Content uploaded by Kevin Yili Hong
Author content
All content in this area was uploaded by Kevin Yili Hong on May 11, 2016
Content may be subject to copyright.
Culture, Conformity and Emotional Suppression in Online Reviews
Yili Hong, Arizona State University
Ni Huang, Temple University
Gord Burtch, University of Minnesota
Chunxiao Li, Arizona State University
Abstract
This study examines the cultural background of consumers as an antecedent of online review
characteristics. We theoretically propose and empirically examine the effect of cultural
background, specifically individualism (versus collectivism), on the tendency of consumers to
conform to prior opinion and the emotionality of the review text. We also examine how
conformity and emotionality relate to review helpfulness. Our hypotheses are tested using a
unique dataset that combines online restaurant reviews from TripAdvisor with measures of
individualism–collectivism values. Our econometric analyses reveal that consumers from a
collectivist culture are less likely to deviate from the average prior rating and to express emotion
in their reviews. Moreover, those reviews that exhibit high conformity and intense emotions are
perceived to be less helpful. We also present several important implications for the management
of online review platforms in light of these findings, which reflect the previously unidentified
drivers of systematic differences in the characteristics of online reviews.
Keywords: culture, online reviews, individualism value, rating deviation, review emotion
Journal of the Association for Information Systems, forthcoming
1
1. Introduction
Online reviews have been the focus of a considerable number of studies in various business
disciplines, particularly Information Systems and Marketing. Several studies have noted that the
effect of online reviews greatly depends on their characteristics. Specifically, negative reviews
tend to be more influential than positive reviews (Chevalier and Mayzlin 2006), whereas the
expression of emotion by an author can affect the perceived helpfulness of his/her review (Yin et
al. 2014) and consumer conversion (Ludwig et al. 2013). Moreover, the disagreement among
prior reviews (e.g., higher variance in star ratings) can have varying effects on product sales and
the characteristics of subsequent reviews (Nagle and Riedl 2014, Sun 2012). Interestingly, only
few studies have explored the characteristics of review authors as possible antecedents of review
content.
To extend prior literature on the antecedents of online reviews (Goes et al. 2014, Huang et al.
2016), this study focuses on the potential role of the cultural background of reviewers
(particularly individualism vs. collectivism values)1. In the process, this study answers the recent
calls for research on the cross-cultural differences in the production of electronic word of mouth
(eWOM) (King et al. 2014). Anecdotal and scientific evidence jointly suggest that cultural
differences have significant potential to explain the variations in review characteristics. By
evaluating the Amazon marketplaces in the United Kingdom (U.K.), Japan, Germany, and the
United States (U.S.), Danescu-Niculescu-Mizil et al. (2009) have observed very “noticeable
differences between reviews” in terms of their average helpfulness and rating variance. Similar
results have been reported in few studies that have examined the cross-cultural differences in the
1 We focus on individualism versus collectivism because this cultural dimension has a particularly strong
relationship with expression of opinion (Huang 2005).
2
production and consumption of online reviews (Chung and Darke 2006, Fang et al. 2013, Koh et
al. 2010). For instance, consumers from collectivist cultures are less likely to write reviews with
low valence (i.e., 1-star ratings) (Fang et al. 2013). Under-reporting biases, which refer to the
tendency of an author to write reviews following extreme experiences, are more prevalent among
consumers from individualist cultures (Koh et al. 2010). Consumers from individualist cultures
are more likely to write reviews for products or services that enable self-expression (Chung and
Darke 2006). However, many questions remain despite these contributions to our understanding
of the role of culture in the review process. According to King et al. (2014, p.175),
“Understanding these differences and being able to adapt the review process to meet these needs
are critical to retailers, so that they can design systems that provide this information in the best
manner possible.”
The majority of the studies on individualism versus collectivism values have focused on their
implications on the tendency of an individual to conform or stand out. Accordingly, we focus on
the following characteristics of online reviews that are directly linked to conformity and are
likely to be influenced by the individualist or collectivist cultural values of an author: i)
conformity to (or deviation from) prior opinion and ii) emotional suppression (or expression).
We address the following questions:
• How does individualism (collectivism) influence deviation from (conformity to) prior
opinion in online reviews?
• How does individualism (collectivism) influence emotional expression (suppression) in
online reviews?
• In turn, how do these cultural influences affect the perceived helpfulness of online
reviews?
3
While Americans say, “the squeaky wheel gets the grease,” the Japanese say, “the nail that
stands out gets pounded down.”2 Such variation in cultural values is not merely anecdotal, as it
has received much attention in the academic literature. For example, researchers in cultural
psychology (Hofstede 2001, House et al. 2004) have systematically documented that individuals
from collectivist cultures are more likely to exhibit conformity to group opinion (Bond and
Smith 1996, Ng et al. 2000) and are less likely to express emotion (Butler et al. 2007, Niedenthal
et al. 2006). These observations suggest that online reviews written by consumers from
collectivist cultures are less likely to deviate from prior opinion and less likely to include
emotional expressions.
This paper empirically evaluates these expectations to extend the literature and answer the
calls for research into cross-cultural differences in eWOM (King et al. 2014). First, our work
builds on the small body of literature that addresses the cultural differences in the production of
online reviews by using data on users from various countries and cultural backgrounds. This
technique contrasts those of the majority of previous studies, which have mostly relied on two-
country designs (e.g., comparing American and Chinese consumers), thereby limiting the
generalizability of their findings (Fang et al. 2013, Koh et al. 2010). Second, previous studies
have considered the role of cultural differences in determining the volume and valence of
reviews in an absolute sense (Fang et al. 2013). We extend such work by considering the self-
group differences in online reviews (i.e., relative valence in terms of deviation from prior
opinion) and emotional expression.
Drawing on the cultural psychology literature, we formulate and evaluate several hypotheses
using a unique dataset that integrates online restaurant reviews from TripAdvisor.com with
2 http://www.nytimes.com/1990/12/25/science/the-group-and-the-self-new-focus-on-a-cultural-rift.html
4
country-level measures of individualism–collectivism values (House et al. 2004). We then
estimate the effects of these values on the measures of review conformity and emotional
suppression. We also examine the subsequent effect of the characteristics of reviews on their
perceived helpfulness. We obtain three key findings. First, consumers from countries with a
higher level of individualism are more likely to deviate from the prior average rating when
writing a review. Second, these reviewers are more likely to express their emotions in the review
text. Third, conformity and emotional expression generally have a negative relationship with
review helpfulness3.
Our work offers important practical implications for online review platforms. First, recent
studies suggest that the approaches being used by many leading review websites to aggregate
reviews (e.g., averaging) tend to ignore reviewer-specific differences in the production of
reviews (Dai et al. 2012). However, our findings reveal previously undocumented systematic
differences in reviewer culture that should be considered by review websites when aggregating
reviews. Second, several features that improve or damage the perceived helpfulness of online
reviews (in terms of “helpful” votes) are more likely to systematically manifest when consumers
come from a particular culture. Therefore, online practitioners, who are cognizant of these issues,
must consider approaches that encourage or deter certain review characteristics. For example,
Yelp offers mobile users with “example” reviews to encourage them to produce longer and
informative content. Based on the location or review history of an individual, a similar strategy
may be proposed to encourage individuals to include or exclude textual features that do or do not
contribute to a “helpful” review.
3 Whenever we distinguish positive from negative emotions, we observe a more nuanced story in which
negative emotions are positively related with helpfulness, which is in line with the literature.
5
The rest of this paper is structured as follows. First, we review the previous studies on online
reviews, particularly those that focus on conformity and emotional suppression. We specifically
focus on the cultural psychology literature that deals with conformity, language use, and
emotional suppression. Second, we propose several hypotheses for empirical examination. Third,
we present the research methodology, report the empirical analyses results, and discuss the
implications and limitations of our work.
2. Literature Review
2.1. Online Reviews
Relative to traditional mass communication, bi-directionality is a unique aspect of online
reviews, thereby emphasizing the need to study the antecedents and consequences of these
reviews (Dellarocas 2003, Goes et al. 2014). Online reviews enable consumers to share their
evaluations and opinions of products or services to an extremely large audience (Dellarocas
2003, Lee and Bradlow 2011, Lu et al. 2013). Following the pioneering works of Ba and Pavlou
(2002) and Dellarocas (2003), many studies from the information systems field have begun to
investigate the downstream effects of reviews in terms of sales (Li and Hitt 2008), helpfulness
(Mudambi and Schuff 2010), and market competition (Kwark et al. 2014). We consider the
antecedents of review characteristics, which have received relatively less attention in the
literature (Goes et al. 2014), by focusing on the textual characteristics of reviews and the
conformity of reviewers to (or deviation from) the prior average.
Recent studies have reported evidence on the broad conformity of reviewers (Muchnik et al.
2013, Lee et al. 2014, Wang et al. 2015). Muchnik et al. (2013) have experimentally
demonstrated that those individuals who are exposed to a positive prior rating have an increased
6
probability of submitting a positive rating. Similarly, Wang et al. (2015) and Lee et al. (2014)
have reported that the opinions of individuals are positively correlated with those of their friends.
However, contrary to reactance theory (Brehm and Brehm 1981), a related stream of research
(Wu and Huberman 2008, Moe and Schweidel 2012, Godes and Silva 2012) has revealed that
some individuals are motivated to “stand out” from the crowd by deviating from the opinions of
others. Conditional on a purchase, a consumer will decide whether to post a review. Wu and
Huberman (2008) have argued that consumers are motivated, at least in part, by the expected
influence of their reviews on the average rating and, implicitly, on the actions or preferences of
others. These researchers have empirically revealed that buyers are most likely to post reviews
when the expected effect is high (i.e., when only few reviews are present or when their
experience extensively deviates from the prevailing average).
Only a few studies have investigated the textual characteristics of reviews, and the
majority of these works have focused on the consequences of textual features. Several textual
features affect review helpfulness and product sales. For example, Goes et al. (2014) have shown
that consumers who are more popular in a review community tend to write highly objective
reviews. Yin et al. (2014) have demonstrated that certain types of negative emotions (i.e.,
anxiety) are likely to be perceived as more helpful than other emotions (i.e., anger). Ahmad et al.
(2015) have considered the relationship between different types of expressed emotions (i.e, hope,
happiness, anxiety, and disgust) and the perceived helpfulness of reviews, and observed
differential effects across each emotion. Ghose et al. (2011) have reported that spelling mistakes
and review subjectivity are negatively associated with helpfulness and product sales. We build
on the review text literature by considering the antecedent of review emotion, namely, the
7
individualism value of the review author. The following section reviews the literature on cultural
values, conformity, and language use.
2.2. Cultural Values, Conformity, and Language Use
National cultural dimensions, such as those introduced by Robert House and Geert Hofstede,
have been used to study various phenomena in information systems (Leidner and Kayworth
2006). However, only few studies have explored the role of cultural values in online reviews
(Chung and Darke 2006, Koh et al. 2010, Fang et al. 2013). These researchers have aimed to
contrast the review authorship or consumption between individuals residing in a collectivist
country and those residing in an individualist country. Chung and Darke (2006) found that self-
relevance has a greater effect on the user-generated content in individualist cultures than that in
collectivist cultures. Koh et al. (2010) found that underreporting is more prevalent among U.S.
customers than among Chinese or Singaporean customers. Fang et al. (2013) reported on a
number of several descriptive differences between American and Chinese reviewers. For
example, Chinese reviewers provide more positive reviews and place a higher weight on
negative reviews.
Although previous studies have explored the differences in the behavior of individuals
from various cultures, two notable aspects are yet to be considered, namely, opinion conformity
and emotional suppression. Both aspects tend to differ across cultures, particularly with respect
to collectivism versus individualism. First, with respect to conformity, many studies have
reported that individuals from collectivist cultures are more likely to conform in judgment and
evaluation (Bond and Smith 1996), behavior (Cialdini et al. 1999), and opinion (Huang 2005).
Second, with respect to emotional expression, several studies have determined that people from
8
individualist cultures are more likely to express emotions (Takahashi et al. 2002), while those
from collectivist cultures are more likely to suppress emotion (Niedenthal 2006), particularly
negative emotion (Butler et al. 2007).
3. Hypothesis Development
We propose several research hypotheses for empirical testing. We divide the research framework
into several components. The antecedents are examined in the first stage, wherein we propose the
formal hypotheses about the effects of the cultural backgrounds of consumers (countries that
exhibit higher levels of individualism versus collectivism) on review characteristics (rating
deviation and review textual characteristics). We empirically examine in the second stage the
potential relationships between review characteristics and perceived review helpfulness. Figure 1
presents the research framework.
Figure 1. Research Framework
3.1. Cultural Background and Review Characteristics
By considering the effects of cultural background on dissenting opinions and emotion
expression, we focus on the distinction between individualism and collectivism values as well as
9
on behaviors that are relevant to i) online review authorship and ii) individualist–collectivist
cultural values.
Collectivist values are generally characterized by a preference for preserving harmony,
avoiding confrontation, and promoting conformity; individual initiatives and deviations from the
dominant opinion of the group are thereby discouraged (Hofstede 2001, House et al. 2004).
Conformity to group pressure has been documented in experiments since the 1950s (Asch 1955).
Through a meta-analysis of 133 conformity studies similar to that of Asch, scholars have also
systematically verified that conformity effects are much stronger among individuals from
collectivist cultures (Bond and Smith 1996). Similarly, other studies have observed greater
conformity among individuals from collectivist cultures in terms of actual behavior (Cialdini et
al. 1999) and opinion formation (Huang 2005). These findings have a direct bearing on our study
context through their suggestions that those reviews written by individuals from collectivist
(individualist) cultures are more likely to conform to (deviate from) prior opinions.
In the online reviews context, when a consumer writes a review about a merchant, he/she
may feel pressured to “conform” (Muchnik et al. 2013, Lee et al. 2014, Wang et al. 2015) to the
group opinion as expressed in previous reviews about the same merchant. This behavior tends to
emerge because the current average rating of a merchant is prominently shown on the webpage
of a review website and may serve as an anchor for subsequent consumers (Adomavicius et al.
2013). Yaveroglu and Donthu (2002) have argued that individuals with from collectivist cultures
(e.g., China and Japan) are more likely to conform to the views of others to fit in, gain social
understanding, and be accepted by others in a group. Those societies that espouse collectivist
values encourage social harmony and bonding within groups (Triandis 1995, Lam et al. 2009).
10
Therefore, we anticipate that consumers from collectivist cultures are more likely to demonstrate
review conformity.
In some cases, a consumer may also observe and deviate from prior opinion (Moe and
Schweidel 2012, Wu and Huberman 2008). As discussed in our review of the cultural
psychology literature, countries with high individualism values encourage individual autonomy
and individualist behavior as well as discourage conformity. Therefore, individuals from
individualist cultures are likely to be more “opinionated” because they want to stand out from the
others or to have their voices heard. Accordingly, individuals from countries with high
individualist values are expected to deviate from prior opinion. Thus, we propose the following:
Hypothesis 1a: On average, those ratings that are submitted by consumers from individualist
(versus collectivist) cultural backgrounds are more likely to deviate from (less likely to
conform to) the prior average rating.
Several studies in the information systems literature have also examined the role of
online review text. Early studies in this line of research have reported the influence of textual
content over and above numerical ratings (Pavlou and Dimoka 2006, Chevalier and Mayzlin
2006). Recent studies have considered the effects of basic textual features, such as readability
and spelling mistakes (Ghose and Ipeirotis 2011, Goes et al. 2014), on review helpfulness. Other
studies have explored highly nuanced features, such as semantic style (Cao et al. 2011) and
objectivity versus subjectivity (Ghose and Ipeirotis 2011).
Scholars have recently examined review texts to identify their emotional and affective
content (Ludwig et al. 2013, Yin et al. 2014). They have revealed that such content can strongly
affect the perceived helpfulness of a review and its influence on customer conversion. One
11
natural extension is to explore the individualist–collectivist cultural background of a review
author as a potential antecedent of emotional content in a review.
The cultural psychology literature includes several studies that suggest a strong
relationship between culture and emotion. The literature has reported that the tendency toward
emotional expression differs according to the cultural background of an individual. Previous
studies have shown that individuals from individualist cultures tend to be more vocal and
expressive, whereas those from collectivist cultures speak in ways intended to maintain harmony
and avoid controversy (e.g., using indirect language) (Holtgraves 1997). Some studies have
demonstrated that people from collectivist cultures tend to suppress or withhold their emotions
when communicating with others (Butler et al. 2007, Niedenthal et al. 2006), whereas those from
individualist cultures are more likely to express their emotions, particularly negative emotions
(Takahashi et al. 2002). These tendencies manifest early in life because children are socialized to
meet the standards of their culture (Friedlmeier et al. 2011).
The public expression of emotions is generally considered acceptable in individualist
cultures, but is generally frowned upon in collectivist cultures. Tsai et al. (2007) have found that
high arousal states, such as excitement and enthusiasm, are typically supported in American
culture because these emotions are more effective in influencing others. By contrast, collectivist
cultures espouse low arousal states, such as calmness, which are better suited to adapting to and
accommodating others. When writing an online review, a consumer expresses an opinion and
performs an evaluation, which often involves a public display of emotion (Yin et al. 2014).
Accordingly, these studies have suggested that the reviews written by individuals from
individualist cultures are more likely to contain emotions. Therefore, we propose the following:
12
Hypothesis 1b: Consumers from individualist cultural backgrounds tend to express more
emotions in their reviews.
3.3. Review Characteristics and Helpfulness
Review helpfulness (generally measured by “helpful” votes) has important implications for both
review curators and consumers. To draw practical implications from our study, one must
understand how the systematic differences in reviewer behavior may be associated with the
perceived helpfulness of reviews. Consumers generally seek different opinions toward the same
restaurant prior to consumption in order to assess whether such establishment can match their
tastes (Sun 2012, Hong et al. 2013, Liu et al. 2014). Ratings that deviate (either positively or
negatively) from prior opinion are likely to stand out and offer unique information by presenting
the “other side of the argument” (Cao et al. 2011). Indeed, previous studies have presented
consistent evidence that negative reviews, in particular, are likely to be perceived as more helpful
because of a “negativity bias,” that is, negative reviews tend to be seen as more informative
(Mudambi and Schuff 2010, Chen and Lurie 2013). Similarly, we anticipate that rating deviation
(i.e., extreme valence relative to past reviews) will result in the higher perceived helpfulness of
the review. Therefore, we propose the following:
Hypothesis 2a: Rating deviation is positively associated with review helpfulness.
Several pioneering studies have also employed text mining techniques vis-à-vis the effect of
review content. Lee et al. (2013) have determined that the presence of informational content in a
message may be more or less useful depending on the type of product that is being considered.
Ghose and Ipeirotis (2006) have found that objective content is more helpful than subjective
content. Considering these past studies and the idea that emotions tend to be perceived as less
13
rational or objective, reviews that contain greater expressions of emotion may be perceived as
less helpful by consumers. Therefore, we propose the following:
Hypothesis 2b: Review emotion is negatively associated with review helpfulness.
4. Research Methodology
4.1. Data
We collected data from several archival data sources (Table 1).
Table 1. Archival Data Sources
Data
Source
Review, Reviewer data
TripAdvisor
Review Emotion
TripAdvisor reviews processed by Linguistic Inquiry and Word
Count (LIWC)
Cultural Values
House et al. (2004); World Value Survey
First, we collected online reviews from a leading review platform, TripAdvisor
(www.tripadvisor.com), spanning the years 2003 to 2014 using a web crawler. Our data included
online reviews for approximately 3,750 restaurants located in six major U.S. cities, namely,
Chicago, Houston, Los Angeles, New York, Phoenix, Philadelphia, and Seattle. We ensured data
accuracy by manually verifying a randomly selected set of 150 reviews. Figure 2 presents a
screenshot of a review in TripAdvisor.
14
Figure 2. Screenshot of a Sample TripAdvisor Review
Our panel was constructed by collecting the entire review history of each restaurant and
ordering the reviews based on their time stamps. From each review, we obtained the star rating,
sequence (order) position, time stamp, and actual review text. We also obtained data on the
characteristics of the review authors, including their historical reviewing activity, website
registration date, and country of residence. We then measured emotional expression by
examining the review text in an automated fashion using the text mining tool, Linguistic Inquiry
and Word Count (LIWC), which will be described further in Section 4.2.
Second, we collected data on cultural values from several sources based on prior
literature. Several scholars and institutions have attempted to measure national cultural values
over the years. The cultural value data collected by Hofstede (2001) have been used extensively
by researchers from various fields; however, these data are subject to severe limitations because
they have been collected from a selected group of IBM employees, thereby introducing biases.
House et al. (2004) have provided a more detailed set of cultural value measures that includes
collectivism versus individualism. These latest measures have been widely used over the recent
years (e.g., Schoorman et al. 2007).
15
Researchers in other disciplines have also operationalized cultural values based on the
World Values Survey (WVS) (e.g., Giannetti and Yafeh 2012, Burtch et al. 2014, Hong and
Pavlou 2014)4. By analyzing the results of WVS, Inglehart and Welzel (2010) have observed that
more than 70% of the variance in responses can be explained by two factors, one of which is
survival versus self-expression (the extent to which a society emphasizes values related to
survival as opposed to self-expression); these factors capture much of the same information as
that captured by the collectivism–individualism measure of Hofstede and House et al. (Inglehart
and Oyserman 2004).
The culture measure of House et al. (2004) has been considered the most up-to-date and
comprehensive because this measure builds on Hofstede (2001), Inglehart (1997), and several
other cultural studies. Therefore, we focused on this measure in our primary analysis and
subsequently performed a series of robustness checks using the measures of Hofstede et al. and
WVS. We assigned a consumer (review author) with an individualism value score based on
his/her self-reported country of residence.
4.2. Key Measures
4.2.1. Dependent Variables:
Rating Deviation: Rating deviation is measured as the absolute difference between the rating of
a focal review and the average prior rating. TripAdvisor uses a half-star average rating system;
therefore, the published average ratings fall within the set (1, 1.5, 2, 2.5, 3, 3.5, 4, 4.5, 5). To
4 The survey was started in 1981 and the latest wave is Wave 6, which consists of nationally
representative surveys that are conducted in almost 100 countries. The respondents include almost 90% of
the world’s population who are surveyed using a common questionnaire. WVS currently includes
interviews with almost 400,000 respondents. We obtained data based on one of the two salient
dimensions on the Inglehart–Welzel Cultural Map (2010): survival versus self-expression values.
16
compute for deviation, we reconstructed the average restaurant rating at the time immediately
before the focal review (nth position in the sequence) as follows: !
"#$
"%$ & !'
"%$
'($ , . Afterward,
we obtained the observed average rating, !
", by rounding !
" to the nearest half star. For example,
for !
"# )*+,, !
"=3 (3 stars); for !
"# )*+-, !
"= 3.5 (three and a half stars); and for !
"# )*.-,
!
"= 4 (four stars). In cases where !
"# )*+/, the value was rounded to 3.5. Rating deviation
(distance between the nth rating !
" and the observed prior average rating !
") can be written as
follows:
[1] 01234567893123:4"#1;< !
"= !
".
As a robustness check, we considered the unrounded prior average and formulated an
alternative measure of deviation; we obtained almost identical results using the unrounded
measure (in terms of the magnitudes and statistical significance of the parameter estimates).
However, this result is expected because the actual and observed (rounded) deviation have a 99%
correlation.
Review Emotion: We utilized LIWC, text analysis software, for identifying sentiment and
emotion in textual content, to obtain the measures of emotion (e.g., happy, cried, and abandon),
positive emotion (e.g., love, nice, and sweet), and negative emotion (e.g., hurt, ugly, and nasty)
(Pennebaker et al. 2001). LIWC has recently attracted frequent use in the information systems
and marketing literature (Sridhar and Srinivasan 2012, Yin et al. 2014, Goes et al. 2014). Before
calculating the textual measures, we cleaned the textual data to remove special characters. Using
LIWC, we operationalized the review emotion measures as the percentage of emotional (overall,
positive, and negative) words out of the total number of words.
17
Review helpfulness: In line with prior literature (Mudambi and Schuff 2010, Chen and Lurie
2013), we measured review helpfulness in terms of the total number of “helpful” votes received
by a review. Given the highly skewed distribution of votes, we used the log transformation of the
raw value in our analyses.
4.2.2. Independent Variables:
Individualism–Collectivism Values: We utilized the collectivism/individualism data from House
et al. (2004). These data, which are based on survey responses from 17,300 individuals, are
highly consistent with the “survival versus self-expression” measure of WVS and the
individualism measure from Hofstede (2001). The collectivism data from House et al. measure
the degree to which individuals express pride, loyalty, and cohesiveness in their organizations or
families. We employed the negative value of collectivism to measure its polar opposite,
individualism. Therefore, higher values of collectivism indicate a greater individualism or lesser
collectivism. We plotted the data from House et al. and the self-expression measure of Inglehart
and Welzel (2010) based on the most recent wave of WVS. Figure 3 shows that Sweden, the
Netherlands, New Zealand, the U.K., Denmark, Germany, and the U.S. rank highly on
individualism, whereas Russia, China, Georgia, Morocco, Zimbabwe, Hungary, and Albania
rank low on individualism. An additional note is warranted here. Cultural values are frequently
assumed as country-level constructs that are inherited by the members of a country. For example,
the Chinese or Japanese are generally less assertive and more prone to conformity than
Americans. However, the cultural values of individuals within a country may vary because of
individual heterogeneity and immigration. Nonetheless, similar to prior literature, we assumed
that the inheritance of cultural values holds for the majority of the residents of a country because
these values are embraced by that society. Therefore, to avoid confusion with respect to the level
18
of measurement and the ecological fallacy, we referred to the cultural backgrounds of the
subjects instead of their cultural values.
Figure 3. Individualism (versus Collectivism) Value by Countries
Travel Experience: We measured the travel experience of a consumer as the number of countries
that he/she has traveled to as reflected in the TripAdvisor data. Travel experience indicates the
exposure of an individual to different cultures. Traveling to different countries allows an
individual to encounter people of different cultural backgrounds, thereby making him/her more
receptive to other cultures and exhibit only few of the systematic differences that we have
hypothesized. Given its skewness, we log-transformed the travel experience variable.
Prior Review Volume: Prior review volume may affect rating deviation because late arrivals (in
terms of the sequence of reviews written for a restaurant) may have different motivations and
preferences than the early adopters. For example, a late arriver may have a higher motivation to
deviate from prior opinion to make his/her review “stand out.”
19
Average Rating: We controlled for the average rating of a consumer because prior research has
noted that some consumers are systematically more positive or negative in their reviewing
behavior (Dai et al. 2012).
Consumer Tenure: We controlled for consumer tenure (number of months since website
registration) for several reasons. First, consumers may grow more positive or negative as they
accumulate review experience. We log-transformed this variable in our analyses because of its
skewed distribution.
Review Age: We controlled for review age (number of days since the review has become live
and available for consumer voting) because older reviews are exposed to more viewers and have
a greater opportunity to accrue helpful votes.
Time Effects: We controlled for time effects by employing monthly dummy variables. The
reviews written at different periods may be systematically different because of unobserved
shocks or trends (e.g., degradation in restaurant quality).
Tables 2 and 3 present the descriptive statistics and correlation matrix of our key variables,
respectively.
Table 2. Descriptive Statistics
Variable
Mean
St.d.
Min
Max
Median
1. Rating deviation
0.79
0.67
0
4.50
0.5
2. Review emotion
8.72
6.52
0
100
7.41
3. Positive emotion
7.96
6.53
0.00
100
6.67
4. Negative emotion
0.74
1.73
0.00
100
0
5. Prior volume
181.46
269.40
1.00
2561
83
6. Individualism
–4.32
0.35
–6.37
–3.46
–4.22
7. Experience
9.52
11.23
1
207
5
8. Average rating
3.77
1.27
1
5
4.1
9. Consumer tenure
28.57
27.79
0
139
21
20
Table 3. Correlation Matrix
Variable
1
2
3
4
5
6
7
8
9
1.Rating deviation
1.00
2.Review emotion
-0.11
1.00
3.Positive emotion
-0.16
0.96
1.00
4.Negative emotion
0.21
0.11
-0.15
1.00
5.Prior volume
-0.05
-0.03
-0.03
-0.01
1.00
6.Individualism
0.01
0.04
0.04
0.02
-0.01
1.00
7.Experience
-0.05
0.01
0.01
0.00
-0.01
-0.14
1.00
8.Average rating
-0.15
-0.07
-0.05
-0.07
0.06
0.00
0.11
1.00
9.Consumer tenure
-0.06
-0.05
-0.05
-0.01
0.05
0.03
0.22
0.22
1.00
4.3. Empirical Model
Although cultural values are operationalized at the country level, assigning these values to
consumers is reasonable in this scenario for several reasons. First, those consumers who are born
and raised in a particular country are likely to inherit the cultural values of that country. Second,
the interaction between country-level cultural values and consumer-level travel experience can
help us further identify the effects of cultural values.
We identified the effects of cultural values, travel experience, and the interaction between
these two by examining within-restaurant variance in reviews via within transformation (i.e., a
standard fixed effect estimation (>?@6AB)) while controlling for time effects via dummy variables
(CD& EDF ). Additionally, we controlled for consumer-level heterogeneity using the
abovementioned controls. We formulated the estimation equations for rating deviation and votes
as follows. In these two equations, i indexes consumers, j indexes restaurants, and t indexes time;
AB is the restaurant fixed effect that controls for restaurant-level, time-invariant unobserved
factors; and CD& EDF is the vector of monthly time dummies. The key parameters of interest
are GH IH 14J6K*6
21
L+M601234567893123:4'BD
# N$&34J393JO1P3<Q'R6 NS&TU6V8WX8!384Y8'Z R N[
&34J393JO1P3<Q &TU 8WX8!384Y8 'R N\&]^_`^a6^bcd`cU'DB R AB
R CD& ED
F
R Y:42!:P'BD R e'BD6
) 608938f6?Q:23:4'BD
# g$&34J393JO1P3<Q'R6 gS&TU6V8WX8!384Y8'Z R g[
&34J393JO1P3<Q &TU6V8WX8!384Y8Z 'R g\&]hd`Ui6j^_`hd`cU'DB R AB
R CD& ED
F
R Y:42!:P'BD R e'BD6
, 6TU k8PXlOP48<< 'B # m$&01234567893123:46'B R6 mS&P4f:!J<'B R m[&
01234567893123:46'B &P4f:!J<'B R m\&08938f6?Q:23:4'B R mn&08938f6o58'B R ABR
p:42!:P'B R e'B
4.4. Estimation Results and Hypotheses Testing
This section reports the estimation results of our main analyses. Following the structure of the
hypothesis development, we began by examining the effects of individualism on rating deviation.
As shown in Table 4, individualism values significantly increase rating deviation, thereby
offering clear support for Hypothesis 1a.
Table 4. Effect of Individualism Value on Rating Deviation
DV:
(1) Rating Deviation
(2) Rating Deviation
(3) Rating Deviation
Individualism
0.026***(0.004)
0.015***(0.005)
0.036***(0.014)
ln(experience)
–0.032***(0.002)
–0.069***(0.022)
Individualism * ln(experience)
–0.009*(0.005)
ln(prior volume)
0.000(0.007)
0.000(0.007)
Average rating
–0.066***(0.002)
–0.066***(0.002)
ln(consumer tenure)
–0.004***(0.001)
–0.004***(0.001)
Review emotion
–0.014***(0.000)
–0.014***(0.000)
Constant
1.498*(0.832)
0.000(0.007)
0.036***(0.014)
Restaurant FE
Yes
Yes
Yes
Time effect
Yes
Yes
Yes
Observations
256,810
256,810
256,810
R-squared (within)
0.015
0.057
0.061
# of restaurants
3,735
3,735
3,735
Notes: Robust standard errors are enclosed in parentheses. Std. Err. is adjusted for clusters in restaurants.
The coefficients are significant at levels *** p < 0.01, ** p < 0.05, and * p < 0.1.
22
Although consumers in countries that promote individualist values tend to deviate from
the prior average rating, a variation may still exist among consumers within the same country.
For example, some consumers from the U.S. may be more conformist, whereas some consumers
from China may be more individualistic. This variation in cultural values may be attributed to
travel experience, which potentially exposes people to different cultural values. Such exposure
makes people more tolerant of other worldviews. Therefore, we further examined the potential
moderating role of consumer travel experience on the relationship between individualism values
and rating deviation. In particular, we calculated the marginal effects and conducted a spotlight
analysis (Spiller et al. 2013) to assess both the main and interaction effects. As can be seen in
Figure 4, first, the main effect of individualism on rating deviation remains positive across the
spectrum of values for different travel experiences. Second, travel experience significantly
moderates the effect of individualism on rating deviation. In sum, as individuals gain travel
experience, they are potentially exposed to different cultures and become less affected by their
own cultural backgrounds.
Figure 4. Spotlight Analysis of the Interaction Effect on Rating Deviation
23
Table 5. Effect of Individualism Value on Review Emotion
(1)
(2)
(3)
(4)
(5)
(6)
(7)
(8)
(9)
DVs:
Overall
Emotion
Overall
Emotion
Overall
Emotion
Positive
Emotion
Positive
Emotion
Positive
Emotion
Negative
Emotion
Negative
Emotion
Negative
Emotion
Individualism
0.518***
0.553***
1.452***
0.440***
0.480***
1.295***
0.086***
0.081***
0.177***
(0.052)
(0.052)
(0.128)
(0.051)
(0.050)
(0.122)
(0.011)
(0.011)
(0.026)
ln(experience)
0.014
–1.585***
–0.002
–1.452***
0.017***
–0.153***
(0.013)
(0.194)
(0.013)
(0.188)
(0.004)
(0.044)
Individualism
–0.375***
–0.340***
–0.040***
*ln(experience)
(0.045)
(0.044)
(0.010)
ln(prior volume)
0.240***
0.241***
0.225***
0.226***
0.018
0.018
(0.046)
(0.046)
(0.045)
(0.045)
(0.011)
(0.011)
Average rating
–0.094***
–0.095***
–0.041***
–0.042***
–0.053***
–0.053***
(0.014)
(0.014)
(0.014)
(0.014)
(0.004)
(0.004)
ln(consumer
–0.151***
–0.150***
–0.160***
–0.160***
0.010***
0.011***
tenure)
(0.009)
(0.009)
(0.009)
(0.009)
(0.002)
(0.002)
Rating deviation
–0.971***
–0.972***
–1.410***
–1.410***
0.439***
0.439***
(0.020)
(0.020)
(0.021)
(0.020)
(0.007)
(0.007)
Constant
13.370***
10.456***
14.286***
12.466***
10.527***
13.997***
0.954***
–0.048
0.360
(0.220)
(1.682)
(1.741)
(0.215)
(1.436)
(1.503)
(0.047)
(0.316)
(0.331)
Restaurant FE
Yes
Yes
Yes
Yes
Yes
Yes
Yes
Yes
Yes
Time effect
Yes
Yes
Yes
Yes
Yes
Yes
Yes
Yes
Yes
Observations
256,810
256,810
256,810
256,810
256,810
256,810
256,810
256,810
256,810
R-squared
(within)
0.042
0.049
0.049
0.039
0.060
0.060
0.013
0.043
0.043
# of restaurants
3,735
3,735
3,735
3,735
3,735
3,735
3,735
3,735
3,735
Notes: Robust standard errors are enclosed in parentheses. Std. Err. is adjusted for clusters in restaurants. The coefficients are
significant at levels *** p < 0.01, ** p < 0.05, and * p < 0.1.
We then examined the effects of cultural background on review emotion, specifically
overall emotion (Column 1 of Table 5), positive emotion (Column 3), and negative emotion
(Column 5). First, consumers from individualistic cultures are more likely to express both
positive and negative emotions, thereby supporting Hypothesis 1b. All these estimated direct
effects of cultural values on review text are attenuated by travel experience, and the main and
interaction effects are visualized in Figure 5. Table 5 and Figure 5 show that consumers from an
individualist cultural backgrounds always express a higher level of overall, positive, and negative
emotions, and the positive effects are attenuated by their travel experience. Interestingly, we
observed that rating deviation is positively correlated with the presence of negative emotion yet
negatively correlated with the presence positive emotion.
24
Figure 5. Spotlight Analysis of the Interaction Effect on Review Emotion
(a) Overall Emotion (b) Positive Emotion (c) Negative Emotion
Table 6. Effect of Review Characteristics on Review Helpfulness
(1)
(2)
(3)
(4)
(5)
Rating deviation
–0.044***
–0.049***
–0.029***
–0.025***
(0.005)
(0.005)
(0.007)
(0.007)
ln(words)
0.048***
0.047***
0.050***
0.051***
(0.002)
(0.002)
(0.002)
(0.002)
Rating deviation * ln(words)
0.022***
0.023***
0.020***
0.019***
(0.001)
(0.001)
(0.002)
(0.002)
Emotion
–0.000**
0.000**
(0.000)
(0.000)
Rating deviation * emotion
–0.001***
(0.000)
Positive emotion
–0.000***
0.001***
(0.000)
(0.000)
Negative emotion
0.003***
–0.001**
(0.000)
(0.001)
Rating deviation * positive emotion
–0.002***
(0.000)
Rating deviation * negative emotion
0.003***
(0.000)
Individualism
0.005*
(0.003)
Review age
0.000***
0.000***
0.000***
0.000***
0.000***
(0.000)
(0.000)
(0.000)
(0.000)
(0.000)
Constant
–0.139***
–0.133***
–0.152***
–0.155***
0.120***
(0.011)
(0.011)
(0.011)
(0.011)
(0.013)
Restaurant FE
Yes
Yes
Yes
Yes
Yes
Observations
298,458
298,458
298,458
298,458
298,458
R-squared
0.070
0.070
0.070
0.071
0.040
# of restaurants
3,747
3,747
3,747
3,747
3,747
Notes: Robust standard errors are enclosed in parentheses, Std. Err. is adjusted for clusters in restaurants.
The coefficients are significant at levels *** p < 0.01, ** p < 0.05, and * p < 0.1.
25
Figure 6. Interaction Effect of Rating Deviation and Review Length on Helpfulness
We then examined the effects of review characteristics (rating deviation and review
emotion) on review helpfulness in terms of “helpful” votes. As shown by the regression results in
Table 6 and the plot in Figure 6, rating deviation increases the perceived helpfulness of a review,
thereby supporting Hypothesis 2a. Figure 6 also shows a positive interaction between rating
deviation and review length, which suggests that deviation exerts a greater influence when the
textual content conveys more information.
Figure 7. Interaction Effect of Rating Deviation and Review Emotion on Helpfulness
(a) Overall Emotion (b) Positive Emotion (c) Negative Emotion
Given the influence of review emotion, we observed that overall emotion has a negative
effect on review helpfulness, thereby supporting Hypothesis 2b. When we further broke down
the positive and negative emotions, we find that positive emotions lead to lower review
helpfulness, whereas negative emotions increases review helpfulness. This finding is consistent
26
with the “negativity bias” in online reviews as demonstrated in prior research (Chen and Lurie
2013). Beyond the main effects, we observed significant interaction effects between rating
deviation and review emotion (Figure 7), which indicates that positive emotion and rating
deviation have a significant negative interaction effect (substitutive effect) on review
helpfulness, and that negative emotion and rating deviation have a positive interaction effect
(complementary effect) on review helpfulness. Column 5 of Table 6 shows that those reviews
written by consumers from individualist cultural backgrounds are generally perceived to be more
helpful.
4.5. Robustness Checks
We validated the robustness of our results in several ways. In Section 4.5.1, we considered
alternative measures of cultural values by re-running our analyses using data from WVS. The
first set of robustness checks aimed to demonstrate that the observed results are not driven by the
measurement of cultural values. In Section 4.5.2, we considered an alternative estimation
approach, namely, seemingly unrelated regression (SUR), by allowing the error terms of
Equations (2) and (3) to be correlated.
4.5.1. Robustness Check 1: Alternative Measures of Cultural Background
We obtained an additional dataset on cultural values from WVS and re-estimated our models to
confirm the stability of our results. We observed a high correlation between the measures of
House et al. (2004) and WVS (ρ = 0.90) in our sample5. Given the lack of temporal variation in
the WVS data, we used the most recent set of survey responses. Our main results remain stable
regardless of our chosen measure.
5 The correlation coefficient between House et al. (2004) and WVS is 0.83 (n = 48).
27
Table 7. Robustness Check: Estimation Using Alternative Measure
DVs:
(1) Rating
Deviation
(2) Overall
Emotion
(3) Positive
Emotion
(4) Negative
Emotion
Individualism
0.027***
0.917***
0.803***
0.117***
(0.010)
(0.110)
(0.107)
(0.019)
ln(experience)
–0.017***
0.446***
0.376***
0.070***
(0.006)
(0.067)
(0.066)
(0.014)
Individualism *
ln(experience)
–0.009**
(0.004)
–0.255***
(0.039)
–0.224***
(0.038)
–0.032***
(0.008)
ln(prior volume)
0.001
0.240***
0.223***
0.019*
(0.007)
(0.046)
(0.045)
(0.011)
Average rating
–0.066***
–0.096***
–0.042***
–0.054***
(0.002)
(0.014)
(0.014)
(0.004)
ln(consumer tenure)
–0.004***
–0.154***
–0.163***
0.010***
(0.001)
(0.009)
(0.009)
(0.002)
Review emotion
–0.014***
(0.000)
Rating deviation
–0.973***
–1.412***
0.440***
(0.020)
(0.020)
(0.007)
Constant
1.663***
6.547***
7.124***
–0.593*
(0.640)
(1.664)
(1.411)
(0.318)
Restaurant FE
Yes
Yes
Yes
Yes
Time effect
Yes
Yes
Yes
Yes
Observations
259,460
259,460
259,460
259,460
R-squared (within)
0.046
0.049
0.060
0.042
# of restaurants
3,736
3,736
3,736
3,736
Notes: Robust standard errors are enclosed in parentheses. Std. Err. is adjusted for clusters in
restaurants. The coefficients are significant at levels *** p < 0.01, ** p < 0.05, and * p < 0.1.
4.5.2. Robustness Check 2: Alternative Estimation Approach
This section reports an additional set of results that were obtained using the SUR model, which
controls for the possibility that review deviation and emotion are co-determined. SUR allows the
error terms of Equations (2) and (3) to be correlated and jointly estimates these equations. The
estimation results, as can be seen in Table 8, are consistent with our main results, thereby
indicating robustness.
28
Table 8. Robustness Check: SUR Estimation
DVs:
(1) Rating
Deviation
(2) Overall
Emotion
(3) Rating
Deviation
(4) Negative
Emotion
(5) Positive
Emotion
Individualism
0.102***
2.238***
0.068***
0.207***
2.045***
(0.012)
(0.099)
(0.012)
(0.027)
(0.099)
ln(experience)
–0.153***
–2.565***
–0.117***
–0.187***
–2.386***
(0.020)
(0.167)
(0.020)
(0.045)
(0.166)
Individualism * ln(experience)
–0.028***
–0.601***
–0.020***
–0.053***
–0.551***
(0.005)
(0.039)
(0.005)
(0.010)
(0.039)
Emotion
–0.029***
(0.000)
Rating deviation
–2.003***
0.843***
–2.758***
(0.016)
(0.004)
(0.016)
Negative emotion
0.145***
(0.001)
Positive emotion
–0.033***
(0.000)
ln(prior volume)
–0.018***
–0.064***
–0.019***
0.016***
–0.078***
(0.001)
(0.009)
(0.001)
(0.002)
(0.009)
Average rating
–0.068***
–0.160***
–0.054***
–0.025***
–0.128***
(0.001)
(0.009)
(0.001)
(0.003)
(0.009)
ln(consumer tenure)
–0.007***
–0.148***
–0.009***
0.011***
–0.158***
(0.001)
(0.009)
(0.001)
(0.002)
(0.009)
Constant
2.303***
19.197***
2.130***
–0.235
19.336***
(0.376)
(3.127)
(0.368)
(0.839)
(3.125)
Time effect
Yes
Yes
Yes
Yes
Yes
Observations
256,810
256,810
256,810
256,810
256,810
R-squared
0.038
0.041
0.050
0.016
0.043
Notes: Robust standard errors are enclosed in parentheses. The coefficients are significant at
levels *** p < 0.01, ** p < 0.05, and * p < 0.1.
5. Discussion
5.1. Key Findings
This study is the first to conceptualize theoretically and test empirically the effect of cultural
values on rating deviation and review emotion in online restaurant reviews. First, we
demonstrate that consumers from an individualist cultural background are more likely to deviate
29
from prior opinion. Second, consumers from an individualist cultural background are more likely
to express emotion in their reviews. Third, these two characteristics of online reviews can have
important implications for review helpfulness.
5.2. Implications
This study offers several theoretical implications. Our work is the first to consider that cultural
differences may affect the tendency for consumers to deviate from (conform to) past reviews.
Recent work has suggested that the present review aggregation approach employed by many
leading platforms (e.g., Yelp) tends to ignore the systematic differences in reviewer behavior and
conformity in the review generation process (Dai et al. 2012). Our findings point to a previously
undocumented driver of the systematic differences in review characteristics. This driver is
related to the cultural background of reviewers, which not only has the potential to exacerbate or
mitigate herding in review generation but also has similar negative implications for the
optimality of existing review aggregation techniques.
Our work is also the first to consider how cross-cultural differences are manifested in
terms of the textual characteristics of reviews beyond a simplistic measure of length. Consumers
from individualist cultural backgrounds express more emotion in their reviews. In turn, both
conformity (lacking rating deviation) and review emotion lead to lower review helpfulness. Our
work is among the first to draw a connection between the cultural background (values) of authors
and the perception of audiences toward the quality of the review content. These results imply that
the operators of online review sites must be cognizant of the systematic, cross-cultural
differences in the content that is being produced, and that they must consider some approaches to
mitigate biases whenever they damage the perceived helpfulness of a review. For instance,
30
review platforms may offer examples of “helpful” reviews to consumers that are tailored based
on the reviewing history or country of residence of the reviewer. Alternatively, review platforms
may seek and solicit reviews from individuals with a particular cultural background to elicit
helpful reviews for others.
Online review aggregators, such as those presented in this study, are likely to be of
greatest use for products and services that cater to various customer segments, namely,
consumers from different cultural backgrounds. If reviews are aggregated based on the reviewing
tendencies of consumers (e.g., weighting reviews based upon the cultural background of authors
and their anticipated likelihood of under- or over-stating divergent opinions), the consumer
search process may be improved, search costs may be reduced, and better purchase decisions
may be expected.
Previous studies that considered cross-cultural differences in online reviews have almost
exclusively employed a two-country design to explore cross-cultural differences in the
production and consumption of reviews (Chung and Darke 2006, Koh et al. 2010, Fang et al.
2013). By contrast, this study leverages a large observational dataset of reviews that were written
by consumers from 52 countries. Therefore, our findings have external validity.
5.3. Limitations and Opportunities
Similar to other studies, our work is subject to some limitations. First, cultural values have been
measured at the national level and then ascribed to individuals based on their country of
residence. A more accurate measure should employ a survey of each consumer based on the
original measures of Hofstede (2001) or House et al. (2004). However, this approach entails
surveying a large number of TripAdvisor users, which is impractical because of our limited
31
access. We acknowledge this limitation and interpret the observed effects as derived from “the
cultural backgrounds of consumers” than from the “cultural values of consumers”6. Nevertheless,
future research may employ a different research design to address this limitation. For example,
researchers may recruit reviewers, survey their cultural values at the individual level, and then
ask them to complete a review task.
Second, a person may be born and raised in one country, and then subsequently
immigrates to another country. Unfortunately, this behavior cannot be observed in our archival
data. Nevertheless, this limitation is unlikely to become a prevalent issue in our sample and will
only introduce noise into our estimations, thereby preventing us from identifying the
hypothesized effects. Given that we have observed significant estimates in our regressions, this
limitation does not pose a significant problem for this study. We infer that our estimates are
conservative.
Third, we cannot measure the dynamics of helpful votes for each review (i.e., we lack
time stamps on helpful votes and can only observe the total number of votes that have accrued as
of the data collection period). Implicitly, our analyses assume that all helpful votes arrive
immediately after a review is published. Ideally, we prefer to analyze the arrival of helpful votes
dynamically because the conformity or deviation of a review will vary over time as other,
subsequent reviews are written. In other words, after its writing, a review may be in high or low
agreement with all prior reviews, but begins to agree with the overall body of opinion as
subsequent reviews appear, thereby affecting the rate at which helpful votes arrive. However,
this limitation does not pose a serious concern for our analyses. First, we have observed a strong
positive correlation (rho = 0.89) between the conformity of the author at the time of authorship
6 We thank an anonymous reviewer for raising this idea.
32
(i.e., agreement with prior reviews) and his/her conformity to the overall body of reviews that
have been authored during the data collection period (i.e., agreement with prior and subsequent
reviews). This finding indicates that review deviation and conformity are relatively static values.
Second, upon repeating our analyses, we have observed similar results in terms of signs and
significance even after limiting our sample of data to those reviews that have been published
within the previous two weeks. Therefore, our results are unlikely to be driven by our inability to
identify the dynamics of helpful vote arrival.
Fourth, we cannot determine the degree to which the variation in online review
characteristics associated with the cultural background of authors is driven by self-selection into
review authorship versus the compositional differences that emerge, conditional on a review
being entered. In other words, consumers from collectivist cultures are more likely to opt out of
reviewing when they are emotionally charged or hold a “different” opinion from prior reviewers.
However, this limitation is a serious concern for our study. First, social psychology presents
evidence that individuals from collectivist cultures are more susceptible to the opinions of peers
and are more likely to conform to such opinions (see Bond and Smith 1996 for a review of this
topic). Therefore, differences in the reviewing behavior may exist over and above the decision of
whether or not to write a review. Second, despite its presence, self-selection does not have
substantive implications for our results or estimates. Our hypotheses and empirical estimations
draw relationships between cultural backgrounds and the characteristics of published reviews
written by individuals from such backgrounds. We have observed systematic differences in the
review content regardless of whether such differences are attributed to deviations in opinion
conditional on authorship or self-selection into authorship.
33
This study offers several opportunities for future research. First, future research may
investigate different U.S. states as a source of heterogeneity to examine the effect of
individualism on online reviews. Second, future studies may examine whether other dimensions
of cultural values (e.g., uncertainty avoidance) can affect consumer behavior in the production or
consumption of online reviews. For example, future research can delve into cross-cultural
differences in review consumption. Third, our analyses of perceived helpfulness abstract away
the possibility that the effects are moderated by the cultural values of the primary audience for a
service provider. For instance, a recent work has provided early evidence of cross-cultural
differences in online review consumption by reporting that individuals from collectivist cultures
place greater value on negative reviews (Fang et al. 2013). Future studies may explore other
differences in perceived helpfulness across cultures, such as whether individuals from collectivist
or individualist cultures exhibit a similar preference for review deviation or conformity.
References
1. Adomavicius, G., Bockstedt, J. C., Curley, S. P., and Zhang, J. 2013. “Do Recommender
Systems Manipulate Consumer Preferences? A Study of Anchoring Effects.” Information
Systems Research, (24:4), pp 956-975.
2. Ahmad, S. N., and Laroche, M. 2015. “How Do Expressed Emotions Affect the Helpfulness
of a Product Review? Evidence from Reviews Using Latent Semantic Analysis.”
International Journal of Electronic Commerce, (20:1), pp 76-111.
3. Asch, S.E. 1955. “Opinions and Social Pressure.” Scientific American, (193), pp 35–35.
4. Ba, S., and Pavlou, P. A. 2002. "Evidence of the Effect of Trust Building Technology in
Electronic Markets: Price Premiums and Buyer Behavior," MIS Quarterly (26:3), pp 243-
268.
5. Bond, R., and Smith, P. B. 1996. “Culture and Conformity: A Meta-Analysis of Studies
Using Asch's (1952b, 1956) Line Judgment Task.” Psychological Bulletin, (119:1), pp. 111–
137.
34
6. Brehm, S., and Brehm, J. W. 1981. Psychological Reactance: A Theory of Freedom and
Control. New York, NY: Academic Press.
7. Burtch, G., Ghose, A., and Wattal, S. 2014. "Cultural Differences and Geography as
Determinants of Online Pro-Social Lending," MIS Quarterly (38:3), pp 773-794.
8. Butler, E. A., Lee, T. L., and Gross, J. J. 2007. “Emotion Regulation and Culture: Are the
Social Consequences of Emotion Suppression Culture-specific?” Emotion, (7), pp. 30–48.
9. Cao, Q., Duan, W. J., and Gan, Q. W. 2011. “Exploring Determinants of Voting for the
‘Helpfulness’ of Online User Reviews: A Text Mining Approach,” Decision Support Systems
(50:2), pp. 511-521.
10. Chen, Z. and Lurie, N. H. 2013. “Temporal Contiguity and Negativity Bias in the Impact of
Online Word of Mouth,” Journal of Marketing Research (50:4), pp 463-476.
11. Chevalier, J., and Mayzlin, D. 2006. "The Effect of Word of Mouth Online: Online Book
Reviews," Journal of Marketing Research (43:3), pp 345-354.
12. Chung, C. M., and Darke, P. R. 2006. “The Consumer as Advocate: Self-relevance, Culture,
and Word-of-mouth,” Marketing Letters (17:4), pp 269-279.
13. Cialdini, R., Wosinska, W., Barrett, D. W., Butner, J., and Gornik-Durose, M. 1999.
"Compliance with a Request in Two Cultures: The Differential Influence of Social Proof and
Commitment/Consistency on Collectivists and Individualists," Personality and Social
Psychology Bulletin (25:10), pp. 1242-1253.
14. Clemons, E.K., Gao, G.G., and Hitt, L.M. 2006. "When Online Reviews Meet
Hyperdifferentiation: A Study of the Craft Beer Industry," Journal of Management
Information Systems (23:2), pp 149-171.
15. Dai, W., Jin, G., Lee, J., and Luca, M. 2012. “Optimal Review Aggregation of Consumer
Ratings: An Application to Yelp.com.” NBER Working Paper.
16. Danescu-Niculescu-Mizil, C., Kossinets, G., Kleinberg, J., and Lee, L. 2009. “How Opinions
are Received by Online Communities: A Case Study on Amazon.com helpfulness votes,” In
Proceedings of the 18th international Conference on World Wide Web (WWW), pp. 141-150.
ACM.
17. De Langhe, B., Puntoni, S., Fernandes, D., and Van Osselaer, S. M. 2011. “The Anchor
Contraction Effect in International Marketing Research.” Journal of Marketing Research
(48:2), pp 366-380.
18. Dellarocas, C. 2003. "The Digitization of Word of Mouth: Promise and Challenges of Online
Feedback Mechanisms," Management Science (49:10), pp 1407-1424.
35
19. Dellarocas, C., Zhang, X.M., and Awad, N.F. 2007. "Exploring the Value of Online reviews
in Forecasting Sales: The Case of Motion Pictures," Journal of Interactive Marketing (21:4),
pp 23-45.
20. Dimoka, A., Hong, Y. and Pavlou, P. A. 2012. "On Product Uncertainty in Online Markets:
Theory and Evidence." MIS Quarterly (36:2), pp 395-426
21. Duan, W., Gu, B., and Whinston, A.B. 2008. "Do Online Reviews Matter? An Empirical
Investigation of Panel Data," Decision Support Systems (45:4), pp 1007-1016.
22. Fang, H., Zhang, J., Bao, Y., and Zhu, Q. 2013. “Towards Effective Online Review Systems
in the Chinese Context: A Cross-cultural Empirical Study,” Electronic Commerce Research
and Applications (12:3), pp 208-220.
23. Friedlmeier, W., Corapci, F., and Cole, P. M. 2011. “Socialization of Emotions in Cross-
cultural Perspective.” Social and Personality Psychology Compass (5), pp. 410–427.
24. Giannetti, M., and Yafeh, Y. 2012. "Do Cultural Differences between Contracting Parties
Matter? Evidence from Syndicated Bank Loans," Management Science (58:2), pp 365-383.
25. Ghose, A., and Ipeirotis, P.G. 2011. “Estimating the Helpfulness and Economic Impact of
Product Reviews: Mining Text and Reviewer Characteristics,” IEEE Transactions on
Knowledge and Data Engineering (23:10), pp. 1498-1512.
26. Godes, D., and Silva, J. C. 2012. “Sequential and Temporal Dynamics of Online Opinion,”
Marketing Science (31:3), pp 448-473.
27. Goes, P. B., Lin, M., & Yeung, C. M. A. 2014. “Popularity Effect” in User-Generated
Content: Evidence from Online reviews. Information Systems Research (25:2) pp 222-238.
28. Hofstede, G.H., Culture's Consequences: International Differences in Work-Related Values.
Thousand Oaks, CA: Sage, 1980 (revised and expanded in 2001).
29. Holtgraves, T. 1997 "Styles of Language Use: Individual and Cultural Variability in
Conversational Indirectness." Journal of Personality and Social Psychology (73:3), pp. 624.
30. Hong, Y., Chen, P. and Hitt, L. 2013, "Measuring Product Type with Dynamics of Online
Review Variance", Proceedings of International Conference on Information Systems,
Orlando, FL.
31. Hong, Y. and Pavlou, P. A. 2014. “Is the World Truly 'Flat'? Empirical Evidence from
Online Labor Markets”. Fox School of Business Research Paper No. 15-045. Available at
SSRN: http://dx.doi.org/10.2139/ssrn.2371748
32. House R.J. et al. (eds.), Culture, Leadership, and Organizations: The GLOBE Study of 62
Societies. Thousand Oaks, CA: Sage, 2004.
36
33. Hu, N., Pavlou, P.A., and Zhang, J. 2006. "Can Online Reviews Reveal a Product's True
Quality? Empirical Findings and Analytical Modeling of Online Word-of-Mouth
Communication," Proceedings of the 7th ACM conference on Electronic commerce, ACM,
pp. 324-330.
34. Huang, H. 2005. "A Cross-cultural Test of the Spiral of Silence." International Journal of
Public Opinion Research (17.3), pp. 324-345.
35. Huang, N., Burtch, G., Hong, Y. & Evan Polman. 2016. “Effects of Multiple Psychological
Distances on Construal and Consumer Evaluation: A Field Study of Online Reviews.”
Journal of Consumer Psychology, forthcoming.
36. Inglehart, R., & Welzel, C. 2010. “Changing Mass Priorities: The Link between
Modernization and Democracy.” Perspectives on Politics, (8:2), pp. 551–567.
37. Inglehart, R., & Oyserman, D. 2004. "Individualism, Autonomy, Self-expression. The
Human Development Syndrome." In: International Studies in Sociology and Social
Anthropology, pp. 74-96.
38. Jupiter Research. 2008. "Online Reviews Most Helpful in Purchase Decisions," Available
online at: http://www.marketingcharts.com/direct/online-product-reviews-most-helpful-in-
purchase-decisions-offline-too-3437/
39. King, R. A., Racherla, P., and Bush, V. D. 2014. “What We know and don't Know about
Online Word-of-mouth: A Review and Synthesis of the Literature,” Journal of Interactive
Marketing (28:3), pp. 167-183.
40. 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), pp 374-385.
41. Kwark, Y., Chen, J., and Raghunathan, S. 2014. “Online Product Reviews: Implications for
Retailers and Competing Manufacturers,” Information Systems Research (25:1), pp 93-110.
42. Lam, D., Lee, A., and Mizerski, R. 2009. “The Effects of Cultural Values in Word-of-Mouth
Communication.” Journal of International Marketing (17:3), pp 55-70.
43. Lee, D., Hosanagar, K. and Nair, H. 2013. “The Effect of Advertising Content on Consumer
Engagement: Evidence from Facebook”. Available at
SSRN: http://ssrn.com/abstract=2290802
44. Lee, T. Y., and Bradlow, E. T. 2011. "Automated Marketing Research Using Online
Customer Reviews," Journal of Marketing Research (48:5), pp. 881-894.
37
45. Lee, Y. J., Hosanagar, K., and Tan, Y. 2015. Do I Follow My Friends or the Crowd?
Information Cascades in Online Movie Ratings. Management Science, (61:9), pp 2241-2258.
46. Leidner, D.E., and Kayworth, T. 2006. "Review: A Review of Culture in Information
Systems Research: Toward a Theory of Information Technology Culture Conflict," MIS
Quarterly (30:2), pp 357-399.
47. Li, X., and Hitt, L. 2007. "Self-selection and Information Role of Online reviews,"
Information Systems Research, (19:4) pp 456-474.
48. Liu, Y., Chen, P., and Hong, Y. 2014. “Value of Multi-dimensional Rating Systems: An
Information Transfer View,” Proceedings of the 35th International Conference on
Information Systems (ICIS), Auckland, New Zealand
49. Lu, X., Ba, S., Huang, L., and Feng, Y. 2013. "Promotional Marketing or Word-of-Mouth?
Evidence from Online Restaurant Reviews," Information Systems Research (24:3), pp. 596-
612.
50. Ludwig, S., De Ruyter, K., Friedman, M., Brüggen, E. C., Wetzels, M., and 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), pp. 87-103.
51. Moe, W., and Schweidel, D. 2012. "Online Product Opinions: Incidence, Evaluation, and
eEvolution." Marketing Science (31:3), pp. 372-386.
52. Muchnik, L., Aral, S., and Taylor, S. J. 2013. “Social Influence Bias: A Randomized
Experiment.” Science, 341(6146), 647-651.
53. Mudambi, S. M., and Schuff, D. 2010. “What Makes a Helpful Online Review? A Study of
Customer Reviews on Amazon.com.” MIS Quarterly, (34:1), pp 185-200.
54. Nagle, F., and Riedl, C. 2014. "Online Word of Mouth and Product Quality Disagreement."
In: Proceedings of the Academy of Management Annual Meeting, Philadelphia, PA.
55. Niedenthal, P., Krauth-Gruber, S., Ric, F. 2006. Psychology of Emotion Interpersonal,
Experimental, and Cognitive Approaches, pp. 5, 305-342. New York, NY: Psychology Press.
56. Pavlou, P., and 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), pp. 392-414.
57. Pennebaker, J. W., Francis, M. E., and Booth, R. J. 2001. Linguistic Inquiry and Word
Count: LIWC 2001. Mahway: Lawrence Erlbaum Associates.
38
58. Schoorman, F. D., Mayer, R. C., and Davis, J. H. 2007. “An Integrative Model of
Organizational Trust: Past, Present, and Future,” Academy of Management Review, (32:2), pp
344-354.
59. Sharman, J. M. 2005. Right of Privacy in State Constitutional Law, The Rutgers LJ, 37, 971.
60. Spiller, S. A., Fitzsimons, G. J., Lynch Jr, J. G., & McClelland, G. H. 2013. Spotlights,
floodlights, and the magic number zero: Simple effects tests in moderated regression. Journal
of Marketing Research, (50:2), 277-288.
61. Sridhar, S. and R. Srinivasan. 2012. "Social Influence Effects in Online Product Ratings."
Journal of Marketing (76:5) pp. 70-88.
62. Sun, M. 2012. “How does the Variance of Product Ratings Matter?” Management
Science, (58:4), pp 696-707.
63. Takahashi, K., Ohara, N., Antonucci, T., Akiyama, H. 2002. “Commonalities and
Differences in Close Relationships among the Americans and Japanese: A Comparison by
the Individualism/Collectivism Concept.” International Journal of Behavioral Development,
(26:5), pp. 453–465.
64. Triandis, H. C. 1995. Individualism and Collectivism. Westview Press.
65. Tsai J.L., Miao F.F., Seppala E., Fung H.H., Yeung D.Y. 2007. “Influence and Adjustment
Goals: Sources of Cultural Differences in Ideal Affect.” Journal of Personality and Social
Psychology, (92:6), pp. 1102–1117
66. Wang, C., Zhang, X., and Hann, I. H. 2015. “Socially Nudged: A Quasi-Experimental Study
of Friends' Social Influence in Online Product Ratings.” Information Systems Research,
forthcoming
67. Wu, F., & Huberman, B. 2008. "How Public Opinion Forms." Internet and Network
Economics. Springer, Berlin, pp. 334-341.
68. Yaveroglu, I. S., and Donthu, N. 2002. “Cultural Influences on the Diffusion of New
Products.” Journal of International Consumer Marketing, (14:4), pp 49-63.
69. Yin, D., S. Bond, H. Zhang. 2014. “Anxious or Angry? Effects of Discrete Emotions on the
Perceived Helpfulness of Online Reviews.” MIS Quarterly, (38:2), pp 539-560.
70. Zhu, F., and Zhang, X. 2010. "Impact of Online Consumer Reviews on Sales: The
Moderating Role of Product and Consumer Characteristics," Journal of Marketing, (74:2), pp
133-148.
