Professional Documents
Culture Documents
Social media may encourage novel ways of signaling that involve different pur-
chase types (experiential vs. material), signaling frequencies (multiple vs. single
signals), and other features unique to social media (e.g., hashtags). This work
examines how purchase signals are received on social media and how these sig-
naling variations affect signal receivers’ perceptions of the authenticity of social
media posts as well as the overall impressions receivers form of the signal sender.
Data collected across six experiments show multiple material purchase signals
lead to more negative impressions compared to multiple experiential purchase sig-
nals. Signal receivers perceive multiple material purchase posts as less authentic,
which dampens their impressions of the signal sender. In line with this mecha-
nism, the impression premium of experiential purchase signals disappears when
receivers use other cues (monetary mentions, other users’ comments, and mar-
keter associations via hashtags) to infer a signal’s lack of authenticity. Additional
data also document downstream consequences on engagement. This work con-
tributes theoretically to research in both signaling and social media and improves
the understanding of substantive situations in which consumers’ objectives of cu-
rating a positive image and creating engagement with their posts, collide with mar-
keters’ objectives of encouraging user-generated content and word of mouth.
430
VALSESIA AND DIEHL 431
状态信号
and signaling has primarily focused on product signaling posters may choose to add certain social media-specific
(e.g., see JCR’s research curation “Products as signals”; elements (e.g., hashtags) that highlight connections with
Moreau 2014). Only recently have some researchers exam- other entities or marketing campaigns (e.g.,
ined different types of status signals (e.g., busyness: #ShotOnIphone). Or, other social media users might pro-
Bellezza, Paharia, and Keinan 2017), possibly, at least in vide feedback to a post, in the form of comments and likes,
part, because social media has broadened the scope of what which may also affect how the post is evaluated. In this
显眼的 can become conspicuous. Product signaling is still possible work, we explore how receivers use various social media-
on social media, as exemplified by the above post about specific cues to infer the authenticity of social media posts.
the iPhone. If anything, social media allows product signal-
the sender (Jung, Song, and Vorderer 2012). Hence, we ex- Moreover, to infer brand authenticity specifically, they rely
amine the effect of using different purchase signals and dif- on observable cues such as the communication style
ferent ways of sharing these signals on the overall (Morhart et al. 2015) or a brand’s origin story (Cinelli and
印象对 impressions signal receivers form of the signal sender. In LeBoeuf 2020). In our case, we expect signal receivers will
用户参与 addition to being highly meaningful to signal senders, 行为 use observable aspects of a signal (a social media post) to
喜欢、转发、 impressions can have important behavioral consequences 后果 infer its authenticity. Since people value authenticity in
评论和点击)
也有重要 for user engagement (likes, retweets, comments, and click- others’ actions (Grandey et al. 2005), we reason that judg-
行为后果 throughs). We further discuss evidence for such down- ing signals as being more authentic may lead to more posi-
FIGURE 1
CONCEPTUAL MODEL.
purchase signals decrease or improve impressions over non- We set our sample size to 150 respondents per cell, that is,
purchase signals. Importantly, because any single signal may 900 across the six-cell design. Of the 903 participants who
be hard to interpret and multiple signals may be seen as completed the survey, 890 unique responses were retained
more diagnostic, any benefit of experiential posts may be based on the criteria described above.
greater, or even only emerge, when sent repeatedly, which Across conditions, we used a total of 12 posts.
we test here. 第一项研究采用了3(体验性购买物质性、购买、非 Specifically, we created three different versions of four
购买)× 2(帖子数量:1vs.4)设计。被试被告知
他们将根据某人在Facebook上的帖子来评估他。他们 posts (see table 1). The three versions used the same word-
Method 看到了一个或四个Facebook帖子,这些帖子都被归属于 ing but mentioned an experiential purchase, a material pur-
他们所认识的人Alex Watson chase, or did not mention any purchase. Purchases were
Study 1 followed a 3 (post: experiential purchase, mate-
matched in terms of price and desirability based on a pre-
rial purchase, non-purchase) 2 (number of posts: 1 vs. 4)
test (see web appendix).
design. Respondents were told they would evaluate a per-
Depending on the condition (1 vs. 4 posts), respondents
son based on their Facebook posts. They saw either one or
either saw all posts of a given type in one of four predeter-
four Facebook posts attributed to their presumed acquain-
mined orders or only one randomly selected post of a given
tance Alex Watson (see web appendix).
type. Next, they evaluated the user who wrote these posts
Participants were recruited from MTurk. Respondents’
by answering: “What is your overall impression of the per-
eligibility was restricted to those residing in the USA with
son who wrote the post?” (1 ¼ very unfavorable, 9 ¼ very
a 95% or higher approval rate and who had completed at
favorable), and “To what extent do you like the person
least 50 hits (Goodman, Cryder, and Cheema 2013). Given
who wrote the post?” (1 ¼ not at all, 9 ¼ a great deal).
the research context, in this and all other studies, partici-
Responses were highly correlated (r ¼ .90) and were aver-
pants were screened to be Facebook users (see screening
aged to compute an impressions measure. All additional
questions in the web appendix). Respondents were also
measures collected in this and other studies (e.g., demo-
asked to complete a captcha verification question in order
graphics) are reported in the web appendix.
to complete the study. Participants on MTurk are known to
misrepresent their characteristics to meet study eligibility
(Chandler and Paolacci 2017). We thus allowed only one
Results
response per MTurk ID. If participants attempted the We estimated a 3 (post type) 2 (number of posts) anal-
captcha or the Facebook screener questions multiple times, ysis of variance (ANOVA) for impressions. This analysis
or entered the study twice for other reasons, only their first revealed a significant main effect of post type (F(2, 884) ¼
answer, even if incomplete, was considered valid. 41.38, p < .001, xp2 ¼ .083) and number of posts (MOne ¼
Subsequent responses were excluded from the analysis. 6.17, SD ¼ 1.60, MFour ¼ 5.74, SD ¼ 2.06; F(1, 884) ¼
VALSESIA AND DIEHL 435
TABLE 1
POSTS STUDY 1
Material posts
“Warm and sunny, the perfect day to spend time outdoors. . .if you have a new GoPro to try! #solucky”
“Today was a good day and is about to get even better! Guess who got a new awesome phone? #happyme”
“Sometimes you just gotta find a way to take care of yourself. . .Got the best cashmere pullover today! #selflove”
“Guys, I finally decided it’s time to be active again! Going to buy my first Fitbit today #excited”
Experiential posts
“Warm and sunny, the perfect day to spend time outdoors. . .if you have seats to watch your team live! #solucky”
13.63, p < .001, xp2 ¼ .014). Planned main effect contrasts In the web appendix, we report an additional study fo-
showed that both material (F(1, 884) ¼ 78.25, p < .001, cusing only on purchase signals. Following prior literature
xp2 ¼ .080) and experiential purchase signals (F(1, 884) ¼ (Bastos and Brucks 2017, Carter and Gilovich 2012,
7.20, p ¼ .007, xp2 ¼ .007) led to less positive impressions Rosenzweig and Gilovich 2012), in that study we test the
than non-purchase signals did (MNon ¼ 6.51, SD ¼ 1.65). robustness of our effect by framing the same purchases as
Further, experiential purchase signals led to more positive either material or experiential, as opposed to comparing
impressions than material purchase signals (MExp ¼ 6.13, different purchases as we did in study 1. We again find that
SD ¼ 1.72 vs. MMat ¼ 5.25, SD ¼ 1.96; F(1, 884) ¼ material versus experiential purchase signals conclusively
38.92, p < .001, xp2 ¼ .041). alter impressions only when multiple signals are present
These main effects were qualified by a significant inter- (MMatFour ¼ 4.60, SD ¼ 2.06; MExpFour ¼ 5.63, SD ¼ 1.94;
action (F(2, 884) ¼ 13.93, p < .001, xp2 ¼ .028). Testing MMatOne¼ 5.77, SD ¼ 1.58; and MExpOne ¼ 6.09, SD ¼
H3 requires the decomposition of the omnibus 2 df interac-
1.61).
tion and the use of a planned 1 df interaction contrast.
In summary, these studies suggest that the number of
Using the interaction contrast, we find the difference be-
posts indeed acts as a moderating factor as suggested by
tween experiential and material purchase signals was mod-
H3. The fact that these studies find only weak support for
erated by the number of posts (F(1, 884) ¼ 13.67, p <
H1a and H1b in the context of single signals and stronger
.001, xp2 ¼ .014). Following up on this predicted interac-
tion, planned simple contrasts indicated the difference be- support in the context of multiple signals suggests that on
tween experiential and material purchase signals was only social media multiple posts may be necessary for differen-
marginally significant for a single post, but emerged ces in purchase type to alter impressions. This is consistent
strongly for multiple posts. The difference between non- with the fact that social media environments are highly am-
purchase signals and material purchase signals was signifi- biguous and receivers might have a hard time drawing di-
cant regardless of the number of posts, while the difference agnostic conclusions about senders’ intentions from a
between non-purchase signals and experiential purchase single post. Ambiguity is resolved as more posts of the
signals was only significant for multiple posts. All con- same type are present, consistent with cue diagnosticity
trasts are presented in table 2. theory.
TABLE 2
STUDY 1 RESULTS
this perception will have a negative effect on the impres- and dependent variable (impressions). The Fornell-Larcker
sions signal receivers form of the signal sender. (1981) criterion for discriminant validity requires the aver-
age variance extracted (AVE) of both constructs to be
Method greater than the squared correlation between them. In this
case, the AVE is 0.90 for impressions and 0.82 for authen-
Study 2 followed a 2 (purchase signal type: experiential
ticity of the post and the squared correlation between them
vs. material) 2 (number of posts: 1 vs. 4) between-
is 0.44,1 meeting this criterion and suggesting discriminant
subject design. Participants were recruited through MTurk.
validity of the measures.
Instead of the captcha question used in study 1, we used a
cultural knowledge check (the same check was used in
studies 2, 3, 4, and 5, while an updated check was used in Results
study 6; see details in the web appendix). We set our sam- Impressions. We estimated an ANOVA with impres-
ple size to 175 respondents per cell. Of the 713 participants sions as the dependent variable, and purchase signal type,
who completed the survey, 700 unique responses were number of posts, and their interaction as independent varia-
retained. Respondents were told they would evaluate a per- bles. Similar to study 1, we found that respondents formed
son, Alex Watson, based on their Facebook post(s). We better impressions if Alex’s posts featured experiential
used the same purchases used in study 1. Changes in the rather than material purchase signals (MExp ¼ 5.98, SD ¼
language of the posts used are reported in table 3 (see stim- 1.76 vs. MMat ¼ 5.35, SD ¼ 1.89; F(1, 696) ¼ 20.91, p <
uli in web appendix). Depending on the condition (1 vs. 4 .001, xp2 ¼ .028). Moreover, there was a negative and sig-
posts), respondents either saw all posts of a given type in nificant effect of the number of posts (MOne¼ 6.21, SD ¼
one of four predetermined orders or saw only one ran- 1.47 vs. MFour ¼ 5.09, SD ¼ 2.03; F(1, 696) ¼ 70.75, p <
domly selected post of a given type. Afterwards they rated .001, xp2 ¼ .091) and a marginally significant interaction
both their impressions of the sender and the authenticity of (F(1, 696) ¼ 3.59, p ¼ .058, xp2 ¼ .004). Replicating the
the post(s). In addition to the two impression questions pattern observed in previous studies, simple effect con-
used in study 1, we added the question: “How do you feel trasts (see table 4) revealed that the difference in impres-
about Alex Watson?” (1 ¼ very negative, 9 ¼ very posi- sions between using experiential and material purchase
tive). Responses to these three impression questions were signals was marginal for a single post but significant for
highly correlated (a ¼ .96) and were averaged to compute multiple posts.
an impressions measure. Adding a third item allows
us to conduct discriminant validity analysis between over- Authenticity of the Post. Next, we estimated a separate
all impressions of the signal sender and the presumed me- ANOVA with authenticity of the post as the dependent var-
diator, authenticity of the post(s) (Fornell and Larcker iable. We found that experiential purchase signals were
1981). perceived as more authentic compared to material ones
Respondents evaluated the authenticity of Alex’s post(s) (MExp ¼ 66.78, SD ¼ 27.85 vs. MMat ¼ 58.86, SD ¼
by rating the extent to which the post/s was/were inauthen- 30.23; F(1, 696) ¼ 13.54, p < .001, xp2 ¼ .018).
tic, disingenuous, and artificial (0 ¼ not at all; 100 ¼ a Moreover, we observed a significant negative effect of the
great deal). Responses to these questions were reverse-
coded and averaged to compute a measure of post authen- 1 Note that the Fornell-Larcker analysis relies on the squared correla-
tion between the weighted factors extracted from the confirmatory fac-
ticity (a ¼ .93). A confirmatory factor analysis supports tor analysis; the squared correlation between factors computed by
discriminant validity between the mediator (authenticity) averaging item scores is .39.
VALSESIA AND DIEHL 437
TABLE 3
POSTS STUDY 2
Material posts
“Guys, I finally decided it’s time to be active again! Got a Fitbit today. #fun”
“Sometimes you just gotta find a way to take care of yourself. Got the best cashmere pullover today! #treatyourself”
“Guess who got a new awesome phone? #happyme”
“So glad I have the opportunity to do the things I like! Got myself a new GoPro! #bestcamera”
Experiential posts
“Guys, I finally decided it’s time to be active again! Got a 2-day ski pass today. #fun”
TABLE 4
STUDY 2 RESULTS
Impressions Experiential 6.38 (1.46) 5.52 (1.95) F(1, 696) ¼ 21.56, p < .001, xp2 ¼ .029
Material 6.03 (1.46) 4.67 (2.03) F(1, 696) ¼ 52.31, p < .001, xp2 ¼ .068
F(1, 696) ¼ 3.70, p ¼ .055, xp2 ¼ .004 F(1, 696) ¼ 20.30, p < .001, xp2 ¼ .027
Post authenticity Experiential 73.50 (26.11) 59.27 (27.89) F(1, 696) ¼ 24.13, p < .001, xp2 ¼ .032
Material 71.19 (26.71) 46.38 (28.48) F(1, 696) ¼ 71.14, p < .001, xp2 ¼ .091
F(1, 696) ¼ .64, p ¼ .423 F(1, 696) ¼ 18.92, p < .001, xp2 ¼ .025
number of posts (MOne¼ 72.39, SD ¼ 26.39 vs. MFour ¼ opposed to material ones. This study again supports the
52.77, SD ¼ 28.88; F(1, 696) ¼ 89.43, p < .001, xp2 ¼ moderating effect proposed in H3 but conclusively shows
.112) and a significant interaction (F(1, 696) ¼ 6.57, p ¼ support for H1b only in the presence of multiple signals,
.011, xp2 ¼ .008). Simple effect contrasts (see table 4) suggesting that multiple purchase signals are necessary to
revealed a similar pattern of results as for impressions. alter impressions. Similarly, receivers only form more neg-
When only one post was available, perceptions of authen- ative authenticity judgments for material purchase signals
ticity did not vary significantly as a function of purchase when multiple signals are present, suggesting the process
signal type. However, when multiple posts were available, proposed in H2 only takes place in the presence of multiple
the authenticity of the post was significantly greater for ex- signals. It appears that in the ambiguous context of social
periential purchase posts. media communication, if individuals post about a purchase
once, receivers might not form strong opinions about their
Mediation Analysis. Post authenticity significantly me-
motivation.
diated the effect of purchase type on overall impressions in
While not part of our predictions, it should be noted that
the presence of multiple purchases (bindirect ¼ .47, 95%
in both study 1 and study 2 we observe a decline in impres-
CI [.70, .25], 10,000 bootstrapped samples), but not
sions for purchase signals in the presence of multiple posts.
when only one purchase was displayed (bindirect ¼ .08,
This decline is always significant for material purchases.
95% CI [.28, .11]). A test of moderated mediation further
However, for experiential purchases, this effect is not con-
revealed the size of the indirect effect of purchase type
sistent across studies. In study 2, we observe a significant
through authenticity was greater in the presence of multiple
decline, as we do in the supplementary study presented in
purchase signals (bdifference ¼ .39, 95% CI [.69, .09],
the web appendix, while such a decline was not significant
PROCESS Model 8, Hayes 2013).
in study 1. Given that the study procedures and samples
were very similar between these studies, these differences
Discussion in results point to a small and likely unreliable effect.
In study 2, we replicated the previous findings that indi- Since purchase type only seems to affect impressions
viduals form more positive impressions of a person who when multiple signals are present, we focus on multiple
posted multiple times about experiential purchases as purchase signals in the remaining studies. Further, H4
438 JOURNAL OF CONSUMER RESEARCH
argues that in the presence of additional diagnostic cues Test of Manipulation. We expected that highlighting
about a signal’s lack of authenticity, receivers will rely less the monetary exchange in the text and via hashtags would
on the type of purchase mentioned when they infer authen- render social media posts less authentic, and we confirmed
ticity. Thus, the type of purchases they post about may this was the case in a separate between-subject test con-
have less impact on the impression receivers form of signal ducted on 198 MTurk participants, who evaluated the au-
senders. We explore this prediction in the next three stud- thenticity of the four posts used in the two conditions (see
ies using different social media-specific cues. web appendix for details on stimuli and questions). Posts
that did not highlight the monetary exchange were per-
TABLE 5
POSTS STUDY 3
TABLE 6
POSTS STUDY 4
Material posts
“I am the proud owner of a bottle of Chateau Lafite. It makes a great addition to my wine collection! #thebestproduct”
“I just got a boxed set of my favorite TV show. I can’t wait to have it on display, it is going to have a place of honor on my DVD shelf!
#bestpurchase”
“I love owning a National Geographic subscription. Looking at all those yellow magazines in my library makes me so happy! #yellowmagazines”
“I recently bought this beautiful coffee table book—a collection of poems and nature pics from all over the world. It looks so amazing.
#beautifulproduct”
STUDY 5—HOW ARE MARKETER- Test of Manipulation. We expected adding the three
DRIVEN PURCHASE POSTS RECEIVED? hashtags would render social media posts less authentic,
which we confirmed in a separate between-subject test
Our studies so far have shown that repeatedly sending conducted on 197 MTurk participants (see full details in
purchase signals (and material purchase signals in particu- web appendix). As intended, the presence of these hashtags
lar) can reduce others’ impressions of the sender.
affected perceptions of authenticity: posts including these
Marketers, however, want consumers to frequently post
hashtags were perceived as significantly less authentic
about their purchases. It is common for marketers to use
compared to ones that did not include these hashtags
TABLE 8
POSTS STUDY 5
Material posts
“I am the proud owner of a bottle of Chateau Lafite. It makes a great addition to my wine collection! #thebestproduct”
“I love owning a National Geographic subscription. Looking at all those yellow magazines in my library makes me so happy! #yellowmagazines”
“My new mountain bike is by far the best product I own. It’s so beautiful—a single chain ring carbon fiber hardtail with 27 in. wheels. #beau-
tifulproduct”
“I just got a boxed set of my favorite TV show. I can’t wait to have it on display, it is going to have a place of honor on my DVD shelf!
#bestpurchase”
significant (bdifference ¼ .29, 90% CI [.01, .56], PROCESS purchase—experiential or material—they would post about
Model 8, Hayes 2013). on Facebook (n ¼ 583). A large portion of respondents
chose a mixture of material and experiential posts (79%)
Discussion (see full results in the web appendix). This suggests that us-
ing mixed signals may indeed be quite widespread.
Study 5 explores a common occurrence on social media:
Interestingly, individuals expected to make equally posi-
consumers sharing marketer-promoted purchase posts. We
tive impressions regardless of how many material or expe-
find that when marketer connections are highlighted, post- riential posts they included (F(4, 577) ¼ 1.05, p ¼ .378).
ing about experiential purchases no longer leads to an im- In other words, everyone believed they had put their best
pression benefit over posting about material purchases. foot forward with whatever combination of signals they
The findings in this study lend additional support to H4 selected.
and to the proposed underlying authenticity-based process In study 6, we explore how different combinations of
(H2). Interestingly, these findings also suggest that when it purchase signals affect impressions. The nature of this
comes to these types of promotions, consumers’ and mar- study is exploratory, and we could a priori make several
keters’ incentives are not always aligned. Marketers want different predictions as to how a combination of experien-
consumers to post WOM and to attribute WOM to specific tial and material signals could be received compared to sig-
promotions (e.g., a sweepstakes); however, consumers may nals of the same nature. On the one hand, we know that
be hurt by associating themselves with marketers in their multiple experiential signals are received more positively
social media posts since this could dampen how others per- compared to multiple material signals. It is therefore possi-
ceive them. Note, however, it is possible not all viewers ble that receivers form more positive impressions of the
clearly perceive these hashtags as being marketer driven, signal sender the more they send experiential purchase sig-
possibly contributing to marginal results on some of our nals. On the other hand, prior literature has identified a
measures, but also suggesting that real-world marketers positive relationship between consistency in behavior and
may benefit from hashtag ambiguity. perceptions of authenticity (Kraus, Chen, and Keltner
2011), so it is also possible that receivers value consistency
STUDY 6—HOW ARE MIXED SIGNALS in a signal sender. In this case, we might find that multiple
RECEIVED? signals of the same nature, whether they are material or ex-
periential, lead to more positive impressions compared to
Earlier studies compared how multiple (vs. single) mate- mixed signals. Finally, a third pattern is also possible. We
rial or experiential purchase signals affect impression for- previously learned that receivers generally form more neg-
mation. In these studies, cumulative signals were all of the ative impressions of senders who send multiple purchase
same nature, either all experiential or all material. In real- signals of a given type (vs. a single signal). It is therefore
ity, however, people might use a combination of experien- also possible that mixed signals might lead to more favor-
tial and material purchase signals. To assess whether able impressions compared to multiple signals of the same
senders would indeed send a combination of multiple sig- type, if receivers focus on the fact that in the case of mixed
nals, we recruited Facebook users from MTurk and asked signals, fewer signals of any given type have been sent. In
them to make four sequential choices as to which study 6, we examine how the mixture of experiential and
VALSESIA AND DIEHL 443
TABLE 9
Marketer association
Impressions Experiential 5.87 (1.85) 5.54 (1.88) F(1, 788) ¼ 2.99, p ¼ .084, xp2 ¼ .003
Material 5.05 (2.03) 5.33 (1.76) F(1, 788) ¼ 2.14, p ¼.144
F(1, 788) ¼ 18.64, p < .001, xp2 ¼ .022 F(1, 788) ¼ 1.27, p ¼ .260
Post authenticity Experiential 58.92 (30.78) 45.56 (31.13) F(1, 788) ¼ 18.04, p < .001, xp2 ¼ .021
Material 51.40 (31.58) 45.58 (31.03) F(1, 788) ¼ 3.49, p ¼ .062, xp2 ¼ .003
F(1, 788) ¼ 5.78, p ¼ .016, xp2 ¼ .006 F(1, 788) ¼ .00, p ¼ .994
material purchases affects how authentic receivers perceive and the dependent variable. In this case, the AVE is 0.91
these posts and how authenticity affects impressions. for impressions and 0.82 for authenticity of the posts, and
the squared correlation between impressions and authentic-
Method ity of the posts is 0.37,4 meeting the Fornell-Larcker crite-
rion and suggesting discriminant validity of the measures.
We set our sample size to 150 participants per condi-
tion.3 Of the 750 MTurk participants who completed the
survey, 737 unique responses were retained. We told
Results
respondents we were interested in their impressions of a Impressions. An ANOVA predicting impressions
Facebook user named Alex Watson. Next, they were ran- revealed a significant omnibus effect of condition (F(4,
domly assigned to one of five conditions, varying the num- 732) ¼ 14.26, p < .001). More importantly, we identified a
ber of material (M) versus experiential (E) posts significant linear trend (F(1, 732) ¼ 46.43, p < .001),
respondents encountered (4M0E, 3M1E, 2M2E, 1M3E, alongside a quadratic term (F(1, 732) ¼ 9.85, p ¼ .002, see
0M4E). Each participant evaluated one of the possible 16 figure 2A). Simple comparisons revealed that posting only
combinations of four Facebook posts using the four mate- about material purchases led to significantly lower impres-
rial and four experiential purchases used in earlier studies sions compared to all other conditions. Further, including
(see table 10 and stimuli in the web appendix). only one experiential purchase led to marginally lower
Respondents in the 4M0E (4 material, 0 experiential) and impressions compared to including two experiential pur-
the 0M4E (0 material, 4 experiential) condition saw only chases and significantly lower impressions compared to in-
material or only experiential purchases in a predetermined cluding three or four experiential purchases. Impressions
order. Respondents in the other conditions saw a set of were not significantly different from each other for two,
posts randomly chosen among possible sets corresponding three, or four experiential purchases (full details in web
to their condition. For instance, respondents in the 3M1E appendix).
(3 material, 1 experiential) condition saw one randomly de- Authenticity of the Posts. An ANOVA predicting au-
termined experiential purchase among three material pur- thenticity of the posts also revealed a significant omnibus
chases. Respondents saw all four posts before answering effect of condition (F(4, 732) ¼ 6.69, p < .001). Again, we
the three impression questions and three authenticity ques- identified a significant linear trend (F(1, 732) ¼ 23.40, p <
tions used in earlier studies. Responses to these items were .001). However, while the overall pattern of results was
averaged to form a measure of impressions (a ¼ .97) and a very similar, the quadratic term was not significant (F(1,
post authenticity measure (a ¼ .93). 732) ¼ 2.38, p ¼ .123). Simple comparisons revealed post-
A confirmatory factor analysis including the items mea- ing only about material purchases or only about one experi-
suring impressions and those measuring authenticity of the ential purchase led to lower authenticity perceptions
posts supports discriminant validity between the mediator
4 Note that the Fornell-Larcker analysis relies on the squared correla-
3 In the web appendix, we also report an earlier version of this study tion between the weighted factors extracted from confirmatory factor
that did not measure authenticity. That study finds a very similar pat- analysis; the squared correlation between factors computed by averag-
tern of results with regard to impressions. ing item scores is .41.
444 JOURNAL OF CONSUMER RESEARCH
TABLE 10
POSTS STUDY 6
Material posts
“Warm and sunny, the perfect day if you have a new GoPro to try! #solucky”
“Today was a good day! Guess who got a new awesome phone? #happyme”
“Sometimes you just gotta do it! Got the best cashmere pullover today! #treat”
“Guys, I finally decided it’s time to be active again. Going to buy my first Fitbit today #excited”
Experiential posts
“Warm and sunny, the perfect day if you have seats to watch your team live! #solucky”
compared to any condition that included at least two expe- accuracy of what senders expect versus what receivers
riential purchases. Authenticity perceptions were not dif- perceive.
ferent from each other for two, three, or four experiential
purchases (see figure 2B; details in the web appendix). GENERAL DISCUSSION
Mediation Analysis. We find authenticity of the post
Using purchases as signals has a long history in practice
significantly mediated the effect of the number of experi- and has been researched extensively. However, the rise of
ential posts on overall impressions (bindirect ¼ .14, 95% CI social media has allowed for novel ways of signaling with
[.08, .20], 10,000 bootstrapped samples, PROCESS model purchases. Across six studies, we find that signal receivers
4, Hayes 2013). question why social media users post about their purchases,
which affects perceptions of post authenticity and, ulti-
Discussion mately, the impression receivers form of the signal sender.
Study 6 replicates our earlier findings that posting about We find the type of purchases senders choose to post about
multiple, exclusively experiential purchases leads to more matters for this process. Receivers perceive multiple mate-
favorable impressions compared to posting about multiple, rial purchase signals as less authentic compared to multiple
exclusively material purchases. Moreover, we find that, experiential purchase signals, and this affects the impres-
generally, including more experiential purchases in one’s sion they form of the signal sender. Importantly, in the so-
posts leads to better impressions. Importantly, we find cial media environment, receivers can use a variety of
these differences are driven by perceptions of the posts’ au- additional cues to infer the sender’s intentions, some of
thenticity, in line with our theorizing. which are under the direct control of the signal sender
Interestingly, sending only experiential purchase signals (e.g., hashtags) and others of which are not (e.g., feedback
by other users). We explored three of these cues (monetary
does not necessarily lead to more positive impressions
mentions, other users’ comments, and marketer associa-
compared to sending mostly experiential purchase signals.
tions via hashtags) and found that receivers do indeed use
Combined with the data discussed in the study’s introduc-
these additional cues to make authenticity inferences, re-
tion, this suggests mixed signals can be equally effective as
ducing the impact of purchase signal type on impressions
only selecting experiential signals, as long as material sig- when such cues allude to inauthentic motives.
nals are kept at a low level. In fact, the data discussed in By exploring how using purchases as signals on social
this study’s introduction find that about half of the senders media affects impressions, we contribute unique instances
choose to post mostly, but not exclusively, about their ex- and a novel understanding of consumers’ reactions to so-
periential purchases (50.77%), seemingly anticipating cial media messages to the well-established literature on
these results, while those who only choose experiential signaling and conspicuous consumption as well as to the
purchase posts (16.98%) may be overestimating how help- burgeoning literature on user-generated social media com-
ful experiential purchase posts are. In addition, there is a munication. Further, by systematically examining how
sizeable portion of the senders who clearly appear miscali- others perceive those that intentionally broadcast purchase
brated (20.24%). They predominantly chose material sig- signals on social media (rather than incidentally observing
nals, which lead to more negative impressions than all ownership, as previously studied), we also contribute to the
other choices, contrary to their stated expectations that literature comparing the various benefits of experiential
what they chose will help them make a good impression on versus material purchases. Notably, we conclusively find
others. Future research may find it interesting to more sys- an experiential benefit only when multiple signals are pre-
tematically explore the role of mixed signals as well as the sent, while prior literature generally observed a variety of
VALSESIA AND DIEHL 445
FIGURE 2
IMPRESSIONS (A) AND AUTHENTICITY OF THE POST PERCEPTIONS (B) AS A FUNCTION OF THE NUMBER OF EXPERIENTIAL
PURCHASE POSTS.
experiences, reducing associations with money being ex- when algorithms prioritize showing certain users consis-
changed. Our findings suggest this could be beneficial for tently in curated feeds. When posts are temporally sepa-
any type of purchase, and for material purchases in particu- rated, other processes such as memory decay may become
lar. An example of how marketers could do that is the more important. Longitudinal investigations of impression
award-winning social media campaign #livelokai that management efforts would be able to examine these issues.
encourages consumers to share pictures of their Lokai bra- Third, we only explore situations in which receivers are
celets on Instagram as a way of sharing about their non- exposed to multiple purchase signals sent in relatively
commercial experiences and their life story. close succession. Future research could explore whether
our effect would be attenuated if more time elapsed be-
way in which consumers can learn about new products and Burrow, Anthony L. and Nicolette Rainone (2017), “How Many
services is by seeing others post about them on social me- Likes Did I Get?: Purpose Moderates Links between Positive
dia. While consumers may not rely on others’ opinions Social Media Feedback and Self-Esteem,” Journal of
Experimental Social Psychology, 69, 232–36.
equally for all purchase types (Dai, Chan, and Mogilner Campbell, Margaret C. and Amna Kirmani (2000), “Consumers’
2020), an interesting question is whether the same features Use of Persuasion Knowledge: The Effects of Accessibility
that influence the impression they form of the signal sender and Cognitive Capacity on Perceptions of an Influence
and their subsequent social media engagement also affect Agent,” Journal of Consumer Research, 27 (1), 69–83.
their interest in the purchase itself. These and many other Caprariello, Peter A. and Harry T. Reis (2013), “To Do, to Have,
or to Share? Valuing Experiences over Material Possessions
Weaknesses of Mechanical Turk Samples,” Journal of Nicolao, Leonardo, Julie R. Irwin, and Joseph K. Goodman
Behavioral Decision Making, 26 (3), 213–24. (2009), “Happiness for Sale: Do Experiential Purchases
Goodman, Joseph K., Selin A. Malkoc, and Brittney L. Make Consumers Happier than Material Purchases?,”
Stephenson (2016), “Celebrate or Commemorate? A Material Journal of Consumer Research, 36 (2), 188–98.
Purchase Advantage When Honoring Special Life Events,” Nunes, Joseph C., Andrea Ordanini, and Gaia Giambastiani
Journal of the Association for Consumer Research, 1 (4), (2021), “EXPRESS: The Concept of Authenticity: What It
497–508. Means to Consumers,” Journal of Marketing, 85 (4), 1–20.
Goodwin, Geoffrey, Jared Piazza, and Paul Rozin (2014), “Moral Nailya, Ordabayeva and Pierre Chandon (2011), “Getting Ahead
Character Predominates in Person Perception and of the Joneses: When Equality Increases Conspicuous
Evaluation,” Journal of Personality and Social Psychology, Consumption among Bottom-Tier Consumers,” Journal of
Van Dijk, E. and M Zeelenberg (2003), “The Discounting of Individuals on Facebook: Are we Known by the Company
Ambiguous Information in Economic Decision Making,” we Keep?,” Human Communication Research, 34 (1), 28–49.
Journal of Behavioral Decision Making, 16 (5), 341–52. Warren, Caleb and Margaret C. Campbell (2014), “What Makes
Vazire, Simine and Samuel D. Gosling (2004), “e-Perceptions: Things Cool? How Autonomy Influences Perceived
Personality Impressions Based on Personal Websites,” Coolness,” Journal of Consumer Research, 41 (2), 543–63.
Journal of Personality and Social Psychology, 87 (1), Watts, Duncan J., Jonah Peretti, and Michael Frumin (2007), Viral
123–32. Marketing for the Real World, Boston, MA: Harvard
Veblen, Thorstein (1899), The Theory of the Leisure Class: An Business School Publishing.
Economic Study of Institutions, New York: The Macmillan Weingarten, Evan and Joseph K. Goodman (2021), “Re-
Company, https://www.google.com/books/edition/The_ Examining the Experiential Advantage in Consumption: A