Content uploaded by Mayukh Dass
Author content
All content in this area was uploaded by Mayukh Dass on May 08, 2019
Content may be subject to copyright.
Decision Sciences
Volume 0 Number 0
May 2019
© 2019 Decision Sciences Institute
The Impact of Online Review Content and
Linguistic Style Matching on New Product
Sales: The Moderating Role of Review
Helpfulness
Omer Topaloglu†
Silberman College of Business, Fairleigh Dickinson University, 1000 River Road, Teaneck,
NJ, 07666, e-mail: otopaloglu@fdu.edu
Mayukh Dass
Rawls College of Business, Texas Tech University, MS 42101, Lubbock, TX, 79409-2101,
e-mail: mayukh.dass@ttu.edu
ABSTRACT
This article investigates the moderating role of review helpfulness on the effects of
online review content and linguistic style matching on new product sales. Using data
from 105,494 online reviews from a popular website, IMDB.com, following 264 movie
releases, the article shows that the impact of review style and content on new product
sales is contingent upon review helpfulness. In particular, results suggest that linguistic
style matching, positive emotion, and cognitive components of online review content
significantly impact new product sales when the reviews are deemed helpful by the
readers. These findings collectively suggest that online review research benefits from
deeper textual analysis that includes review content and linguistic style compared to
traditional methods that rely solely on numerical ratings. Also, review helpfulness plays
a critical role in consumer decision-making considering a rapidly increasing amount of
online information is now available to consumers. [Submitted: March 31, 2017. Revised:
April 1, 2019. Accepted: April 3, 2019.]
Subject Areas: Linguistic Style Matching, LIWC, Online Review Content,
and Review Helpfulness.
INTRODUCTION
As the Internet has become ubiquitous, consumers’ use of peer-generated online
reviews is becoming more prevalent than ever before as a part of the product
information search, a key step in the consumer decision-making process (Wang,
Guo, Zhang, Wei, & Chen, 2016). From a customer’s decision-making perspective,
online reviews enhance customer value and increase decision efficiency (Kozinets,
De Valck, Wojnicki, & Wilner, 2010), and are considered to be more reliable and
less biased than firm-generated product information (Brown, Broderick, & Lee,
†Corresponding author
1
2The Impact of Online Review Content and Linguistic Style Matching on New Product Sales
2007). From a firm’s standpoint, online reviews are shown to boost online and
offline sales (Chintagunta, Gopinath, & Venkataraman, 2010), product awareness
(Liu, 2006), and word-of-mouth (Berger & Milkman, 2012). Therefore, as a ubiq-
uitous, peer-to-peer, and mostly text-based communication tool, online reviews are
vital for both consumer and firm decision-making, and therefore deserve greater
scrutiny of their outcomes.
As online review generation has become mainstream, review websites have
started to host a huge number of online reviews, which to an extent creates infor-
mation overload for users. For instance, in Amazon.com, electronics and computer
categories average more than 4,000 reviews per product, whereas the beauty prod-
ucts category averages around 400 (Anderson, 2015). Similarly, the books category
averages several hundred reviews with some top-selling titles attracting thousands
of reviews (Cao, Duan, & Gan, 2011). Therefore, the amount of online information
available may be more than what consumers need and, thus lead to inefficiency
in decision-making (Jones, Ravid, & Rafaeli, 2004). To help consumers overcome
this information overload issue, websites now allow users to rate helpfulness of
these reviews and share them with the readers.
Online reviews, therefore, communicate three types of information to the
consumers, including a numerical rating, a text content, and a helpfulness rating
of the review. Out of these three components, the majority of existing research on
the effects of online reviews on consumer purchase behavior has predominantly
focused on the effects of numerical review characteristics, which include volume
(number of reviews posted), valence (numerical rating) and dispersion (variance
in the product rating; e.g., Chen & Xie, 2008; Chintagunta et al., 2010; Liu, 2006).
More recently, researchers have started focusing on understanding how the two
inseparable components of text content, the review sentiment and the linguistic
style (Huffaker, Swaab, & Diermeier, 2011), play a significant role in consumer
decision-making (Gundecha & Liu, 2012; Ludwig et al., 2013). As recent studies
suggest that consumers go beyond considering numerical ratings of the reviews
and examine other review characteristics during the decision-making process (e.g.,
Hu, Koh, & Reddy, 2014), our understanding on the effects of the text content and
helpfulness rating of reviews on consumer decision making (i.e., product sales)
is limited. Therefore, to address this literature gap, we focus on two research
questions: (i) How do review style and content affect new product sales? and (ii)
How does review helpfulness moderate these effects?
In this article, we focus on the effects of online review content and style
on new product sales in the motion picture industry. The review content refers to
nouns, verbs, and adjectives that communicate emotions and logic; the linguistic
style, however, refers to the use of function words such as pronouns (Tausczik
& Pennebaker, 2010). In particular, this study examines the moderating role of
review helpfulness on the link between content and style of online reviews and
new product sales. We posit that because users are likely to undervalue, disre-
gard, or completely miss online reviews deemed unhelpful, effects of reviews on
consumers’ decision-making will rely heavily upon review helpfulness. We use a
dictionary-based, computerized text mining approach to analyze 105,494 online
reviews from a popular website, IMDB.com, following 264 movie releases. Text
analytics is pertinent to the purpose of this study because it allows us to uncover
Topaloglu and Dass 3
the conceptual content and linguistic style in unstructured textual data (Kulkarni,
Apte, & Evangelopoulos, 2014). In addition, the motion picture industry provides
an ideal context for this study because movies are experiential products, and thus
inability to evaluate movies before purchase leads to heavy reliance on online
reviews to make decisions (Chintagunta et al., 2010).
Our study attempts to contribute to the electronic word-of-mouth (eWOM)
literature in three ways. First, drawing upon communication accommodation the-
ory (Giles & Smith, 1979), we extend the previous work by studying linguistic style
matching (LSM) in conjunction with textual content. Although social psychology
research documents the role of content and style in the behavioral outcomes of
written communication (Ireland & Pennebaker, 2010), eWOM literature presents
a gap in this more comprehensive approach. Similar to the content, style of an
online review or the use of function words, including pronouns, prepositions, arti-
cles, conjunctions, and auxiliary verbs, conveys meaning, and creates an impact on
readers’ behavioral reactions to the written communication. Second, most previous
work on online reviews focuses on existing product sales and thus uses numerical
measurement proxies. We attempt to address this gap by implementing a hierar-
chical linear model (Singer & Willet, 2003) to study the link between the change
in deeper textual characteristics and the change in new product sales with a lon-
gitudinal perspective. Deep textual characteristics refer to affective and cognitive
components used to communicate a user’s positive and negative emotions as well
as his or her logic. Third, we contribute to the extant eWOM theory by document-
ing how review helpfulness as a key variable interacts with review content and
style, and leads to changes in new product sales.
The rest of the article is presented as follows. In the next section, we discuss
the extant literature on eWOM, linguistic style matching and review helpfulness,
and present the related hypotheses. The third section discusses the data and the
methodology used to test the hypotheses. Finally, we offer a discussion of the results
and a conclusion presenting the contributions, limitations, and future direction of
the studies.
LITERATURE REVIEW
Electronic WOM refers to online product or service information generated by
users not associated with a company (Brown et al., 2007; Godes et al., 2005).
EWOM takes various forms, such as product reviews (Chen & Xie, 2008; Cheng
& Ho, 2015; Zhu & Zhang, 2010), virtual communities (Guo, Pathak, & Cheng,
2015), blogs (Kozinets et al., 2010; Senecal & Nantel, 2004), and microblogs
(Hennig-Thurau, Wiertz, & Feldhaus, 2015; Topaloglu, Dass, & Kumar, 2017).
Extant research suggests that eWOM has a positive impact on both online (Cheva-
lier & Mayzlin, 2006) and offline sales (Chintagunta et al., 2010; Liu, 2006;
Srinivasan, Rutz, & Pauwels, 2016) because consumers find it to be more reliable
than seller-controlled online information (Brown et al., 2007) and tend to access it
right before making a purchase (Chen & Xie, 2008). EWOM also has an effect on
product awareness (Liu, 2006) and subsequent eWOM (Berger & Milkman, 2012;
Kozinets et al., 2010). EWOM becomes particularly important during new product
introductions (Liu, 2006) because it influences firms’ marketing communication
4The Impact of Online Review Content and Linguistic Style Matching on New Product Sales
strategy such that managers use online product reviews as complements or
substitutes to seller-created product information (Chen & Xie, 2008). Further,
product and consumer characteristics affect the relationship between online
product reviews and sales (Zhu & Zhang, 2010). For instance, recommendations
for experience products (i.e., movies) are significantly more influential than
recommendations for search products (Liu, 2006; Senecal & Nantel, 2004).
Existing studies mostly use quantitative indicators of eWOM, including va-
lence, dispersion, and volume (see Table 1). Valence, described as the degree of
attraction or aversion expressed toward a product, may lead to both normative and
informative consequences (Filieri, 2015). A positive or negative post can have “a
persuasion effect,” which leads to behavioral outcomes, whereas a neutral post can
have “an informative effect” (Liu, 2006). Dispersion on the other hand, is defined
as “the extent to which product-related conversations are taking place across a
broad range of communities” (Godes & Mayzlin, 2004). Past studies investigating
the behavioral outcomes of these different eWOM metrics present equivocal find-
ings. Srinivasan et al. (2016) find a positive relationship between eWOM valence
and sales. Chevalier and Mayzlin (2006) also demonstrate a causal relationship
between numerical ratings and sales after a highly anticipated book release. In a
similar vein, accounting for the sequential release of new products, Chintagunta
et al. (2010) find that valence, not volume, drives sales. Some earlier works, how-
ever, find that volume (Duan, Gu, & Whinston, 2008; Liu, 2006) or dispersion
(Godes & Mayzlin, 2004), not valence of eWOM, affects sales. One potential
explanation for these findings is the undermined role of textual characteristics,
including emotion or reason-based word choice and linguistic styles, in eWOM
formation (Ludwig et al., 2013).
HYPOTHESES
Linguistic Style Matching
Recent advances in text analytics allow researchers to analyze large volumes of
verbatim review data, which creates an opportunity to extend eWOM theory by
investigating both review style and content (Gundecha & Liu, 2012). In any social
interaction, individuals automatically mimic others’ verbal and nonverbal behav-
iors (Ireland & Pennebaker, 2010). Nonverbal mimicry ranging from aligning
postures to matching breathing patterns (LaFrance, 1985; McFarland, 2001) is a
way to communicate emotions and attitudes (Ramanathan & McGill, 2007). Verbal
mimicry, however, is evident in oral and written communication. Past research
on verbal mimicry makes a clear distinction between language style and language
content (Ireland & Pennebaker, 2010). Style and content of online reviews provide
important decision inputs for users (Huffaker et al., 2011). Text content refers to
regular nouns, verbs, or adjectives that communicate internal feelings or objective
reasoning. However, because linguistic style permeates content, focusing solely
on content is insufficient in text-based communication. Meaning is also conveyed
through style or function words, including pronouns, prepositions, articles,
conjunctions, and auxiliary verbs (Tausczik & Pennebaker, 2010). For instance,
using a formal linguistic style compared to an informal style could lead to vastly
Topaloglu and Dass 5
Table 1: Overview of major empirical studies in the literature.
Article Key Variables Context Key Finding
Online Review Characteristics Sales
Numeric Sentiment Linguistic
Style
Helpfulness
Senecal & Nantel, 2004 √✗✗ ✗ ✗Electronics Online reviews are more influential than expert
reviews in affecting product choice
Chevalier & Mayzlin, 2006 √✗✗ ✗ ✗Books Review ratings impact sales-ranking
Liu, 2006 √√ ✗✗ √Movies Volume (not valence) of reviews predicts sales
Duan et al., 2008 √✗✗ ✗ √Movies Valence (not volume) of reviews predicts sales after
accounting for endogeneity
Chen & Xie, 2008 √✗✗ ✗ ✗Electronics Marketing strategy can be adjusted based on online
reviews to increase sales
Zhu & Zhang, 2010 √✗✗ ✗ √Video Games Reviews are less effective for popular games
Cao et al., 2011 √√ ✗√✗Software Semantic characteristics are more influential than
others in affecting review helpfulness
Ludwig et al., 2013 √√ √ ✗√Books Affective content and linguistic style affect sales
Hu et al., 2014 √√ ✗√√
Books Ratings affect sales only through review sentiment
Srinivasan et al., 2016 √√ ✗✗ √FMCG Traditional (not online) marketing is the main driver
of sales
Gao et al., 2017 √✗✗✗ ✗Physician
Services
Low quality physicians are less likely to be rated
online.
This article √√ √ √ √
Movies Review helpfulness moderates the relationship
between review characteristics and sales
6The Impact of Online Review Content and Linguistic Style Matching on New Product Sales
different interpretation of the same sentiment. In an online setting, where there is
a high level of author anonymity, LSM serves as a significant criterion to qualify
the source of a message. Thus, the use of similar function words, irrespective of
content, influences the message interpretation (Ireland & Pennebaker, 2010). In
fact, based on communication accommodation theory (Giles & Smith, 1979), the
level of matching between two conversations also leads to behavioral outcomes
similar to sales or subsequent online reviews (Ireland & Pennebaker, 2010).
Relying on social psychology theories, communication accommodation the-
ory posits that when interacting with others, people adjust their verbal and non-
verbal communication (speech, voice, gestures, etc.) to accommodate social ex-
pectations for the conversation (Giles & Smith, 1979). This phenomenon is also
valid in written communication. Writers’ linguistic styles or use of function words
either highlight or reduce the social distance between their readers and themselves
(Giles, 2009). LSM is an automatic, trait-like individual difference. Interestingly,
writers fail to adjust their degree of style matching even when they are instructed
to do so (Ireland & Pennebaker, 2010). The well-known strength-of-the-weak-ties
hypothesis purports that weak ties among distant social network members are more
influential than the strong ties among close network members (Granovetter, 1973).
As the linguistic style diverges from a social norm, readers perceive a diverse social
identity. Therefore, when an online product review is written in a style different
than what an online user is familiar with, the style itself, that is, the function
word usage matching the social norm, irrespective of the content, will have an
impact on the reader’s evaluation and judgment (Pornpitakpan, 2004). Extending
the strength-of-the-weak-ties framework to the eWOM context, Godes and May-
zlin (2004) find that when a broad range of communities generate eWOM, related
TV show viewership increases because weak ties across dissimilar communities
enable effective and result-oriented information to spread. In other words, the so-
cial structure has an impact on the way eWOM is interpreted. Further, past studies
find a strong relationship between review content variance and new product sales
(Duan et al., 2008). Moviegoers prefer different and diverse information when they
make a decision to watch a movie (Phillips, Mannix, Neale, & Gruenfeld, 2004).
Therefore, we propose that low LSM, which suggests that online reviews are gen-
erated by users from diverse communities, is likely to have a positive impact on
new product sales.
H1: Low LSM in online reviews will have a positive impact on new product sales.
The Role of Content
The dual processing theories in the social psychology literature suggest that there
are two distinct processes in human cognition that influence consumer attitudes
(Chaiken & Trope, 1999). One is a fast and associative information processing
based on low-effort heuristics, and the other is a slow rule-based processing based
on high-effort systematic reasoning (Cheng & Ho, 2015). On the one hand, valence
and arousal of consumer emotions directly influence attitudes (Clore & Storbeck,
2006). For example, altruism, self-enhancement, and reciprocity initiate positive
WOM; in contrast, anxiety reduction and vengeance trigger negative WOM
(Berger & Milkman, 2012). Similarly, positive affective cues in a website lead
Topaloglu and Dass 7
to positive behavioral intentions (Li, Browne, & Chau, 2006). On the other hand,
a distinct, deliberate, and conscious cognitive process also influences attitude
formation (Chaiken & Trope, 1999). Online reviewers diligently considering
the true merits of a product and its objective attributes go through a detailed
cognitive processing before generating a review. These affective and cognitive
processes work together to shape online consumer attitudes (Massey, Khatri, &
Montoya-Weiss, 2007; Pappas, Kourouthanassis, Giannakos, & Chrissikopoulos,
2015). Generating an online review about a new product is an expression of such
attitudes, and the mental process that users go through when creating it is reflected
in their linguistic styles as well as the cognitive and affective cues woven in the
content of their message. Using text analytics facilitates capturing positive and
negative tone and cognitive content of online reviews, and determining how these
cues affect purchasing behavior. It is expected that a text with meager affective
and cognitive content does not have the desired impact on behavioral outcomes.
Prior research shows that both emotion-based and reason-based expressions
in online reviews trigger similar processes in readers’ cognition and lead to behav-
ioral outcomes (Pappas et al., 2015). While firms want customers to disseminate
positive or at least neutral information, negative information tends to travel faster
(Godes et al., 2005) and is found more credible (Chevalier & Mayzlin, 2006).
Negative evaluators are seen as more intelligent and competent than positive ones
(Hennig-Thurau et al., 2015), and strongly opinionated information tends to be
more viral than neutral information (Berger & Milkman, 2012). Further, Srini-
vasan et al. (2016)) find that in Facebook posts, the affective component measured
in positive and negative dimensions as well as the cognitive component separately
impact online sales. Therefore, we conclude that positive and negative sentiments
in online reviews should have an impact on new product sales. It is the richness of
an online content, its affective and cognitive substance, which creates an impact
on purchase behavior. We propose that:
H2A: Positive affective content of online reviews will have a positive impact on
new product sales.
H2B: Negative affective content of online reviews will have a negative impact on
new product sales.
H2C: Cognitive content in online reviews will have a positive impact on new
product sales.
Review Helpfulness
Online information search behavior is a fundamental research stream in the e-
commerce literature (Boyer & Hult, 2005; Wang et al., 2016). How web-based
information is presented impacts information overload, decision-making, and in-
formation usability (Oztekin, 2011; Speier, Vessey, & Valacich, 2003). Users utilize
low-effort heuristics or decision-making cues when there are too many options to
choose from (Chakravarti, Janiszewski, & ¨
Ulk¨
umen, 2006) or the involvement
with the choice decision is low (Chaiken & Trope, 1999). Although online re-
views contribute to the efficiency of consumer decision-making, large volumes of
available online reviews create information overload for consumers (Cao et al.,
8The Impact of Online Review Content and Linguistic Style Matching on New Product Sales
2011). Product categories, such as books, DVDs, or electronics, average several
hundred reviews per product in popular online retailers (Cao et al., 2011), and
this trend is projected to grow (Kulkarni et al., 2014). For online retailers, review
helpfulness emerges as an attempt to provide value to online users and reduce their
cognitive overload. A helpful online review is defined as “a peer-generated product
evaluation that facilitates the consumer’s purchase decision process” (Mudambi
& Schuff, 2010). Therefore, online users are likely to use review helpfulness as a
simple decision-making cue.
Review extremity, review depth, and product type affect the perceived help-
fulness of online reviews (Mudambi & Schuff, 2010). An online review produced
by an expert reviewer makes readers feel that the review is more useful (Cheng
& Ho, 2015) and influential (Zhu & Zhang, 2010), and the website is more trust-
worthy (Fuller, Serva, & Benamati, 2007). Prior research finds that the reviews
considered helpful have a stronger impact on book sales than the other reviews
do (Chen & Xie, 2008). Helpfulness ratings not only work as a decision-making
cue but also help in increasing a review’s exposure as websites often rank reviews
based on their helpfulness ratings. The majority of users may not even see most
reviews with low helpfulness ratings. Therefore, review helpfulness plays a highly
critical role in online retailing and has the potential to increase the effects of review
style and content on new product sales. Therefore, we propose that:
H3: Positive impact of low LSM on new product sales will accentuate when review
helpfulness is high.
H4A: Positive impact of positive affective content on new product sales will
accentuate when review helpfulness is high.
H4B: Negative impact of negative affective content on new product sales will
accentuate when review helpfulness is high.
H4C: Positive impact of cognitive content on new product sales will accentuate
when review helpfulness is high.
The research framework of our study is presented in Figure 1.
METHOD
Data
This section details our sample, sources, and procedure for data collection. We
collect data on 760,175 online reviews from 876 movies during the period June
2005 to December 2009 from IMDB. Apart from the reviews, we also collect
information such as the total revenue, budget, launch date, distribution, rating,
genre, and award nominations for each movie. Following previous research (Liu,
2006), we use production budget as a proxy for advertising budget, which typically
amounts to 50% of the movie production budget (Vogel, 2001, p. 96). We use
the genre description from Yahoo! Movies (Sci-Fi, Thriller, Children, Romance,
Comedy, Action, Drama) and three MPAA ratings (PG,R, and others). For each
review, we extract the date, star rating, total number of views of the review, and
number of viewers identifying it as a helpful review.
Topaloglu and Dass 9
Figure 1: Research framework of this study.
Below, we explain the specific problems we encountered and the rules we
used to resolve them. First, we exclude movies launched before the start of the
sample period or in theaters for fewer than 8 weeks (Eliashberg & Shugan, 1997).
Second, we exclude movies for which consumer reviews are not available from the
day of launch. Third, we exclude movies if information on any of the following
characteristics is not available: revenue, budget, launch date, distribution, rating,
genre, and award nominations. And finally, we exclude movies if we did not have
the helpfulness score. Based on these rules, our final sample reduces to 105,494
reviews from 264 movies.
Dependent Variable
The dependent variable of our study is box office sales. We performed univariate
analysis of the variable and found it to be not normally distributed. Therefore, we
performed a log transformation of the variable, and use it as the dependent variable
in this study.
Text Mining of Reviews
One of the goals of this article is to investigate how the review content and
language style affect the sales of new products. Therefore, we use natural language
processing tools to text mine online reviews. Our text mining objective faces
two challenges. First, movie reviews consist of unstructured semantic analysis
of movies. They are written by movie viewers who are expressing their views
without following any predetermined specific writing format. This calls for a text
10 The Impact of Online Review Content and Linguistic Style Matching on New Product Sales
mining tool that has a broad data dictionary to uncover the underlying emotions
and cognitions of the reviewers. Second, the level of domain knowledge varies
across the reviewers. Therefore, it is possible that reviewers have different writing
styles while posting their comments. Given these two challenges in our context, we
decided to use a text mining software called the Linguistic Inquiry and Word Count
(LIWC, 2007 edition) to uncover emotions, cognitions, and linguistic styles in the
reviews (Pennebaker, Francis, & Booth, 2001). LIWC uses a broad dictionary of
more than 4,500 words to determine 22 standard linguistic dimensions and 32
psychological constructs (Pennebaker, Chung, Ireland, Gonzales, & Booth, 2007),
and therefore is suitable enough to explore the movie reviews. The dictionary
includes 116 words for its pronoun category (e.g., I, them, itself), which we used
to measure linguistic style matching. Nine hundred and fifteen words are classified
under the affect category, including regular words (e.g., amazing, disappointing),
misspellings, and internet slang (e.g., LOL, ROFL). Out of these affective words,
406 are positive emotion words (e.g., love, nice, sweet), and 499 are negative
emotion words (e.g., hurt, ugly, nasty). Similarly, the dictionary has around 730
words for cognitive components such as need, consequence, cause, know, etc. The
software reads each review one word at a time. As each word is processed, the
dictionary file is searched; if there is a match, the word category score, which is
found by dividing the number of matching target words by the number of total
words in a review, is incremented. The internal reliability of the subcategories
used in our research are shown to be sufficiently high (αࣘ0.91; Pennebaker et al.,
2007).
The initial purpose of the LIWC system was to identify a group of words
that tapped basic affective and cognitive dimensions (Pennebaker et al., 2007). The
number of word categories, however, has expanded over time. While developing
these categories, researchers have first generated sets of words. For example, the
emotion or affective sub-dictionaries were based on words from several sources,
including common emotion rating scales, such as the PANAS (Watson, Clark,
& Tellegen, 1988), Roget’s Thesaurus, and standard English dictionaries. This
process was followed by extensive brainstorming sessions. One of the first validity
tests of the LIWC scales was conducted by Pennebaker and Francis (1996), who
found that LIWC successfully measures positive emotions, negative emotions,
and cognitive elements. The validity of LIWC dictionaries has been confirmed
in more than 100 studies including textual analyses of online reviews (Goes,
Lin, & Au Yeung, 2014), blogs (Cohn, Mehl, & Pennebaker, 2004), social media
sentiment (Topaloglu et al., 2017), newspaper articles (Humphreys, 2010), and
instant messaging (Slatcher & Pennebaker, 2006). In particular, researchers use
LIWC to generate scores for affective and cognitive states based on the count
and usage of preselected words. Goes et al. (2014)), for instance, utilize LIWC
to extract positive and negative sentiments from online reviews and explore the
relationship between the number of viewers and review objectivity.
We measured helpfulness as the sum total number of viewers who rated a
review helpful. As the helpfulness information on the websites are given as “15
out of 30 people found this review helpful,” we did consider helpfulness as a ratio
between the two numbers, that is, 15/30 =0.50. However, we realized that the
ratios are in percentages, and therefore bear limited meaning. For example, our
Topaloglu and Dass 11
model would have considered “15 out of 30” to have the same level of helpfulness
as “5 out of 10.” Furthermore, as we are studying the level of helpfulness, it made
more sense to use total number of viewers who rated a review helpful as our score,
and not just the ratio. Because the helpfulness depends on the number of views of
the reviews, we control for “views” in our model with “distribution” covariate.
Controls
In line with prior research, we include the following control variables that may
affect online activity—advertising spend, distribution, movie genre and awards.
Advertising Spend (BUDGET): Extant literature suggests that advertising
expenditure plays a fundamental role in the box office revenue generation (Elberse
& Eliashberg, 2003; Prag & Casavant, 1994; Zufryden, 1996). As consumers rely
on movie advertisements as a source of information, it is important that we control
movies’ total advertising budget in our analysis.
Movie Distribution (DISTRIBUTION): The number of screens on which a
movie is released is the most important influence on viewership (Neelamegham &
Chintagunta, 1999) as higher distribution increases visibility, driving up box office
revenue.
Movie Genre: Some genres are more likely to have higher buzz than others.
For example, movies in the action, thriller, or romance genres generally attract
more attention. Similarly, movies with R-rating, or movies with stars are more
likely to be reviewed by a larger number of critics and to succeed in box of-
fice. Therefore, we included MPAA ratings (R RATED, PG RATED) and genres
(COMEDY, ACTION, DRAMA) in the model.
Movie Awards (AWARDS): Movies that are nominated for awards receive
high visibility and greater press coverage, thus resulting in greater box office
revenue. Therefore, we included AWARDS as a control in our model. Table 2
includes model variables.
Model
As our dataset is a panel longitudinal data, we used a mixed model to control for
the random effects of time, and examine the fixed effects related to our hypotheses.
The model is specified as following:
Log(BO)it =β0+β1×LSMi,t−1+β2×VOLUMEi,t −1
+β3×VALENCEi,t−1+β4×DISPERSIONi,t−1
+β5×BUDGETi+β6×DISTRIBUTIONi,t−1
+β7×POSITIVE EMOi,t−1+β8×NEGATIVE EMOi,t−1
+β9×COGNITIONi,t−1+β10 ×HELPFUL
+β11 ×RRATED +β12 ×PG RATED +β13 ×COMEDY
+β14 ×ACTION +β15 ×DRAMA +β16 ×AWARDSt−1
+β17 ×LSMi,t−1×HELPFUL +β18 ×VALENCEi,t−1
×HELPFUL +β19 ×POSITIVE EMOi,t−1×HELPFUL
12 The Impact of Online Review Content and Linguistic Style Matching on New Product Sales
Table 2: Model variables.
Variable Operationalization Source
Dependent Variable
New Product Sales (BO) Box office sales at week tIMDB
Review Characteristics
Linguistic Style Matching (LSM) The ratio of the total number of function words in a specific online review to the
average number of function words across all reviews for a particular movie
(Ireland & Pennebaker, 2010). LSMij =Fij
N
i=1Fij
Nj
;F
ij =number of function
words in review i for movie j over number of words in review i for movie j; Nj
=number of reviews for movie j.
IMDB
Volume The total number of reviews for a movie posted at week tIMDB
Valence The average rating given by the reviewers for a movie at week tIMDB
Dispersion The standard deviation of the ratings given by the reviewers for a movie at week tIMDB
Helpfulness The sum total number of viewers who rated a review helpful IMDB
Positive Emotion (POSITIVE EMO) The average positive emotion of the reviews computed by LIWC for a movie at
week t
IMDB
Negative Emotion (NEGATIVE EMO) The average negative emotion of the reviews computed by LIWC for a movie at
week t
IMDB
Cognitive Component (COGNITION) The average cognitive component of the reviews computed by LIWC for a movie
at week t
IMDB
Movie Characteristics
Budget An indicator variable created by median split where BUDGET =1whenthe
advertising budget of the movie is higher than the median, and BUDGET =0
otherwise.
IMDB
Distribution (THEATER) The number of theaters that showed the movies at week t−1. IMDB
MPAA rating The movie rating included in the model as two indicator variables, PG RATED (
=1, else 0) and RRATED (=1, else 0).
Yahoo! Movies
Movie Genre The movie genre included in the model as three indicator variables COMEDY (=
1, else 0), ACTION (=1, else 0), and DRAMA (=1, else 0).
Yahoo! Movies
Award Nominations (AWARD) The number of nominations of the movies announced before week t. Yahoo! Movies
Topaloglu and Dass 13
+β20 ×NEGATIVE EMOi,t−1×HELPFUL
+β21 ×COGNITIONi,t−1×HELPFUL +α1×TIMEi+ei(1)
To illustrate the significance of the helpfulness as the moderator, we estimated
the model (Equation (1)) with and without the interaction terms. The model without
the interactions terms is shown below:
Log(BO)it =β0+β1×LSMi,t−1+β2×VOLUMEi,t −1+β3×VALENCEi,t−1
+β4×DISPERSIONi,t−1+β5×BUDGETi
+β6×DISTRIBUTIONi,t−1+β7×POSITIVE EMOi,t −1
+β8×NEGATIVE EMOi,t−1+β9×COGNITIONi,t−1
+β10 ×HELPFUL +β11 ×RRATED
+β12 ×PG RATED +β13 ×COMEDY +β14 ×ACTION
+β15 ×DRAMA +β16 ×AWARDSt−1+α1×TIMEi+ei(2)
The above models are estimated using PROC MIXED in SAS.
RESULTS
In this section, we first discuss our data and then present the findings from our
analysis. Table 3 presents the static data description of our data. Overall, we have
2,876 weeks of usable data from 264 movies. The average LSM =0.76 (σ=0.11)
suggests that on average, review writing style is 75% similar to the writing style
of other reviews. The average helpfulness is 0.15 (σ=0.59), suggesting that on
average, only a small number of reviews are considered helpful. Next, we illustrate
the dynamic data description of our data. Figure 2 presents model variable dynam-
ics with a descriptive purpose using functional data analysis (Dass & Shropshire,
2012). It illustrates how the average values of different variables calculated across
all the movies in our sample change over time. The x-axis represents time in weeks
after the movie release and ends with the tenure of the movies at the theater. As
expected, average sales decreases following a short increase in the early weeks of
movies. Average LSM score, however, fluctuates. It increases in the early weeks
of a movie release followed by a temporary decrease. This curve shape tells us that
linguistic styles and social identities of those who post reviews either early or late
during the tenure of a movie in the theaters are similar to each other. In the middle,
a more diverse group of reviewers chimes in. This is consistent with the previous
research that finds the impact of eWOM dispersion declines over time (Godes &
Mayzlin, 2004). Review content displays an insightful evolution. Average positive
sentiments in reviews show an S shape decrease. However, negative sentiments and
cognitive components of the reviews increase over time, suggesting that as weeks
progress, reviews on average become more negative and more rational. Finally,
review helpfulness shows a sharp decline after the early weeks of a movie release,
suggesting that the earlier reviews are deemed more helpful than later reviews. This
14 The Impact of Online Review Content and Linguistic Style Matching on New Product Sales
Table 3: Descriptive statistics and correlation matrix.
Var i a b l e s NMean SD Max Min 1 2 3 4 5 6789
1. New product sales 3551 5,483,438.64 13,227,266.53 238,615,211 0 1
2. LSM 2876 .76 .11 .99 0 −.04*1
3. Helpfulness 2876 0.155 0.597 1 0 .33** .04 1
4. Positive emotion 2876 5.27 1.70 19.02 0 .52** .12** .49** 1
5. Negative emotion 2876 2.19 .99 10.81 0 −.06** .00 .01 −.29** 1
6. Cognitive component 2876 16.39 1.97 26.71 8 −.05** .05*.02 −.02 .03 1
7. Volume 2876 14.66 41.77 1,152 1 .40** .03 .79** −.06** .03 .02 1
8. Valence 2876 6.19 2.05 10 0 .00 .12 .02 .18** −.24** −.10** −.02 1
9. Dispersion 2876 2.13 1.45 27.09 0 .09** −.00 .09** −.11** .12** .07** .15** −.21** 1
*p <.1; **p <.05.
Topaloglu and Dass 15
Figure 2: Average variable evolution over time.
16 The Impact of Online Review Content and Linguistic Style Matching on New Product Sales
Table 4: Model results.
Var i a b l e Eq . ( 2 )
Estimates (no
interactions)
Eq. (1)
Estimates (with
interactions)
Intercept 14.683*** 14.594***
LSM −0.489** −0.371**
Positive emotion −0.019 −0.026*
Negative emotion −0.026 −0.029
Cognitive component 0.004 −0.002
Helpfulness 0.659
LSM*Helpfulness −4.481**
Valence*Helpfulness −0.068**
Positive emotion*Helpfulness 0.242***
Negative emotion*Helpfulness 0.181
Cognitive component*Helpfulness 0.102**
Control variables
Volume 0.003*** 0.005***
Valence 0.006 0.009
Dispersion 0.028 0.054**
Budget 0.864*** 0.859***
Distribution 0.793*** 0.796***
RRated −0.498** −0.475**
PG Rated 0.273 0.256
Comedy 0.424** 0.417**
Action 0.149 0.124
Drama −0.097 −0.079
Award nominations 0.015** 0.016**
Random Effects
Time −0.464*** −0.459***
*p <.1; **p <.05; ***p<.001; AIC =5807.8; AIC =5781; BIC =5875.7; BIC =
5869.2.
emphasizes the importance of studying the moderating role of review helpfulness
as early rectifications on management strategies may improve sales of lackluster
movies. (Appendix A presents dynamic plots of these variables for each movie.)
The results of the two-level mixed model provide partial support for the
proposed hypotheses (Table 4). We proposed that when there is a low LSM between
a review and all the reviews about a particular movie, readers are more likely to
encounter a review by a reviewer similar to themselves. Thus, they feel more
comfortable with the purchase decision, and new product sales are positively
affected. The results of the analysis show a significant effect in this direction (β=
−0.489,p<.05), thus supporting H1. In addition to the impact of review style, the
richness of review content is also expected to play a critical role in the behavioral
outcomes of online reviews. However, we find that positive emotions (β=−0.019,
n.s.), negative emotions (β=−0.026, n.s.), and cognitive components (β=
−0.004, n.s.) in the content of online reviews do not have a significant direct effect
on box office revenue. Therefore, H2A, H2B, and H2C are not supported. When
we look at the interaction between review helpfulness and review style and content,
Topaloglu and Dass 17
the results show significant differences. We observe a significant interaction effect
of linguistic style and review helpfulness on box office revenue (β=−4.481, p<
.05). The impact of content generated by reviewers from diverse communities on
new product sales is accentuated when the reviews are considered helpful. Thus,
H3 is supported. Further, a helpful review with high positive emotional content
has a significant impact on new product sales (β=0.242,p<.001). Thus, H4A is
supported. Helpful reviews with negative emotional content, however, do not have
a significant positive impact on new product sales (β=0.181, n.s.). Therefore,
H4B is not supported. Finally, helpful reviews with high cognitive content have a
positive impact on box office revenue (β=0.102,p<.05). This finding supports
H4C. Collectively, these results provide strong support for the notion that review
helpfulness works as a critical decision-making cue and that the effect of review
style and content on new product sales largely depends on review helpfulness.
Of the control variables, consistent with prior research (Godes & Mayzlin,
2004), the effects of volume (β=0.005,p<.001) and dispersion (β=0.054,
p<.05) are positive and significant. However, valence does not have a significant
effect on sales (β=−0.009, n.s.). This finding is consistent with the argument
that online product ratings are not an ideal metric to study review characteristics
(Liu, 2006), and our results suggest that contextual analysis and linguistic styles
are viable alternatives. Finally, various movie characteristics, including budget
(β=0.859,p<0.001), distribution (β=0.796,p<.001), award nomination
(β=0.016,p<.05), R rating (β=−0.475,p<.05), comedy genre (β=0.417,
p<.05), and action genre (β=0.453,p<.05), affect box office revenue.
Robustness Checks
We performed multiple checks on the robustness of the model implemented in the
study. First, we address the issue of reverse causality and potential endogeneity
issue of movie revenue at week tand movie rating at week t−1. Extant studies
suggest that movie distribution plays a significant role on the movie evaluation
and its ratings (e.g., Eliashberg & Shugan, 1997). Therefore, movie distribution,
in terms of number of movie theaters at week t−1 may be considered as an
instrument variable to control for endogeneity. Second, the findings of our analysis
suggest that diversity in opinion and different linguistic styles can help increase
sales of new products as the diversity can stimulate curiosity among potential
consumers. To verify this assertion, we included an interaction between dispersion
and LSM in the model, but unfortunately, the term is found to be insignificant.
Third, we tested the U-shaped relationship between review content (posi-
tive emotions, negative emotions, and cognitive components) and new product
sales. These new variables were found to be insignificant, and the main results
unchanged. Fourth, extant literature suggests that the movie reviews affect movie
sales differently during the tenure of the movie. For instance, Basuroy, Chatterjee,
and Ravid (2003) suggest that positive reviews are positively effective in the early
eight weeks, while the effects of negative reviews on movie sales diminish over
time. To verify this hypothesis, we split our data set into two sets: (i) movies that
ran for eight weeks or less and (ii) movies that ran more than eight weeks, and
estimate our model (Equation 1) using each subsample separately. Results suggest
18 The Impact of Online Review Content and Linguistic Style Matching on New Product Sales
that the moderating role of review helpfulness on the effects of valence and cog-
nitive components on sales is significant in the early eight weeks. However, the
moderating role on the effects of LSM on sales is significant after eight weeks of
movie tenure. Finally, to validate our use of a mixed model to test our hypotheses,
we ran the fixed effects model as a single layer multiple regression model without
the control variables but with time as a predictor. Results presented in the Appendix
B suggest that only volume, dispersion and time are significant. On further testing
for multicollinearity on the single layer model, we only found “time” to have a
VIF factor higher than 9 (Maruyama, 1998), highlighting the validity of the mixed
model to test the hypotheses.
DISCUSSION
This study highlights the impact of online review content and linguistic style
matching on new product sales and the key role review helpfulness plays in this
relationship in the context of the motion picture industry. Relying on cognitive
psychology literature (Chaiken & Trope, 1999), we hypothesize that affective
and cognitive elements in review content should have an impact on behavioral
outcomes of online reviews. Although prior research investigates online reviews
mostly based on quantitative metrics such as volume, valence, or dispersion, recent
advances in data collection and text analytics allow researchers to dig deeper
into qualitative characteristics of online reviews. For instance, valence mostly
measured by online product ratings denotes a degree of attraction or aversion
toward a product (Chevalier & Mayzlin, 2006). This approach, however, presents
room for improvement for three reasons. First, online ratings suffer from high
ceiling effects and low variability; this is evident in the accumulation of scores
toward the boundaries of a rating scale (Resnick & Zeckhauser, 2002). Second,
looking beyond the numerical ratings and analyzing the content provide a richer
understanding of customer opinions (Kozinets et al., 2010). Third, users generating
long product reviews usually do so after an extensive elaboration, and their posts
may include pros and cons of the product. Analyzing the multidimensional content
of large data sets becomes a challenge for researchers (Godes et al., 2005) and
leads them to use numerical product ratings (Chevalier & Mayzlin, 2006). After
all, valence may take two values, positive or negative, and it is a daunting task to
determine whether a long and detailed product review is favorable or critical. In
many instances, it can be both. Using a text mining approach, however, enables
us to extract both positive and negative emotions woven into the fabric of online
reviews.
Social psychology theories purport that style and content are inseparable
components in oral and written communication (Huffaker et al., 2011). Commu-
nication accommodation theory suggests that when interacting with others, people
adjust their communication style to accommodate social expectations (Giles &
Smith, 1979) and that readers feel more comfortable when an online review style
matches a certain standard they come to expect. When LSM is low, reviews are
generated by users from diverse backgrounds. Therefore, we hypothesize a di-
rect negative impact of LSM on new product sales. In addition, we show that
review helpfulness, the likelihood that a review facilitates others’ decision-making
Topaloglu and Dass 19
process, moderates the relationship between online reviews and new product sales,
considering the difficulty users face in processing the huge amount of available
online information. Table 5 provides a summary of the research findings.
Implications for Research
This study provides unique insights into eWOM theory. First, our analysis shows
that the impact of online review content on new product sales strongly depends
on review helpfulness. This finding also supports the theory that when elabora-
tion likelihood is low, customers opt for a peripheral route that includes low-
effort decision-making cues, not a central route that involves careful consideration
of the true merits of presented information (Chaiken & Trope, 1999). In other
words, when a large amount of online information is available, especially for low-
involvement products, customers are unlikely to process most of the information
and review helpfulness functions as a critical decision-making cue. Positive emo-
tions and cognitive components in the content of helpful reviews have a significant
positive impact on new product sales.
Furthermore, the results reveal that LSM has a significant negative impact on
new product sales. Prior research argues that when reviewers’ writing styles match
an expected norm, readers feel more comfortable and are more likely to take an
action (Ireland & Pennebaker, 2010). This is also consistent with the strength of
the weak-ties hypothesis that argues that people dissimilar to each other are more
influential over each other than people who are similar to each other (Granovetter,
1973). To our knowledge, Ludwig et al.’s study (2013) is the only one to test this
phenomenon for online reviews, and they find a positive impact of LSM in online
reviews on book sales. Contrary to their work, however, we find that as stylistic
match between an online review and all the reviews regarding the same product
goes down, sales increase. Social dynamics in an online review setting may differ
from the contexts of existing studies, including personal correspondence, academic
writing, or poetry (Ireland & Pennebaker, 2010). Style inconsistency possibly hints
at the reviewer variety that attracts a wider range of customers, which subsequently
brings in more new product sales. In addition, our data show an interaction between
LSM and review helpfulness. That is, helpful reviews generated by diverse users
exert a significant influence over new product sales.
Implications for Practice
This study also has implications for practitioners and decision makers. Managing
the online sentiment after a product launch, encouraging customers to generate
online reviews, and running viral marketing campaigns are increasingly popu-
lar strategic elements decision makers consider (Gunnec & Raghavan, 2017). In
this respect, our results suggest that product managers should encourage their
customers to generate helpful online reviews that are also rich in terms of their
affective and cognitive content. These types of reviews, especially early in the
product life cycle, are shown to be more effective than high product ratings. Mu-
dambi and Schuff (2010) find that extreme ratings are considered less helpful than
moderate ratings for experience goods and that the length of a review has a positive
impact on helpfulness perceptions. Therefore, when decision makers prompt their
20 The Impact of Online Review Content and Linguistic Style Matching on New Product Sales
Table 5: Summary of the findings.
Expectations Results Implications
H1: Low LSM in online reviews will have a positive impact on
new product sales.
Supported Decision makers should encourage users from diverse
backgrounds to post reviews.
H2A: Positive affective content of online reviews will have a
positive impact on new product sales.
Not supported The amount of online information and low elaboration
likelihood undermines the impact of review content on sales.
H2B: Negative affective content of online reviews will have a
negative impact on new product sales.
Not supported The amount of online information and low elaboration
likelihood undermines the impact of review content on sales.
H2C: Cognitive content in online reviews will have a positive
impact on new product sales.
Not supported The amount of online information and low elaboration
likelihood undermines the impact of review content on sales.
Moderating Effects of Review Helpfulness
H3: Positive impact of low LSM on new product sales will
accentuate when review helpfulness is high.
Supported Decision makers should encourage users from diverse
backgrounds to post helpful reviews.
H4A: Positive impact of positive affective content on new
product sales will accentuate when review helpfulness is
high.
Supported Decision makers should encourage users to post helpful
reviews that include positive affective components.
H4B: Negative impact of negative affective content on new
product sales will accentuate when review helpfulness is
high.
Not supported Helpful reviews are preferable even with negative affective
content because they do not exert negative influence on new
product sales.
H4C: Positive impact of cognitive content on new product
sales will accentuate when review helpfulness is high.
Supported Decision makers should encourage users to post helpful
reviews that include logical reasoning.
Topaloglu and Dass 21
customers to post an online review following a purchase, they should encourage
customers to express their opinions in detail and at length.
In addition, our results show that stylistic inconsistencies in online reviews
lead to an increase in new product sales. This finding stands in contrast to conven-
tional understanding in eWOM research that the comments of a select few opinion
leaders are most effective in moving the products (Helm, M¨
oller, Mauroner, &
Conrad, 2013) because these influential opinion leaders or category experts are
expected to have a more consistent linguistic style. On the contrary, our results
reveal that reviewers from diverse backgrounds whose linguistic styles are not
necessarily consistent with each other prove to be more valuable in viral marketing
campaigns.
Limitations and Future Directions
This research presents some limitations. Similar to the results of all dictionary-
based text mining studies, the findings of this research depend heavily on the
validity and reliability of the dictionary used. Although all of these types of dic-
tionaries present an improvement potential, the number of studies and contexts in
which the LIWC dictionary has been successfully tested gives us confidence in our
results (Tausczik & Pennebaker, 2010). Further, future text-mining studies in this
realm should consider changing writing habits in online communication. In partic-
ular, the role of emojis as a new generation of emoticons in online reviews should
be investigated. Because emojis are shorthand tools to express emotions and ideas
(Novak, Smailovi´
c, Sluban, & Mozetiˇ
c, 2015), they may strengthen the impact of
affective and cognitive components on behavioral outcomes. In addition, future re-
search should consider variables other than review helpfulness that may potentially
function as an online decision-making cue given the increasingly large amount of
online information. One potential alternative for future research is to study the
possibility of creating a match between the personal characteristics of reviewers
and readers. Customers are becoming less likely to read and heed all product
reviews. Similar to the way websites offer different products to their customers
based on customers’ previous buying habits, they may display matching reviews to
their customers and save them time and hassle. Although our study tried to include
as many controls as possible (number of screens (DISTRIBUTION), number of
awards (AWARDS), marketing budget (BUDGET), MPAA ratings (R RATED,
PG RATED), and genres (COMEDY, ACTION, DRAMA)), future studies may
also consider adding information from third-party endorsements such as articles
written in popular press in the analysis to provide another layer of robustness. As
the online platforms are constantly evolving, it is possible that the style, content,
and type of reviews are changing. Therefore, future studies may consider examin-
ing whether such changes are occurring, and if so, how they change the findings
reported in this article.
CONCLUSION
In this article, we explore the impact of online review content and style on new
product sales and the moderating role of review helpfulness in this relationship.
22 The Impact of Online Review Content and Linguistic Style Matching on New Product Sales
We collect a large number of online reviews from the motion picture industry and
utilize a dictionary-based text mining approach to test the proposed relationships.
The results reveal a significant interaction between online review content and
review helpfulness. In particular, positive emotions and cognitive components in
the content of helpful online reviews impact new product sales. Further, LSM
of online reviews exerts a negative impact on new product sales. Overall, these
findings highlight the importance of studying textual characteristics of online
reviews and the impact of review helpfulness as a decision-making cue in this
research realm.
SUPPORTING INFORMATION
Additional supporting information may be found online in the Supporting Infor-
mation section at the end of the article.
Appendix A: Data Description using Functional Data Analysis
Appendix B: Results from Single level Model
REFERENCES
Anderson, K. (2015). What is the ecommerce industry average for # of
product reviews per product sold? accessed January 1, 2018, available
at https://www.quora.com/What-is-the-ecommerce-industry-average-for-of-
product-reviews-per-product-sold-Per-product-available.
Basuroy, S., Chatterjee, S., & Ravid, S. A. (2003). How critical are critical reviews?
The box office effects of film critics, star power, and budgets. Journal of
Marketing,674, 103–117.
Berger, J., & Milkman, K. L. (2012). What makes online content viral? Journal of
Marketing Research,492, 192–205.
Boyer, K. K., & Hult, G. T. (2005). Customer behavior in an online ordering
application: A decision scoring model. Decision Sciences,364, 569–598.
Brown, J., Broderick, A. J., & Lee, N. (2007). Word-of-mouth communication
within online communities: Conceptualizing the online social network. Jour-
nal of Interactive Marketing,213, 2–20.
Cao, Q., Duan, W., & Gan, Q. (2011). Exploring determinants of voting for the
“helpfulness” of online user reviews: A text mining approach. Decision
Support Systems,502, 511–521.
Chaiken, S., & Trope, Y. (1999). Dual-process theories in social psychology.New
York, NY: Guilford Press.
Chakravarti, A., Janiszewski, C., & ¨
Ulk¨
umen, G. (2006). The neglect of prescreen-
ing information. Journal of Marketing Research,434, 642–653.
Chevalier, J., & Mayzlin, D. (2006). The effect of word of mouth: Online book
reviews. Journal of Marketing Research,433, 345–354.
Topaloglu and Dass 23
Chen, Y., & Xie, J. (2008). Online consumer review: Word-of-mouth as a new
element of marketing communication mix. Management Science,543, 477–
491.
Cheng, Y., & Ho, H. (2015). Social influence’s impact on reader perceptions of
online reviews. Journal of Business Research,684, 883–887.
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,295, 944–
957.
Clore, G. L., & Storbeck, J. (2006). Affect as information about liking, efficacy,
and importance. New York, NY: Psychology Press.
Cohn, M., Mehl, M. R., & Pennebaker, J. W. (2004). Linguistic indicators of psy-
chological change after September 11, 2001. Psychological Science,1510,
687–93.
Dass, M., & Shropshire, C. (2012). Introducing functional data analysis to man-
agerial science. Organizational Research Methods,154, 693–721.
Duan, W., Gu, B., & Whinston, A. B. (2008). Do online reviews matter? An
empirical investigation of panel data. Decision Support Systems,454, 1007–
1016.
Elberse, A., & Eliashberg, J. (2003). Demand and supply dynamics for sequentially
released products in international markets: The case of motion pictures.
Marketing Science,223, 329–354.
Eliashberg, J., & Shugan, S. M. (1997). Film critics: Influencers or predictors?
Journal of Marketing,612, 68–78.
Filieri, R. (2015). What makes online reviews helpful? A diagnosticity-adoption
framework to explain informational and normative influences in e-WOM.
Journal of Business Research,686, 1261–1270.
Fuller, M. A., Serva, M. A., & Benamati, J. S. (2007). Seeing is believing: The tran-
sitory influence of reputation information on e-commerce trust and decision
making. Decision Sciences,384, 675–699.
Gao, G. G., Greenwood, B. N., Agarwal, R., & McCullough, J. S. (2017). Vo-
cal minority and silent majority: How do online ratings reflect population
perceptions of quality. MIS Quarterly,393, 565–589.
Giles, H. (2009). The process of communication accommodation. In N. Coup-
land & A. Jaworksi (Eds.), The new reader in sociolinguistics. Basingstoke,
England: Macmillan, 276–286.
Giles, H., & Smith, P. (1979). Accommodation theory: Optimal levels of conver-
gence. Baltimore, MD: University Park Press.
Godes, D., & Mayzlin, D. (2004). Using online conversations to study word-of-
mouth communication. Marketing Science,234, 545–560.
Godes, D., Mayzlin, D., Chen, Y., Das, S., Dellarocas, C., Pfeiffer, B., et al. (2005).
The firm’s management of social interactions. Marketing Letters,163, 415–
428.
24 The Impact of Online Review Content and Linguistic Style Matching on New Product Sales
Goes, P. B., Lin, M., & Au Yeung, C. M. (2014). “Popularity effect” in user-
generated content: Evidence from online product reviews. Information Sys-
tems Research,252, 222–238.
Granovetter, M. (1973). The strength of weak ties. American Journal of Sociology,
786, 1360–1380.
Gundecha, P., & Liu, H. (2012). Mining social media: A brief introduction. In new
directions in informatics, optimization, logistics, and production. Tutorials
in Operations Research,1, 17.
Gunnec, D., & Raghavan, S. (2017). Integrating social network effects in the
share-of-choice problem. Decision Sciences,486, 1098–1131.
Guo, H., Pathak, P., & Cheng, H. K. (2015). Estimating social influences from
social networking sites—Articulated friendships versus communication in-
teractions. Decision Sciences,461, 135–163.
Helm, R., M¨
oller, M., Mauroner, O., & Conrad, D. (2013). The effects of a lack of
social recognition on online communication behavior. Computers in Human
Behavior,293, 1065–1077.
Hennig-Thurau, T., Wiertz, C., & Feldhaus, F. (2015). Does Twitter matter? The
impact of microblogging word of mouth on consumers’ adoption of new
movies. Journal of the Academy of Marketing Science,433, 375–394.
Hu, N., Koh, N. S., & Reddy, S. K. (2014). Ratings lead you to the product, reviews
help you clinch it? The mediating role of online review sentiments on product
sales. Decision Support Systems,57, 42–53.
Huffaker, D. A., Swaab, R., & Diermeier, D. (2011). The language of coalition
formation in online multiparty negotiations. Journal of Language and Social
Psychology,301, 66–81.
Humphreys, A. (2010). Megamarketing: The creation of markets as a social pro-
cess. Journal of Marketing,742, 1–19.
Ireland, M. E., & Pennebaker, J. W. (2010). Language style matching in writing:
Synchrony in essays, correspondence, and poetry. Journal of Personality and
Social Psychology,993, 549–71.
Jones, Q., Ravid, G., & Rafaeli, S. (2004). Information overload and the message
dynamics of online interaction spaces: A theoretical model and empirical
exploration. Information Systems Research,152, 194–210.
Kozinets, R. V., De Valck, K., Wojnicki, A. C., & Wilner, S. J. S. (2010). Networked
narratives: Understanding word-of-mouth. Journal of Marketing,74(March),
71–89.
Kulkarni, S. S., Apte, U. M., & Evangelopoulos, N. E. (2014). The use of latent
semantic analysis in operations management research. Decision Sciences,
455, 971–994.
LaFrance, M. (1985). Postural mirroring and intergroup relations. Personality and
Social Psychology Bulletin,11, 207–217.
Li, D., Browne, G. J., & Chau, P. Y. K. (2006). An empirical investigation of web
site use using a commitment-based model. Decision Sciences,373, 427–441.
Topaloglu and Dass 25
Liu, Y. (2006). Word of mouth for movies: Its dynamics and impact on box office
revenue. Journal of Marketing,703, 74–89.
Ludwig, S., De Ruyter, K., Friedman, M., Br¨
uggen, E. C., Wetzels, M., & Pfann,
G. (2013). More than words: The influence of affective content and linguistic
style matches in online reviews on conversion rates. Journal of Marketing,
771, 87–103.
Maruyama, G. M. (1998). Basics of structural equation modeling. Thousand Oaks,
CA: Sage Publications.
Massey, A. P., Khatri, V., & Montoya-Weiss, M. M. (2007). Usability of online
services: The role of technology readiness and context. Decision Sciences,
382, 277–293.
McFarland, D. H. (2001). Respiratory markers of conversational interaction. Jour-
nal of Speech, Language, and Hearing Research,44, 128–143.
Mudambi, S. M., & Schuff, D. (2010). What makes a helpful online review? A
study of customer reviews on Amazon.com. MIS Quarterly,341, 185–200.
Neelamegham, R., & Chintagunta, P. K. (1999). A Bayesian model to forecast
new product performance in domestic and international markets. Marketing
Science,182, 115–136.
Novak, P. K., Smailovi´
c, J., Sluban, B., & Mozetiˇ
c, I. (2015). Sentiment of emojis.
PloS One,1012, e0144296.
Oztekin, A. (2011). A decision support system for usability evaluation of web-
based information systems. Expert Systems with Applications,383, 2110–
2118.
Pappas, I. O., Kourouthanassis, P. E., Giannakos, M. N., & Chrissikopoulos, V.
(2015). Explaining online shopping behavior with fsQCA: The role of cogni-
tive and affective perceptions. Journal of Business Research,692, 794–803.
Pennebaker, J. W., Chung, C. K., Ireland, M., Gonzales, A., & Booth, R. J. (2007).
The development and psychometric properties of LIWC 2007. LIWC.net:
Austin, TX.
Pennebaker, J. W., & Francis, M. E. (1996). Cognitive, emotional, and language
processes in disclosure. Cognition and Emotion,10, 601–626.
Pennebaker, J. W., Francis, M. E., & Booth, R. J. (2001). Linguistic inquiry and
word count (LIWC): LIWC2001. Mahwah, NJ: Lawrence Erlbaum Asso-
ciates.
Phillips, K. W., Mannix, E. A., Neale, M. A., & Gruenfeld, D. H. (2004). Diverse
groups and information sharing: The effects of congruent ties. Journal of
Experimental Social Psychology,404, 497–510.
Pornpitakpan, C. (2004). The persuasiveness of source credibility: A critical review
of five decades’ evidence. Journal of Applied Social Psychology,342, 243–
281.
Prag, J., & Casavant, J. (1994). An empirical study of the determinants of rev-
enues and marketing expenditures in the motion picture industry. Journal of
Cultural Economics,18, 217–35.
26 The Impact of Online Review Content and Linguistic Style Matching on New Product Sales
Ramanathan, S., & McGill, A. L. (2007). Consuming with others: Social influ-
ences on moment-to-moment and retrospective evaluations of an experience.
Journal of Consumer Research,34, 506–524.
Resnick, P., & Zeckhauser, R. (2002). Trust among strangers in Internet trans-
actions: Empirical analysis of eBay’s reputation system. In The economics
of the Internet and e-commerce. Bingley, UK: Emerald Group Publishing
Limited, 127–157.
Senecal, S., & Nantel, J. (2004). The influence of online product recommendations
on consumers’ online choices. Journal of Retailing,802, 159–169.
Singer, J. D., & Willett, J. B. (2003). Applied longitudinal data analysis.New
York, NY: Oxford University Press.
Slatcher, R. B., & Pennebaker, J. W. (2006). How do I love thee? Let me count the
words the social effects of expressive writing. Psychological Science,178,
660–664.
Speier, C., Vessey, I., & Valacich, J. S. (2003). The effects of interruptions, task
complexity, and information presentation on computer-supported informa-
tion decision-making performance. Decision Sciences,344, 771–797.
Srinivasan, S., Rutz, O. J., & Pauwels, K. (2016). Paths to and off purchase:
Quantifying the impact of traditional marketing and online consumer activity.
Journal of the Academy of Marketing Science,444, 440–453.
Tausczik, Y. R., & Pennebaker, J. W. (2010). The psychological meaning of words:
LIWC and computerized text analysis methods. Journal of Language and
Social Psychology,291, 24–54.
Topaloglu, O., Dass, M., & Kumar, P. (2017). Does who we are affect what
we say and when? Investigating the impact of activity and connectivity on
microbloggers’ response to new products. Journal of Business Research,
77(August), 23–29.
Vogel, H. (2001). Entertainment industry economics: A guide for financial analysis.
(5th ed.). Cambridge, UK: Cambridge University Press.
Wang, H., Guo, X., Zhang, M., Wei, Q., & Chen, G. (2016). Predicting the incre-
mental benefits of online information search for heterogeneous consumers.
Decision Sciences,475, 957–988.
Watson, D., Clark, L. A., & Tellegen, A. (1988). Development and validation of
brief measures of positive and negative affect: The PANAS scales. Journal
of Personality and Social Psychology, 546, 1063.
Zhu, F., &Zhang, X. (2010). Impact of online consumer reviews on sales: The mod-
erating role of product and consumer characteristics. Journal of Marketing,
742, 133–148.
Zufryden, F. (1996). Linking advertising to box office performance of new film
releases: A marketing planning model. Journal of Advertising Research,364,
29–41.
Omer Topaloglu is an assistant professor of marketing in the Silberman Col-
lege of Business at Fairleigh Dickinson University. He holds a Ph.D. in Business
Topaloglu and Dass 27
Administration from Texas Tech University, an M.B.A. from Montclair State Uni-
versity, and a B.A. in Economics from Bogazici University. His research areas
include digital marketing, service marketing, and brand management. He teaches
courses in principles of marketing, social media marketing, and sales.
Mayukh Dass is J.B. Hoskins Professor of Marketing at the Rawls College of
Business, Texas Tech University. He is currently serving as the Associate Dean
of Graduate Programs and Research, and as the Program Director of the Rawls
Business Leadership Program at Rawls College of Business, Texas Tech Univer-
sity. His areas of expertise include networks, dynamic economies, analytical and
mathematical models.
