Content uploaded by Edward C Malthouse
Author content
All content in this area was uploaded by Edward C Malthouse on Jan 17, 2018
Content may be subject to copyright.
Too good to be true: the role of online reviews’
features in probability to buy
Ewa Maslowska
a
*, Edward C. Malthouse
a
and Stefan F. Bernritter
b
a
Medill IMC Spiegel Digital & Database Research Center, Northwestern University, 1845 Sheridan
Road, Evanston, IL 60208, USA;
b
Amsterdam School of Communication Research, University of
Amsterdam, Nieuwe Achtergracht 166, 1018 WV, Amsterdam, The Netherlands
(Received 14 August 2015; accepted 11 May 2016)
Online consumer reviews are broadly believed to be a necessary and powerful
marketing tool, and as such they have attracted considerable attention from both
marketers and academics. However, previous research has not sufficiently focused on
the effects of various review features on sales but rather used proxy measures such as
consumers’ purchase intention or perceived helpfulness of reviews. Hence, the aim of
this study was to investigate the effect of review valence and volume on actual sales.
We use data from three different e-commerce websites and study light bulbs, women’s
athletic shoes, natural hair care products, and herbal vitamins. The results show that,
contrary to popular belief, more positive ratings do not simply result in higher sales.
We find that the effect can be nonlinear, where the probability of purchase increases
with rating to about 4.2!4.5 stars, but then decreases. Also, although the majority of
extant research suggests that larger numbers of reviews bring more positive outcomes,
we show that it is not always the case.
Keywords: online reviews; probability of purchase; online ratings; review valence;
review volume
Introduction
Online consumer reviews have changed the marketing reality in which advertising has
traditionally operated as one-way communication from companies to consumers via mass
communication channels (Campbell et al. 2011). Although brands still advertise in these
traditional ‘paid’ media such as television, radio, and print, new advertising channels are
now available to them (Edelman and Salsberg 2011). Companies can utilize ‘owned’
media to contact customers directly to, for example, compel them to share their experien-
ces or visit the company’s website, or their ‘earned’ media by providing customers with
space where they can promote the company to other consumers. These new communica-
tion channels have become especially important for companies, especially in the context
of consumer endorsements being an important advertising strategy (Bernritter, Verlegh,
and Smit, 2016; Lee, Park, and Han 2011).
Consumers do not rely solely on advertising messages anymore, but direct their atten-
tion to other sources of information, especially online reviews. Surveys show that more
than half of consumers consults online reviews (CMA 2015; Mintel 2015). That is
because the majority of consumers trust recommendations from others more than tradi-
tional forms of advertising (Nielsen, 2012). Online reviews help consumers make
*Corresponding author. Email: ewa.maslowska@northwestern.edu
!2016 Advertising Association
International Journal of Advertising, 2016
http://dx.doi.org/10.1080/02650487.2016.1195622
Downloaded by [Northwestern University], [Ewa Maslowska] at 08:34 24 June 2016
decisions such as purchasing products, watching movies, or joining a sports club. They
have become a major driving force in marketing (Cui, Lui, and Guo 2012) and are a com-
mon feature on many websites. Information from other consumers, such as online
reviews, is thought to be more persuasive and trustworthy, and is especially important in
the online environment (Ba and Pavlou 2002; Willemsen, et al. 2011). Consumers buying
products online have to rely on the information provided on the website and do not have
the ability to try out the product (Lee 1998). Due to their alleged persuasiveness and prev-
alence, online reviews have attracted substantial attention from both researchers and
practitioners.
However, while online product reviews may affect consumers’ purchase decisions
(Zhu and Zhang 2010), even overshadowing other relevant product information such as
price (de Langhe, Fernbach, and Lichtenstein 2016), past research has mostly neglected
the effect of review features on actual sales and rather focused on other outcomes such as
review helpfulness, credibility, attitude, and intention (e.g., Jim!
enez and Mendoza 2013;
Purnawirawan, Dens, and De Pelsmacker 2014; Wang, Cunningham, and Eastin 2015),
or used proxy measures of sales (Wang et al. 2015) such as sales rank (e.g., Amblee and
Bui 2011; Chevalier and Mayzlin 2003) or the number of reviews (e.g.,
€
O#
g€
ut and Onur
Ta¸s2012; Ye, Law, and Gu 2009). Also, a large body of consumer review research is
built on experimental- or survey-based work (e.g., Purnawirawan, de Pelsmacker, and
Dens 2012), and assumes that effects of online reviews are linear. One exception is a
recent meta-analysis conducted by Purnawirawan et al. (2015), in which the authors dem-
onstrate that valence has a curvilinear effect on usefulness and a ceiling effect on atti-
tudes. They do not look at actual sales though. Previous research has also looked into
different review features such as valence, length, variation, and volume, but as we will
show in the remainder of this paper, the results are mixed (King, Racherla, and, Bush
2014; Zhang et al. 2010; Zhu and Zhang 2010). We aim to address these shortcomings by
studying the effect of two main features of online reviews, namely the average star rating
of a product (i.e., valence) and the number of reviews (i.e., volume) on actual sales in
three different online shops. We are able to do this using purchase data from three
companies.
These inconsistencies and gaps in the literature emphasize that there is still much to be
known about the workings and limits of consumers’ online reviews (Kimmel and Kitchen
2014). In addressing these shortcomings, we are contributing to the literature in at least
three important ways. First, by investigating the nonlinear nature of the effects of review
valence and volume on sales, we are providing an explanation for the often-mixed results
of studies that treated these effects as linear. Second, by examining the effects of online
product reviews on actual sales, we are able to show whether often-mixed findings that
are based on proxy measures of sales also hold for actual purchases. Finally, by testing
the effects of online review features among a diverse set of product categories, we
acknowledge the need for the comparison of the effects of review features among a broad
spectrum of products (Zhu and Zhang 2010). This is essential since previous research has
primarily focused on one product category (e.g., entertainment products such as movies
or books) (Cui, Lui, and Guo 2012) or compared two categories of products (Zhu and
Zhang 2010).
The role of reviews’ features
The valence of online product reviews is usually summarized by the average number of
stars rated on a five-point scale. Most online retailers also display the number of reviews
2E. Maslowska et al.
Downloaded by [Northwestern University], [Ewa Maslowska] at 08:34 24 June 2016
next to the valence. The amount of information that can influence customers’ decisions is
limited. Reviews can be seen as the most reliable and accessible indicators of product
quality and customer experiences (Simonson 2015), and hence, affect customers simplify-
ing decision-making process. Because consumers are cognitive misers with limited cogni-
tive resources, and as such they often use heuristics instead of elaboration, they may rely
on heuristic cues in the shopping environment. Product reviews have been shown to influ-
ence consumers via heuristic processing (Forman, Ghose, and Wiesenfeld 2008) during
which consumers use peripheral cues !rather than the content of reviews !to form
opinions. This is especially the case when consumers face an information overload (For-
man, Ghose, and Wiesenfeld 2008) or in low-risk situations when consumers apply sim-
ple search and evaluation techniques. Thus, cues such as the average number of stars
(valence hereafter) and the number of reviews (volume hereafter) are perfectly suited as a
summary for consumers processing product information based on heuristics. The advan-
tage of these kinds of summaries is that they cannot be misinterpreted, in contrast to infor-
mation provided in the text of a review. We also focus on volume and valence because
they are commonly shown next to the product price and thus all customers are exposed to
them, even those who do not navigate to the ‘review tab.’ This makes them two key
(Wang et al. 2015, 73) and broadly studied (e.g., Mahajan, Muller, and Kerin 1984; Liu
2006) metrics.
Review valence
Valence is widely believed to positively influence attitudes toward the brand and purchase
behaviors. Many companies even want to remove negative reviews because they fear that
they will discourage potential customers. However, as Table 1 shows, the results of past
studies are equivocal (King, Racherla, and Bush 2014; Zhang et al. 2010). A positive
effect of valence has been shown for book sales rank and box office performance of mov-
ies (Chevalier and Mayzlin 2006; Chintagunta, Gopinath, and Venkataraman 2010;
Zhang and Dellarocas 2006). Also, Gopinath, Thomas, and Krishnamurthi (2014) find
that valence has a direct effect on sales of cell phones. Chevalier and Mayzlin (2006)
show that 1-star reviews can negatively affect sales rank on Amazon.com and Clemons,
Gao, and Hitt (2006) find that while high ratings can predict sales, bad ratings do not pre-
dict poor sales. Finally, Kim et al. (2015) demonstrate that negative reviews are associ-
ated with a decrease in readers’ spending levels. Valence has been also shown to affect
other variables such as consumers’ attitude, perceived source credibility, and purchase
intention (Wang, Cunningham, and Eastin 2015).
Others demonstrate that valence does not matter. In their study of the online movie
reviews on Yahoo.com, Duan, Gu, and Whinston (2008a) show that valence has no signif-
icant effect on box office sales, which is in line with the findings of Liu (2006) for box
office revenue. Similarly, Chen, Wu, and Yoon (2004) show that online reviews do not
affect book sales rank on Amazon. Finally, Amblee and Bui (2011) find that the valence
does not predict purchases of digital microproducts. Still, the majority of studies demon-
strate valence effects.
Some studies suggest that a disproportional number of positive online reviews may
cause consumers to discount positive reviews as not reliable (Chevalier and Mayzlin
2006) and therefore may negatively affect sales. In accordance with this reasoning, Bos-
man, Boshoff, and van Rooyen (2013) show that valence significantly affects review
credibility and that for every additional star, credibility decreases on average by 2.39%
(if all other factors remain unchanged). This suggests that a review with a poor rating is
International Journal of Advertising 3
Downloaded by [Northwestern University], [Ewa Maslowska] at 08:34 24 June 2016
Table 1. Overview of some relevant literature on the effects of valence and volume of reviews.
Study Findings of valence Findings on volume Context
Amblee and Bui (2011) No effect CSales rank
Short stories
Basuroy, Chatterjee, and Ravid
(2003)
C
The effect of negative reviews diminishes
over time
Negative reviews hurt more than positive
ones help
Box office revenue
Bosman, Boshoff, and van Rooyen
(2013)
¡
For every additional star, the review
credibility decreases by 2.39%
Credibility of book reviews on Amazon.
com and Barnesandnoble.com
Chen, Wu, and Yoon (2004) No effect CAmazon.com books
Chevalier and Mayzlin (2006)CFor Amazon
No effect for BN
Cfor Amazon Books sales Amazon and Barnes and
Noble
Chintagunta, Gopinath, and
Venkataraman (2010)
CNo effect Box office sales (movies)
Chiou and Cheng (2003)CNegative messages hurt low-image
brands
CMessage repetition helps high-
image brands
Brand evaluation and attitude toward the
Web owner
Clemons, Gao, and Hitt (2006)CNo effect Beer sales
Cui, Lui, and Guo (2012)C
Stronger effect than volume for search
products
Stronger effect of negative reviews
than positive ones
C
More important for experience
products
The effect decreases over time
Sales rank data Amazon
New products (electronics and video
games)
Differences between search and experience
products
Dellarocas, Zhang, and Awad (2007)CCBox office sales (movies)
Dhar and Chang (2009)CSales of music albums
Duan, Gu, and Whinston (2008a) No effect CBox office sales (movies)
Duan, Gu, and Whinston (2008b)COn volume CBox office performance
(continued)
4E. Maslowska et al.
Downloaded by [Northwestern University], [Ewa Maslowska] at 08:34 24 June 2016
Table 1. (Continued )
Study Findings of valence Findings on volume Context
Floyd et al. (2014)C
Sales elasticities based on valence are
significantly higher than those based on
review volume
CSales Elasticities Meta-analysis
Ghose and Ipeirotis (2011)CFor audio video only CFor DVDs and digital
cameras, but not for audio
video
Sales rank
Gopinath, Thomas, and
Krishnamurthi (2014)
CNo effect Cell phones
Gu, Park, and Konana (2012) No effect for retailer-hosted reviews
CFor external reviews
CFor both retailer-hosted and
external reviews
Cameras Amazon sales rank
Ho-Dac, Carson, and Moore (2013)CFor weak brands
No effect for strong brands
Blue-Ray and DVD players sales
Differences between strong and weak
brands
Jang, Prasad, and Ratchford (2012)C
A unit increase in the mean of product
reviews is worth $45 on average
C
Less important
Experiment on hotel choice
Kostyra et al. (2016)CDoesn’t have a direct effect, but
moderates the effect of
valence
A choice-based conjoint experiment Effect
on willingness-to-pay
Kusumasondjaja, Shanka, and
Marchegiani (2012)
¡
When the reviewer’s identity is not
disclosed, there is no significant
difference between positive and
negative reviews
Credibility
Lee and Youn (2009)CModerated by review platform Willingness to recommend
Liu (2006) Valence does not offer explanatory power CBox office revenue
€
O#
g€
ut and Onur Ta¸s(2012)CNumber of reviews as a proxy for hotel
rooms sales on Booking.com
(continued)
International Journal of Advertising 5
Downloaded by [Northwestern University], [Ewa Maslowska] at 08:34 24 June 2016
Table 1. (Continued )
Study Findings of valence Findings on volume Context
Park and Lee (2009) Greater effect of negative reviews than
positive reviews, established websites
than unestablished websites,
experience goods than search goods
eWOM effectiveness
Purnawirawan et al. (2015) Curvilinear effect on usefulness, stronger
for experience than search productsC
belongs to the ceiling effect on
attitudes
Perceived usefulness, attitude
Wang et al. (2015) moderated by volume Hotel rooms
Yang et al. (2012)COnly for niche movies No effect for
niche movies with greater box office
revenue
C
Greater for mass movies
Box office revenues
Ye, Law, and Gu (2009)CA 10% improvement in reviewers’
rating can increase sales by 4.4%
Number of reviews as a proxy of the
number of hotel room bookings
Zhang et al. (2010)CCRestaurants popularity
Zhu and Zhang (2010)CFor less popular online games CFor both popular and less
popular online games
Sales of console games
Popularity of games
Online and offline games
6E. Maslowska et al.
Downloaded by [Northwestern University], [Ewa Maslowska] at 08:34 24 June 2016
perceived as more trustworthy. In line with this notion, O’Reilly and Marx (2011) show
that consumers are skeptical of reviews that are too positive. In other words, consumers
who see only 5-star reviews become suspicious (Dholakiya 2014). The theory that exces-
sively positive reviews can have negative effects is also supported by the research of
Mudambi and Schuff (2010) who find that moderate reviews are better than extreme
reviews for experience goods.
There are several reasons why moderate reviews may be more appreciated and exces-
sively positive reviews may be perceived as unreliable. First, when consumers rate a
product on a five-point scale, a mean rating of three may indicate variance in reviewers
suggesting that both positive and negative product attributes are discussed, which builds
credibility and reduces reporting bias. Previous research has shown that two-sided mes-
sages can enhance source credibility and brand evaluation. For example, in advertising,
Kamins and Assael (1987) show that one-sided versus two-sided messages increase dero-
gation of the advertiser. Similarly, Eisend (2006) finds that two-sided advertising can
enhance persuasion by indicating that it is trustworthy. Second, positive product charac-
teristics may be perceived as less diagnostic of product quality (Herr, Kardes, and Kim
1991; Mizerski 1982). Reviews are often posted by either extremely satisfied or extremely
disappointed customers (Hennig-Thurau et al. 2004), which may diminish their useful-
ness. In addition, positive reviews are more prominent (Meli!
an-Gonz!
alez, Bulchand-
Gidumal, and Gonz!
alez L!
opez-Valc!
arcel 2013), which (1) causes a contrast effect with
negative reviews, enhancing consumers attention to negative reviews, and (2) may be
responsible for the fact that consumers seek out negative reviews to get assurance the
company is not hiding anything (O’Neil 2015). Third, very positive reviews may be seen
as not reliable because of consumers’ naive theories about the sources of positive infor-
mation (Chen and Lurie 2013). For instance, consumers may write positive reviews to
signal competence or for self-presentation reasons (Chen 2014). Finally, consumers also
know that firms can alter or remove reviews, or stimulate positive reviews with financial
rewards to create high ratings (Li and Hitt 2008). Hence, having some extremely positive
reviews (i.e., 5-star reviews) is not problematic as long as there are some moderate ones
(or even fewer negative ones) to average things out. Such reasoning is in line with the
findings of Doh and Hwang (2009) showing that a few negative messages can enhance
attitudes toward website and review credibility.
To disentangle extant findings, we build on Mudambi and Schuff (2010) who show
that extreme reviews of experience goods are less helpful than moderate ones, producing
an inverted U-shape effect, and Schindler and Bickart (2012) who find that the proportion
of positive evaluative statements in a review has an inverted U-shape effect on review
value. Also, Purnawirawan et al (2015) in their meta-analysis find that the effects of
review valence are nonlinear: Valence has a curvilinear effect on usefulness and a ceiling
effect on attitudes.
The phenomenon that too much of a good thing can have negative consequences has
been acknowledged by a stream of research in various disciplines. For instance, it has
been demonstrated that an overabundance of users’ connections with friends on Facebook
harms how others rate these users’ attractiveness and extraversion, compared to a rather
intermediate amount of connections (Tong, van der Heide, Langwell, and Walther 2008).
Another study found that consumers’ trust in brands could decrease if they have experi-
enced too many satisfying transactions with the brand previously (Vlachos, Vrechopou-
los, and Pramatari 2011). In the same vein, it has been shown that too much positive
affect can decrease proactive behavior (Lam, Spreitzer, and Fritz 2014). Therefore, we
propose that the relationship between valence and sales is nonlinear. We expect that
International Journal of Advertising 7
Downloaded by [Northwestern University], [Ewa Maslowska] at 08:34 24 June 2016
higher valence will lead to more positive effects (i.e., higher purchase probability), but
extremely good average star ratings will lead to less positive outcomes, because they
imply a lack of variation and induce suspicion. Hence, we hypothesize that:
H1: The relationship between valence and purchase probability is nonlinear, where
higher valence leads to higher purchase probability, but extremely high valence leads to
lower purchase probability.
Review volume
The presence of reviews is expected to bring positive outcomes. For example, Amblee
and Bui (2011) demonstrate that digital microproducts with reviews sell significantly bet-
ter than products without reviews. However, extant research does not agree on the effect
of the number of reviews (Table 1).
For example, Bazaar (2015) and Matfield (2011) show that volume is generally posi-
tively correlated with sales volume and revenues, regardless of valence. Also, other past
studies have shown that volume affects market outcomes, such as box office sales (Duan,
Gu, and Whinston 2008b; Liu 2006), sales rank of electronic products (Cui, Lui, and Guo
2012; Gu, Park, and Konana 2012), but also other measures such as consumer attention
(Godes and Mayzlin 2004; Liu 2006). Chiou and Cheng (2003) find that consumers
exposed to 12 posts (vs. 6) about cell phones evaluate the product more positively, have a
higher overall attitude, and show stronger overall liking. Although the majority of studies
confirms the positive belief about the effect of volume (King, Racherla, and Bush 2014),
there are also studies suggesting that volume on its own is not enough.
For example, Gopinath, Thomas and Krishnamurthi (2014) show that the volume of
online word of mouth (WOM) does not impact sales of cell phones in a significant way.
Similarly, Chintagunta, Gopinath, and Venkataraman (2010) find that volume does not
have a significant impact on box office performance. One study even shows that, for attri-
bute-focused reviews, a large number of reviews create information overload for consum-
ers (Park and Lee 2008), which might result in negative effects on consumers’ purchase
behavior. This suggests an important distinction between the displayed number of
reviews, which is a peripheral cue signaling popularity and credibility, and the actual
reviews. We are looking at the displayed number of reviews and hence do not expect
such an effect. For consumers to experience overload, they would need to read all the
reviews. In addition, as discussed by Simonson (2015), too much information does not
have to lead to cognitive overload, because consumers are usually offered a way to
choose information they want to process. Summarizing, although the results of previous
research seem mixed, most of the extant research suggests a positive effect of volume.
As stated earlier in the introduction, the number of reviews constitutes a heuristic cue.
Such a cue can trigger more positive responses (Chiou and Cheng 2003). According to
the elaboration likelihood model (Cacioppo et al. 1986), consumers use a shortcut and fol-
low the peripheral route of processing, which means they focus on heuristics when they
are not motivated to process a message. Hence, they look for cues that could signal the
value of a message, such as volume.
The idea behind the alleged positive effect is that volume, as a peripheral cue, can sig-
nal the popularity of a product and intensity of WOM (Duan, Gu, and Whinston 2008a;
Cui, Lui, and Guo 2012; Liu 2006; Park and Lee 2008). More reviews can elicit consum-
ers’ interest and increase product awareness (Chen, Wu, and Yoon 2004), that is, make it
more salient in the consumer’s mind. The number of online product reviews may indicate
8E. Maslowska et al.
Downloaded by [Northwestern University], [Ewa Maslowska] at 08:34 24 June 2016
the product’s popularity because consumers can assume that the number of reviews is
related to the number of consumers who have bought the product, and hence, purchase
intention increases with the number of reviews (Park, Lee, and Han 2007). Also, higher
volume suggests more information and hence can reduce consumers’ uncertainty and
strengthen their confidence in a product, leading to a greater willingness to pay for it
(Brynjolfsson and Smith 2000). More reviews may be perceived as more objective and
thus more trustworthy (Wang et al. 2015). Finally, higher volume can suggest that con-
sumers care about the brand or product, because they take their time to write a review
about it (Chiou and Cheng 2003). Thus, volume can also be indicative of consumers’
enthusiasm about the product (Duan, Gu, and Whinston 2008a).
We thus hypothesize that:
H2: The relationship between volume and purchase probability is positive with higher
volume leading to higher purchase probability.
Method
To test our hypotheses, we use data from three different Internet-only retailers selling
products from different categories. Since none have physical stores, all purchases can be
recorded and linked to the reviews that were shown at the time of ordering. We are thus
able to link online reviews to real purchase behavior. Our focus is on whether or not an
item was purchased rather than predicting the quantity or purchase amount. Retailer 1
sells different types of light bulbs, retailer 2 sells health and beauty care products, and
retailer 3 is an Internet retailer in apparel, jewelry, electronics, home, toy, health, and
beauty.
We have six weeks of sales data from 31 August 2014 through 11 October 2014 for
retailer 1, and 15 weeks of sales data from 29 Jun 2014 to 11 October 2014 for retailers 2
and 3. We include four product categories: light bulbs, women’s athletic shoes, natural
hair care products, and herbal vitamins. We restrict our attention in this study to the six
largest brands or brands that were bought at least 1.000,00 times. For retailer 1, we also
focus on the five largest categories of light bulbs, that is, LED, halogen, incandescent, lin-
ear, and compact fluorescent. We included dummy variables indicating the category as
control variables, but do not report results for these variables since they are not part of
our hypotheses. Table 2 gives summary statistics for the categories. For each exposure
we know the following about the reviews to which the customer was exposed: (1) the
average number of stars across the reviews for the stock keeping units (SKU) (stars), (2)
the number of reviews for the SKU (num), and (3) the price.
Purchases are complicated decision processes that depend on factors beyond review
traits. Previous research suggests that information on prices should be included when dis-
cussing purchase decisions. In addition, brand awareness, which is a result of marketing
Table 2. Descriptive statistics.
Product category Number of SKUs Number of orders Number of displays Buy rate
Light bulbs 919 2608 118,891 2.19%
Women’s athletic shoes 107 438 49,958 0.88%
Natural hair care 26 915 14,838 6.2%
Herbal vitamins 81 1701 21,512 7.9%
International Journal of Advertising 9
Downloaded by [Northwestern University], [Ewa Maslowska] at 08:34 24 June 2016
communication efforts, should also be included. Some previous studies have shown that
the effect of reviews may differ for different brand familiarity and price (e.g., Ba and Pav-
lou 2002; Berger, Sorensen, and Rasmussen 2010; Chiou and Cheng 2003). We include
the brand and the price in our model to control for such effects. Notice that brand is a
proxy for ad spend. We create a dummy variable indicating brands in order to control for
the effect of brands’ advertising. Our data are summarized at the SKU level, which corre-
sponds to a web page used to display the SKU. For each SKU (web page) we know the
volume, valance and price for the item, the number of times the page was viewed (expo-
sures), and the number of times the item was included in an order. Thus, a simple estimate
of purchase probability is the number of orders divided by the number of exposures. Our
model attempts to explain these probabilities with volume and valence, after controlling
for price and brand. Volume and price are logged because they are right skewed with
outliers.
We estimate the following logistic regression using generalized additive models
(GAMs) (Hastie and Tibshirani 1990):
log p
1!p
!"
Dbrand Clog price CsvalenceðÞCslog volumeðÞ
where pis the probability of purchase, sis a univariate smoothing spline with three
degrees of freedom representing the nonlinear effect of valence and volume on log odds
of purchase. GAMs extend standard linear models by permitting nonlinear functions of
each of the independent variables. In such models each linear component is replaced with
a nonlinear function. They are called additive, because they estimate a separate function
for each variable and then add their contributions together (James et al. 2013). SAS
PROC GAM uses smoothing splines, which are flexible, nonparametric functions that can
approximate any continuous univariate function. The flexibility of each function is con-
trolled by a penalty term. GAM software compares the fit from the nonlinear spline model
with the fit from a linear model to determine whether nonlinearity is necessary. When the
effect of the spline is significant, the nonlinear function explains significantly more vari-
ance than the linear term.
Results
Table 3 provides the analysis of deviance results for the different terms in the model,
which indicates whether there are significant nonlinear contributions from the variables.
Each smoothing effect in the model has a chi-square test comparing the deviance between
the full model and the model without this variable. Table 4 gives parameter estimates of
the linear terms.
Concerning the first hypothesis, the nonlinear effect of valence (i.e., average stars) is
significant (p<.05) for all four product categories (Table 3, Spline (avgstar)), indicating
that the spline explains significantly more variance than a linear term. Figure 1 shows the
effect of the number of stars on the log odds ratio, where the shaded bands indicate 95%
confidence intervals for the mean prediction. As discussed in the Methods section, with
additive models each term affects the logit of buying in an additive way. Consider, for
example, Figure 1(a) and the average star rating. The spline has an ‘effect’ of about 0 for
a product with 2.5 stars, and the effect is about 0.4 for 4.2 stars. Thus, the logit of buying
is about 0.4 higher for a 4.2 star bulb compared with a 2.5 star bulb. Likewise the logit of
buying Brand D is 0.56 higher than base-category brand (Table 4). A logit of 0
10 E. Maslowska et al.
Downloaded by [Northwestern University], [Ewa Maslowska] at 08:34 24 June 2016
corresponds to a purchase probability of 50% D1/(1 Cexp(¡0)). Likewise a logit of ¡1
gives a probability of 1/(1 Cexp(¡¡1)) D27%, a logit of ¡2 gives a probability of
12%, a logit of ¡3 gives a probability of 4.7%, and a logit of ¡4 gives a probability of
1.8%.
For light bulbs and hair care products (Figure 1(a) and 1(c)), the number of stars
seems to have no effect between 1 and 3, but the function increases after 3 stars, indi-
cating that, for example, products with 4-star reviews are more likely to be purchased
than those with an average of 3 stars. The function achieves a maximum around 4.2,
and decreases somewhat thereafter. For athletic shoes and herbal vitamins (Figure 1
(b) and 1(d)), the number of stars has an increasingly positive effect up to about 3.8
and 4 stars, respectively. Then the effect flattens and later decreases. This suggests
that a SKU with an average of 3.8–4.2 stars is more likely to be purchased than one
with 5 stars, confirming H1.
With regard to the second hypothesis, Table 3 (Spline (logreview)) shows that the
nonlinear effect of volume (i.e., number of reviews) is significant (p<.05) for all product
categories except light bulbs (pD.14). This means that adding nonlinear term to our
models significantly reduces deviance for the three product categories, but not for light
bulbs. Also, the linear effect of volume on purchase probability of bulbs (Table 4, linear
logreview) is not significant (pD.45), which means that volume does not matter for
bulbs. Figure 1 demonstrates that the function of volume increases for light bulbs, but it
is quite flat and slightly decreasing for natural hair care products and herbal vitamins, sug-
gesting that a larger number of reviews are associated with a lower probability to buy
these products. For women’s athletic shoes, the function of volume increases slightly and
then flattens as well. Such results are in line with H2 for women’s athletic shoes, but not
for other categories. Hence, we cannot support H2.
Price and brand also play roles in predicting consumers’ probability to buy products.
Not surprisingly, price has a negative effect on purchase probability of bulbs. However,
its effect on purchase probability of athletic shoes and vitamins is nonsignificant (pD
.92, pD.19, respectively). Interestingly, price has a positive effect on purchase probabil-
ity of hair care products. We will discuss the reasons for and implications of our findings
in the next section.
Table 3. Analysis of deviance.
Source df Sum of squares Chi-square p
Light bulbs
Spline (avgstar) 2 28.058908 28.0589 <.0001
Spline (logreview) 2 3.904887 3.9049 0.1419
Women’s athletic shoes
Spline (avgstar) 2 7.952682 7.9527 0.0188
Spline (logreview) 2 92.309074 92.3091 <.0001
Natural hair care
Spline (avgstar) 2 14.322789 14.3228 0.0008
Spline (logreview) 2 6.055067 6.0551 0.0484
Herbal vitamins
Spline (avgstar) 2 6.390388 6.3904 0.0410
Spline (logreview) 2 25.969568 25.9696 <.0001
International Journal of Advertising 11
Downloaded by [Northwestern University], [Ewa Maslowska] at 08:34 24 June 2016
Table 4. Parameter estimates.
Light bulbs Natural hair care
Parameter Estimate SE tp Parameter Estimate SE t p
Intercept ¡3.59619 0.15914 ¡22.60 <.0001 Intercept ¡4.22661 0.50173 ¡8.42 <.0001
Brand A ¡0.24638 0.10744 ¡2.29 0.0218 Brand A 0.38659 0.22914 1.69 0.0916
Brand B ¡0.27761 0.12946 ¡2.14 0.0320 Brand B 0.53540 0.23519 2.28 0.0228
Brand C ¡0.15320 0.09970 ¡1.54 0.1244 Brand C 0.23020 0.19495 1.18 0.2377
Brand D 0.56069 0.15348 3.65 0.0003 Brand D 0.50228 0.19139 2.62 0.0087
Brand E ¡0.3736 0.13838 ¡2.72 0.0065 Brand E 0.24193 0.13334 1.81 0.0696
Linear log price ¡0.14578 0.03237 ¡4.50 <.0001 Linear log price 0.49267 0.17257 2.85 0.0043
Linear avg. stars 0.08213 0.02664 3.08 0.0020 Linear avg. stars 0.09737 0.09764 1.00 0.3187
Linear log review 0.01991 0.02619 0.76 0.4471 Linear log review ¡0.11643 0.03501 ¡3.33 0.0009
Women’s athletic shoes Herbal vitamins
Intercept ¡5.15012 0.83974 ¡6.13 <.0001 Intercept ¡2.41724 0.31787 ¡7.60 <.0001
Brand A ¡0.61653 0.10067 ¡6.12 <.0001 Brand A ¡0.13324 0.11701 ¡1.14 0.2548
Linear log price 0.01553 0.15626 0.10 0.9208 Brand B ¡0.51356 0.10543 ¡4.87 <.0001
Linear avg. stars 0.11419 0.11744 0.97 0.3309 Brand C ¡0.33159 0.14221 ¡2.33 0.0197
Linear log review 0.31602 0.02659 11.88 <.0001 Brand D ¡0.64769 0.19225 ¡3.37 0.0008
Linear log price ¡0.08559 0.06545 ¡1.31 0.1910
Linear avg. stars 0.15663 0.05730 2.73 0.0063
Linear log review ¡0.08713 0.02303 ¡3.78 0.0002
12 E. Maslowska et al.
Downloaded by [Northwestern University], [Ewa Maslowska] at 08:34 24 June 2016
Discussion
Despite the substantial interest in the workings and limits of online reviews, there are still
gaps in the literature. The aim of this study was to address some of them by investigating
the effect of review valence and volume on actual sales across different product catego-
ries. For the categories studied, we found that products with the average star rating of 4.5
through 5 are less likely to be purchased than those between 4 and 4.5 stars. The reason
for that may be that consumers perceive such reviews as too good to be true. They may
also think that such reviews were possibly generated by representatives from the company
or their public relations agency. This suggests that it is important to have some fraction of
nonperfect reviews, which supports the findings of Doh and Hwang (2009), who suggest
that a few negative messages can increase attitudes and perceived credibility. Hence, we
agree with Zhang, Craciun, and Shin (2010) and Zhang, Li, and Chen (2012) that compa-
nies should not censor negative reviews. Moreover, if consumers learned that negative
reviews were being censored, then reviews would lose credibility. It is not in the interest
of the retailer to censor negative reviews.
With regard to the volume of reviews, the results are more complicated and show that
a higher volume is not always good. Volume seems to have a small positive effect for ath-
letic shoes, but the effect is quite flat. We also find that for some product categories vol-
ume does not matter (i.e., light bulbs), while for others (i.e., natural hair care products
and herbal vitamins) an increase in volume may lead to slightly lower probability of pur-
chase. Such results are in line with some of the extant research finding no effects of vol-
ume, but in contrast to previous research showing that volume has a positive effect. An
explanation for our findings may be found in the work of Park and Lee (2008) who argue
that being exposed to too many reviews may make consumers experience cognitive over-
load, ‘[…] the phenomenon of too much information overwhelming a consumer, causing
adverse judgmental decision making’ (Park and Lee 2008, p. 388).
Figure 1. The effect of valence and volume on purchase probability.
Note: Clockwise from top left: (a) light bulbs, (b) women’s athletic shoes, (c) natural hair care prod-
ucts, (d) herbal vitamins.
Source: Author
International Journal of Advertising 13
Downloaded by [Northwestern University], [Ewa Maslowska] at 08:34 24 June 2016
The negative effect of volume could also be due to an omitted variable bias !whether
or not the consumer reads the reviews. Perhaps in some categories consumers are more
likely to read reviews, and so the effect is negative. In categories where consumers rely
on the heuristic (volume), the effect should be positive. Such reasoning would be in
accordance with the cognitive overload theory. It is important to notice that even though
we find an effect for reviews, the detected effect is almost flat. That would suggest that
the number of reviews does not have such a strong effect, which would be more in line
with previous studies finding no effects of volume (e.g., Chintagunta, Gopinath, and
Venkataraman 2010).
In addition, the positive effects of volume demonstrated by some previous studies are
mainly grounded in the view that volume increases product awareness (cf., Dellarocas,
Awad, and Zhang 2004). However, this idea may only hold true for reviews posted on
external review websites or blogs, but not on retailer websites (Duan, Gu, and Whinston
2008a; Kostyra et al. 2016). In line with such a theory, online reviews posted on a
retailer’s internal website have been shown to have a limited influence on sales rank of
high-involvement goods (Gu, Park, and Konana 2012). In our study, we are dealing with
retailer websites, which suggests that consumers are aware of products offered there and
thus volume does not play a big role in their decision process.
Due to the complicated nature of purchase decisions, we included in our analysis price
and brand. These two product characteristics can serve as cues signifying product quality
and hence they can decrease uncertainty and simplify decision process (Kostyra et al.
2016). The analysis shows that the effect of price differs per product category. Price has a
negative effect for light bulbs, which means that more expensive products have a lower
probability of purchase. Such a result is in line with some previous studies, which have
also demonstrated a negative effect of price in the context of online reviews (e.g., Gu,
Park, and Konana 2012; Ye, Law, and Gu 2009). Price also has a negative but nonsignifi-
cant effect on the purchase of vitamins. On the other hand, price has a positive effect on
purchase of natural hair care products and athletic shoes (although the effect is nonsignifi-
cant). Hence, customers choose more expensive products. It is plausible that customers
buying natural products are looking for quality and treat price as an indicator of it. These
results suggest that price !as a heuristic cue !has a different meaning for high- versus
low-involvement products.
Limitations, alternative explanations, future research, and implications
Our findings bring significant implications, but it is important to notice that we used real-
life data, which have limitations in terms of internal and external validities. Our sample
comes from three different retailers in different categories, but we could not include all
the possible categories. In addition, our data may not represent other populations or situa-
tions, such as with higher priced products. Consumers differ in their attention to, and reli-
ance on, reviews and their previous experience. Also, some consumers are more likely to
write online reviews (Zhu and Zhang 2010). Consumers also have pre-existing product
knowledge, attitudes toward brands and stores. However, we did not have a chance to
control for such variables. As Chiou and Cheng (2003) argue, the effect of valence may
depend on pre-existing brand image. The effect of reviews is also contingent on custom-
ers’ previous loyalty and intensity of information search (V!
azquez-Casielles, Su!
arez-
!
Alvarez, and del R!
ıo-Lanza 2013). Although we did control for the effect of brand, we
did not know consumers’ individual characteristics, such as brand preferences. Therefore,
these personal factors could be included in the future studies.
14 E. Maslowska et al.
Downloaded by [Northwestern University], [Ewa Maslowska] at 08:34 24 June 2016
Furthermore, consumers are aware that companies can moderate or censor reviews, as
well as motivate positive reviews with rewards (Li and Hitt 2008). This may diminish the
trustworthiness of reviews. Although we did not observe such patterns in our data (e.g.,
only 0.10% of reviews of light bulbs and cosmetics were rejected), such an assumption
may motivate customers to also look for external information about the product. In order
to address this problem, some retailers syndicate reviews, believing it can make them
more credible. Although we did not examine such practices in our data, future studies
could include customers’ knowledge about such practices, perceived review trustworthi-
ness, and the effect of external reviews.
Correspondingly, based on research into advertising message processing, we
assumed that customers may perceive negative reviews as more diagnostic and infor-
mative and hence useful. However, customers may also consider positive product
reviews to be ambiguous, which may lead them to look for information elsewhere
(Lee and Youn, 2009). Research is needed to gain a more qualitative insight into
how customers perceive reviews and how they combine them with other sources of
information.
In our study we looked at four product categories. These products were all used
offline and were rather commonly purchased goods, yet we found some mixed results
with regard to volume and price. That may be because even though similar, the cate-
gories may have differed with respect to involvement and motivations driving their
purchase. Light bulbs are a highly utilitarian product category, while natural hair care
may be perceived as both utilitarian and emotional choice. We can expect that the
results may differ also for other categories, for example, for goods that are used online
(Zhu and Zhang 2010)ordependonproductlifetime(Chen,Wang,andXie2011). As
discussed by Allsop, Bassett, and Hoskins (2007), customers turn to different sources
for different information, and the effect of WOM depends on the specific situation.
For some product categories, reviews are more prevalent and useful. For example,
reviews may be more likely to affect customers interested in higher price and highly
coveted products (Riegner, 2007). Also, the effect of valence seems to be conditional
upon the type of product: search versus experience (Willemsen et al. 2011). Future
studies should look at the effects of reviews across different categories of products, as
also suggested by Purnawirawan et al. (2015). The role of such product characteristics
as utilitarian versus hedonic nature of the purchase or product involvement should be
included.
Our results concerning valence also bring important implications for future research.
For example, as Bazaar’s report (2015)shows,reviewswithdifferentstarratingscon-
tain different features, with 1–3 star reviews containing product shortcomings and 3–4
star reviews suggesting product improvements. Investigating the effect of these differ-
ences in content might be worthwhile for futureresearch.Correspondingly, content of
the reviews can influence customers and the way they process heuristic cues. As men-
tioned above, whether customers read or not reviews can explain the negative effect of
volume. It can also moderate the effect of valence. Nevertheless, it was not the focus of
our study for several reasons. We saw in our data that the overwhelming majority of cus-
tomers did not consider the content (e.g., for natural hair products only between 0.28%
and 16% of customers looked at the content of reviews). This may be explained by the
low-involvement character of the product category studied. In addition, when customers
did consult the reviews, we do not know whichspecificreviewsandtowhatdegreethey
read them. Therefore, future studies should investigate the role of review content in rela-
tion heuristic cues.
International Journal of Advertising 15
Downloaded by [Northwestern University], [Ewa Maslowska] at 08:34 24 June 2016
Moreover, different characteristics of reviews, such as review length, valence, and
volume, can affect each other. For example, volume may be a moderator of valence
(cf., Kostyra et al. 2016). According to the selective attention theory (Treisman
1969), people are not able to process and respond to all the stimuli, because their
processing capacity is limited. In order to simplify the processing, individuals rely on
heuristic cues and filter out information thatislessrelevant.Wecanexpectthispro-
cess to be especially at play when involvement with the decision process is low.
Hence, in our case, we expected customers to rely on heuristic cues (i.e., volume and
valence), which was supported by our data showing that the majority of customers did
not read the reviews. These cues may interact with each other, especially when pro-
viding ambiguous information. Only a few studies have looked at the interaction of
different review characteristics, and we do not know whether it is better to have, for
example, 10 moderate reviews (i.e., with an average of 3) or 2 negative ones. Because
we were interested in nonlinear effects, we applied GAMs, which are not suitable for
testing interaction effects. Including interactions would require splitting the data and
running several additional models for different values of volume, which would make
the results complicated. In addition, some previous studies have shown that volume
has an effect regardless of valence (e.g., Bazaar 2015), which supported our approach.
However, we do acknowledge that various interaction effects between review charac-
teristics require further investigation.
The effectiveness of influence efforts depends on environmental and contextual fac-
tors and interactions between them (Hansen 1976). Hence, other information sources may
affect consumers’ decisions. Purchase decisions may be influenced by person-related fac-
tors, but also factors such as price, brand, product information, product quality and value,
advertising, promotion, or the reference price (e.g., Chang and Wildt 1994). For example,
increased marketing efforts can diminish the effect of valence (Yang et al. 2012). To con-
trol for some of those variables we included price in our analysis, but future studies
should investigate how price influences the effects of online reviews. For example, the
role of price has been shown to decrease when online reviews are present (Kostyra et al.
2016), but it is also plausible that the importance of online reviews differs depending on
the product price.
We also controlled for the effect of brand. Although brands are extremely important in
marketing communication, the relationship between brands and online reviews has
received little attention (Lovett, Peres, and Shachar 2013). Past research has shown that
the effect of brand diminishes when reviews are included (Kostyra et al. 2016). In addi-
tion, the effect of online reviews has been shown to differ for different brands. For exam-
ple, Berger, Sorensen, and Rasmussen (2010) suggest that negative reviews can increase
sales of unestablished brands by increasing their awareness, Chiou and Cheng (2003)
show that the effect of reviews differs for low-image vs. high-image brands, and Ho-Dac,
Carson, and Moore (2013) show that online reviews matter less for strong brands. There-
fore, future research should not only control for the effect of brand, but should examine
how different brands are affected by online reviews, and marketers should measure what
the effects of reviews for their brand are.
To conclude, this study demonstrates how the valence and volume of online consumer
reviews affect purchases. It is important to stress that !contrary to popular belief !bet-
ter reviews do not always have a more positive effect. Since products that are rated on
average higher than 4.5 out of 5 stars can be less likely to be purchased than those with an
average around 4!4.5 stars, retailers should not delete too negative reviews in order to
only display perfect reviews of their products.
16 E. Maslowska et al.
Downloaded by [Northwestern University], [Ewa Maslowska] at 08:34 24 June 2016
Acknowledgements
This research was supported by the IMC Medill Spiegel Digital & Database Research Center,
Northwestern University and in part through the computational resources and staff contributions
provided for the Social Sciences Computing cluster (SSCC) at Northwestern University. Recurring
funding for the SSCC is provided by Office of the President, Weinberg College of Arts and Scien-
ces, Kellogg School of Management, the School of Professional Studies, and Northwestern Univer-
sity Information Technology.
Disclosure statement
No potential conflict of interest was reported by the authors.
Notes on contributors
Ewa Maslowska is a postdoctoral research associate at the Medill IMC Spiegel Research Center on
Digital and Database Marketing (SRC). Her research is grounded in consumer behavior and focuses
on data-driven analyses investigating the relationship between various forms of consumer engage-
ment and purchase behaviors. She earned her PhD degree in 2013 in persuasive communication
from the Amsterdam School of Communication Research (ASCoR), University of Amsterdam. Her
dissertation explores how personalized marketing communication influences consumers’ cognitive
processes, attitudes, and behaviors. Prior to her studies at UvA, Ewa completed her MA degree in
psychology at Jagiellonian University.
Edward C. Malthouse is the Theodore R and Annie Laurie Sills Professor of Integrated Marketing
Communications and Industrial Engineering at Northwestern University and the research director
for the Spiegel Institute on digital and database marketing. He was the co-editor of the Journal of
Interactive Marketing between 2005 and 2011. He earned his PhD degree in 1995 in computational
statistics from Northwestern University and completed a postdoc at the Kellogg marketing depart-
ment. His research interests center on engagement, media marketing, new media, integrated market-
ing communications, customer lifetime value models, predictive analytics, and unsupervised
learning. He is the author of Segmentation and Lifetime Value Models Using SAS and the co-editor
of Medill on Media Engagement.
Stefan F. Bernritter is an assistant professor in Marketing Communication at the Amsterdam School
of Communication Research (ASCoR), University of Amsterdam. He wrote his dissertation about
the antecedents and consequences of consumers’ intentional public online affiliations with brands
(i.e., consumer endorsements) at ASCoR. Before his PhD degree at ASCoR, he completed an MSc
(research) degree in behavioral science and an MSc degree in communication science at Radboud
University Nijmegen. His research interests lie in consumers’ brand-related social media use, con-
sumer identity, and mobile marketing
ORCID
Stefan F. Bernritter http://orcid.org/0000-0002-4291-7824
References
Allsop, D.T., B.R. Bassett, and J.A. Hoskins. 2007. Word-of-mouth research: principles and appli-
cations. Journal of Advertising Research 47, no. 4: 398!411.
Amblee, N., and T. Bui. 2011. Harnessing the influence of social proof in online shopping: the
effect of electronic word of mouth on sales of digital microproducts. International Journal of
Electronic Commerce 16, no. 2: 91!114. http://doi.org/10.2753/JEC1086-4415160205
Ba, S., and P.A. Pavlou. 2002. Evidence of the effect of trust building technology in electronic mar-
kets: price premiums and buyer behaviour. MIS Quarterly 26, no. 3: 243!268.
Basuroy, S., S. Chatterjee, and S.A. Ravid. 2003. How critical are critical reviews? The box office
effects of film critics, star power, and budgets. Journal of Marketing 67, no. 4: 103!117.
International Journal of Advertising 17
Downloaded by [Northwestern University], [Ewa Maslowska] at 08:34 24 June 2016
Bazaar. 2015. The conversation index.Vol.8[Online].http://www.bazaarvoice.com/au/conversation-
index/
Berger, J., A.T. Sorensen, and S.J. Rasmussen. 2010 March. Positive effects of negative publicity:
when negative reviews increase sales. Marketing Science 29: 815!827. http://doi.org/10.1287/
mksc.1090.0557
Bernritter, S.F., P.W.J. Verlegh, and E.G. Smit. 2016. Why nonprofits are easier to endorse on social
media: the roles of warmth and brand symbolism. Journal of Interactive Marketing 33: 27!42.
Bosman, D.J., C. Boshoff, and G.J. van Rooyen. 2013. The review credibility of electronic word-of-
mouth communication on e-commerce platforms. Management Dynamics 22, no. 3: 29!44.
Brynjolfsson, E., and M.D. Smith. 2000. Frictionless commerce? A comparison of internet and con-
ventional retailers. Management Science 46, no. 4: 563!585. http://doi.org/10.1287/
mnsc.46.4.563.12061
Cacioppo, J.T., R.E. Petty, C.F. Kao, and R. Rodriguez. 1986. Central and peripheral routes to per-
suasion: an individual difference perspective. Journal of Personality and Social Psychology 51,
no. 5: 1032!1043. http://doi.org/10.1037/0022-3514.51.5.1032
Campbell, C., L.F. Pitt, M. Parent, and P.R. Berthon. 2011. Understanding consumer conversations
around ads in a Web 2.0 World. Journal of Advertising 40, no. 1: 87!102.
Chang, T.-Z., and A.R. Wildt. 1994. Price, product information, and purchase intention: an empiri-
cal study. Journal of the Academy of Marketing Science 22, no. 1: 16!27. http://doi.org/
10.1177/0092070394221002
Chen, P.-Y., S.-Y. Wu, and J. Yoon. 2004. The impact of online recommendations and consumer
feedback on sales. The International Conference on Information Systems,ICIS 2004, December
12!15, 2004, Washington, DC.
Chen, Y., Q. Wang, and J. Xie. 2011. Online social interactions: A natural experiment on word of
mouth versus observational learning. Journal of Marketing Research 48, no. 2: 238!254.
Chen, Z. 2014. Social considerations in online word of mouth. A Dissertation Presented to the Aca-
demic Faculty. In Partial Fulfillment of the Requirements for the Degree Doctor of Philosophy
in the Scheller College of Business, Georgia Institute of Technology.
Chen, Z., and N.H. Lurie. 2013. Temporal contiguity and negativity bias in the impact of online word
of mouth. Journal of Marketing Research 50, no. 4: 463!476. http://doi.org/10.1509/jmr.12.0063
Chevalier, J.A., and D. Mayzlin. 2006 August. The effect of word of mouth on sales: online book
reviews. Journal of Marketing Research 43: 345!354. http://doi.org/10.1509/jmkr.43.3.345
Chevalier, J.A. and D. Mayzlin, 2003. The effect of word of mouth on sales: Online book reviews.
Yale SOM Working Paper No’s. ES-28 & MK-15. Available at SSRN: http://ssrn.com/
abstractD432481
Chintagunta, P.K., S. Gopinath, and S. Venkataraman. 2010 March. The effects of online user
reviews on movie box office performance: accounting for sequential rollout and aggregation
across local markets. Marketing Science 29: 944!957. http://doi.org/10.1287/mksc.1100.0572
Chiou, J.S., and C. Cheng. 2003. Should a company have message boards on its web sites? Journal
of Interactive Marketing 17, no. 3: 50!61. http://doi.org/10.1002/dir.10059
Clemons, E.K., G.G. Gao, and L.M. Hitt. 2006. When online reviews meet a study of the hyperdif-
ferentiation: craft beer industry. Journal of Management Information Systems 23, no. 2:
149!171. http://doi.org/10.2753/MIS0742-1222230207
Competition & Markets Authority. 2015. Online reviews and endorsements. https://www.gov.uk/
government/uploads/system/uploads/attachment_data/file/436238/Online_reviews_and_en
dorsements.pdf
Cui, G., H.-K. Lui, and X. Guo. 2012. The effect of online consumer reviews on new product sales.
International Journal of Electronic Commerce 17, no. 1: 39!58. http://doi.org/10.2753/
JEC1086-4415170102
de Langhe, B., P.M. Fernbach, and D.R. Lichtenstein. 2016. Navigating by the stars: investigating
the actual and perceived validity of online user ratings. Journal of Consumer Research 42:
817!833.
Dellarocas, C., N. Awad, and X. Zhang. 2004. Exploring the value of online reviews to organiza-
tions: implications for revenue forecasting and planning. Proceedings of International Confer-
ence on Information Systems 21: 1!34. http://doi.org/10.1002/dir.20087
Dellarocas, C., X. M. Zhang, and N. F. Awad. 2007. Exploring the value of online product reviews
in forecasting sales: The case of motion pictures. Journal of Interactive Marketing, 21, no. 4:
pp. 23!45.
18 E. Maslowska et al.
Downloaded by [Northwestern University], [Ewa Maslowska] at 08:34 24 June 2016
Dhar, V., and E.A. Chang. 2009. Does chatter matter? The impact of user-generated content on
music sales. Journal of Interactive Marketing 23, no. 4: 300!307. http://doi.org/10.1016/j.
intmar.2009.07.004
Dholakiya, P. 2014. Are excessively positive customer reviews hurting your conversion rates? [Study].
http://unbounce.com/conversion-rate-optimization/customer-reviews-conversion-rates/.
Doh, S.-J., and J.-S. Hwang. 2009. How consumers evaluate eWOM (electronic word-of-mouth)
messages. Cyberpsychology & Behavior : The Impact of the Internet, Multimedia and Virtual
Reality on Behavior and Society 12, no. 2: 193!197. doi: 10.1089/cpb.2008.0109
Duan, W., B. Gu, and A.B. Whinston. 2008a. Do online reviews matter? An empirical investigation
of panel data. Decision Support Systems 45, no. 4: 1007!1016. http://doi.org/10.1016/j.
dss.2008.04.001
Duan, W., B. Gu, and A.B. Whinston. 2008b. The dynamics of online word-of-mouth and product
sales !an empirical investigation of the movie industry. Journal of Retailing 84: 233!242.
http://doi.org/10.1016/j.jretai.2008.04.005
Edelman, D., and B. Salsberg. 2011. Beyond paid media: marketing’s new vocabulary. http://www.
mckinsey.com/business-functions/marketing-and-sales/our-insights/beyond-paid-media-market
ings-new-vocabulary
Eisend, M. 2006. Two-sided advertising: a meta-analysis. International Journal of Research in
Marketing 23, no. 2: 187!198.
Floyd, K., R. Freling, S. Alhoqail, H.Y. Cho, and T. Freling. 2014. How online product reviews
affect retail sales: a meta-analysis. Journal of Retailing 90, no. 2: 217!232. http://doi.org/
10.1016/j.jretai.2014.04.004
Forman, C., A. Ghose, and B. Wiesenfeld. 2008. Examining the relationship between reviews and
sales: the role of reviewer identity disclosure in electronic markets. Information Systems
Research 19, no. 3: 291!313. http://doi.org/10.1287/isre.1080.0193
Ghose, A., and P.G. Ipeirotis. 2011. Estimating the helpfulness and economic impact of product
reviews: Mining text and reviewer characteristics. IEEE Transactions on Knowledge and Data
Engineering 23, no. 10: 1498!1512.
Godes, D., and D. Mayzlin. 2004. Using online conversations to study word-of-mouth communica-
tion. Marketing Science 23, no. 4: 545!560. http://doi.org/10.1287/mksc.1040.0071
Gopinath, S., J.S. Thomas, and L. Krishnamurthi. 2014. Investigating the relationship between the
content of online word of mouth, advertising, and brand performance. Marketing Science 33,
no. 2: 241!258. http://doi.org/10.1287/mksc.2013.0820
Gu, B., J. Park, and P. Konana. 2012. The impact of external word-of-mouth sources on retailer
sales of high-involvement products. Information Systems Research 23, no. 1: 182!196. http://
doi.org/10.1287/isre.1100.0343
Hansen, F. 1976. Psychological theories of consumer choice. Journal of Consumer Research 3,
no. 3: 117!142.
Hastie, T.J., and R.J. Tibshirani. 1990. Generalized additive models. Vol. 43. Boca Raton, FL:
Chapman & Hall/CRC.
Hennig-Thurau, T., K.P. Gwinner, G. Walsh, and D.D. Gremler. 2004. Electronic word-of-mouth
via consumer-opinion platforms: what motivates consumers to articulate themselves on the
internet? Journal of Interactive Marketing 18, no. 1: 38!52.
Herr, P.M., F.R Kardes, and J. Kim. 1991. Effects of word-of-mouth and product-attribute informa-
tion on persuasion: an accessibility-diagnosticity perspective. Journal of Consumer Research
17, no. 4: 454!462.
Ho-Dac, N.N., S.J., Carson, and W.L. Moore. 2013 November. The effects of positive and negative
online customer reviews : do brand strength and category maturity matter? Journal of Market-
ing 77: 37!53. http://doi.org/10.1509/jm.11.0011
James, G., D. Witten, T. Hastie, and R. Tibshirani. 2013. An Introduction to Statistical Learning
(Vol. 112). New York: Springer.
Jang, S., A. Prasad, and B.T. Ratchford. 2012. How consumers use product reviews in the purchase
decision process. Marketing Letters 23, no. 3: 825!838. http://doi.org/10.1007/s11002-012-
9191-4
Jim!
enez, F.R., and N. Mendoza. 2013. Too popular to ignore: the influence of online reviews on
purchase intentions of search and experience products. Journal of Interactive Marketing 27, no.
3: 226!235. http://doi.org/10.1016/j.intmar.2013.04.004
International Journal of Advertising 19
Downloaded by [Northwestern University], [Ewa Maslowska] at 08:34 24 June 2016
Kamins, M.A., and H. Assael. 1987 February. Two-sided versus one-sided appeals: a cognitive per-
spective on argumentation, source derogation, and the effect of disconfirming trial on belief
change. Journal of Marketing Research 24: 29!39.
Kim, S.J., R.J. Wang, E. Maslowska, and E.C. Malthouse. 2015. “Understanding a fury in your
words”: the effects of posting and viewing electronic negative word-of-mouth on purchase
behaviors. Computers in Human Behavior 54: 511!521. http://doi.org/10.1016/j.
chb.2015.08.015
Kimmel, A.J., and P.J. Kitchen. 2014. WOM and social media: presaging future directions for
research and practice. Journal of Marketing Communications 20, no. 1!2: 5!20. http://doi.
org/10.1080/13527266.2013.797730
King, R.A., P. Racherla, and V.D. Bush. 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, no. 3:
167!183. http://doi.org/10.1016/j.intmar.2014.02.001
Kostyra, D.S., J. Reiner, M. Natter, and D. Klapper. 2016. Decomposing the effects of online cus-
tomer reviews on brand, price, and product attributes. International Journal of Research in
Marketing 33, no. 1: 11!26. http://doi.org/10.1016/j.ijresmar.2014.12.004
Kusumasondjaja, S., T. Shanka, and C. Marchegiani. 2012. Credibility of online reviews and initial
trust: the roles of reviewer’s identity and review valence. Journal of Vacation Marketing 18,
no. 3: 185!195. http://doi.org/10.1177/1356766712449365
Lam, C.F., G. Spreitzer, and C. Fritz. 2014. Too much of a good thing: curvilinear effect of positive
affect on proactive behaviors. Journal of Organizational Behavior 35, no. 4: 530!546.
Lee, H.G. 1998. Do electronic marketplaces lower the price of goods? Communications of the ACM
41, no. 1: 73!80.
Lee, J., D.H. Park, and I. Han. 2011. The different effects of online consumer reviews on consum-
ers’ purchase intentions depending on trust in online shopping malls: an advertising perspective.
Internet Research 21, no. 2: 187!206.
Lee, M., and S. Youn. 2009. Electronic word of mouth (eWOM). How eWOM platforms influence
consumer product judgement? International Journal of Advertising 28, no. 3: 473!499. http://
doi.org/10.2501/S0265048709200709
Li, X., and L.M. Hitt. 2008 March. Self-selection and information role of online product reviews.
Information Systems Research 19: 456!474. doi:10.1287/isre.1070.0154
Liu, Y. 2006. Word of mouth for movies: its dynamics and impact on box office revenue. Journal of
Marketing 70, no. 3: 74!89. http://doi.org/10.1509/jmkg.70.3.74
Lovett, M.J., R. Peres, and R. Shachar. 2013 August. On brands and word of mouth. Journal of
Marketing Research 50: 427!444. http://doi.org/10.1509/jmr.11.0458
Mahajan, V., E. Muller, and R.A. Kerin. 1984. Introduction strategy for new products with positive
and negative word-of-mouth. Management Science 30, no. 12: 1389!1404. http://doi.org/
10.1287/mnsc.30.12.1389
Matfield. 2011. The impact of customer reviews and ratings on conversion rates. http://www.smartin
sights.com/conversion-optimisation/product-page-optimisation/reviews-conversion-rate-impact/
Meli!
an-Gonz!
alez, S., J. Bulchand-Gidumal, and B.G. L!
opez-Valc!
arcel. 2013. Online customer
reviews of hotels as participation increases, better evaluation is obtained. Cornell Hospitality
Quarterly 54, no. 3: 274!283.
Mintel. 2015. Seven in 10 Americans seek out opinions before making purchases. http://www.min
tel.com/press-centre/social-and-lifestyle/seven-in-10-americans-seek-out-opinions-before-mak
ing-purchases
Mizerski, R.W. 1982. An attribution explanation of the disproportionate influence of unfavorable
information. Journal of Consumer Research 9, no. 3: 301!310. http://doi.org/10.1086/208925
Mudambi, S.M., and D. Schuff. 2010. What makes a helpful online review? A study of customer
reviews on Amazon.com. MIS Quarterly 34, no. 1: 185!200.
Nielsen. 2012. Consumer trust in online, social, and mobile advertising grows. http://www.nielsen.
com/us/en/insights/news/2012/consumer-trust-in-online-social-and-mobile-advertising-grows.html
€
O#
g€
ut, H., and B.K. Onur Ta¸s. 2012. The influence of internet customer reviews on the online sales
and prices in hotel industry. The Service Industries Journal 32, no. 2: 197!214. http://doi.org/
10.1080/02642069.2010.529436
O’Neil, T. 2015. Survey confirms the value of reviews, provides new insights. http://www.powerre
views.com/blog/survey-confirms-the-value-of-reviews/
20 E. Maslowska et al.
Downloaded by [Northwestern University], [Ewa Maslowska] at 08:34 24 June 2016
O’Reilly, K., and S. Marx. 2011. How young, technical consumers assess online WOM credibility.
Qualitative Market Research: An International Journal 14: 330!359. http://doi.org/10.1108/
13522751111163191
Park, C. and T.M. Lee. 2009. Information direction, website reputation and eWOM effect: A moder-
ating role of product type. Journal of Business Research 62, no. 1: 61!67.
Park, D.H., and J. Lee. 2008. eWOM overload and its effect on consumer behavioural intention
depending on consumer involvement. Electronic Commerce Research and Applications 7, no.
4: 386!398. http://doi.org/10.1016/j.elerap.2007.11.004
Park, D.-H., J. Lee, and I. Han. 2007. The effect of on-line consumer reviews on consumer purchas-
ing intention: the moderating role of involvement. International Journal of Electronic Com-
merce 11, no. 4: 125!148. http://doi.org/10.2753/JEC1086-4415110405
Purnawirawan, N, P. De Pelsmacker, and N. Dens. 2012. Balance and sequence in online reviews :
how perceived usefulness affects attitudes and intentions. Journal of Interactive Marketing 26,
no. 4: 244!255. http://doi.org/10.1016/j.intmar.2012.04.002
Purnawirawan, N, M. Eisend, P. De Pelsmacker, and N. Dens. 2015. A meta-analytic investigation
of the role of valence in online reviews. Journal of Interactive Marketing 31: 17!27. http://doi.
org/10.1016/j.intmar.2015.05.001
Purnawirawan, N., Nathalie Dens, and P. De Pelsmacker. 2014. Expert reviewers beware! The
effects of review set balance, review source and review content on consumer responses to
online reviews. Journal of Electronic Commerce Research 15, no. 3: 162!177.
Riegner, C. 2007. Word of mouth on the web: the impact of Web 2.0 on consumer purchase deci-
sions. Journal of Advertising Research 47, no. 4: 436!447.
Schindler, R.M., and B. Bickart. 2012. Perceived helpfulness of online consumer reviews: the role
of message content and style. Journal of Consumer Behavior 11: 234!243. doi: 10.1002/cb
Simonson, I. 2015. Mission (largely) accomplished: what’s next for consumer BDT-JDM research-
ers. Journal of Marketing Behavior 1: 9!35.
Tong, S.T., B. van der Heide, L. Langwell, and J.B. Walther. 2008. Too much of a good thing? The
relationship between number of friends and interpersonal impressions on Facebook. Journal of
Computer!Mediated Communication 13, no. 3: 531!549.
Treisman, A.M. 1969. Strategies and models of selective attention. Psychological Review 76, no. 3:
282!299.
V!
azquez-Casielles, R., L. Su!
arez-
!
Alvarez, and A.-B. del R!
ıo-Lanza. 2013. The word of mouth
dynamic: how positive (and negative) WOM drives purchase probability an analysis of interper-
sonal and non-interpersonal factors. Journal of Advertising Research 53, no. 1: 43!60. http://
dx.doi.org/10.2501/JAR-53-1-043-060
Vlachos, P.A., A.P. Vrechopoulos, and K. Pramatari. 2011. Too much of a good thing: curvilinear
effects in the evaluation of services and the mediating role of trust. Journal of Services Market-
ing 25, no. 6: 440!450.
Wang, S., N.R. Cunningham, and M.S. Eastin. 2015. The impact of eWOM message characteristics
on the perceived effectiveness of online consumer reviews. Journal of Interactive Advertising
15, no. 2: 151!159. doi: 10.1080/15252019.2015.1091755
Wang, M., Q. Lu, R.T. Chi, and W. Shi. 2015. How word-of-mouth moderates room price and hotel
stars for online hotel booking. Journal of Electronic Commerce Research 16, no. 1: 72!80.
Willemsen, L.M., P.C. Neijens, F. Bronner, and J.A. de Ridder. 2011. “Highly recommended!” The
content characteristics and perceived usefulness of online consumer reviews. Journal of Com-
puter-Mediated Communication 17, no. 1: 19!38. http://doi.org/10.1111/j.1083-
6101.2011.01551.x
Yang, J., W. Kim, N. Amblee, and J. Jeong. 2012. The heterogeneous effect of WOM on product
sales: why the effect of WOM valence is mixed? European Journal of Marketing 46, no. 11:
1523!1538. http://doi.org/10.1108/03090561211259961
Ye, Q., R. Law, and B. Gu. 2009. The impact of online user reviews on hotel room sales. Interna-
tional Journal of Hospitality Management 28, no. 1: 180!182. http://doi.org/10.1016/j.
ijhm.2008.06.011
Zhang, J.Q., G. Craciun, and D. Shin. 2010. When does electronic word-of-mouth matter? A study
of consumer product reviews. Journal of Business Research 63, no. 12: 1336!1341. http://doi.
org/10.1016/j.jbusres.2009.12.011
Zhang, X., and C. Dellarocas. 2006. The lord of the ratings: is a movie’s fate is influenced by
reviews ? International Conference on Information Systems 117: 1959!1978.
International Journal of Advertising 21
Downloaded by [Northwestern University], [Ewa Maslowska] at 08:34 24 June 2016
Zhang, Z., X. Li, and Y. Chen. 2012. Deciphering word-of-mouth in social media. ACM Transac-
tions on Management Information Systems 3, no. 1: 1!23. http://doi.org/10.1145/
2151163.2151168
Zhang, Z., Q. Ye, R. Law, and Y. Li. 2010. The impact of e-word-of-mouth on the online popularity
of restaurants: a comparison of consumer reviews and editor reviews. International Journal of
Hospitality Management 29, no. 4: 694!700. http://doi.org/10.1016/j.ijhm.2010.02.002
Zhu, F., and X. (Michael) Zhang. 2010 March. Impact of online consumer reviews on sales: the
moderating role of product and consumer characteristics. Journal of Marketing 74: 133!148.
http://doi.org/10.1509/jmkg.74.2.133
22 E. Maslowska et al.
Downloaded by [Northwestern University], [Ewa Maslowska] at 08:34 24 June 2016
