Journal of Social and Clinical Psychology, Vol. 33, No. 8, 2014, pp. 701-731
© 2014 Guilford Publications, Inc.
The authors would like to sincerely thank the following individuals for proofreading
and/or providing suggestions on how to improve our article: Darren Steers, Alﬁe
Steers, Jennifer J. Pruess, and Chelsie M. Young.
Address correspondence to Mai-Ly N. Steers, Department of Psychology, University
of Houston, 126 Heyne Bldg., Houston, TX 77204-5022; E-mail: firstname.lastname@example.org
STEERS et al.
SEEING EVERYONE ELSE’S HIGHLIGHT REELS:
HOW FACEBOOK USAGE IS LINKED TO
MAI-LY N. STEERS
University of Houston
ROBERT E. WICKHAM
Palo Alto University
LINDA K. ACITELLI
University of Houston
Two studies investigated how social comparison to peers through computer-me-
diated interactions on Facebook might impact users’ psychological health. Study
1 (N = 180) revealed an association between time spent on Facebook and de-
pressive symptoms for both genders. However, results demonstrated that making
Facebook social comparisons mediated the link between time spent on Facebook
and depressive symptoms for men only. Using a 14-day diary design (N = 152),
Study 2 found that the relationship between the amount of time spent on Face-
book and depressive symptoms was uniquely mediated by upward, nondirec-
tional, and downward Facebook social comparisons. Similarly, all three types of
Facebook social comparisons mediated the relationship between the number of
Facebook logins and depressive symptoms. Unlike Study 1, gender did not mod-
erate these associations. Both studies provide evidence that people feel depressed
after spending a great deal of time on Facebook because they feel badly when
comparing themselves to others.
702 STEERS ET AL.
Comparison is the thief of joy.
Over forty years ago, communication theorist Marshall McLuhan
(1964) coined the phrase, “the medium is the message.” He did not
mean to imply that individuals should ignore messages commu-
nicated through a particular medium; but rather, people should
not only be cognizant of a medium’s obvious properties but also
be aware of how it subtly inﬂuences culture. He argued that im-
portant technological advances have the potential to become exten-
sions of the people using them and, in turn, may redeﬁne human in-
teractions. Moreover, McLuhan envisioned that technology would
someday provide people with the tools to create a global village.
Thus, technological media has the power to organize societies,
and can profoundly change interpersonal relationships for better or
worse. For example, with more than a billion monthly active users
worldwide (Facebook, 2012), the social networking site Facebook
has brought McLuhan’s idea of a global village to fruition, funda-
mentally altering the dynamics of human interaction. Prior research
has tied Facebook use to positive effects such as fulﬁllment of ego
needs (Toma & Hancock, 2013), greater subjective well-being (Kim
& Lee, 2011), and higher relationship quality for those in a romantic
relationship (Steers, Øverup, Brunson, & Acitelli, 2014). However,
for some individuals the results of such cyber exchanges may be
more dystopian than utopian.
For instance, internet addiction, which is deﬁned as using the in-
ternet to an excessive degree, has been associated with depressive
symptoms among young people as well as older adults (e.g., Mor-
rison & Gore, 2010). Moreover, an analysis of 200 college students’
Facebook status updates, a mechanism by which individuals often
divulge information en masse to their Facebook friends, revealed
that 25% of the Facebook proﬁles evidenced depressive symptoms
through their status updates over the past year (although only 2.5%
exhibited major depressive episode criteria; Moreno et al., 2011).
Another study surveyed 425 Facebook users and found that indi-
viduals who possessed a Facebook account over a longer period
(i.e., for several years) tended to perceive that others are happier
and life is unfair (Chou & Edge, 2012). Moreover, individuals who
spent more hours per week and those who befriended strangers
were more likely to believe that others on Facebook had better lives
than they did. Finally, a related study found that people tend to
FACEBOOK USAGE 703
underestimate others’ negative emotions, which often leads to emo-
tional pluralistic ignorance (Jordan, Monin, Dweck, Lovett, John, &
Gross, 2011). That is, those afﬂicted with emotional difﬁculties may
fail to recognize others’ internal struggles, which may compound
feelings of loneliness and isolation. The researchers reasoned that
this occurs because people publically portray themselves as being
happier than they actually are (Jordan et al., 2011).
The present work builds on these established ﬁndings by examin-
ing the extent to which spending time on Facebook encourages in-
dividuals to compare their lives to others’. If people portray them-
selves as happier than they actually are, then perceptions of the
happiness and well-being of one’s Facebook friends are likely to be
distorted. However, the underlying mechanism behind what moti-
vates individuals on Facebook to make such comparisons and how
this relates to their sense of well-being has yet to be elucidated. In
our proposed model, increased social comparisons stemming from
spending time on Facebook leads to greater depressive symptomol-
ogy among users.
Social comparisons occur when people automatically contrast
themselves with others on abilities or attributes they deem impor-
tant. Leon Festinger (1954) ﬁrst theorized that individuals have an
innate desire to socially compare themselves to others as a way to
evaluate their own opinions and abilities and that people usually
selectively choose whom to compare themselves to on the basis of
perceived similarity. That is, people tend to compare themselves to
peers or friends on self-relevant issues or concepts.
Prior work has established a relationship between social com-
parisons and mental well-being in normal populations (e.g., Gil-
bert, Allan, Brough, Melley, & Miles, 2002; Troop, Allan, Treasure,
& Katzman, 2003). Speciﬁcally, research has found that making up-
ward social comparisons, seeing oneself as inferior to others, are as-
sociated with negative health outcomes, such as greater depressive
symptoms, lower self-esteem, and negative self-evaluations (i.e.,
Allan & Gilbert, 1995; Tesser, Millar, & Moore, 2000). Conversely,
downward social comparisons, seeing oneself as better off than
or superior to others, has been commonly associated with posi-
tive health outcomes such as less anxiety, positive self-esteem, and
704 STEERS ET AL.
greater positive affect (e.g., Allan & Gilbert, 1995; Amoroso & Wal-
ters, 1969; Wills, 1981).
However, other literature suggests that the relationships between
social comparison, affective responses, and, consequently, well-be-
ing, may be more complex than simply being inherent to the direc-
tion of the social comparison. For instance, Buunk, Taylor, Dakof,
Collins, and VanYperen (1990) found evidence that individual dif-
ferences such as self-esteem, perceived control over circumstances,
and feelings of dissatisfaction, may moderate whether individuals
feel positive or negative affect following upward or downward so-
cial comparisons. In recent years, researchers have expanded upon
the idea that affective responses may be independent of direction of
social comparison by suggesting that it is the act of frequently socially
comparing oneself to others rather than the direction of social compari-
son that is related to long-term destructive emotions (White, Langer,
Yariv, & Welch, 2006). Thus, any beneﬁts gained from making social
comparisons may be temporary whereas engaging in frequent so-
cial comparisons of any kind may be linked to lower well-being.
GENDER AND SOCIAL COMPARISON
Research has also indicated that there are gender differences in
social comparison at both an individual and group level. In their
studies on uniqueness bias, Goethals, Messick, and Allison (1991)
consistently found that males differentiated themselves from oth-
ers more often than females. That is, men believed they were more
intelligent, athletic, creative, and smarter than others. Conversely,
women viewed themselves as at the same or below others on most
levels. Women only tended to exhibit self-other differentiation on
In addition, the literature has consistently demonstrated that, due
to perceived similarity, individuals prefer same-sex social compari-
sons (e.g., Major & Forcey, 1985; Suls, Gaes, & Gastorf, 1979) and that
the genders experience differing effects as a result of their compari-
sons. Researchers found that men reported lower self-esteem when
engaging in upward social comparison with other males. However,
women reported lower self-esteem when making upward social
comparisons with males, but not with other females (Martinot, Re-
dersdorff, Guimond, & Dif, 2002). There were two primary reasons
for this phenomenon: females viewed themselves as subordinate
FACEBOOK USAGE 705
to males, which negatively impacted their self-esteem, and women
protected their self-esteem when making upward social compari-
sons with other females by considering them part of their ingroup
(Martinot & Redersdorff , 2003). Researchers found that men expe-
rienced lower levels of self-esteem only when engaging in upward
social comparisons with women in traditionally female-oriented
domains (Redersdorff, 2002).
FACEBOOK SOCIAL COMPARISON
Although social comparison processes have been examined at
length in traditional contexts (i.e., face-to-face), to our knowledge,
the literature has yet to examine social comparison in online social
networking settings. However, this may be an important avenue to
explore due to the fact that online interactions, speciﬁcally those on
the popular social networking site, Facebook, would likely involve
social comparisons that may be associated with health outcomes.
Facebook users spend over 700 billion minutes per month on Face-
book (Facebook, 2010). In fact, a recent report estimated that mem-
bers devote an overwhelming 16 percent of their total internet time
to Facebook-related activities in the U.S. alone (Davis & Angelova,
2011). Thus, an investigation is needed to better understand how
Facebook activities relate to well-being.
After logging on, Facebook users are exposed to a continual
stream of information (i.e., status updates, viewing newly upload-
ed pictures, friends posting on each other’s walls, liking of other
people’s status updates). Thus, these Facebook activities may serve
as stimuli for individuals to automatically engage in frequent non-
directional, upward, and/or lateral social comparisons, especially
for those who spend longer amounts of time/frequently view Face-
book. Because the Facebook platform promotes self-disclosure, us-
ers may reveal highly personal information, which they normally
would not divulge (Gross & Acquisti, 2005). Therefore, Facebook
users are often privy to information about their Facebook friends
that they might not have known otherwise.
More generally, an individual may engage in social comparisons
on Facebook by comparing the number of likes or comments other
people have posted to their status updates relative to their peers.
However, individuals may also make speciﬁc social comparisons
after viewing a particular friend’s pictures or status updates. For
706 STEERS ET AL.
example, a recent divorcée might feel worse about being single af-
ter seeing an acquaintance’s recent engagement photos posted on
Facebook (upward social comparison). People might also engage in
social comparisons on Facebook in order to feel better about them-
selves. For instance, a man may temporarily feel more conﬁdent
after reading a status update about his friend’s failing grade on an
exam for which he earned an “A” (downward social comparison).
Users are often unable to anticipate what their friends will post.
Thus, they often cannot control what they will the view when they
log on or what will information will serve as the impetus for them
to make social comparisons. However, we hypothesize that more
time spent on Facebook will provide Facebook users with greater
opportunities to socially compare themselves to their friends.
THE CURRENT RESEARCH
Previous research has demonstrated that people’s goals, motives,
and interests often remain the same whether the interactions are
made online or face-to-face (Mckenna & Bargh, 2000). Thus, the
goal of comparing oneself to others may be just as strong whether
one is on Facebook or interacting face-to-face. Across two studies,
we tested the primary hypothesis that social comparisons on Face-
book would mediate the association between time on Facebook and
Study 1 used a cross-sectional design to examine whether mak-
ing nondirectional Facebook social comparisons, deﬁned as asking
people whether they compare themselves to others rather than in
which direction, mediated the relationship between Facebook usage
and depressive symptomology. To further investigate the proposed
mediation model, a daily diary design was employed in Study 2.
Moreover, Study 2 examined different types of social comparison
(upward, downward, and nondirectional) as potential mediators of
the relationship between time on Facebook and depressive symp-
toms. We further explored whether these three types of Facebook
social comparisons would also serve as mediators between num-
ber of Facebook logins (how many times people view Facebook
per day as opposed to length of time) and depressive symptoms.
We hypothesized that Facebook logins might function similarly to
Facebook time. In addition, because previous studies have found
gender differences in social comparison (e.g., Goethals et al., 1991;
FACEBOOK USAGE 707
Martinot et al., 2002) and women are more than twice as likely to
be depressed as men (i.e., Piccinelli & Wilkinson, 2000), we exam-
ined gender as a potential moderator of the relationship between
Facebook social comparisons and depressive symptoms across both
studies (moderated mediation).
We predicted that active Facebook users would be higher in general
social comparison orientations because they have more opportuni-
ties to compare themselves to their friends than nonactive or non-
Facebook users. Thus, we ﬁrst considered the possibility that active
Facebook users would be signiﬁcantly higher in general social com-
parison orientations than nonactive Facebook users or individuals
who do not have a Facebook account. That is, active Facebook us-
ers would be more likely to engage in social comparisons (on and
ofﬂine) than nonactive Facebook users or individuals who do not
have a Facebook account (H1).
Previous research has demonstrated that Facebook use or exces-
sive internet use predicted depression (Moreno et al., 2011; Morri-
son & Gore, 2010). Hence, we hypothesized that time on Facebook
would be positively associated with depressive symptoms (H2).
Moreover, the amount of time on Facebook was expected to be pos-
itively related to nondirectional Facebook social comparison, such
that, the more time an individual spends on Facebook the more he
or she is likely to socially compare (H3). Additionally, we anticipat-
ed that nondirectional Facebook social comparisons would serve as
a mediator between time on Facebook and depressive symptoms
(H4). Finally, we explored gender as a moderator of the relationship
between nondirectional Facebook social comparisons and depres-
sive symptoms (H5).
Study 1 was conducted with 180 students (39 males, 141 females)
from a large southwestern university. Having a Facebook account
was not a prerequisite; however, most participants did have an
account (92%). Participants were ethnically diverse (17% African
708 STEERS ET AL.
American, 17% Asian American, 26% Hispanic, 32% Caucasian,
3% Middle Eastern, 4% Multiracial, and 1% Native American) and
ranged in age from 19 to 57 years (M = 24.41, SD = 5.88). Partici-
pants who reported having a Facebook account were asked if they
considered themselves to be active Facebook users. An active Face-
book user was deﬁned as someone who checks his or her Facebook
account on a regular basis. If a participant rarely checked their
Facebook account, had deactivated their account, and/or did not
regard themselves an active member, they were considered a non-
active Facebook user. There were 133 active users (26 males, 107
females), 33 nonactive Facebook users (7 males, 26 females), and
14 participants who did not have a Facebook account (6 males, 8
females). Because of the small sample size, those who did not have
a Facebook account were combined with nonactive Facebook users
(47 nonactive Facebook users).
Participants were recruited from undergraduate psychology class-
es and were told the study would be exploring internet use and
personality. Respondents accessed the online questionnaire via a
research website and were asked to complete demographic infor-
mation, social comparison measures, and depressive symptomol-
ogy measures. In addition, active Facebook users were directed to
Facebook-related questions (i.e., average amount of time per day
they spent on Facebook) and measures adapted for Facebook use.
Upon successful completion, participants were compensated with
Social Comparison. General social comparisons were measured
through the Iowa-Netherlands Comparison Orientation Measure
(INCOM; Gibbons & Buunk, 1999). The INCOM gauges partici-
pants’ tendencies to socially compare themselves to others using 11
items (e.g., I always pay a lot of attention to how I do things com-
pared with how others do things) on a 5-point Likert scale, ranging
from 1 (I disagree strongly), to 5 (I agree strongly; α = .86).
In addition, we adapted the Iowa-Netherlands Comparison Ori-
entation Measure (Gibbons & Buunk, 1999) to a Facebook context
(nondirectional COM-F) to determine social comparison tendencies
FACEBOOK USAGE 709
on Facebook (e.g., When I am on Facebook, I always pay a lot of at-
tention to how well I have done something compared to how others
do things.). As previously mentioned, the measure is nondirectional
in that it does not measure whether people are engaging in upward
or downward social comparison, but simply asks people whether
they compare themselves to others on a 5-point Likert scale (α =
Depressive Symptomology. Depressive symptoms were measured
by the Center for Epidemiological Studies Depression Scale (CES-
D; Radloff, 1977). The CES-D was developed to diagnose depressive
symptoms in normal populations using 20 self-report items (e.g., I
felt like I could not shake off the blues even with help from my fam-
ily or friends). Participants report on how they have felt during the
past week and the scale is rated on a scale of 0 (Rarely or none of the
time; less than 1 day) to 3 (Most or all of the time; 5–7days). Scores
are summed and total possible scores ranging between 0-60. Higher
scores indicate more depressive symptoms (α = .93).
Time on Facebook. The amount of time participants spent on Face-
book was assessed through one item which asked participants,
“How long on average do you spend per day on Facebook?” Re-
spondents could choose from the following response choices: Less
than 5 minutes, From 5–30 minutes, From 30 minutes-1 hour, Be-
tween 1–2 hours, Between 2–3 hours, Between 3–4 hours, and 4+
RESULTS AND DISCUSSION
Means, standard deviations, and zero-order correlations for Study
1 are presented in Table 1. Women scored over 14 on average on the
CES-D, indicating normal levels of depressive symptoms; however,
the average CES-D for men exceeded the clinical threshold indicat-
ing a mild level of depressive symptoms (Ensel, 1986; Zich, Attkis-
son, & Greenﬁeld, 1990).
An independent samples t-test revealed no signiﬁcant differences
between active Facebook users (M = 36.55, SD = 7.40) and nonactive
Facebook users (M = 34.77, SD = 8.21); t (178) = 1.32, p = .19, d = .23
on the general measure of social comparison (INCOM) were found.
Thus, H1 was not supported. This ﬁnding may suggest that that
people do not compare themselves to others more in online con-
710 STEERS ET AL.
texts, relative to face-to-face interactions. However, due to the fact
that we do not know the full scope of participants’ online behaviors
and that other potential confounds may exist, this interpretation
should be viewed with caution. Additionally, no signiﬁcant differ-
ences were observed between male (M = 3.46, SD = 1.77) and female
(M = 3.36, SD = 1.58); t (35) = .28, p = .76, d = .06 participants for the
total amount of time spent on Facebook.
The moderated mediation hypotheses (H2–H5) were examined
using multiple group path analysis in Mplus (Muthén & Muthén,
2012). Moreover, the ab products method described by MacKin-
non and colleagues (Mackinnon, Fairchild, & Fritz, 2007; MacKin-
non, Lockwood, Hoffman, West, & Sheets, 2002), along with boot-
strapped conﬁdence intervals for the indirect effects (Shrout & Bol-
ger, 2002) were used to test the signiﬁcance of the mediated effects.
Figure 1 displays the results for the analysis of H2-H5.
Our analysis revealed that time on Facebook was positively re-
lated to depressive symptoms for both males (β = .36, p < .01) and
females (β = .32, p < .01) (H2). Furthermore, time on Facebook was
positively related to nondirectional COM-F for males (β = .51, p <
.01) and (β = .22, p < .05) for females (H3). Thus, there was evidence
to support H2 and H3. However, nondirectional COM-F was only
signiﬁcantly related to depressive symptoms scores for males (β =
.43, p < .01), whereas it was not signiﬁcant for females (β = .00, p
=.986). Hence, the b path was signiﬁcant for males only. The test of
the indirect effects showed that the mediated effect of nondirection-
al COM-F on the relationship between time on Facebook and de-
pressive symptoms was signiﬁcant for males (β = .219, 95% CI: .026,
.413; p < .05) but not for females (β = .00, 95% CI: -.046, .045, p =.986;
H4). Consistent with expectations, nondirectional COM-F served as
TABLE 1. Correlations Among Study 1 Major Variables
1. Time on Facebook — .15 .22 .57**
2. General Social Comparison (INCOM) .22 — .60** .51**
3. Non-directional COM-F .51** .63** — .61**
4. Depressive Symptomology (CESD) .32** .04 .07 —
Mean (SD) Males 3.46 (1.44) 33.62 (6.97) 28.85 (8.74) 16.54 (12.45)
Mean (SD) Females 3.36 (1.59) 30.94 (9.12) 30.94 (9.12) 14.55 (10.89)
Note. Correlations for Females (N = 107) are presented below the diagonal. Correlations for Males (N =
26) are presented above the diagonal. * p < .05; **p < .01.
FACEBOOK USAGE 711
a mediator between time on Facebook and depressive symptoms
for males only (H5).
Equality constraints were imposed for men’s and women’s a-paths
(i.e., aMen = aWomen) and b-paths (i.e., bMen = bWomen), and Chi-square dif-
ference tests revealed that the a-path between time on Facebook and
COM-F was not signiﬁcantly different for men and women, χ2(1,
133) = 1.62, p = 0.20. However, the b-path between nondirectional
COM-F and depressive symptoms was signiﬁcantly different for
men and women χ2 (1, 133) = 5.42, p < 0.05. These ﬁndings suggest
that spending more time on Facebook is associated with greater
Facebook social comparison, which in turn predicts greater depres-
sive symptoms, but only among men (H4 and H5).
A previous study found that men were signiﬁcantly more likely
than women to use these social networking sites for dating pur-
poses (Raacke & Bonds-Raacke, 2008). Thus, from an evolutionary
perspective, it is possible that the more time men spend on Face-
book the more likely they are to compete with other males (possibly
for mates) and feel inadequate after comparing themselves to their
peers. Time spent on Facebook did not predict women’s outcomes
in the same way. We reasoned that women might use Facebook as
FIGURE 1. Mediational Model for Study 1.
Note. †p < .10; *p < .05; **p < .01.
712 STEERS ET AL.
a way to maintain connections with friends or to bond with other
women rather than compete with them.
On the other hand, perhaps depressed individuals spend more
time on Facebook and consequently, make more social compari-
sons. We examined the plausibility of this alternative interpretation
and found the mediated effect was not signiﬁcant for either gender.
Speciﬁcally, the a-path leading from depressive symptoms to social
comparisons was signiﬁcant for among men (p < .01), but not wom-
en (p = .48). However, the b-path leading from depressive symp-
toms to time on Facebook was nonsigniﬁcant for men (p = .15), but
signiﬁcant among women (p < .05). As a result, the bootstrapped
CIs for the indirect effect contained zero (i.e., nonsigniﬁcant) for
both genders. Although no parametric test is available to compare
the ﬁt of these models, the presence of a signiﬁcant indirect effect
(among men) in the hypothesized direction, but not in the opposite
direction, provides additional support for our process model. The
next study uses a diary design to provide further evidence that ex-
posure to social media leads to increased social comparisons, and in
turn greater depressive symptoms.
In study 2, we conducted a 14-day interval-contingent diary focus-
ing solely on Facebook users in order to gauge a more accurate as-
sessment of how much time participants spend on Facebook and
what types of social comparisons they make. This approach limits
bias that can occur through the administration of global cross-sec-
tional measures and provides greater statistical power. In addition
to the nondirectional COM-F measure, Study 2 contained questions
adapted from the INCOM to measure upward and downward so-
cial comparison (upward COM-F and downward COM-F, respec-
tively). Participants ﬁlled out short questionnaires each night.
STUDY 2 HYPOTHESES
The previous hypotheses (H2 –H5) regarding time on Facebook in
Study 1 were retained and comprise the ﬁrst four hypothesis of
Study 2 (H1–H4). All analyses were conducted at the within-per-
sons level. However, Study 2 provides a more rigorous test of the
FACEBOOK USAGE 713
previous mediation hypothesis by simultaneously examining the
different types of social comparison (upward, downward, and non-
directional COM-F) since all three types of social comparison were
signiﬁcantly correlated with one another (see Table 2). Furthermore,
we assessed Facebook activities using both the amount of time spent
on Facebook, and the number of Facebook logins each day.
Hypotheses 5–8 mirror the ﬁrst four hypotheses for Study 2, but
with frequency of logins as the predictor, rather than amount of time
spent on Facebook. We theorized that frequency of logins might
function as a proxy for time on Facebook. Based on the results from
Study 1 and evidence suggesting that individuals who frequently
engage in social comparisons experience negative consequences
(White, Langer, Yariv, & Welch, 2006), we expected that all three
types of social comparison (upward, downward, or nondirectional)
would serve as a signiﬁcant mediator in both mediational models.
Predictions for Facebook logins were as follows. We hypothesized
that participants’ daily frequencies of logins would be positively
associated with daily depressive symptoms (H5). Moreover, dai-
ly frequencies of logins were expected to be positively related to
daily Facebook social comparisons (H6). Additionally, we expected
daily Facebook social comparisons would mediate the association
between daily frequency of logins and daily depressive symptoms
(H7). Finally, we explored gender as a moderator of the relationship
between Facebook social comparisons and depressive symptoms
TABLE 2. Within-Persons Correlations Among Study 2 Major Variables
1. Time on Facebook — .50** .06† .15** –.12** –.01
2. Facebook Logins .43** — .04 .11** –.10** –.06
3. Upward COM-F .11** .08** — .37** –.35** .16**
4. Non-directional COM-F .20** .08** .36** — –.49** .08*
5. Downward COM-F –.15** –.08** –.25** –.57** — .07*
.02 -.01 .24** .06* .06† —
Mean (SD) Males 4.01 (1.42) 6.39 (3.65) 4.67 (2.13) 4.81 (2.61) 14.29 (2.66) 13.98 (5.77)
Mean (SD) Females 3.97 (1.54) 6.81 (4.27) 4.67 (2.49) 4.97 (2.88) 13.91 (2.88) 14.47 (6.84)
Note. Correlations for Females (N = 93) are presented below the diagonal. Correlations for Males (N =
59) are presented above the diagonal. †p < .10; *p < .05; **p < .01.
714 STEERS ET AL.
In total, 154 individuals (95 female, 59 male) from a large south-
western university completed 2,035 of the 2,156 possible diary en-
tries (94% completion rate) across the fourteen days. Participants
ranged in age from 18 to 42 years old (M = 22.55, SD = 4.22) and the
sample was again ethnically diverse (15% African American, 22%
Asian Americans, 31% Hispanic, 25% Caucasian, 2% Middle East-
ern, 4% Multiracial, and 1% Native American). Only students 18
years or older who logged into their Facebook account on a daily
basis were eligible.
Study 2 consisted of two phases. During phase 1, participants com-
pleted a questionnaire containing demographic information and
attended an orientation session designed to familiarize them with
the diary procedure. Phase 2 consisted of an interval-contingent di-
ary report which was completed for 14 days following orientation.
During orientation, a trained research assistant reviewed the diary
form with participants and explained that one diary record was to
be completed online each night before bed. If they failed to com-
plete an entry at night, participants were instructed to complete the
survey the following morning. Participants without internet access
on a given night were instructed to ﬁll out hard copies.
Furthermore, during orientation, special emphasis was placed
upon clarifying the open-ended question pertaining to Facebook
logins. Additionally, participants were advised to consider only the
amount of time they were actively viewing Facebook when estimat-
ing their total amount of time spent on Facebook. Upon successful
completion, participants were compensated with extra credit.
Participants responded to the following items each night of the 14-
day diary collection phase.
Facebook Time/Logins. Participants reported on the number of
times they logged onto Facebook using an open-ended format (i.e.,
FACEBOOK USAGE 715
How many times did you check your Facebook account today?).
Participants were instructed to consider any time they clicked on
Facebook and/or read an automated e-mail/text/smartphone alert
from Facebook as a view. If they ignored the automated e-mail/
text/smartphone alert from Facebook or kept Facebook running on
their browser but did not look at it, this was not considered a view.
As in Study 1, participants were asked to estimate approximately
how long they spent on Facebook during the day using the follow-
ing response choices: Less than 5 minutes, 5–15 minutes, 16–29 min-
utes, 30 minutes to an hour, Between 1–1½ hours, Between 1½–2
hours, Between 2–2½ hours, Between 2½–3 hours, Between 3–3½
hours, Between 3½–4 hours, Between 4–4½ hours, Between 4½–5
hours, and Between 5+hours.
Facebook Social Comparison. Six items from the Iowa-Netherlands
Comparison Orientation Measure (Gibbons & Buunk, 1999) were
adapted to measure nondirectional, downward, and upward social
comparisons on Facebook. All items contained the common stem:
“TODAY, when I was on Facebook….” Upward social comparison
items included: “… I felt less conﬁdent about what I have achieved
compared to other people,” and “… I concluded I am not as popular
as other people.” Nondirectional items included: “…I paid a lot of
attention to how I do things compared to how others do things,”
and “… if I wanted to ﬁnd out how well I have done something,
I compared what I have done with how well others have done.”
Finally, downward social comparison items included: “…I paid at-
tention to how I do things versus how others do things and felt my
way was better,” and “… I believed that I had accomplished more
than other people had.” All items were measured on a 9-point Lik-
ert scale ranging from I disagree strongly to I agree strongly.
Depressive Symptomology. To minimize participant burden, depres-
sive symptoms were measured using a subset of ﬁve items from
the CES-D (Radloff, 1977). An exploratory factor analysis of Study
1 data revealed the strongest loadings for items 3, 6, 8, 12, and 18,
which led us to include these items in the diary portion of Study 2.
Participants responded to these items using a 9-point Likert scale
items ranging from none of the time today to most of the time today,
and item responses were summed to create a total score ranging
from 5 to 45 (α = .86).
716 STEERS ET AL.
In Study 2, we examined a mediational model based on daily diary
(level 1) responses. The a-paths in our mediational model describe
the association between daily Facebook usage, operationalized
by login frequency or amount of active time spent on Facebook,
and three forms of social comparison: upward, nondirectional,
and downward. Because the experience sampling (diary) method
produces a hierarchical data structure with daily diary responses
nested within individuals, multilevel modeling can be used to ac-
count for the non-independence among diary responses and pro-
vide unbiased signiﬁcance tests (West, Ryu, Kwok, & Cham, 2011).
The a-paths in our multilevel mediational model are described by
the following set of level 1 equations:
(1) upwardij = β0j + β1j * (FBusageij) + eupwardij
(2) nondirectionalij = β2j + β3j * (FBusageij) + enondirij
(3) downwardij = β4j + β5j * (FBusageij) + edownwardij
in which each of the social comparisons measures (i.e., upwardij,
nondirectionalij, downwardij) are regressed on a measure of
Facebook usage (i.e., FBusageij). The presence of ij subscripts for the
social comparisons outcomes, Facebook predictor, and residuals
(i.e., eupwardij, enondirij, edownwardij) signals that these terms vary across both
persons (j) and diary reports (i). Finally, the β0j – β5j coefﬁcients
represent the intercepts (β0j, β2j, β4j) and slopes (β1j, β3j, β5j), which vary
randomly across persons. Using the slopes-as-outcomes formulation
for describing multilevel models, each of these level 1 coefﬁcients is
represented as an outcome variable in a level 2 equation. For the
sake of brevity, we only include the level 2 equations for upward
(4) β0j = γ00 + γ01 * (FBusage.j) + u0j
(5) β1j = γ10 * (FBusageij) + u1j* (FBusageij)
In equation 4, γ00 represents the ﬁxed component of the intercept or
the grand mean of upwardij across all persons and diary responses,
γ01 is the cross-level effect of a person’s average Facebook usage
(FBusage.j), and u0j is the random component representing person-
FACEBOOK USAGE 717
speciﬁc deviations from this overall mean. Similarly, γ10 reﬂects
the ﬁxed or average regression coefﬁcient for FBusageij across all
persons, and u1j is the random (person-speciﬁc) component of this
slope. The level 2 equations for nondirectionalij and downwardij
have an identical structure to equations 4 and 5.
Turning to the second stage of our mediational model, the b - and
c’-paths describe the regression of daily depressive symptoms (CES-
Dij) on the mediators (upwardij, nondirectionalij, downwardij), and
the exogenous predictor (FBusageij), respectively.
(6) CESDij = β6j + β7j * (upwardij) + β8j * (nondirectionalij) + β9j *
(downwardij) + β10j * (FBusageij) + eCESDij
In equation 6, β6j represents the random intercept for daily
depressive symptoms scores, whereas β7j- β9j represent random
slopes for the mediators (b-paths), and β10j reﬂects the random slope
for the exogenous predictor (c’-path). Finally, eCESDij carries the level
1 residual. Each of the coefﬁcients in equation 6 can be expressed
as a level 2 equation; however, to avoid presenting redundant
information, only the equations for the ﬁrst two coefﬁcients are
(7) β6j = γ60 + γ61 * (upward.j) + γ62 * (Genderj) + u6j
(8) β7j = γ70 * (upwardij) + γ71 * (upwardij * Genderj) + u7j* (upwardij)
In equation 7, γ60 represents the ﬁxed component of the intercept
or the grand mean of all depressive symptoms scores across all
persons and diary responses. γ61 describes the cross-level effect of a
person’s average level of upward social comparisons across all diary
responses (upward.j), γ62 expresses the difference in daily depressive
symptoms as a function of participant gender (men = 0; women
= +1), and u6j represents the random, person-speciﬁc deviation of
depressive symptoms from the grand mean. Turning to equation
8, γ70 is the ﬁxed or average regression of CESDij on upwardij across
all persons, γ71 carries the cross-level interaction which describes
the extent to which the average slope changes as a function of the
person’s gender, and u7j describes the random effect for the slope.
The level 2 equations for β8j-β10j are identical in structure to equation
8 (β7j) in that they include cross-level interactions terms for Genderj,
and random effects.
718 STEERS ET AL.
In all models, level 1 predictor variables were centered within-
persons, and the person-level aggregate predictors were grand-
mean centered (West et al., 2011). This centering scheme removes
all of the person-speciﬁc variability from level 1 predictors, and the
level 2 aggregate re-introduces this person-speciﬁc variability as a
distinct predictor. Under this centering scheme, level 1 coefﬁcients
represent the within-persons or daily effect of the predictor, and the
level 2 aggregate represent between-persons or aggregate effects.
Given that we were interested in the day-to-day impact of Facebook
FIGURE 2. Mediational Model for Study 2 .
Note. *p < .05; **p < .01
FACEBOOK USAGE 719
usage, we focused on the level 1 coefﬁcients in the present analysis.
More speciﬁcally, we aimed to test a multiple-mediator version of
the lower-level or 1-1-1 mediation model described by Bauer and
colleagues (Bauer, Preacher, & Gil, 2006). Parameters were estimat-
ed using SAS PROC MIXED (SAS Institute, 2011) with restricted
maximum likelihood estimation. In addition to the person-speciﬁc
non-independence modeled by the random intercept, non-indepen-
dence among temporally adjacent level 1 outcomes (Wickham &
Knee, 2013) was modeled by ﬁtting an autoregressive structure to
the level 1 residual covariance matrix. The lagged and concurrent
× lagged (sensitization-satiation) predictors were also examined
(Wickham & Knee, 2013), but none of the effects reached signiﬁ-
cance and were dropped from the ﬁnal model.
Participant-level variables were manually screened for irregulari-
ties and responses provided by two female participants were ex-
cluded because they reported unusually high number of logins per
day (108 and 365 logins on average per day) relative to the overall
mean of 6.93 logins per day (SD = 8.01). This left a total of 152 par-
ticipants (93 females, 59 males).
A multivariate random intercept model (Mehta & Neale, 2005;
Raudenbush, 1995) was examined using SAS PROC MIXED (SAS
Institute, 2011) for the Facebook usage, social comparisons, and
depressive symptoms measures. This approach allows for the
estimation of a within-persons correlation matrix from diary data,
which is presented in Table 2, along with the means and standard
deviations for these variables. A series of mixed-effects models
revealed no signiﬁcant gender differences on any of the diary
measures (all ps > .51).
Null Random Intercept Models. Null random intercept models
were examined for the three social comparison mediators
(upward, nondirectional, downward) and the outcome variable
720 STEERS ET AL.
(depressive symptoms). These unconditional models provide
an initial estimate of the variability at the within- and between-
persons level, that are useful for computing the proportion variance
explained by predictors in the conditional models. For upward
social comparisons, the level 2 variance was τ200 = 7.08, the level 1
variance was σ2 = 5.70, and the raw temporal carryover parameter
was .83 (ρ = .14). For nondirectional social comparisons, the level
2 variance was τ200 = 7.24, the level 1 variance was σ2 = 7.92, and
the raw temporal carryover parameter was 1.32 (ρ = .17). For
downward social comparisons, the level 2 variance was τ200 = 9.43,
the level 1 variance was σ2 = 8.04, and the raw temporal carryover
parameter was 1.21 (ρ = .14). Finally, for depressive symptoms, the
level 2 variance was τ2 00 = 22.30, the level 1 variance was σ2 = 43.46,
and the raw temporal carryover parameter was 9.86 (ρ = .23). Intra-
class correlations (ICC) ranged from .48 to .55 across the 3 mediators
variables, suggesting that the total variance in social comparisons
was approximately equally distributed between levels 1 (within-
persons) and 2 (between-persons). In contrast, the ICC of .34 for
depressive symptoms suggests that the majority of variability in
depressive symptomology was within-persons.
Mediation Model for Facebook Time. H1-H4 and H5-H8 were
components of the moderated mediation hypotheses and each
separate hypothesis represented a mediational pathway. H1
states that the daily amount of time spent on Facebook would be
positively related to depressive symptoms at the daily level (the
c path). Facebook time was entered as a predictor of depressive
symptomology (Level 1), but the level 1 coefﬁcient failed to reach
signiﬁcance, γ101 = .064, t (151) = .48, p = .63. Thus, time on Facebook
appeared to be unrelated to depressive symptoms. However, Kenny
and colleagues (Kenny, Kashy, & Bolger, 1998) point out that a
signiﬁcant association between the predictor and outcome variables
is not necessary to establish an indirect effect from the predictor
to the outcome via mediating variables. In fact, the statistical test
of this is often underpowered relative to the tests for the a- and
b-paths, as well as the test of the indirect effects.
In H2, we predicted that time spent on Facebook would be related
to daily Facebook social comparisons (the a paths). As reﬂected in
equations 1–5, a multilevel regression model was speciﬁed in which
both the Level 1 and Level 2 (aggregate) time on Facebook (FBtime)
along with the gender main effect and interaction terms were en-
FACEBOOK USAGE 721
tered as predictors into the equations with daily Facebook social
comparisons (Level 1) as the criterion variables. None of the gender
main-effects or interactions reached signiﬁcance, so these predictors
were dropped to increase the precision of the FBtime coefﬁcients
(Snjiders & Bosker, 2012) and the models were re-estimated. Signiﬁ-
cant level 1 effects emerged across all three models, as illustrated in
Figure 2A. Time spent on Facebook was positively related to both
upward, γ10 = .145, t(151) = 2.94, p < .01, and nondirectional social
comparisons, γ30 = .347, t (151) = 6.12, p < . 01, suggesting that indi-
viduals experienced more frequent upward and nondirectional so-
cial comparisons on days when they spent more time of Facebook. A
signiﬁcant level 1 coefﬁcient was also observed for FBtime predict-
ing downward social comparisons, but in the opposite direction,
γ50 = -.249, t (151) = -4.27, p < .01. Comparing the level 1 variance
estimates from the null models to the conditional models revealed
that FBtime explained some degree of variability in daily reports
of upward (σ2cond = 5.32, pseudo R2 = .07), nondirectional (σ2cond =
7.36, pseudo R2 = .07), and downward (σ2cond = 7.53, pseudo R2 = .06)
social comparisons. Finally, all of the random slope variances were
signiﬁcant (all ps < .01), suggesting that the magnitude of these a-
paths varied across participants. On the whole, these ﬁndings sug-
gest that on days where individuals reported spending more time
on Facebook, they tended to report engaging in more nondirection-
al and upward Facebook social comparisons and fewer downward
In the second half of our mediational model (b-paths), the rela-
tionship between upward, nondirectional, and downward social
comparisons and depressive symptomology was examined while
controlling for daily time spent on Facebook (i.e., equations 6–8).
Analyses revealed signiﬁcant positive associations between daily
depressive symptoms and upward, γ70 = .612, t(137) = 7.25, p < .01,
nondirectional, γ80 = .183, t(129) = 2.66, p < .01, and downward γ90
= .402, t(128) = 5.15, p < .01 social comparisons. In contrast to the
previous model, only upward and downward social comparisons
regression coefﬁcients exhibited heterogeneity across participants,
as evidenced by signiﬁcant random slope variances (both ps < .01).
Collectively, the social comparisons predictors explained a notewor-
thy proportion of variability in daily depressive symptoms (σ2cond =
37.34, pseudo R2 = .14).
As in Study 1, the indirect effect of FBtime on depressive symp-
toms via social comparisons was assessed using the test of the ab
722 STEERS ET AL.
products. Unlike normal regression, the ab products in multilevel
modeling are not equivalent to the c-c’ estimates, but rather they
represent an exclusive mediated effect; however, the Sobel approach
for testing the signiﬁcance of the ab products (Sobel, 1982) remains a
valid approach (Krull & MacKinnon, 1999). There was a signiﬁcant
indirect effect from time on Facebook to depressive symptoms via
upward (Z = 2.72, p < .01), nondirectional (Z = 2.44, p < .01), and
downward social comparisons (Z = 2.62, p < .01).
Mediation Model for Facebook Logins. H5–H8 also predicted
moderated mediation (similar to H1–H4) but with Facebook views/
logins as the predictor and gender as the moderator. As previously
mentioned, the only difference between the two models is that
number of logins on Facebook served as the predictor variable
rather than the amount of time spent on Facebook. Thus, all analyses
to test the mediation model with Facebook logins as the predictor
were identical to those for Hypotheses 1–4. Moreover, the pattern
of results, as illustrated in Figure 2b, was identical to the previous
model. Sobel tests also conﬁrmed the presence of signiﬁcant
indirect effects from Facebook login frequency to daily depressive
symptoms, via upward (Z = 2.71, p < .01), nondirectional (Z = 2.28, p
< .05), and downward (Z = 2.81, p < .01) social comparisons.
Alternative Models. As with study 1, alternative (reverse causation)
models were examined to evaluate the possibility that people with
depressive symptoms were more likely to spend more time on
Facebook making social comparisons. The a-paths leading from
depressive symptoms to all social comparison measures were
signiﬁcant or marginal (all ps < .07). However, only two of the
possible six b-paths leading from social comparisons to Facebook
usage were signiﬁcant or marginally so. These two paths included
nondirectional social comparisons to time spent on Facebook, γ =
.078, t(137) = 5.27, p < .01, and to Facebook logins, γ = .074, t(137)
= 1.86, p < .07. The remaining four b-paths from upward and
downward social comparisons to Facebook time and logins were all
nonsigniﬁcant (all ps > .12). These alternative models also suggest
that of the six reversed indirect effects, only one was marginally
signiﬁcant (nondirectional social comparisons, Z = 1.89, p < .06, to
Facebook time). All ﬁve of the remaining indirect effects failed to
reach signiﬁcance. These ﬁndings provide additional evidence for
our hypothesized process model.
FACEBOOK USAGE 723
In sum, most of the hypotheses for Study 2 were supported, with
the exception of the gender hypothesis. After controlling for the dif-
ferent types of social comparisons across all participants, all three
types of social comparisons (upward, downward, and nondirec-
tional) were uniquely found to be signiﬁcant mediators of the re-
lationship between time on Facebook and depressive symptoms.
These results were consistent with the ﬁndings from Study 1 which
utilized a nondirectional social comparison measure. Thus, overall
results revealed that spending a great deal of time on Facebook (or
viewing Facebook more frequently) is positively related to compar-
ing one’s self to others, which in turn is associated with increased
Moreover, in Study 2, we found an identical pattern of all three
types of Facebook social comparisons uniquely serving as signiﬁ-
cant mediators between frequency of viewing Facebook and de-
pressive symptoms. These results provide further evidence for our
original hypotheses. That is, frequently viewing Facebook appears
to be functionally equivalent to spending greater amounts of time
on Facebook. Perhaps, more Facebook views and/or spending a
greater amount of time on Facebook on a daily basis both allow
participants greater opportunity to spontaneously socially compare
themselves to their peers, which in turn is associated with an in-
crease in daily depressive symptoms.
There are several factors that may be contributing to this consis-
tent mediated effect across the two studies. Previous research has
found that people often display their idealized or hoped-for possi-
ble selves on Facebook, through various modes of identity construc-
tion (e.g., posts, pictures, status updates; Zhao, Grasmuck, & Mar-
tin, 2008). That is, many individuals on Facebook may be sharing
only positive and/or self-enhancing news but not fully disclosing
their daily struggles in order to appear more socially desirable. Al-
though these Facebook self-presentations appear to have a positive
effect on the subjective well-being of those constructing their online
identities (Kim & Lee, 2011), frequently viewing these portrayals
may intensify other people’s negative cognitions behind the scenes.
This may be due to the fact that people often think they are alone
in feeling negative emotions (Jordan et al., 2011). This emotional
pluralistic ignorance combined with Facebook social comparisons
724 STEERS ET AL.
based upon their friend’s highlight reels, could potentially provoke
or exacerbate negative emotions and cognitions, and thus, contrib-
ute to greater depressive symptoms. Moreover, this positive asso-
ciation between Facebook social comparison and depressive symp-
toms was consistent for all three types of social comparison (b path).
The only major difference between how the different types of so-
cial comparison function was that the relationship between time on
Facebook and Facebook logins were both signiﬁcantly and nega-
tively associated with making downward social comparisons (a
path). That is, on days that individuals spent more time on Facebook
(or frequently viewed Facebook), they tended to make fewer down-
ward social comparisons (e.g., feel they are more accomplished than
their Facebook peers). By contrast, the relationship between time
on Facebook (and Facebook logins) was signiﬁcantly and positively
associated with upward and nondirectional social comparisons (a
path). Due to the fact that we controlled for the other types of social
comparison, results revealed that participants may have been com-
paring themselves to others (nondirectional social comparisons)
and/or perhaps feeling inferior to their peers (upward social com-
Nevertheless, as expected, daily downward social comparisons
were still positively associated with daily depressive symptoms (b
path). This result might be surprising given the literature suggest-
ing that downward social comparisons are often linked to positive
effects (e.g., Allan & Gilbert, 1995; Amoroso & Walters, 1969; Wills,
1981). However, our ﬁndings are consistent with other literature
suggesting that engaging in frequent social comparisons of any
kind may be deleterious to one’s mental well-being (White et al.,
2006). Furthermore, our results may differ from previous studies in
that we controlled for the other two types of social comparisons in
Our study also provides evidence that engaging in downward
social comparisons may be indicative of defensiveness. Consis-
tent with previous research, individual differences, such as low
self-esteem, may be moderating whether downward social com-
parisons elicit negative affect (Buunk et al., 1990). That is, perhaps
individuals with low self-esteem might be engaging in downward
social comparisons on Facebook in order to improve or bolster their
self-worth (a defensive mechanism); however, after doing so, they
actually feel worse. Thus, participants who make any type of so-
FACEBOOK USAGE 725
cial comparisons on Facebook on a given day appeared more de-
pressed. This is supported by the result that nondirectional social
comparisons predicted depressive symptoms in both studies with
two different samples. Future research should explore additional
moderators of this relationship.
Although it did not affect the results of the mediational analysis, a
notable difference between the two studies was that the relationship
between time on Facebook and depressive symptoms was not sig-
niﬁcant in Study 2. In fact, depressive symptoms were uncorrelated
with time on Facebook at the daily diary level (see Table 2) whereas
these two variables were signiﬁcantly correlated (r = .32 p < .01 for
females; r = .57 p < .01 for males) in Study 1 (see Table 1). Given the
fact that two of the mediators have positive effects (upward and
nondirectional social comparisons) and that the other mediator has
a negative effect on the a path (downward social comparison), this
may account for why the total effect for the relationship between
time on Facebook and depressive symptoms was nonsigniﬁcant.
The second major distinction between the two studies was that
gender was not found to be a moderator at the within-persons level
for Study 2. In Study 1, results demonstrated that making nondi-
rectional social comparisons on Facebook mediated the association
between time spent on Facebook and depressive symptoms for men
only. However, because there were fewer men than women who
participated, and the ones who did reported on average a mild level
of depressive symptoms, the gender differences for Study 1 may
not be generalizable. Diary methodology is generally considered to
be a more precise representation of everyday behavior than cross-
sectional studies due to a decrease in retrospective bias.
Finally, in both studies, we tested the possibility that the predic-
tor (time on Facebook) and the outcome (depressive symptoms)
might be reversed. That is, highly depressed people might spend
more time on Facebook, perhaps in an effort to bond with others,
and therefore, might be more prone to making social comparisons.
However, after testing this alternative model in Study 1, we found
that the mediated effect of nondirectional social comparison was
nonsigniﬁcant. Likewise, in Study 2, the reversed indirect effect was
marginally signiﬁcant for nondirectional social comparisons with
Facebook time as the outcome variable only (but not for Facebook
logins as the outcome). Moreover, upward and downward social
comparison indirect effects failed to reach signiﬁcance. These results
render additional support for our hypothesized process model that
726 STEERS ET AL.
people who spend more time on Facebook on a daily basis people
are more likely to compare themselves to others and in turn report
greater daily depressive symptoms (regardless of gender).
Limitations and Future Directions
In light of the current studies’ strengths, there are several limita-
tions which need to be examined. One major limitation of both
studies was that they were correlational. Therefore, causality can-
not be inferred with as much conﬁdence as experiments. Addition-
ally, extraneous inﬂuences on participants could not be controlled
because participants were ﬁlling out reports online at their leisure
rather than in the laboratory. Thus, potential third-variable con-
founds may be evident. For instance, other daily events may have
impacted participants’ Facebook usage and/or their reports of de-
pressive symptoms on a given day. Future research should incor-
porate open-ended questions in order to control for these possible
Another potential limitation is that the consequences of social
comparison were included in the items measuring downward and
upward social comparison items (i.e., Today, when I was on Face-
book, I believed that I had accomplished more than other people
had.) in Study 2. Upward and downward social comparisons are,
by deﬁnition, feeling better than or worse than others with whom
one compares one’s self to, and thus there could be some affective
component embedded in the directional items that overlaps with
depressive symptoms. However, if the affective component of the
directional items were driving our results, we would expect that
upward social comparison would be positively related to depres-
sive symptoms, downward social comparison would be negatively
related to depressive symptoms, and nondirectional social compari-
son would be unrelated to depressive symptoms. On the contrary,
our results indicate that any kind of social comparison (upward,
downward, and nondirectional) was related to depressive symp-
toms regardless of the direction.
As previously mentioned, other studies have found that excessive
internet use is associated with depressive symptoms (e.g., Morrison
& Gore, 2010). Although the design of the present study did not
examine general internet use, it is possible that the ﬁndings may be
FACEBOOK USAGE 727
attributable to excessive internet usage (e.g., turning to the internet
to alleviate loneliness). Follow-up studies could easily address this
inferential limitation by obtaining an independent measure of over-
all internet usage.
Along these lines, another major limitation of both studies was
that we did not gain access to participants’ Facebook accounts but
rather both studies solely relied on self-reports. Accessing partici-
pants’ Facebook accounts might provide greater speciﬁcity as to
what types of Facebook postings (pictures, status updates, etc.)
provoke Facebook social comparisons and offer an explanation as
to why such activities elicit comparisons. Furthermore, self-reports
might be subject to biases, possibly by participants’ level of depres-
sive symptoms. Future studies might consider requesting that par-
ticipants use PDAs to access Facebook in order to gain a more ob-
jective measure of time spent on Facebook, frequency of Facebook
logins, and to more accurately monitor participants’ daily Facebook
Despite the aforementioned limitations, the diary design is still
considered to be more discriminating (and accurate) than a one-time
report on general frequencies of Facebook use and general feelings
about Facebook experiences. Furthermore, a major strength of the
diary design is that it allows for inclusion of Level 2 aggregates (the
between-persons effects) to parse out within- and between-effects.
Thus, it was possible to examine the pure within-persons effect to
see how engaging in daily social comparisons may be inﬂuencing
depressive symptoms over time. Moreover, because social compari-
sons cannot be assessed by merely examining a participant’s Face-
book page, diary methodology provides the most practical way to
measure participants’ Facebook social comparisons.
Finally, another possible limitation is that participants were
pooled from a college population. Because college students are
transitioning from living under the rules and guidance of their
parents to becoming more inﬂuenced by peers, they may be more
susceptible to Facebook social comparison. Thus, the results of this
study may not be generalizable to older or younger Facebook users.
Future research should explore these populations of Facebook users
to see if Facebook social comparison inﬂuences differing age groups
in the same manner. Moreover, although Study 2 utilized a diary
design, we only assessed participants’ responses over a two-week
728 STEERS ET AL.
period. Future longitudinal studies could investigate whether the
relationships between Facebook usage, Facebook social compari-
son, and well-being remain relatively stable or change signiﬁcantly
McLuhan (1964) may be correct in his assertion that the medium is
the message. That is, new media come with central, obvious mes-
sages but also often hold unforeseen, deleterious consequences.
Facebook’s message is clear—it is a medium designed to connect
people to one another. However, the negative health outcomes as-
sociated with Facebook use may not be inherent to the platform, but
rather are unintended consequences related to how people choose
to use this medium. Speciﬁcally, certain individuals may be more
susceptible to comparing themselves to others’ Facebook highlight
reels on dimensions they feel are personally relevant, whereas other
people viewing the same information may not respond in the same
A major contribution of the present research is that it provides
evidence that computer-mediated interactions on Facebook may
indeed negatively impact users’ psychological health. Moreover,
these studies found that spending more time on Facebook and/or
viewing Facebook more frequently, provides people with the op-
portunity to spontaneously engage in Facebook social comparisons
(of any kind), which in turn, is associated with greater depressive
symptoms. This pattern of higher depressive symptoms after en-
gaging in Facebook social comparisons may be especially true for
college students since they may still be struggling to establish their
identities apart from their families, and, consequently, may be more
susceptible to peer inﬂuences. Thus, the current research holds
important implications for general populations and, in particular,
college students who are depressed and might also be addicted to
Facebook. Future interventions might target the reduction of Face-
book use among those at risk for depression.
FACEBOOK USAGE 729
Allan, S., & Gilbert, P. (1995). A social comparison scale: Psychometric properties
and relationship to psychopathology. Personality and Individual Differences, 19,
Amoroso, D. M., & Walters, R. H. (1969). Effects of anxiety and socially mediated
anxiety reduction on paired-associate learning. Journal of Personality and Social
Psychology, 11, 388–396.
Bauer, D. J., Preacher, K. J., & Gil, K. M. (2006). Conceptualizing and testing random
indirect effects and moderated mediation in multilevel models: New proce-
dures and recommendations. Psychological Methods, 11, 142–163.
Buunk, B. P., Taylor, S. E., Dakof, G. A., Collins, R. L., & VanYperen, N. W. (1990). The
affective consequences of social comparison: Either direction has its ups and
downs. Journal of Personality & Social Psychology, 59, 1238–1249.
Chou, H.-T. G., & Edge, N. (2012). “They are happier and having better lives than I
am”: The impact of using Facebook on perceptions of others’ lives. CyberPsy-
chology, Behavior & Social Networking, 15, 117–121.
Davis, N., & Angelova, K. (2011, September 26). Time spent on Facebook is growing
at an astonishing rate. Business Insider. Retrieved from http://www.busines-
Ensel, W. (1986). Measuring depression: The CES-D scale, social support, life events
and depression. In N. Lin, A. Dean, & W. Ensel (Eds.), Social support, life events,
and depression (pp. 51–70). New York: Academic Press.
Facebook. (2010). Statistics. Retrieved from http://newsroom.fb.com/content/de-
Facebook. (2012). Statistics. Retrieved from http://newsroom.fb.com/content/de-
Festinger, L. (1954). A theory of social comparison processes. Human Relations, 7,
Gibbons, F. X. (1986). Social comparison and depression: Company’s effect on mis-
ery. Journal of Personality and Social Psychology, 51, 140–148.
Gibbons, F., & Buunk, B. (1999). Individual differences in social comparison: Devel-
opment of a scale of social comparison orientation. Journal of Personality and
Social Psychology, 76, 129 –142.
Gilbert, P., Allan, S., Brough, S., Melley, S., & Miles, J.N.V. (2002). Relationship of
anhedonia and anxiety to social rank, defeat, and entrapment. Journal of Af-
fective Disorders, 71, 141–151.
Goethals, G., Messick, D., & Allison, S. (1991). The uniqueness bias: Studies of con-
structive social comparison. In J. Suls & T. A. Wills (Eds.), Social comparison re-
search: Contemporary theory and research (pp. 149–176). Hillsdale, NJ: Lawrence
Gross, R., & Acquisti, A. (2005). Information revelation and privacy in online social net-
works. In Proceedings of the 12th Annual Workshop on Privacy in the Elec-
tronic Society. New York: ACM Press.
Jordan, A. H., Monin, B., Dweck, C. S., Lovett, B. J., John, O. P., & Gross, J. J. (2011).
Misery has more company than people think: Underestimating the preva-
lence of others’ negative emotions. Personality and Social Psychology Bulletin,
730 STEERS ET AL.
Kenny, D. A., Kashy, D. A., & Bolger, N. (1998). Data analysis in social psychology. In
D. T. Gilbert, S. T. Fiske, & G. Lindzey (Eds.), The handbook of social psychology,
Vols. 1 and 2 (4th ed., pp. 233–265). New York: McGraw-Hill.
Kim, J., & Lee, J.-E.R. (2011). The Facebook paths to happiness: Effects of the number
of Facebook friends and self-presentation on subjective well-being. Cyberpsy-
chology, Behavior, and Social Networking, 14, 359–364.
Krull, J. L., & MacKinnon, D. P. (1999). Multilevel mediation modeling in group-
based intervention studies. Evaluation Review, 23, 418–444.
MacKinnon, D. P., Fairchild, A. J., & Fritz, M. S. (2007). Mediation analysis. Annual
Review of Psychology, 58, 593–614.
MacKinnon, D. P., Lockwood, C. M., Hoffman, J. M., West, S. G., & Sheets, V. (2002).
A comparison of methods to test mediation and other intervening variable
effects. Psychological Methods, 7, 83–104.
Major, B., & Forcey, B. (1985). Social comparisons and pay evaluations: Preferences
for same-sex and same-job wage comparisons. Journal of Experimental Social
Psychology, 21, 393–405.
Martinot, D., & Redersdorff, S. (2003). Impact of comparisons with out-group mem-
bers on women’s self-esteem: Role of the stereotypical connotation of the per-
formance context. International Journal of Psychology, 38, 348–358.
Martinot, D., Redersdorff, S., Guimond, S., & Dif, S. (2002). Ingroup versus out-
group comparisons and self-esteem: The role of group status and ingroup
identiﬁcation. Personality and Social Psychology Bulletin, 28, 1586–1600.
McKenna, K.Y.A., & Bargh, J. A. (2000). Plan 9 from cyberspace: The implications
of the Internet for personality and social psychology. Personality and Social
Psychology Review, 4, 57–75.
McLuhan, M. (1964). Understanding media: The extensions of man. New York: Mc-
Mehta, P. D., & Neale, M. C. (2005). People are variables too: Multilevel structural
equations modeling. Psychological Methods, 10, 259–284.
Moreno, M. A., Jelenchick, L. A., Egan, K. G., Cox, E., Young, H., Gannon, K. E., &
Becker, T. (2011). Feeling bad on Facebook: Depression disclosures by college
students on a social networking site. Depression and Anxiety, 28, 447–455.
Morrison, C. M., & Gore, H. (2010). The relationship between excessive internet
use and depression: A questionnaire-based study of 1,319 young people and
adults. Psychopathology, 43, 121–126.
Muthén, L. K., & Muthén, B. O. (1998-2012). Mplus user’s guide (7th ed.). Los Ange-
les: Muthén & Muthén.
Piccinelli, M., & Wilkinson, G. (2000). Gender differences in depression. Critical re-
view. British Journal of Psychiatry, 177, 486–492.
Raacke, J., & Bonds-Raacke, J. (2008). MySpace and Facebook: Applying the uses
and gratiﬁcations theory to exploring friend-networking sites. CyberPsychol-
ogy & Behavior, 11, 169–174.
Radloff, L. (1977). Center for Epidemiologic Studies Depression Scale. Applied Psy-
chological Measurement, 1, 385–401.
Raudenbush, S. W. (1995). Maximum likelihood estimation for unbalanced mul-
tilevel covariance structure models via the EM algorithm. British Journal of
Mathematical and Statistical Psychology, 48, 359–370.
FACEBOOK USAGE 731
Redersdorff, S. (2002). Strategies de protection de l’estime de soi dans un context de com-
paraison intergroupe: une approache experimentale du role du statut du groupe. Un-
published Doctoral Dissertation, Laboratoire de Psychologie Sociale et Cog-
nitive, Universite Blasé Pascal, Clermont-Ferrand.
SAS Institute. (2011). SAS documentation, Version 9.3. Cary, NC: SAS Publications.
Shrout, P. E., & Bolger, N. (2002). Mediation in experimental and nonexperimen-
tal studies: New procedures and recommendations. Psychological Methods, 7,
Snjiders, T.A.B., & Bosker, R. J. (2012). Multilevel analysis: An introduction to basic and
advanced multilevel modeling (2nd Ed.). Thousand Oaks, CA: Sage.
Sobel, M. E. (1982). Asymptotic intervals for indirect effects in structural equations
models. In S. Leinhart (Ed.), Sociological methodology 1982 (pp. 290–312). San
Steers, M.-L. N., Øverup, C. S., Brunson, J. A., & Acitelli, L. K. (2014). Love online:
How relationship awareness on Facebook relates to relationship quality. Manuscript
submitted for publication.
Suls, J., Gaes, G., & Gastorf, J. W. (1979). Evaluating a sex-related ability: Compari-
son with same-, opposite-, and combined-sex norms. Journal of Research in
Personality, 13, 294–304.
Tesser, A., Millar, M., & Moore, J. (2000). Some affective consequences of social com-
parison and reﬂection processes: The pain and pleasure of being close. In E. T.
Higgins & A. W. Kruglanski (Eds.), Motivational science: Social and personality
perspectives (pp. 60–75). New York: Psychology Press.
Toma, C. L., & Hancock, J. T. (2013). Self-Afﬁrmation underlies Facebook use. Per-
sonality and Social Psychology Bulletin, 39, 321–331.
Troop, N. A., Allan, S., Treasure, J. L., & Katzman, M. (2003). Social comparison
and submissive behaviour in eating disorder patients. Psychology and Psycho-
therapy, 76, 237–249.
West, S. G., Ryu, E., Kwok, O., & Cham, H. (2011). Multilevel modeling: Current and
future application s in personality research. Journal of Personality, 79, 2–50.
White, J., Langer, E., Yariv, L., & Welch, J.I.V. (2006). Frequent social comparisons
and destructive emotions and behaviors: The dark side of social compari-
sons. Journal of Adult Development, 13, 36–44.
Wickham, R. E., & Knee, C. R. (2013). Examining temporal processes in diary stud-
ies. Personality and Social Psychology Bulletin, 39, 1184–1198.
Wills, T. A. (1981). Downward comparison principles in social psychology. Psycho-
logical Bulletin, 90, 245–271.
Zhao, S., Grasmuck, S., & Martin, J. (2008). Identity construction on Facebook: Digi-
tal empowerment in anchored relationships. Computers in Human Behavior,
Zich, J. M., Attkisson, C. C., & Greenﬁeld, T. K. (1990). Screening for depression in
primary care clinics: The CES-D and the BDI. International Journal of Psychiatry
in Medicine, 20, 259–277.