Computers in Human Behavior 61 (2016) 271e279

Contents lists available at ScienceDirect

Computers in Human Behavior

journal homepage: www.elsevier.com/locate/comphumbeh

Full length article

Facebook use depending on age: The influence of social comparisons*

Phillip Ozimek*, Hans-Werner Bierhoff
Ruhr-Unversity Bochum, Department of Psychology, Germany

a r t i c l e i n f o a b s t r a c t

Article history: Facebook use is analyzed depending on age and occurrence of social comparisons. The hypotheses state
Received 27 October 2015 that age is negatively associated with Facebook use and that this association is mediated by social
Received in revised form comparison orientation. Data collection was realized via the Internet. The online-questionnaire included
15 January 2016
information on Facebook use on the basis of a behavioral report (cf., McAndrew & Jeong, 2012), an in-
Accepted 11 March 2016
Available online 21 March 2016
ventory for measuring social comparison orientation by Gibbons and Buunk (1999), and demographic
questions. Results are based on 335 participants. They confirm both the negative association between age
and frequency of Facebook activities and the mediation of this association by comparison orientation.
Keywords:
Facebook
These results are interpreted on the basis of evolutionary theory. This study offers new insights on the
Age mediating role of social comparisons in the relationship between age and Facebook use. In the discussion
Social comparison suggestions for further routes of research on the link between age, comparison orientation, and Facebook
Comparison orientation use are outlined.
Social networking sites © 2016 Elsevier Ltd. All rights reserved.

1. Introduction narcissism - compared with moderate and low scorers - reveal

Facebook, which has existed since 2004, has become the biggest
social online platform. It offers millions of users a platform for daily
communication and interaction including uploading pictures,
sharing contributions and commenting on them, interacting within
“Facebook groups”, or chatting with other users. Facebook provides
new research opportunities for social scientists. We focus on two
research questions that were suggested by Wilson, Gosling, and
Graham (2012): who is using Facebook and why do people use
Facebook?
Research shows that self-esteem is enhanced by considering
and editing one's own profile (compared with a condition of
objective self-awareness; Gonzales & Hancock, 2011) although
negative mood effects on readers of Facebook profiles also occur
(Lee, 2014). In addition, people scoring highly on narcissism ques-
tionnaires may use Facebook as a medium for their self-
aggrandizement because they are provided with the opportunity
to present an ideal profile of themselves to others (Brailovskaia &
Bierhoff, 2016; Mehdizadeh, 2010). People who score highly on
*
We would like to thank Andrea Führer for help with the data analysis and Fiona
Baer for improving the English.
* Corresponding author. Ruhr-University Bochum, Social Psychology, GAFO 03/
926, Universit€
atsstraße 150, 44801 Bochum, Germany.
E-mail addresses: phillip.ozimek@rub.de (P. Ozimek), hans.bierhoff@rub.de
(H.-W. Bierhoff).
more intimate and more detailed information including more self-
promotion in status updates (Winter et al., 2014). Finally, frequency
of Facebook use is positively related to social comparison frequency
(Lee, 2014). Therefore, the conclusion is justified that social com-
parison constitutes an important motive for Facebook use.
Another focus of research is on description in terms of de-
mographic variables of Facebook users including age and gender.
For example, Stefanone, Lackaff, and Rosen (2011) found that
women participate in larger social networks than men. In this study
we focus on the age variable and on the activities of Facebook users.
Specifically, we investigate differences in Facebook activity in-
tensity between older and younger users. Previous research in-
dicates that age and Facebook use intensity are negatively related
(McAndrew & Jeong, 2012). In the same vein, younger participants
report that they disclose more personal information on Facebook
than older participants (Christofides, Muise, & Desmarais, 2012).
This study is not only planned as a replication of results which
indicate that Facebook use intensity decreases with age, but also as
a test of a plausible explanation of this inverse association which is
based on interest in social comparison. We assume the existence of
a trajectory from age via interest in social comparison to Facebook
use intensity. The major theoretical perspective on which the
explanation of the age effect on Facebook use is based is evolu-
tionary psychology (cf. Buss, 2004; Miller, 2000). In the discussion
section we will mention practical implications of the results of our
research.

http://dx.doi.org/10.1016/j.chb.2016.03.034
0747-5632/© 2016 Elsevier Ltd. All rights reserved.
272 P. Ozimek, H.-W. Bierhoff / Computers in Human Behavior 61 (2016) 271e279
1.1. Behavior ¼ f(Age)
The age variable serves as a proxy of developmental change. In
addition, it is very popular in developmental and experimental
psychology because it is highly visible, easy to measure, and
because it accounts of substantial amounts of variance in many
dependent variables (Wohlwill, 1970, 1973). These are some of the
reasons why age is important to study.
The basic paradigm Behavior ¼ f(Age) encourages research
which typically is descriptive rather than hypothesis testing. But
the distinction between description and explanation vanishes if
underlying variables which may cause the developmental change
observed are taken into account. Because studies that employ the
paradigm Behavior ¼ f(Age) tend to be descriptive it is important to
identify intervening variables which mediate the variation of
behavior with age (Wohlwill, 1970).
From their overview of research on Facebook use Wilson et al.
(2012) derive the suggestion that Facebook use decreases with
age and that younger users have much more Facebook friends than
older users. This review was based on publications including the
year 2010. In correspondence with their suggestion McAndrew and
Jeong's (2012) research revealed an inverse relationship between
age and the amount of Facebook activity. In addition, they
contributed to a more comprehensive description of Facebook use
by distinguishing nine different Facebook activities (e.g., “seeking
personal information about others”, “photo impression manage-
ment”, and “linking”) and by developing a questionnaire for the
assessment of the frequency of these different activities. In corre-
spondence with the Behavior ¼ f(Age) paradigm they obtained
behavioral reports about the occurrence of these activities.
1.2. Social comparison processes
Social comparisons are based on the observation of others. They
frequently occur in the domains of achievement, interpersonal re-
lationships, and health. They serve the reduction of subjective
uncertainty with respect to achievements, opinions, values, and
problems (Lee, 2014). Social comparison theory which originally
was developed by Festinger (1954) assumes that similarity of the
comparison person constitutes an important criterion for the in-
formation value which is derived from social comparisons (cf.,
Bierhoff & Frey, 2011; Goethals & Darley, 1977; Mussweiler &
Strack, 2000). In addition, social comparisons elicit repercussions
on self-evaluation (Collins, 1996) and may be associated with either
positive or negative feelings. For example, Lee (2014) reported a
positive correlation between frequency of social comparison on
Facebook and the elicitation of negative feelings. More specifically,
people who often compare with others on Facebook tend to come
to the conclusion that they are relatively worse than the compari-
son person. This effect is, among other things, due to the fact that
Facebook users mainly emphasize their positive traits on their
profile pages (cf., Gonzales & Hancock, 2011). Recipients may infer
from the comparison with other Facebook users that they are less
‘perfect’ than the other users who serve as comparison persons.
Facebook provides users a platform for self-presentation
(Gonzales & Hancock, 2011). On social network sites people track
how others present themselves, what they do, and how they
interact with others (Lee, 2014). In this context it is likely that social
comparison processes are elicited. This assumption corresponds
with the suggestion of Mussweiler, Rüter, and Epstude (2006) that
social comparison processes are always triggered when people are
confronted with information about other people (see also
Mussweiler & Strack, 2000). Therefore, Facebook is likely to elicit
many social comparisons, because users constantly receive per-
sonal information, status updates, photos and news of their friends.
The aim of this study is to replicate the inverse relationship
between age and amount of Facebook activity. In addition, based on
the assumption that interest in social comparisons decreases with
age the proposition is examined that interest in social comparisons
mediates the negative correlation between age and amount of
Facebook activity.
1.3. Evolutionary psychology
Evolutionary psychology is based on Charles Darwin's hypoth-
eses about natural and sexual selection (Darwin, 1871), which were
developed further by biologists and psychologists like David Buss,
William Hamilton, Robert Trivers and Edward Wilson (cf., Buss,
2004, 2007). Principles of evolutionary psychology were applied
to many real life problems. For example, the spread of gossip may
be explained with propositions which are derived from evolu-
tionary psychology (McAndrew, Bell, & Garcia, 2007).
Furthermore, Piazza and Bering (2009) applied the theoretical
framework of evolutionary psychology to internet behavior. They
concluded that evolutionary thinking is relevant in four internet
domains: mating and sexual competition, kinship, social exchange,
and personal information management. All of these domains are
covered by behavior on online social network sites.
The fact that evolutionary psychology refers to genetic de-
terminants of behavior does not mean that environmental in-
fluences are ignored or considered as less important because
evolutionary adaptations always take place in response to envi-
ronmental challenges. In addition, what an individual does in a
specific environment is not pre-programmed by genes but the
result of a complex interplay of the shared human genetic makeup,
individual traits, social learning, and perception of the immediate
social circumstances (cf., Buss, 2004).
McAndrew and Jeong (2012) assumed that social comparison
processes have an evolutionary basis because it might be valuable
for social actors to collect information about competitors. Because
mating takes place in the context of sexual competition, it is
important to fulfill expectations of potential mates and to infer
information about own standing relative to competitors on di-
mensions, which are favorably evaluated by potential mates. This
reasoning is in accordance with Darwin's sexual selection theory
(Darwin, 1871; Miller, 2000), which focuses on adaptations, which
serve successful mating. Darwin's theory of sexual selection iden-
tifies two selection processes: intrasexual competition and inter-
sexual selection (Buss, 2007). The first process refers to rivalry
among mates of the same gender, whereas the second process re-
fers to choice preferences: preferential mate choices, which are
based on consensus of members of the opposite gender regarding
desired traits, enhance the reproduction success of mates whose
traits correspond with the preferences of the opposite gender
(Buss, 2007). We assume that individuals who focus on social
comparison have a mate advantage in the process of sexual selec-
tion because they have better chances to win the intrasexual
competition and to attract choices of the opposite gender (inter-
sexual selection).
In addition, we assume that adolescents and young adults
experience more sexual competition and selection pressures than
older adults. In correspondence with this reasoning, Miller (2007)
suggests that the cognitive processing behind mating mecha-
nisms peak in young adulthood although some mental fitness in-
dicators peak later.
In accordance with this theorizing, McAndrew and Jeong (2012)
interpreted the negative correlation between age and amount of
Facebook activity on the basis of evolutionary psychology by
assuming that a decrease in evolutionary pressure, which is pre-
sumably caused by sexual competition, leads to less interest in
 P. Ozimek, H.-W. Bierhoff / Computers in Human Behavior 61 (2016) 271e279
performing social comparisons which in turn leads to less Facebook
use.
2. Derivation of hypotheses
Due to the assumed age-related decrease in evolutionary pres-
sure, we assume that older people compare themselves less with
their peers than younger people. Because people in the thirties or
forties are likely to have found a romantic partner the pressure on
them to create conditions of reproduction success is reduced. Note
that evolutionary pressures related to intrasexual competition and
intersexual selection may continue to exert their influence because
rivals may challenge the current relationship by tactics of mate
poaching or because separation and divorce of the romantic part-
ners occur. But in general, these evolutionary pressures are likely to
be less relevant for older than younger people. On the basis of this
reasoning, we assume that interest in social comparison is the
mediating factor in the inverse relationship between age and fre-
quency of Facebook use. This assumption is supported by empirical
findings, which indicate that age is negatively correlated with fre-
quency of social comparison (McAndrew & Jeong, 2012), which in
turn is positively correlated with Facebook use intensity (Lee, 2014).
Note that McAndrew and Jeong (2012) did not empirically test this
mediation hypothesis. They only speculated that it could explain
the inverse relationship between age and frequency of Facebook
use. In the current study this hypothesis, which goes beyond
description of Facebook use depending on different age groups, is
investigated for the first time.
From this reasoning two hypotheses are derived. The first hy-
pothesis implies an inverse relationship between age and Facebook
activity (cf., McAndrew & Jeong, 2012).
H1. Respondents' age and their frequency of Facebook activities
are inversely related.
McAndrew and Jeong (2012) argued that the negative associa-
tion between age and frequency of Facebook activities, which is
postulated in H1, is mediated by interest in social comparison (see
above). Younger participants are more inclined to compare them-
selves with others and are therefore more active on Facebook than
older participants.
H2. Comparison orientation (CO) mediates the effect of age on
Facebook activity intensity (FB): Age / () CO / (þ)FB.
3. Method
3.1. Participants
The only requirement for participation was the active use of a
Facebook account. From the 335 respondents who provided com-
plete data sets 99 (29.6%) were male and 236 (70.4%) female. The
age range of participants was between 16 and 56 years. The average
age was 24.37 years (SD ¼ 6.8). The male participants were on
average somewhat older (M ¼ 26.33 years, SD ¼ 7.5) than the fe-
male participants (M ¼ 23.43 years, SD ¼ 6.01). 86.6% were in the
age range from 16 to 29 years and 80.9% were high school gradu-
ates. 73.7% were students (50.4% psychology students). 95.2% of
participants were of German nationality. Further nationalities of
the participants included Ukrainian, Italian, Russian, Croatian,
Turkish, and Moldavian. Finally, 63.3% of participants indicated that
they were more than 3 months together with a partner.
273
3.2. Material
Besides demographic data (age, highest education, nationality,
relationship status, and study subject) behavioral reports of the
frequency of Facebook use were obtained. In addition, social com-
parison orientation was measured as a personality trait. The order
of behavioral report and personality measure was randomized.
3.2.1. Facebook activity
Instead of a global behavioral assessment McAndrew and Jeong
(2012) employed differentiated behavioral reports, which distin-
guished nine behavioral domains. Each domain was represented by
several items, which were translated into German. For the German
scale all items were unified in format. For each item a frequency
assessment was obtained on 5-point Likert scales ranging from
never (1) to very often (5).
The nine behavioral domains were originally derived from re-
sults of factor analysis. Specifically, the following behavioral do-
mains were distinguished: Social Comparison (SC; e.g. “I'm looking
at other's relationship status”), Photo Activity (PA; e.g., “I'm posting
photographs”), Seeking Personal Information About Others (PI; e.g.,
“I'm looking at others' profiles”), Group Interaction (GI; e.g., “I'm
creating groups”), Photo Impression Management (PI; e.g., “ In
photos I'm posing like a model”), Social Interaction (SI; e.g., “I'm
reading posts on my own wall”), Linking (LI; e.g., “I'm posting links
or video clips on my own profile”), Photo Avoidance (PA; e.g., “My
pictures show the background and location”) and Family Activity
(FA; e.g., “ I'm looking at pages of my relatives?”).
Some indirect evidence is available which indicates that the
scales show good construct validity. For example, the hypothesis
was confirmed that females are more concerned with photograph-
related activity than males (McAndrew & Jeong, 2012). But the
authors did not report internal consistencies of their scales. In
contrast, we calculated coefficients of internal consistency and
performed a confirmatory factor analysis (CFA) using Mplus
(Muthe n & Muthe n, 2012). We employed the mean and variance-
adjusted weighted least-squares method (WLSMV), which is the
default estimator for models that contain categorical outcomes. The
WLSMV is reported to be robust against model violations (Kline,
2005). The model fit was assessed by three statistics, including
(a) the chi-square test statistic to test the hypothesis that the re-
lationships proposed in the model provide a plausible explanation
of those that exist in the data (note that most models are either
slightly misspecified or do not account for all measurement error,
when sample sizes are large, and that in such cases a nonsignificant
chi-square is rarely obtained; Bentler & Bonett, 1980), (b) the
Comparative Fit Index (CFI; an acceptable fit is inferred if the CFI is
0.90 or higher), and (c) the Root Mean Square Error of Approxi-
mation (RMSEA; an acceptable fit is inferred if the RMSEA is equal
to 0.08 or smaller). For further information regarding structure
equation modeling see Kline (2005).
In the first step we assumed that the behavioral reports are
based on nine factors in correspondence with the nine subscales
included. But CFA-results revealed that the nine-factor model is
insufficient, chi-square p < 0.05, CFI ¼ 0.805, RMSEA ¼ 0.085. Next,
by using an exploratory factor analysis, we derived a new factor
model with only three factors. Only 30 of the 34 items displayed
substantial loadings on the three factors. In the third step we
assumed that Facebook Activity constitutes a latent second order
factor, which summarizes the three factors derived from factor
analysis. A CFA revealed an acceptable fit of this model including
the latent second order factor Facebook Activity, chi-square
p < 0.05; CFI ¼ 0.907; RMSEA ¼ 0.058. Note that we had to allow
for the occurrence of dependencies among some residuals in order
to achieve an acceptable fit (sc1 with sc2; gi1 with gi3; gi1 with gig;
274 P. Ozimek, H.-W. Bierhoff / Computers in Human Behavior 61 (2016) 271e279
sc3 with sc5; fa2 with fa1 and si1 with si3; cf., Appendix A).
Thus, the German version of the Facebook activity scale consists
of 30 items, which are distributed on three factors: Watching (11
items; e.g., “I'm looking at other's relationship status”), Impressing
(6 items; e.g., “I'm struggling to decide which profile picture I
would like to post”) and Acting (13 items; e.g., “ I'm posting pho-
tographs”). The three-factor solution not only is favored by statis-
tical criteria, but it is also meaningful as a subdivision of the items
which are either related to inspecting information from others, or
related to prepare for self-presentation on Facebook, or related to
perform an activity on Facebook. Internal consistencies were
satisfactory to good for each subscale (awatching ¼ 0.832;
aimpressing ¼ 0.791; aacting ¼ 0.765). The items and their factor
loadings of the 9-factor model and the 3-factor model appear in
Appendix A and B.
3.2.2. Comparison orientation
Interest in social comparison was conceptualized as a person-
ality trait by Gibbons and Buunk (1999) who devised the Iowa-
Netherlands Comparison Orientation Measure (INCOM). It was
measured in our study by the German version of the INCOM by
Schneider and Schupp (2011). Individual differences in comparison
orientation are quite stable over time with a retest reliability of 0.60
across one year. Gibbons and Buunk (1999) emphasize that this
moderate stability over time is due to the fact that comparison
orientation is sensitive to situational demands. Note that the
INCOM correlates positively with social comparison frequency on
Facebook (Lee, 2014). In addition, other studies confirm the link
between INCOM and social comparison frequency (e.g., Buunk,
Zurriaga, Peíro, Nauta, & Gosalvez, 2005). Therefore, the predic-
tive validity of the INCOM scale as a measure of frequency of social
comparison on Facebook is high. Further evidence corroborates the
validity of the scale. For example, high scorers in comparison to low
scorers on the INCOM are more interpersonally oriented than
introspectively oriented and are more sensitive to the behavior of
others (Gibbons & Buunk, 1999).
The response scale of the 11-items questionnaire ranges from “I
do not agree” (1) to “I fully agree” (5). The items refer to general
interest in social comparison with others. The INCOM consists of
two subscales named Ability (6 items; e.g., “I often compare how I
am doing socially”) and Opinion (5 items; e.g., “I always like to know
what others in a similar situation would do”). Assessing the
viability of the two-factor model showed sufficient model fit
indices in the original study (Gibbons & Buunk, 1999). The two
subscales are highly correlated and correlate similarly with other
concepts (except negative affectivity; Gibbons & Buunk, 1999). The
internal consistency of the Dutch, German, and American versions
of the questionnaire turned out to be satisfactory in the original
study. Cultural comparisons indicated that comparison orientation
was higher among American than among Dutch respondents
(Gibbons & Buunk, 1999). In the Dutch adult sample comparison
orientation correlated slightly negatively with age (r ¼ 0.20;
Gibbons & Buunk, 1999).
We conducted a confirmatory factor analysis for examining the
validity of the two-factor structure of the INCOM in our sample
which revealed good model fit indices, chi-square p < 0.05;
CFI ¼ 0.979; RMSEA ¼ 0.075. The internal consistencies of both
subscales were satisfactory to good (aAbility ¼ 0.874;
aOpinion ¼ 0.798).
In addition, pairwise t-Tests were calculated for the statistical
examination of order effects. They indicated that no significant
order effects occurred (for CO-Ability t(333) ¼ 1.768, p > 0.05; for
CO-Opinion t(333) ¼ 0.228, p > 0.05; for FB-Watching
t(333) ¼ 1.145, p > 0.05; for FB-Impressing t(333) ¼ 1.682,
p > 0.05; for FB-Acting t(333) ¼ 0.187, p > 0.05).
4. Results
4.1. Preliminary analyses
Table 1 summarizes the descriptive statistics of the three Face-
book Activity subscales and the two INCOM subscales. The high
mean of the Facebook-Impressing scale indicates that impression
management is very popular among Facebook users. On average,
comparison orientation is moderately high.
In general, males' orientation regarding comparing abilities and
comparing opinions was lower for both scales than females'
orientation, for comparing abilities M ¼ 2.88 (SD ¼ 0.88) vs.
M ¼ 3.11 (SD ¼ 0.89), t (333) ¼ 2.157, p < 0.05, for comparing
opinions M ¼ 3.53 (SD ¼ 0.92) vs. M ¼ 3.76 (SD ¼ 0.75), t
(333) ¼ 2.444, p < 0.05. In contrast, Facebook activity subscales
did not vary systematically between males and females, p > 0.05.
Relationship status was unrelated to comparison orientation
scales, Facebook-Impressing and Facebook-Acting, p > 0.05. How-
ever, singles scored higher on Facebook-Watching than people in a
relationship, M ¼ 2.72 (SD ¼ 0.60) vs. M ¼ 2.55 (SD ¼ 0.60), t
(333) ¼ 2.565, p < 0.05.
Table 2 displays the correlation of the scales including age. From
an inspection of Table 2 it is clear that age is negatively correlated
with all domains of Facebook activity and comparison orientation.
In addition, the Facebook activity subscales correlate moder-
ately positively with each other. For example the correlation of FB-
Watching and FB-Impressing is r ¼ 0.363, p < 0.001 (see Table 2).
The comparison orientation subscales are also correlated positively.
Finally, the comparison orientation domains are positively corre-
lated with both Facebook-Watching and Facebook-Impressing, but
not with Facebook-Acting.
4.2. Examination of hypotheses
H1 refers to the negative association between age and Facebook
activity intensity, whereas H2 assumes that comparison orientation
mediates the effect of age on Facebook activity intensity. Both hy-
potheses were examined by structure equation modeling. The
program Mplus by Muthe n and Muthe n (2012) was employed. We
specified the mediation model with age as endogenous variable,
which is connected by a direct path with Facebook Activity. The
latter variable, which constitutes the dependent variable in the
model, is derived as a second order factor from Facebook-Watching,
Facebook-Impressing, and Facebook-Acting. In addition, an indirect
path from age to Facebook Activity via both comparison orientation
subscales was specified. Therefore, Comparison-Ability and
Comparison-Opinion are conceptualized as mediating variables.
We used the mean and variance-adjusted weighted least-squares
method (WLSMV) for categorical outcomes, which is robust
against violations of statistical requirements.
For the application of mediation models, Hayes (2013) outlines
three preconditions: homoscedasticity, independence of data, and
normal distribution of residuals. Because we used robust estimators
Table 1
Descriptive statistics of scales.
Measure Number of items Range of scale M SD
FB-watching 11 1e5 2.61 0.60
FB-impressing 6 1e5 4.15 0.67
FB-acting 13 1e5 2.24 0.51
CO-ability 7 1e5 3.04 0.89
CO-opinion 4 1e5 3.70 0.89
Note. M ¼ Mean, SD ¼ Standard derivation, CO ¼ Comparison Orientation,
FB ¼ Facebook.
 P. Ozimek, H.-W. Bierhoff / Computers in Human Behavior 61 (2016) 271e279 275

Table 2 the proposed path model reaches acceptable model fit indices,

Correlations between age, Facebook activity, and comparison orientation. validating the model structure specified, and both hypotheses
1 2 3 4 5 6 receive empirical support within the validated model.
Age (1) e
FB-watching (2) 0.198** e 5. Discussion
FB-impressing (3) 0.261*** 0.363*** e
FB-acting (4) 0.129* 0.416*** 0.394*** e
The aim of this study was to describe developmental change in
CO-ability (5) 0.234*** 0.219*** 0.206*** 0.069 e
CO-opinion (6) 0.170** 0.208*** 0.200*** 0.092 0.478*** e
Facebook activity intensity among users in the age range of from 18
to 55 years and to explain the developmental change by age-
Note. CO ¼ Comparison Orientation, FB ¼ Facebook. dfs ¼ 333. *p < 0.05 **
p < 0.01
*** dependent changes in interest in social comparison. We hypothe-
p < 0.001.
sized that age is inversely related to Facebook activity intensity. The
expected negative association between age and frequency of
and calculations (i.e., the WLSMV and the bootstrapping- Facebook activities was confirmed. These results replicate findings
technique) these requirements need not be fulfilled in our by McAndrew and Jeong (2012). In correspondence with our results
application. they found negative associations with age for several domains of
The structure equation model is summarized in Fig. 1. It achieves Facebook activity including social comparisons via Facebook, photo
acceptable model fit indices, chi-square test with p < 0.05; activities, group interaction, seeking personal information, and
CFI ¼ 0.920; RMSEA ¼ 0.005. Specifically, the model shows that impression management. The negative association between age
both the direct path and the indirect path are significant although and frequency of Facebook activities does occur across an age range
the direct effect of 0.245 is bigger than the indirect effect of from 16 to 56 years. The binomial effect size display (Rosenthal &
0.071. To illustrate the magnitude of the effects, the binomial ef- Rubin, 1982) indicates that the real world importance of the asso-
fect size display is informative (Rosenthal & Rubin, 1982). The direct ciation is considerably high.
effect increases the success rate from 0.3775 to 0.6225. In addition, The second hypothesis focused on explanation. It suggested that
the indirect effect increases the success rate from 0.4695 to 0.5355. comparison orientation is the mediating variable between age and
The Sobel-Z-Test reveals in correspondence with the mediation Facebook activity intensity. The statistical analyses, which were
model in Fig. 1 that the indirect effect is significant, Z (333) ¼ 2.26; based on structure equation modeling, Sobel-Z-Test, and boot-
p ¼ 0.024. Note that only partial mediation e not full mediation e strapping, confirmed the hypothesis although only partial media-
occurs because the significant direct effect does not vanish after tion was detected. McAndrew and Jeong (2012) also suggested that
including the indirect effect in the model. Furthermore, boot- older people express a lesser need for comparison with others. This
strapping with 1000 replications indicates that the confidence in- suggestion was based on the assumption that evolutionary pres-
terval of the indirect effect of 0.071 does not include zero, sure to compare with others is stronger in young adults than in
CI95 ¼ 0.15, CI95þ ¼ 0.02. Note that gender, educational back- older adults. Such an interpretation seems viable because sexual
ground and relationship status are not significant when they are selection issues including intrasexual competition and intersexual
added as covariates to the mediation model (p > 0.05). In summary, selection seem to be more pressing for younger people than older
three statistical tests indicate unanimously that the indirect effect people (cf., Miller, 2007). The results also replicate findings from
is significant. Therefore, H2 is supported by the data. the study of Lee (2014) who reported a positive correlation be-
Because the negative path from age to Facebook activity is sig- tween interest in social comparison and Facebook activity intensity.
nificant, H1 is supported by the data. In addition, the verification of According to the proposed mediation, with increasing age the
the mediation, Age / () CO / (þ) FB, supports H2. In summary, interest in social comparisons decreases and this decreased interest
in turn is related to a lower level of Facebook activity. This trajectory

Fig. 1. Structure equation model summarizing direct and indirect effects of age on Facebook Activity.
276 P. Ozimek, H.-W. Bierhoff / Computers in Human Behavior 61 (2016) 271e279
may be summarized by the formula Age / () CO / (þ) FB where
CO ¼ comparison orientation and FB ¼ Facebook use intensity. The
indirect effect, which summarizes the mediation via comparison
orientation, is quite small, although the binomial effect size display
(Rosenthal & Rubin, 1982) indicates that it is nevertheless quite
remarkable. The increase in success rate is notable because Face-
book use intensity is presumably caused by a plurality of factors
beyond comparison orientation. Therefore, the confirmation of the
second hypothesis and the magnitude of the indirect effect via
comparison orientation are quite impressive. For example, the
fulfillment of the interest in social comparison may compete with
other leisure activities in the offline world and other online activ-
ities like viewing videos on YouTube. In addition, real life data,
which were collected in this study, are less controlled than exper-
imental data, which are controlled for confounding. But real life
data promises more practical relevance than experimental data.
The evidence of McAndrew and Jeong (2012) and Lee (2014) as well
as the confirmation of our hypotheses suggest that comparison
orientation is a valid mediator between age and Facebook activity
intensity.
From a theoretical viewpoint we emphasized ideas from
evolutionary psychology as an explanation of the decrease of in-
terest in social comparison depending on age. More specifically,
processes of sexual selection including intrasexual competition and
intersexual selection, which presumably instigate interest in social
comparison, are likely to peak in young adults (Miller, 2007).
Two additional results correspond with this reasoning. The
pattern of correlations in Table 2 shows that Facebook-Impressing
displays a higher negative correlation with age than Facebook-
Watching and Facebook-Acting. Facebook-Impressing represents
self-presentation, which presumably is closely related to sexual
selection processes in terms of mate-attraction tactics and mate-
deception adaptations (Buss, 2007, p. XIV). It is likely that
younger adults more than older adults present themselves posi-
tively on Facebook in order to impress potential mates. Another
piece of evidence, which fits into evolutionary thinking, is the result
that singles more than people in relationships emphasize
Facebook-Watching. Facebook-Watching seems to be very func-
tional for singles who are looking for a mate.
The survey included a personality questionnaire, behavioral
reports, and demographic variables. The reliability and validity of
the comparison orientation questionnaire is high and it is only
marginally related to social desirability (Gibbons & Buunk, 1999).
Interest in social comparison was operationalized by the compari-
son orientation questionnaire (INCOM) that measures individual
differences with respect to the extent in which people perform
social comparisons (Gibbons & Buunk, 1999) and exhibits predic-
tive validity with respect to comparison frequency and importance
of social comparison on Facebook (Lee, 2014). In accordance with
Gibbons and Buunk (1999) we distinguished between abilities and
opinions. Both dimensions of comparison orientation may mediate
the influence of age on Facebook use intensity, but in the current
study only the association with opinions was significant.
Behavioral reports are considered as good behavior measures
(Kaiser, Byrka, & Hartig, 2010). They refer to the performance of
behavior and are therefore more direct indicators of behavior than
attitudes or behavioral intentions. The structure of the behavioral
reports about frequency of Facebook activities turned out to be
three-dimensional. The three subscales, which we derived from
psychometric analyses of the items, were labeled Watching,
Impressing, and Acting. Each subscale achieved appropriate inter-
nal consistency. The subscales were correlated moderately posi-
tively with each other. Therefore the three subscales may serve as
markers of the overall level of Facebook activity, which was
conceptualized as a second order factor. In fact, the contribution of
each subscale to the second order factor was approximately equal
indicating that each subscale is a good indicator of Facebook ac-
tivity intensity.
The practical relevance of this research is far-reaching. It in-
cludes information about effective advertisement and marketing
strategies as well as advice for potential internet users with respect
to their interest in social comparison. The empirical confirmation of
both hypotheses suggests that communication barriers may exist
between younger and older adults because they in general rely
differently on online social network sites and, more specifically,
because they differ in their interest in social comparison.
Insofar as the Facebook activity is driven by comparison orien-
tation, the more active user could be described as a high comparer
who is characterized by both high sensitivity to others and uncer-
tainty about the self (Gibbons & Buunk, 1999). Such high comparers
presumably are motivated by the need to reduce their uncertainty
and/or self-enhancement (Collins, 1996). In addition, people are
using Facebook for heightening their self-esteem even though it is
not clear whether online friendships are really capable of
increasing one's own self-esteem effectively (Ellison, Steinfield, &
Lampe, 2007; Steinfield, Ellison, & Lampe, 2008; Valkenburg,
Peter, & Schouten, 2006).
6. Limitations and future research
Males and 30e56 year olds are underrepresented in our sample.
In addition, most participants (80.9%) were high school graduates
and the portion of psychology students is substantial (50.4%).
Although it is unlikely that these sample characteristics influenced
the main results, a more balanced recruitment of participants with
special emphasis on older participants would be desirable in future
research. In general, rigorous sampling during the survey with an
emphasis on equipartition of age groups would be more
appropriate.
In our sample as well as in the Dutch sample of Gibbons and
Buunk (1999) comparison orientation correlated negatively with
age. This statement applies to all subscales of comparison orien-
tation. This pattern of results constitutes a confirmation of the first
hypothesis, which assumed a negative association between age of
participants and their Facebook use intensity.
But because the frequency assessments were made retrospec-
tively, subjective bias may distort the memory of the respondents
(Fahrenberg, Myrtek, Pawlik, & Perrez, 2007; Pohl, 2004). To reduce
this problem, ambulatory assessments may be used (Fahrenberg
et al., 2007) which is facilitated by the application of computer-
assisted data collection. Further studies might use ambulatory
assessment to complement the retrospective approach, which was
taken in this study. For example, an application for mobile devices
such as smartphones could be employed which asks participants
several times a day to complete a short questionnaire on the per-
formance of different Facebook activities and the occurrence of
specific context variables. On the basis of such a design it would be
possible to infer under what circumstances participants engage in
which Facebook activity.
Note that a statistical mediation effect not necessarily implies
true mediation in reality. From the confirmation of the second
hypothesis one can only conclude that the idea of mediation is
consistent with the given data set. The alternative hypothesis of a
reversal of independent variable and mediator is in principle viable
(Fiedler, Schott & Meiser, 2011; Hayes, 2013). But given the fact that
age is not influenced by interest in social comparison, the reversal
between independent variable and mediator is not plausible.
However, unmeasured causal variables, which are correlated with
age, might also influence Facebook use intensity. In general, age is
associated with many changes in developmental tasks. For
 P. Ozimek, H.-W. Bierhoff / Computers in Human Behavior 61 (2016) 271e279 277

example, older adults might be busier because of professional ob-
ligations than younger adults. Being busy is a causal factor, which
likely reduces Facebook use intensity. Therefore, threats to internal
validity which are related to confounding effects in correlational
studies have to be taken into account.
In addition, the causal relationship between comparison
orientation and Facebook activity intensity is open to interpreta-
tion. Because comparison orientation represents a stable person-
ality trait it is more likely that comparison orientation causes
frequency of Facebook use than vice versa. However, the issue of
causality cannot be definitely answered without experimental ev-
idence, which shows that increasing interest in social comparison
heightens Facebook activities.
To achieve this, a priming study is in preparation, which com-
pares participants primed with a high social comparison orienta-
tion with control participants regarding their need for using
Facebook, postulating that a high comparison orientation triggers
Facebook use. Two distinct priming groups are compared with a
control group: Participants will be presented images depicting
persons taken from the IAPS database (Lang, Bradley & Cuthbert,
1999). They have to rate these images regarding the depicted per-
sons' wealth (i.e., experimental group 1; material comparison
frame) or intelligence (i.e., experimental group 2; social compari-
son frame) compared to themselves on a 9-point-Likert scale. The
control group will not perform the priming task, which is covered
as a task for assessing people's evaluating competence. The
dependent variables include direct and indirect questions on
Facebook activity intensity. In sum, we hypothesize that partici-
pants express a higher interest in using Facebook after performing
the priming task than participants in the control condition. Addi-
tionally, the priming effect is expected to be stronger in the ma-
terial comparison priming group than in the social comparison
priming group.
7. Conclusions
The present study not only provided evidence of the replica-
bility of the decline of frequency of Facebook activities depending
on age, but also provided new insights into the social-psychological
processes, which lie behind this decline. One mediating factor that
we focused on was comparison orientation. Results indicated
convincingly that partial mediation is ensured. But because the
statistical mediation was only partial, it is likely that additional
mediators may be relevant.
While more than 500 studies have been published on Facebook
users (Wilson et al., 2012), the research is still in its infancy. This
study joins a series of studies on the relationships between per-
sonality characteristics and frequency of Facebook activities. It
provides new evidence on the interplay between age, social per-
sonality, and the performance of Facebook activities. Behavioral
reports were combined with comparison orientation as a person-
ality trait in order to understand better the relationship between
age and Facebook use intensity.
Appendix A. : Item structure and factor loadings of the
9-factor Facebook activity scale.

Scale Item Factor loading

Social Comparison (SC) I'm reading other's educational background 0.778

I'm reading others' work/career information 0.740
I'm reading others' interests/activities 0.611
I'm looking at other's relationship status 0.574
I'm reading others' favorite music, TV, books, quotes 0.447
Photo Activity (PA) I'm commenting on photographs 0.766
I'm posting photographs 0.750
I'm tagging/untagging photographs 0.684
I'm reading comments on photographs of others 0.589
I'm reading comments on my own photographs 0.561
Seeking personal information about others (PI) I'm looking at others' profiles 0.877
0.732
I'm looking at others' photo albums 0.746
I'm reading posts on others' walls 0.458
I'm reading others' mini-feed
Group interaction (GI) I'm creating events or invitations 0.868
I'm creating groups 0.600
I'm responding to events or invitations 0.562
I'm looking at or interacting with groups 0.318
Photo impression management (IM) I'm struggling to decide which profile picture I would like to post 0.808
In my profile picture I want to present the best of me 0.623
I'm graphically editing my profile photos 0.580
In photos I'm posing like a model 0.389
In photos I'm making faces (funny, cute, etc.) 0.354
Social interaction (SI) I'm reading private messages from others 0.808
I'm reading posts on my own wall 0.626
I'm sending private messages to others 0.623
Linking (LI) I'm looking at links (e.g. YouTube) or video clips on others' profiles 0.664
I'm posting links or video clips on my own profile 0.593
Photo avoidance (AV) My pictures are portraits (i.e. background hardly visible) 1.00
The pictures of my face are only “headshots” 0.550
My pictures show the background and location 0.378
My pictures show me in action (e.g. playing sports or working) 0.126
Family activity (FA) I'm in photographs with my family 0.895
I'm looking at pages of my relatives 0.451
278 P. Ozimek, H.-W. Bierhoff / Computers in Human Behavior 61 (2016) 271e279

Appendix B. : Item structure and factor loadings of the

3-factor Facebook activity scale.

Scale Item Factor loading

Watching (W) I'm looking at others' profiles 0.890

I'm reading posts on others' walls 0.772
I'm looking at others' photo albums 0.736
I'm reading comments on photographs of others 0.713
I'm looking at other's relationship status 0.659
I'm reading others' interests/activities 0.542
I'm reading other's educational background 0.521
I'm reading others' mini-feed 0.511
I'm reading others' work/career information 0.484
I'm reading others' favorite music, TV, books, quotes 0.415
I'm looking at pages of my relatives 0.304
Impressing (I) I'm reading comments on my own photographs 0.831
I'm struggling to decide which profile picture I would like to post 0.814
In my profile pictures I want to present the best of me 0.792
I'm reading posts on my own wall 0.748
I'm reading private messages from others 0.547
I'm sending private messages to others 0.462
Acting (A) I'm posting photographs 0.828
I'm commenting on photographs 0.794
I'm tagging/untagging photographs 0.790
In photos I'm making faces (funny, cute, etc.) 0.551
I'm graphically editing my profile photos 0.515
I'm looking at links (e.g. YouTube) or video clips on others' profiles 0.471
I'm posting links or video clips on my own profile 0.430
I'm creating events or invitations 0.399
I'm responding to events or invitations 0.365
My pictures show me in action (e.g. playing sports or working) 0.361
I'm creating groups 0.346
My pictures show the background and location 0.339
I'm in photographs with my family 0.236

References
Bentler, P. M., & Bonett, D. G. (1980). Significance tests and goodness of fit in the
analysis of covariance structures. Psychological bulletin, 88, 588.
Bierhoff, H. W., & Frey, D. (2011). Sozialpsychologie - Individuum und soziale Welt.
Go€ttingen: Hogrefe.
Brailovskaia, J., & Bierhoff, H. W. (2016). Cross-cultural narcissism on Facebook:
relationship between self-presentation, social interaction and the open and
covert narcissism on a social networking site in Germany and Russia. Computers
in Human Behavior, 55, 251e257.
Buss, D. M. (2004). In Evolutionary psychology (2nd ed.). Boston, MA: Pearson
Education.
Buss, D. M. (2007). Forward: the future of mating intelligence. In G. Geher, &
G. Miller (Eds.), Mating intelligence. Sex, relationships and the mind's reproductive
system (pp. IXeXIX). New York: Psychology Press.
Buunk, B. P., Zurriaga, R., Peíro , J. M., Nauta, A., & Gosalvez, I. (2005). Social com-
parisons at work as related to a cooperative social climate and to individual
differences in social comparison orientation. Applied Psychology, 54, 61e80.
Christofides, E., Muise, A., & Desmarais, S. (2012). Hey Mom, what's on your Face-
book? Comparing Facebook disclosure and privacy in adolescents and adults.
Social Psychological and Personality Science, 3, 48e54.
Collins, R. L. (1996). For better or worse: the impact of upward social comparison on
self-evaluation. Psychological Bulletin, 119, 70e94.
Darwin, C. (1871). The descent of men and selection in relation to sex. London:
Murray.
Ellison, N. B., Steinfield, C., & Lampe, C. (2007). The benefits of Facebook ‘friends’:
social capital and college students' use of online social network sites. Journal of
Computer-Mediated Communication, 12, 1143e1168.
Fahrenberg, J., Myrtek, M., Pawlik, K., & Perrez, M. (2007). Ambulantes assessment-
Verhalten im Alltagskontext erfassen, 22 Psychologische Rundschau, 58(1),
12e23.
Festinger, L. (1954). A theory of social comparison processes. Human Relations, 7,
117e140.
Fiedler, K., Schott, M., & Meiser, T. (2011). What mediation analysis can (not) do.
Journal of Experimental Social Psychology, 47, 1231e1236.
Gibbons, F. X., & Buunk, B. P. (1999). Individual differences in social comparison:
development of a scale of social comparison orientation. Journal of Personality
and Social Psychology, 76, 129e142.
Goethals, G. R., & Darley, J. M. (1977). Social comparison theory: an attributional
approach. In J. M. Suls, & R. M. Miller (Eds.), Social comparison processes (pp.
259e278). Washington, DC: Hemisphere.
Gonzales, A., & Hancock, J. (2011). Mirror, mirror on my Facebook Wall: effects of
exposure to Facebook on self-esteem. Cyberpsychology, Behavior, and Social
Networking, 14(1e2), 79e83.
Hayes, A. F. (2013). Introduction to mediation, moderation, and conditional process
analysis: A regression-based approach. New York: Guilford Press.
Kaiser, F. G., Byrka, K., & Hartig, T. (2010). Reviving Campbell's paradigm for attitude
research. Personality and Social Psychology Review, 14, 351e367.
Kline, R. B. (2005). In Principles and practice of structural equation modeling (2nd ed.).
New York: Guilford Press.
Lang, P. J., Bradley, M. M., & Cuthbert, B. N. (1999). International affective picture
system (IAPS): Technical manual and affective ratings. NIMH Center for the Study
of Emotion and Attention, 39e58.
Lee, S. Y. (2014). How do people compare themselves with others on social network
sites?: The case of Facebook. Computers in Human Behavior, 32, 253e260.
McAndrew, F. T., Bell, E. K., & Garcia, C. M. (2007). Who do we tell, and whom do we
tell on? Gossip as a strategy for status enhancement. Journal of Applied Social
Psychology, 37, 1562e1577.
McAndrew, F. T., & Jeong, H. S. (2012). Who does what on Facebook? Age, sex, and
relationship status as predictors of Facebook use. Computers in Human Behavior,
28, 2359e2365.
Mehdizadeh, S. (2010). Self-presentation 2.0: narcissism and self-esteem on Face-
book. Cyberpsychology, Behavior, and Social Networking, 13, 357e364.
Miller, G. F. (2000). The mating mind. How sexual choice shaped the evolution of
human nature. New York: Doubleday.
Miller, G. F. (2007). Mating intelligence: frequently asked questions. In G. Geher, &
G. Miller (Eds.), Mating intelligence. Sex, relationships and the mind's reproductive
system (pp. 367e393). New York: Psychology Press.
Mussweiler, T., Rüter, K., & Epstude, K. (2006). The why, who, and how of social
comparison: a social cognition perspective. In S. Guimond (Ed.), Social com-
parison and social psychology: Understanding cognition, intergroup relations, and
culture (pp. 33e54). New York: Cambridge University Press.
Mussweiler, T., & Strack, F. (2000). Consequences of social comparison: selective
accessibility, assimilation, and contrast. In J. Suls, & L. Wheeler (Eds.), Handbook
of social comparison (pp. 253e270). New York: Kluwer.
Muthe n, L. K., & Muthe n, B. O. (2012). Mplus version 7.0 user's guide. Los Angeles, CA:
Muthen & Muthen.
Pohl, R. F. (2004). In Cognitive illusions: A handbook on fallacies and biases in thinking,
judgement and memory. New York: Psychology Press.
Piazza, J., & Bering, J. M. (2009). Evolutionary cyber-psychology: applying an
evolutionary framework to internet behavior. Computers in Human Behavior, 25,
1258e1269.
 P. Ozimek, H.-W. Bierhoff / Computers in Human Behavior 61 (2016) 271e279
Rosenthal, R., & Rubin, D. B. (1982). A simple, general purpose display of magnitude
of experimental effect. Journal of Educational Psychology, 74, 166e169.
Schneider, S., & Schupp, J. (2011). The social comparison scale: testing the validity,
reliability, and applicability of the Iowa-Netherlands comparison orientation
measure (INCOM) on the German population. DIW Data Documentation, 360.
Stefanone, M. A., Lackaff, D., & Rosen, D. (2011). Contingencies of self-worth and
social-networking-site behavior. Cyberpsychology, Behavior, and Social
Networking, 14, 41e49.
Steinfield, C., Ellison, N. B., & Lampe, C. (2008). Social capital, self-esteem, and use of
online social network sites: a longitudinal analysis. Journal of Applied Develop-
mental Psychology, 29, 434e445.
Valkenburg, P. M., Peter, J., & Schouten, A. P. (2006). Friend networking sites and
279
their relationship to adolescents' well-being and social self-esteem. CyberP-
sychology & Behavior, 9, 584e590.
Wilson, R. E., Gosling, S. D., & Graham, L. T. (2012). A review of Facebook research in
the social sciences. Perspectives on Psychological Science, 7, 203e220.
Winter, S., Neubaum, G., Eimler, S. C., Gordon, V., Theil, J., Herrmann, J., et al. (2014).
Another brick in the Facebook walleHow personality traits relate to the content
of status updates. Computers in Human Behavior, 34, 194e202.
Wohlwill, J. F. (1970). The age variable in psychological research. Psychological Re-
view, 77, 49e64.
Wohlwill, J. F. (1973). The study of behavioral development. New York: Academic
Press.