically, including price controls (price_control) and using
lagged eWOM variables in the model (lagged_indep) often led to a larger volume-based
elasticity estimate, whereas incorporating other eWOM independent variables (other_eWOM)
decreased estimates. Another model specification variable, indep_vars, was also estimated to be
significant, and its negative coefficient indicates that including more independent variables
would reduce the size of volume-based elasticity.
5. Conclusion and Discussion
In light of the proliferation of online consumer review websites and the nature of tourism and
hospitality products, eWOM has been recognized as one of the most important information
sources in the industry. It not only affects consumers’ decision-making process but also
influences hotel performance. Hence, several scholars have attempted to estimate the effects of
eWOM on hotel performance (e.g., Xie et al., 2014). Interestingly, their findings have been
inconsistent and inconclusive. This study provides generalizable evidence, demonstrating the
effects of online consumer reviews on eWOM elasticity according to volume- and valence-based
metrics. Specifically, a meta-analysis using an HLM model was applied by synthesizing 25
studies on eWOM elasticity in the tourism and hospitality fields to identify factors explaining
elasticity variation. This research thus offers important and insightful contributions to academics
and practitioners in the tourism and hospitality industry.
This study resolves existing conflicts in the literature and sheds lights on generalized
effects of eWOM volume and valence metrics by accounting for research setting, data structure,
variable measurement, model specification, and research output as five aspects that influence
eWOM elasticity of hotel performance. As a result, this research synthesizes extant research on
eWOM elasticity. To the authors’ knowledge, this is the first systematic attempt to compare the
degree to which eWOM elasticities are different between different measurements. Indeed, the
results of this study revealed that mean eWOM valence elasticity (0.888) is twice as large as its
counterpart (0.417) in the general marketing field as reported by You et al. (2015), highlighting
how necessary it is for hotels to maintain a high eWOM valence level to attract customers. This
result can be explained by the experiential nature of hotel products: no one knows the quality of
the product until consuming it. Customers therefore require supplemental independent reviews to
make a decision and reduce the risks associated with purchase. However, the mean eWOM
volume elasticity (0.055) is lower than its counterpart (0.236) in the general marketing field (You
et al., 2015). One possible explanation is that, as one of the first industries to embrace online
review platforms (Ong, 2012), eWOM valence elasticity is more substantial in the tourism and
hospitality industries than in others. Therefore, effective reputation management may be
especially rewarding for the tourism and hospitality business.
Second, these results underscored higher eWOM elasticity for mid-scale and high-end
hotels, suggesting that these hotels should place high priority on monitoring and managing
eWOM on various platforms in a timely manner. This finding falls into line with Blal and
Sturman’s (2014) and Lu et al.’s (2014) studies. Third, because TripAdvisor-based volume
elasticity was found to be more pronounced than others, hoteliers should motivate their guests to
post reviews on TripAdvisor over other platforms. Hotel managers’ engagement with consumers’
feedback on TripAdvisor is therefore of paramount importance. As shown in Table 2,
approximately 60% of the sample have investigated TripAdvisor as a key platform for their
research. Xiang, Du, Ma, and Fan (2017) have indicated a consistent conclusion that TripAdvisor
has been widely perceived as a premier data source in the hotel industry due to the number of
reviews available, wide distribution of review sentiment, adequate length of consumer reviews,
credibility, and helpfulness. Finally, the results from this regression-based meta-analysis provide
flexible estimates of eWOM elasticity-based, context-specific circumstances. Therefore, hotel
chains and groups can calibrate property-specific eWOM returns based on the characteristics of
each hotel and propose more effective strategies to prioritize reputation management in some of
This study provides some methodological implications and guidelines for researchers
when investigating the eWOM–performance relationship in tourism and hospitality. First and
foremost, it is inappropriate to use review volume to proxy hotel performance/sales in the
empirical model because this results in a biased estimate of eWOM effect, based on this study’s
meta-analysis results. Checking Appendix A, around 30% of articles have considered the number
of consumer reviews to measure hotel performance. Use of indicators that allow researchers to
estimate actual performance, rather than proxy measurements, is suggested. Second, it is highly
recommended that researchers collect panel data to understand the eWOM–performance
relationship, if available. The estimates from panel data (60% of the sample) were found to be
significantly different from others (40% of the sample used cross-section data). Due to several
inherent advantages of panel data, this type of data is particularly useful to understand a causal
relationship. Third, model specification is important to ensure a reliable estimate of eWOM
effect. More relevant independent variables—in particular, the price variable—should be
incorporated into the empirical model to alleviate the potential omitted variable bias. Lastly,
since the results suggested that models using the lagged independent variables do not yield
different results, use of the unlagged independent variables in empirical models is recommended
to keep more observations in the sample. This finding is partially consistent with You et al.’s
(2015) study concluding that the omission of a lagged term did not have significant influence on
Some limitations of this study should be noted. First, due to sample size concerns, other
eWOM metrics were not included, such as eWOM variance and online ranking. The estimated
elasticities based on these metrics may present different patterns from those identified here.
Second, eWOM elasticities in non-hotel hospitality settings were not considered. Lastly, since
the standard error of elasticities was not coded due to data limitation, it was not possible to
conduct some formal statistical procedures (e.g., funnel plot, fail-safe N, and standard error
control in regression) to test the issue of publication bias rigorously (Orwin, 1983; Stanley &
Doucouliagos, 2012). Therefore, it is recommended that future studies investigate alternative
eWOM metrics and elasticities in other tourism and hospitality sectors, such as restaurants.
Abrate, G., Capriello, A., & Fraquelli, G. (2011). When quality signals talk: Evidence from
the Turin hotel industry. Tourism Management, 32(4), 912-921.
Anderson, C. K. (2012). The impact of social media on lodging performance. Cornell Hospitality
Report, 12(15), 4-11.
Anderson, C. K., & Lawrence, B. (2014). The influence of online reputation and product
heterogeneity on service firm financial performance. Service Science, 6(4), 217-228.
Babić Rosario, A., Sotgiu, F., De Valck, K., & Bijmolt, T. H. (2016). The effect of electronic
word of mouth on sales: A meta-analytic review of platform, product, and metric
factors. Journal of Marketing Research, 53(3), 297-318.
Banerjee, A. V. (1992). A simple model of herd behavior. The Quarterly Journal of
Economics, 107(3), 797-817.
Baltagi, B. (2008). Econometric Analysis of Panel Data (4th ed.). Chichester, UK: John Wiley
Beaman, J. G., Huan, T. C., & Beaman, J. P. (2004). Tourism surveys: sample size, accuracy,
reliability, and acceptable error. Journal of Travel Research, 43(1), 67-74.
Blal, I., & Sturman, M. C. (2014). The differential effects of the quality and quantity of
online reviews on hotel room sales. Cornell Hospitality Quarterly, 55(4), 365-375.
Card, N. A. (2015). Applied Meta-Analysis for Social Science Research. New York: Guilford
Chaiken, S., & Maheswaran, D. (1994). Heuristic processing can bias systematic processing:
effects of source credibility, argument ambiguity, and task importance on attitude
judgment. Journal of Personality and social Psychology, 66(3), 460–473.
Chintagunta, P. K., Gopinath, S., & Venkataraman, S. (2010). The effects of online user
reviews on movie box office performance: Accounting for sequential rollout and
aggregation across local markets. Marketing Science, 29(5), 944-957.
De Maeyer, P. (2012). Impact of online consumer reviews on sales and price strategies: A
review and directions for future research. Journal of Product & Brand
Management, 21(2), 132-139.
Dickersin, Kay (2005), “Publication Bias: Recognizing the Problem, Under-standing its
Origins and Scope and Preventing Harm,” in Publication Biasin Meta-analysis:
Prevention, Assessment, and Adjustments, Rothstein Han-nah H., Sutton Alexander J.
and Borenstein Michael, eds. Hoboken, NJ: JohnWiley & Sons
Dickinger, A. (2011). The trustworthiness of online channels for experience-and goal-
directed search tasks. Journal of Travel Research, 50(4), 378-391.
Duan, W., Gu, B., & Whinston, A. B. (2008). The dynamics of online word-of-mouth and
product sales—An empirical investigation of the movie industry. Journal of Retailing,
Duverger, P. (2013). Curvilinear effects of user-generated content on hotels’ market share: a
dynamic panel-data analysis. Journal of Travel Research, 52(4), 465-478.
Edeling, A., & Fischer, M. (2016). Marketing’s Impact on Firm Value: Generalizations from
a Meta-Analysis. Journal of Marketing Research, 53(4), 515-534.
Filieri, R. (2016). What makes an online consumer review trustworthy?. Annals of Tourism
Research, 58, 46-64.
Floyd, K., Freling, R., Alhoqail, S., Cho, H. Y., & Freling, T. (2014). How online product
reviews affect retail sales: A meta-analysis. Journal of Retailing, 90(2), 217-232.
Greene, W. H. (2007). Econometric Analysis (6th ed.). Upper Saddle River, NJ: Prentice
Gujarati, D. N., & Porter, D. C. (2010). Essentials of Econometrics. New York: McGraw-
Herr, P. M., Kardes, F. R., & Kim, J. (1991). Effects of word-of-mouth and product-attribute
information on persuasion: An accessibility-diagnosticity perspective. Journal of
Consumer Research, 17, 454–462
Hofstede, G. (2001). Culture’s consequences: Comparing values, behaviours, institutions and
organizations across nations (2nd ed). Thousand Oaks, CA: Sage.
Hox, J. J., Moerbeek, M., & van de Schoot, R. (2010). Multilevel Analysis: Techniques and
Applications (2nd ed.). New York: Routledge.
Iyengar, R., Van den Bulte, C., & Choi, J. (2012). Distinguishing among mechanisms of
social contagion in new product adoption: Framework and illustration. In Working
paper: The Wharton School.
Kahneman, D., & Tversky, A. (1979). Prospect theory: an analysis of decision under risk.
Econometrica, 47(2), 263–292.
Kim, J., & Gupta, P. (2012). Emotional expressions in online user reviews: How they
influence consumers' product evaluations. Journal of Business Research, 65(7), 985-
Kim, W. G., Lim, H., & Brymer, R. A. (2015). The effectiveness of managing social media
on hotel performance. International Journal of Hospitality Management, 44, 165-171.
Levy, S. E., Duan, W., & Boo, S. (2013). An analysis of one-star online reviews and
responses in the Washington, DC, lodging market. Cornell Hospitality Quarterly, 54(1),
Lewis, G., & Zervas, G. (2016). The Welfare Impact of Consumer Reviews: A Case Study of
the Hotel Industry. Working paper,
Litvin, S. W., & Dowling, K. M. (2016). TripAdvisor and hotel consumer brand
loyalty. Current Issues in Tourism, 1-5.
Liu, Y. (2006). Word of mouth for movies: Its dynamics and impact on box office
revenue. Journal of marketing, 70(3), 74-89.
Liu, Z., & Park, S. (2015). What makes a useful online review? Implication for travel product
websites. Tourism Management, 47, 140-151.
Lu, Q., Xiao, L., & Ye, Q. (2012). Investigating the impact of online word-of-mouth on hotel
sales with panel data. Paper presented at the Management Science and Engineering
(ICMSE), 2012 International Conference.
Lu, Q., Ye, Q., & Law, R. (2014). Moderating effects of product heterogeneity between online
word-of-mouth and hotel sales. Journal of Electronic Commerce Research, 15(1), 1–12.
Mathwick, C., & Rigdon, E. (2004). Play, flow, and the online search experience. Journal of
consumer research, 31(2), 324-332.
McKercher, B. (2016). Towards a taxonomy of tourism products. Tourism Management, 54,
Melo, P. C., Graham, D. J., & Noland, R. B. (2009). A meta-analysis of estimates of urban
agglomeration economies. Regional Science and Urban Economics, 39(3), 332-342.
Miguéns, J., Baggio, R., & Costa, C. (2008). Social media and tourism destinations:
TripAdvisor case study. Advances in tourism research, 26(28), 1-6.
Mintel (2016). Holiday Planning and Booking Process, Mintel, London.
Mudambi, S. M., & Schuff, D. (2010). What makes a helpful online review? A study of
customer reviews on Amazon.com. MIS Quarterly, 34(1), 185–200.
Nieto, J., Hernández-Maestro, R. M., & Muñoz-Gallego, P. A. (2014). Marketing decisions,
customer reviews, and business performance: The use of the Toprural website by
Spanish rural lodging establishments. Tourism Management, 45, 115-123.
Öğüt, H., & Onur Taş, B. K. (2012). The influence of internet customer reviews on the online
sales and prices in hotel industry. The Service Industries Journal, 32(2), 197-214.
Ong, B. S. (2012). The perceived influence of user reviews in the hospitality
industry. Journal of Hospitality Marketing & Management, 21(5), 463-485.
Orwin, R. G. (1983). A fail-safe N for effect size in meta-analysis. Journal of Educational
Statistics, 8(2), 157-159.
Qi, L., Lei, X., & Qiang, Y. (2012, 2012). Investigating the impact of online word-of-mouth
on hotel sales with panel data. Paper presented at the Management Science and
Engineering (ICMSE), 2012 International Conference.
Park, S., & Nicolau, J. L. (2015). Asymmetric effects of online consumer reviews. Annals of
Tourism Research, 50, 67-83.
Pavlou, P. A., & Dimoka, A. (2006). The nature and role of feedback text comments in online
marketplaces: Implications for trust building, price premiums, and seller
differentiation. Information Systems Research, 17(4), 392-414.
Phillips, P., Barnes, S., Zigan, K., & Schegg, R. (2016). Understanding the impact of online
reviews on hotel performance: an empirical analysis. Journal of Travel Research, 56(2),
Purnawirawan, N., Eisend, M., De Pelsmacker, P., & Dens, N. (2015). A meta-analytic
investigation of the role of valence in online reviews. Journal of Interactive
Marketing, 31, 17-27.
Reisinger, Y., & Crotts, J. C. (2010). Applying Hofstede’s national culture measures in
tourism research: Illuminating issues of divergence and convergence. Journal of Travel
Research, 49(2), 153-164.
Rianthong, N., Dumrongsiri, A., & Kohda, Y. (2016). Improving the multidimensional
sequencing of hotel rooms on an online travel agency web site. Electronic Commerce
Research and Applications, 17, 74-86.
Rindfleisch, A., Malter, A. J., Ganesan, S., & Moorman, C. (2008). Cross-sectional versus
longitudinal survey research: Concepts, findings, and guidelines. Journal of
Marketing Research, 45(3), 261-279.
Schamel, G. (2012). Weekend vs. midweek stays: Modelling hotel room rates in a small
market. International Journal of Hospitality Management, 31(4), 1113-1118.
Schuckert, M., Liu, X., & Law, R. (2015). Hospitality and tourism online reviews: Recent
trends and future directions. Journal of Travel & Tourism Marketing, 32(5), 608-621.
Stanley, T. D., & Doucouliagos, H. (2012). Meta-Regression Analysis in Economics and
Business. New York: Routledge.
Wang, M., Lu, Q., Chi, R. T., & Shi, W. (2015). How word-of-mouth moderates room price
and hotel stars for online hotel booking an empirical investigation with Expedia
data. Journal of Electronic Commerce Research, 16(1), 72.
Xiang, Z., Du, Q., Ma, Y., & Fan, W. (2017). A comparative analysis of major online review
platforms: Implications for social media analytics in hospitality and tourism. Tourism
Management, 58, 51-65.
Xie, K. L., Chen, C., & Wu, S. (2016). Online consumer review factors affecting offline hotel
popularity: evidence from tripadvisor. Journal of Travel & Tourism Marketing, 33(2),
Xie, K. L., Zhang, Z., & Zhang, Z. (2014). The business value of online consumer reviews
and management response to hotel performance. International Journal of Hospitality
Management, 43, 1-12.
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., Mueller, N. J., & Croes, R. R. (2016). Market accessibility and hotel prices in the
Caribbean: The moderating effect of quality-signaling factors. Tourism
Management, 56, 40-51.
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.
Ye, Q., Law, R., Gu, B., & Chen, W. (2011). The influence of user-generated content on
traveler behavior: An empirical investigation on the effects of e-word-of-mouth to
hotel online bookings. Computers in Human Behavior, 27(2), 634-639.
You, Y., Vadakkepatt, G. G., & Joshi, A. M. (2015). A meta-analysis of electronic word-of-
mouth elasticity. Journal of Marketing, 79(2), 19-39.
Figure 1. Conceptual framework of determinants of eWOM elasticity on hotel performance.
Figure 2. Histogram and kernel density of calibrated elasticity from extant literature.
Table 1. Coding scheme and description of independent variables.
midpoint of research period (in year)
time of study
1 = hotel properties in China; 0 = other
1= all types, 2 = low, 3 = mid, 4 = high
classes of hotels sampled
1 = panel data; 0 = other
whether data cover multiple time periods
log of sample size in the model
size of data sample
1= weekly, 2 = monthly (reference category), 3 =
quarterly, 4 = yearly
frequency of data observations
1 = TripAdvisor-based eWOM metric; 0 = other
platform effect of eWOM
1 = number of reviews to proxy hotel sales as a
performance measure; 0 = other
measurement of sales
1= valence-based, 2 = volume-based
type of eWOM elasticity
1 = including price or average room rate as a
control variable; 0 = other
omitted variable problem associated with
1 = specifying lag of independent variables; 0 =
dynamic specification of independent
1 = including other online eWOM measures in
the empirical model; 0 = other
omitted variable problem associated with
other eWOM metrics
1 = linear function form (lev-lev); 0 = other
functional form of the model
number of independent variables in the empirical
complexity of the model
1 = published in a research journal with ISSN; 0
Table 2. Descriptive statistics of continuous independent variables.
Table 3. Descriptive statistics of categorical independent variables.
class = all
class = low
class = mid
class = high
freq = weekly
freq = monthly
freq = quarterly
freq = yearly
elasticity_type = valence
elasticity_type = volume
Table 4. Estimation results of HLM meta-analysis.
class = low
class = mid
class = high
freq = weekly
freq = quarterly
freq = yearly
# of observations
# of studies
(Notes: *** indicates significance at the 0.01 level, ** indicates significance at the 0.05 level, * indicates
significance at the 0.1 level. Robust standard errors are presented in parentheses. † indicates that the sample
excludes elasticities using review counts as hotel performance proxy.)