we’re looking at research on whether or not CMC has a negative or positive effect on people. There is a popular narrative in our society that smartphones and the social media that they carry are bad for us and our children. Kindly Analyse the following articles, Find similarites and differences and real life examples of how the subject matter affects us .
See discussions, stats, and author profiles for this publication at:
Urban Studies Change in the Social Life of Urban Public Spaces
Data · November 2015
3 authors, including:
Keith Neil Hampton
Michigan State University
All content following this page was uploaded by Keith Neil Hampton on 24 November 2015.
The user has requested enhancement of the downloaded file.
Urban Studies
2015, Vol. 52(8) 1489–1504
� Urban Studies Journal Limited 2014
Reprints and permissions:
DOI: 10.1177/0042098014534905
Change in the social life of urban
public spaces: The rise of mobile
phones and women, and the decline
of aloneness over 30 years
Keith N Hampton
Rutgers University, USA
Lauren Sessions Goulet
Facebook, USA
Garrett Albanesius
University of Pennsylvania, USA
This study illustrates that over the past 30 years, Americans have become less socially isolated while
using public spaces. Based on content analysis of films from four public spaces over a 30-year period,
the behaviour and characteristics of 143,593 people were coded. The most dramatic changes in the
social life of urban public spaces have been an increase in the proportion of women and a corre-
sponding increase in the tendency for men and women to spend time together in public. Despite the
ubiquity of mobile phones, their rate of use in public is relatively small. Mobile phone users appear
less often in spaces where there are more groups, and most often in spaces where people might oth-
erwise be walking alone. This suggests that, when framed as a communication tool, mobile phone
use is associated with reduced public isolation, although it is associated with an increased likelihood
to linger and with time spent lingering in public. We argue that public spaces are an important com-
ponent of the communication system that provides exposure to diverse messages, brings people into
contact to discuss their needs and interests, and helps people recognise their commonalities and
accept their differences. The increased tendency to spend time in groups while in public contrasts
with evidence from other research that suggests a decline in American public life, and that mobile
phones have increased social isolation in public spaces. The increase in group behaviour, women and
lingering in public may have positive implications for engagement within the public sphere.
civic engagement, gender, mobile phones, public sphere, social isolation
Received October 2013; accepted April 2014
Corresponding author:
Keith Hampton, Rutgers University, Department of
Communication, 4 Huntington Street, New Brunswick,
New Jersey, 08901, USA.
at RUTGERS UNIV on April 22, 2015usj.sagepub.comDownloaded from
A number of studies in the USA have found
that people are increasingly likely to live
alone, to engage with smaller social circles,
to disengage from civic institutions and to
spend time in private spaces (Hampton et
al., 2011c; Klinenberg, 2012; Lofland, 1998;
Putnam, 2000). These shifts are often attrib-
uted to large-scale social change, such as the
movement of women into the paid labour
force, and technological change, such as the
rise of home computing, the internet and
mobile phones. These studies have primarily
examined shifts that have taken place within
institutions and private spheres of interac-
tion. However, it has generally been assumed
that these shifts have consequences for con-
tact in public spaces. One common scenario
suggests that opportunities for private
engagement lead to a withdrawal from pub-
lic life (Sennett, 1977). Technologies such as
the mobile phone may further undermine
public life by increasing the opportunity for
people to spend time in private while in pub-
lic spaces (Turkle, 2011). A shift in the social
life of urban public spaces, toward aloneness,
might have very negative consequences for
individuals and society: higher rates of lone-
liness and depression (Matias et al., 2011)
and a general decline in trust and exposure
to social diversity (Sennett, 1977). However,
to our knowledge, no study has attempted to
measure change in social interaction in pub-
lic places over time. Whether people are
more alone in public, and amongst less
diverse companions than in the past is an
open question. We explore this question with
a longitudinal study of public spaces and
change in the composition of individuals and
groups in these spaces over the past 30 years.
The public
A place where people come together, face-to-
face. The [city] center is the place for news and
gossip, for the creation of ideas, for marketing
them and swiping them, for hatching deals,
for starting parades. This is the stuff of the
public life of the city – by no means wholly
admirable, often abrasive, noisy, contentious,
without apparent purpose. But this human
congress is the genius of the place, its reason
for being, its great marginal edge. (Whyte,
2009 [1988]: 341)
Public spaces are a component of the public
sphere (Habermas, 1989). The public sphere
is where strangers meet; it stands in contrast
to the private sphere, where close relation-
ships, such as the family flourish (Sennett,
1977). Like other components of the public
sphere; such as the mass media, civic institu-
tions, and informal civil behaviours; we con-
ceptualise public spaces as an opportunity
for the exchange of messages with diverse
others. Public spaces include a city’s streets,
sidewalks, parks, and plazas to which all
persons have legal access (Lofland, 1973).
Thus, the distinguishing feature that sepa-
rates public space from private space is that
it minimises the segregation of people based
on lifestyles, such as their opinions, income,
gender and race (Strauss, 1961). One recent
study found that three visits to public spaces
per week was associated with having a net-
work of contacts one half standard deviation
higher in diversity when compared with the
average; similar in magnitude to civil and
civic behaviours, such as the difference
between knowing most versus no neigh-
bours, and the difference associated with
belonging to two different voluntary organi-
sations (Hampton et al., 2011b).
Opportunities for public engagement can
vary by individual, place and time. A place
that is public for one person, in that the pro-
portion of copresent others clearly leans
towards the unfamiliar, may simultaneously
be a private place for another who is sur-
rounded by an entourage of close friends. A
truly public space brings people from diverse
backgrounds and classes into contact (Low
et al., 2005). Although this contact can be
1490 Urban Studies 52(8)
at RUTGERS UNIV on April 22, 2015usj.sagepub.comDownloaded from
informal and fleeting, such interactions con-
trast with the homogeneity of close friend-
ship groups, which tend to minimise
opportunity for encounters with diverse oth-
ers (Lofland, 1998).
Serendipity, chance encounters and peo-
ple watching are important subsets of the
interactions that take place in public spaces.
Indeed, much of the activity that takes place
in public might be viewed as non-purposeful.
That is, people chatting informally, or hang-
ing around a place with little apparent pur-
pose. While this might be negatively
characterised as loitering, it is better
described as lingering. The urbanist William
H. Whyte (2001 [1980]) argued that public
spaces should be designed to encourage peo-
ple to linger, as it provides for conversation
and chance encounters. In one study, one in
six people interviewed across a variety of
public places reported that, in their history
of use of that place, they had met someone
new and continued that relationship to form
a long-term friendship (Hampton et al.,
2010). Whyte’s contemporaries, such as Jane
Jacobs (1961), similarly noted the role of
people who linger for the opportunities in
sidewalk life that they provide for interac-
tion and surveillance.
Nevertheless, serendipitous encounters
are the minority of all public interactions.
Public spaces are primarily a forum for
interacting with friends rather than strangers
(Demerath and Levinger, 2003). Thus, pub-
lic space can be ‘a discursive space where
individuals and groups congregate to discuss
matters of mutual interest’ (Hauser, 1998) as
well as a social and spatial semiotic (Ravelli
and Stenglin, 2008). That is, public spaces
shape public opinion by affording delibera-
tion, and through meaning-making that
results from observing the context of the
space, and the artefacts and people within.
Although contact in public spaces is likely to
be incomplete when compared with more
formal forms of political deliberation
(Fishkin, 2000), influence need not involve
persuasion, or manipulation, but can take
the form of imitation or contagion
(Hamilton, 1971). The meaning and mes-
sages contained within a public space might
act directly on an individual’s opinion, or, as
with other modes of communication in the
public sphere, fit into a multistep flow of
opinion formation (Katz and Lazarsfeld,
1955). Public spaces are an important com-
ponent of the communication system that
provides exposure to diverse messages,
brings people into contact to discuss their
needs and interests, and helps people recog-
nise their commonalities and accept their
Beyond democratising effects, the contact
that takes place in public spaces has other,
well-established benefits. Walking on public
streets in the company of others, as opposed
to walking alone, is associated with revitali-
sation and reduced levels of anxiety and
depression (Staats and Hartig, 2004). Time
spent in public spaces has been found to
increase attachment and sense of commu-
nity, lead to higher levels of perceived health,
and reduce feelings of loneliness (Cattell et
al., 2008; Kweon et al., 1998). A shift toward
higher levels of isolation while in public may
be tied to other large-scale, social trends,
such as increased treatment for depression
and anxiety disorders (Comer et al., 2011;
Marcus and Olfson, 2010) and declines in
generalised trust, empathic concern and per-
spective taking (Konrath et al., 2011; Wilkes,
The shift toward aloneness
Some studies indicate that people are more
isolated and removed from public spaces
than in the past. Interactions with social ties
may be undertaken increasingly within the
confines of private spaces (Lofland, 1998;
Popenoe, 1985). This trend is not new; priva-
tism has its roots in the rise of capitalism,
Hampton et al. 1491
at RUTGERS UNIV on April 22, 2015usj.sagepub.comDownloaded from
industrialisation and urbanisation (Tönnies,
1887; Wirth, 1938). The responsibility for
this shift is often charged to technological
change. The infrastructure of the city; water,
sewage and electric systems; highways; and
the telephone all enable a separation of
home and work that permits people to
reduce the time they spend in public. For
example, refrigerators and freezers reduce
the need for daily visits to the market; air
conditioners remove people from the stoop;
and television reduces the need to visit the
theatre (Lofland, 1998; Putnam, 2000).
When it is necessary to travel through public
space, the automobile makes it possible to
enclose oneself in a bubble of private space
(Lofland, 1973).
The rise of new, digital technologies, such
as home computing and the internet, have
similarly been tied to the ability of people to
spend leisure and work time within the con-
fines of the home (Graham and Marvin,
1996). It is easy to infer that when people
have access to technologies that afford the
opportunity to spend time in private, they
will do so. However, it has not been demon-
strated empirically that home centredness
comes at the expense of time spent with
acquaintances in public spaces. Indeed, there
is some evidence to suggest that technologies
that afford home centredness, such as the
television, enhance, rather than displace time
spent in public (Robinson, 2011). The inter-
net may also offer a new type of online pub-
lic space (Papacharissi, 2002). A number of
scholars have pointed out that new mobile
technologies, such as the mobile phone,
extend this trend in a new way by allowing
people to create a cocoon of private interac-
tion in urban public spaces, which, like the
automobile, shields them from those around
them (Hampton and Gupta, 2008; Ito et al.,
2008). The mobile phone can transform pub-
lic companions, what Goffman (1971) called
‘Withs’, into a ‘Single’ (Humphreys, 2005).
Some scholars have argued that new mobile
technologies have resulted in public spaces
that are no longer communal spaces; fewer
traditional in-person interactions in public;
and people in public spaces engaged through
technology with someone miles away rather
than with someone in the same space
(Turkle, 2011: 155). Not only may people be
spending more time alone in public, but the
availability of close social ties through
mobile devices may lead to intense participa-
tion in networks of close relationships at the
expense of exposure to diverse others
(Gergen, 2008).
A growing literature to suggest a rise in
the related concept of individualism has
accompanied evidence of a rise in privatism.
Individualism, described by de Tocqueville
in his reflections on American democracy, is
the tendency for man to ‘isolate himself
from the mass of his fellows and withdraw
into the circle of family and friends; with this
little society formed to his taste, he gladly
leaves the greater society to look after itself’
(Tocqueville, 2007 [1835]: 281). Moreover,
de Tocqueville felt that individualism was of
‘democratic origin and threatens to grow as
conditions get more equal’ (Tocqueville,
2007: 281).
Data from the US General Social Survey
(GSS) suggest that individualism in America
may be increasing. McPherson et al. (2006)
found that the core networks of Americans
– their closest circle of friends and family –
have become smaller and more closed. This
contraction has come at the expense of
diversity – the maintenance of core ties out-
side of the home. In 1985, approximately
64% of American adults reported discussing
an important matter with someone outside
of their family; by 2008, this number had
dropped to 45% (Hampton et al., 2011c).
Although some have disputed the validity of
the 2004 GSS data (Fischer, 2009), three
subsequent replications have found average
network sizes and distributions that closely
mimic the GSS (Brashears, 2011; Hampton
1492 Urban Studies 52(8)
at RUTGERS UNIV on April 22, 2015usj.sagepub.comDownloaded from
et al., 2011a, 2011c).
New technologies were
also targeted as a possible cause for this
trend, however, Hampton et al. (2011a,
2011c) largely excluded an association
between the use of the internet, mobile
phones and related technologies and smaller
or less diverse core networks. (In fact, much
the opposite seems to be true.) Whatever the
cause, this trend may have negative conse-
quences for individual social support as well
as opportunities to engage with diverse oth-
ers (Hampton and Ling, 2013).
More together
In this paper we question whether the grow-
ing literature on aloneness can be extended
from private and institutional settings to
public life? In questioning whether public
life is less diverse and more alone than in the
past, we also ask if recent technological
change is the most significant large-scale
social change to have affected public life?
The focus on technological change in the lit-
erature on privatism and individualism has
drawn attention away from other sources of
large-scale change that may have had the
same or a larger impact on interaction in
public spaces. The list of social changes that
may have affected public interactions over a
time period that coincides with the rise of
new digital technologies is long and includes
trends that the urban literature has treated
extensively. They include the privatisation
and ‘Disneyfication’ of public spaces
(Hannigan, 1998; Kohn, 2004; Zukin, 1995)
and those that have received less attention,
such as restrictions on tobacco use in the
workplaces that have pushed smokers into
public doorways and sidewalks (Kaufman
et al., 2010). However, one major social
change stands out as particularly important
– increased gender equity.
We anticipate that a shift in the public
and private lives of women has had major
implications for the use of public spaces. It
is no secret that over the last three decades,
women’s participation in the labour
force has grown sharply, whereas men’s par-
ticipation has fallen over the same period
(Inglehart, 2003; US Congress Joint
Economic Committee, 2010). In the USA,
the number of women in the workforce has
increased by 44.2% since 1984, with nearly
all growth occurring by 2000 (US Congress
Joint Economic Committee, 2010). Women
are spending much more time out of the
home than in the past (Jacobs and Gerson,
2001). This trend combines with related
trends, such as women staying in school lon-
ger (Peter and Horn, 2005), a decline in
occupational segregation (Blau et al., 2013),
an increase in the average age of marriage
and child bearing (Goldin and Katz, 2002),
and the movement away from the segrega-
tion of women’s activities into private spaces
and men’s activities into public spaces (Bott,
1955). There is little doubt that the move-
ment of women’s activities outside of the
home is one of the most significant social
changes of recent decades.
Scholars have not consistently interpreted
increased gender equity as positive for par-
ticipation in the public sphere. McPherson
et al. (2006) suggest that much of the recent
shift in core network diversity can be attrib-
uted to a tendency, as labour force equity
increases, for men to shrink the number of
non-kinship ties rather than for women to
increase the number of ties outside of the
home. Similarly, Putnam (2000) argues that
women’s increased labour force participa-
tion has reduced civic and civil behaviours
and may share responsibility for the decline
in social capital over the last half century.
Although these and other scholars have
focused on the implications of increased gen-
der equity for participation in civic institu-
tions and in private spaces, to our
knowledge no one has considered how the
recent shift in women’s activities outside the
home has affected participation in public
Hampton et al. 1493
at RUTGERS UNIV on April 22, 2015usj.sagepub.comDownloaded from
spaces. This deficit is likely based on the
assumption that most women enter their cars
at home, exit at their place of work, and do
not have meaningful opportunities to engage
with public spaces. However, this assumption
ignores the role of the street and public
spaces in general as places for walking, lin-
gering, watching and socialising. Gender is
one of many possible sources of diversity in
these spaces. Participation in public spaces is
as much a part of the public sphere as is the
mass media, civic institutions and civil beha-
viours. In contrast to speculation that the
increased participation of women in the
labour force has driven down participation
in the public sphere (McPherson et al., 2006;
Putnam, 2000), we anticipate that women
have fundamentally reshaped interactions in
public spaces over the past 30 years. Women
of the late-20th and early-21st centuries have
better access to public spaces, in America
and in most Western countries, than women
of the century before them (Bondi and
Domosh, 1998).
This paper provides the results from a longi-
tudinal study of interaction patterns in pub-
lic spaces that cover a 30-year period. It
analyses the behaviour and characteristics of
143,593 people in four public spaces, based
on the content of time-lapse films created in
1979 and 1980 and videos of the same spaces
shot between 2008 and 2010.
The time-lapse films used in this analysis
are from an archive held by the Project for
Public Spaces (PPS). PPS is a non-profit
organisation founded by Fred Kent, who
worked as a research assistant to the urba-
nist William H Whyte. Whyte and his assis-
tants used Super 8 film to inform the Street
Life Project (Whyte, 2001 [1980]), which was
started in 1968 in response to new zoning
regulations in New York City that gave
incentives to builders to include plazas and
other public spaces as part of the construc-
tion of large, commercial buildings. Whyte
used a variety of methods, including time-
lapse films, to assess variation in pedestrian
behaviour and the use of public spaces in
New York City and around the world. The
result of this work was a comprehensive
amendment to New York City zoning laws
in 1975 and, in 1980, a summary of findings
published as The Social Life of Small Urban
Spaces (Whyte, 2001 [1980]). PPS works to
apply and expand Whyte’s work.
PPS created and archived more than 3600
canisters of Super 8 film. Our team spent
more than 3000 hours digitalising and cata-
loguing these films to serve as a baseline for
comparing public life 30 years ago with
urban life today. The corpus of the film
archive was narrowed for our study, based
on the visibility of pedestrians, the duration
of film available, the consistency of camera
angles and similarity in time period. We
recognised that local, historical factors, such
as changes in neighbourhood characteristics
(e.g. crime and physical design), were likely
to influence activity in any public place.
Although it would be impossible to control
completely for these threats to historical
validity, we attempted to minimise error as a
result of external factors by sampling from a
range of locations. Four locations, all filmed
between 1979 and 1980, provided a relatively
large number of films. They were taken from
a stationary view point, using a camera angle
that provided a view of pedestrians that
would allow us to identify group activity and
some individual characteristics. The four
locations were Chestnut Street (between 10th
and 11th Streets in Philadelphia, PA),
Downtown Crossing (Boston, MA), Bryant
Park (northwest corner sidewalk in New
York City) and the steps of the Metropolitan
Museum of Art (New York City) (see online
supplement, Movie S1: Bryant Park 1980).
These four small, urban public places have
distinct characteristics:
1494 Urban Studies 52(8)
at RUTGERS UNIV on April 22, 2015usj.sagepub.comDownloaded from
Chestnut Street: Located in Center City
Philadelphia, this area is within one block of
a subway station and provides access to a
number of low-rise office towers, a hospital
and a small number of retail establishments.
There are no benches or other seating, and
there are no public parks or significant resi-
dential areas within a four-block radius.
This area might best be described as an ‘in-
between’ space, serving as a pedestrian tran-
sit point between destinations.
Downtown Crossing: A shopping district
located within Boston’s downtown, one
block east of Boston Common, and a num-
ber of blocks west of the main financial dis-
trict. Adjacent to a subway station and
closed to vehicular traffic, pedestrians can
walk freely in the streets to access a major
department store, restaurants and other
retail establishments.
Bryant Park: Located in Midtown
Manhattan, outside of the northwest corner
of the park at the corner of West 42nd and
the Avenue of Americas. Bryant Park is a
major destination. The northwest corner is
located within a block of two subway sta-
tions and provides easy access to a number
of a large office towers and retail establish-
ments. It is one block from Times Square.
Metropolitan Museum of Art: Located on 5th
Avenue on the eastern border of Central
Park in New York’s Upper East Side.
Granite steps lead to the entrance of the
museum and are positioned between two
fountains. The steps are a well-known public
place and a destination for people to meet
and eat. A ten-minute walk from the nearest
subway station and a popular destination for
people who live and work in nearby low-rise
residential, office and retail establishments.
In 2008 and 2010, we returned to these four
locations to re-film pedestrian life at compa-
rable times of day, day of the week, and in
weather conditions similar to the original
time-lapse films (see online supplement,
Movie S2: Metropolitan Museum of Art
2010). The original films were typically
obtained from the vantage point of a win-
dow or rooftop. We were not able to secure
permission to access the same filming posi-
tion, but were able to reproduce a similar
vantage point through the use of a 16-foot
cine stand (a long pole with support legs).
Our video unit was stationed outside of
pedestrian flow and camouflaged by posi-
tioning the apparatus next to a building or
light post. The video unit received little
notice from pedestrians. Security guards
located in Bryant Park and Downtown
Crossing briefly interrupted our observa-
tions to ask that we request permission from
their private management companies to set
up our video unit, which we did.
The original and new films totalled more
than 38 hours of footage. Films were
sampled at 15-second intervals for a total of
9173 observation periods. The original films
were often made using time-lapse tech-
niques. To standardise our film sample at 15
seconds of standard film speed, we took
advantage of one of Whyte’s common tech-
niques: he often recorded a stopwatch at the
start of his films before adjusting the frame
rate for time-lapse. A comparison of the film
before and after the camera was adjusted for
time-lapse allowed us to calculate the correct
sampling interval.
The coding procedure involved taking a
screen shot of the sampled video frame and
dividing the frame into a number of prede-
fined coding areas. Coding areas were stan-
dardised so that the same geographic space
was coded in both the new and original
videos. The position of our video camera
was at a slightly lower altitude than that of
the camera in the older films. As a result,
our videos typically captured a smaller geo-
graphic space than what was captured in the
original. So that the space represented by
the older films matched the location and size
of our current day videos, we cropped the
Super 8 films to match our modern vantage
Hampton et al. 1495
at RUTGERS UNIV on April 22, 2015usj.sagepub.comDownloaded from
point. In this way, we coded a space that
was consistent in size and location across
The quality of the Super 8 films as a
result of their vantage point, the characteris-
tics of the film, and their general detrition
after being warehoused for three decades
limited the amount of information that we
could reliably code. For example, it was dif-
ficult to determine personal characteristics,
such as race, other than gender. Researchers
coded individuals in each area for four char-
acteristics: gender, group size, lingering and
mobile phone use. Each variable was coded
as a dichotomy. The focus on few phenom-
ena and the use of dichotomous units simpli-
fied the coding instructions. Individuals
were coded as being members of a group if
observed walking, sitting, or standing in
close proximity of another individual. If a
coder was uncertain if an individual was part
of a group or merely in close proximity, the
coder reviewed the video immediately before
and after a sampled frame to verify that the
individuals represented a collective unit (e.g.
it was relatively easy to conclude that two
people sitting immediately next to each other
on a long bench were part of a group).
Additional indicators that an individual was
part of a collective unit included physical
touching, apparent talking and collective
locomotion. Lingering was defined as an
individual’s presence in two or more consec-
utive film samples (inhabiting the same area
for 15 seconds or more). Mobile phone use
was captured through the observation of a
handset held to an ear or mouth, or the typ-
ing of a text message.
On average, each observation of a film
frame required 13 minutes of content coding
(a total of 2000 coding hours). For consis-
tency, a single researcher was responsible for
training and accessing the reliability of all
coders. The reliability of each coder was
assessed informally during training, in a
series of pilot tests, and through subsequent
formal assessments. During training and
pilot testing, using Krippendorff’s Alpha as
an indices of reliability, pairs of coders
achieved reliability on each of the four vari-
ables .0.90. In subsequent formal assess-
ments, coders maintained reliability for
measures of gender, group size and linger-
ing, but Krippendorff’s Alpha for mobile
phone use was lower (0.75). The lower level
of reliability for mobile phone use is consis-
tent with a pattern within binary observa-
tions where one of the values is relatively
rare (mobile phone use); Krippendorff’s
Alpha is lower in this instance even when
coders made few mistakes. Indeed, the use
of mobile phones with headsets (such as
Bluetooth devices) is likely underreported,
because coders may have had difficulty
observing this activity.
We are not alone together
The results show that over the past 30 years
in the majority of public spaces, there has
been a decline in the tendency for people to
spend time alone and a corresponding
increase in the proportion of people in
groups (Table 1).
In 1979, 32% of people
who visited the steps in front of the
Metropolitan Museum (Met Steps) did so
alone. In 2010, only 24% of visitors were
alone; a percentage decline of 24% in the
presence of singletons. In Downtown
Crossing, the proportion of people observed
walking alone declined from 69% in 1979 to
53% in 2010; a change of 24%. Similarly,
the proportion of people walking alone on
the sidewalk outside of Bryant Park dropped
from 72% in 1980 to 66% in 2008, a propor-
tional change of 8%. In only one setting, on
Chestnut Street in Philadelphia, was there
an increase in the number of people walking
alone. In 1979, 66% of people on Chestnut
Street walked alone; in 2010 this had grown
1496 Urban Studies 52(8)
at RUTGERS UNIV on April 22, 2015usj.sagepub.comDownloaded from












Hampton et al. 1497
at RUTGERS UNIV on April 22, 2015usj.sagepub.comDownloaded from
to 74%, an increase of 12% in the propor-
tion of people who were alone.
The finding of increased group activity in
three of the four field sites is a good indica-
tor of change in the direction of reduced
activity spent alone in public spaces. At the
very least, it refutes the counter hypothesis
that there has been a large, widespread
social change in favour of people spending
time alone in public spaces. We have no
definitive explanation for the increase in the
proportion of singletons on Chestnut Street.
However, we can infer that it is related to
the unique character of the space as a transi-
tion point between destinations. This section
of Chestnut Street lacks much of the diver-
sity in leisure and commercial activity pres-
ent in the other three spaces. If, pedestrian
traffic in this area is disproportionately
accounted for by transits to and from the
workplace, the observations of Whyte (2001
[1980]) suggest that such public spaces are
less likely to provide opportunities for social
engagement. A closer examination reveals
that the finding may be tied to a large, social
trend: the presence of women accounts com-
pletely for the increase in people walking
alone on Chestnut Street. Of those who were
alone, there was a 9.70% increase in the pro-
portion of women, whereas there was a
6.54% decrease in the proportion of men.
Gender equity in public spaces
A substantial shift in the composition of
groups and the presence of women accompa-
nied the decline in the number of people
walking alone in public. Men and women
increasingly socialised together in public,
and women represented a larger proportion
of people in public. At the Met Steps, the
proportion of women increased by 33%; in
Bryant Park by 18%; and on Chestnut Street
by 2%. Only in Downtown Crossing was
there a decline in the proportion of women –
a decline of 15%. The proportion of dyadic
groups that were homophilious based on
gender declined by 21% at the Met Steps.
There was a similar decline in group homo-
phily of 23% in Bryant Park and a decline of
31% on Chestnut Street. The only place
where groups became less diverse was in
Downtown Crossing, where same-sex dyads
increased by 15%.
Equalisation in participation in public
spaces has accompanied changes in women’s
participation in the labour force. Women
represent a larger proportion of people
observed in public in three of the four field
sites. That the proportion of women in
Downtown Crossing has decreased over time
is somewhat surprising. First, it is counter to
the broader trend toward greater equity in
public, although, with the exception of the
Met Steps, men are still the dominant pres-
ence in all the observed public spaces.
Second, the decrease is counter to expecta-
tions that shopping is a ‘feminine’ activity,
that women would have a greater presence
in a public space that is dominated by retail
opportunities (Falk and Campbell, 1997),
and that as women’s incomes have increased,
they would increasingly be responsible for
purchasing decisions. However, this outlier
may also be related to equity. Although
there was an overall increase in same-sex
dyads within Downtown Crossing, male-
only dyads increased by 51%, whereas the
presence of female-only dyads declined by
329%. The decline of women within this set-
ting could be interpreted as a shift in gender
roles, consistent with other reports (Otnes
and McGrath, 2001), that men may increas-
ingly be taking on an activity that was tradi-
tionally regarded as feminine. The 27% drop
in men who are alone at Downtown
Crossing, compared with the 19% drop in
the percent of women, may suggest that men
no longer view shopping as more instrumen-
tal than social (Falk and Campbell, 1997).
1498 Urban Studies 52(8)
at RUTGERS UNIV on April 22, 2015usj.sagepub.comDownloaded from
Mobile phone use in public spaces
Despite anecdotal perceptions of the ubi-
quity of mobile phone use in public, the rate
of observed mobile phone use was relatively
low and limited primarily to use by people
who were not with co-located companions.
The rate of public mobile phone use ranged
from a low of 3% on the Met Steps to a high
of 10% of people observed outside Bryant
Park. On Chestnut Street, 96% of mobile
phone users were alone, 94% of mobile
phone users in Bryant Park were alone and
88% of mobile phones users in Downtown
Crossing were alone. Only on the Met Steps
– the location of the lowest overall propor-
tion of mobile phone users, were mobile
phone users more likely to be in a group
than to be alone (43%).
Mobile phones users appear less often in
spaces where there are more groups, and
most often in spaces where people might oth-
erwise be walking alone. Mobile phone use
may support gender equity in public space.
On Chestnut Street (the only space where we
observed an increase in people walking
alone, and only for women), 11.74% of
women who did not have a collocated com-
panion were using a mobile phone, com-
pared with only 6.26% of men. The mobile
phone may provide women with a means to
balance paid work, unpaid work and ‘net
work’, as well as reduce the vulnerability that
women experience as a result of being alone
in public (Goffman, 1977).
One argument, from the study of mobile
phone use, extends Goffman’s (1971) obser-
vations that there are two types of individu-
als in public spaces, ‘Singles’ and ‘Withs’, to
suggest that mobile phone can transform
Withs into Singles once a companion starts
using a mobile phone (Humphreys, 2005).
The argument that mobile phone use dis-
tracts from co-present interactions within
public spaces is dominant within the study of
mobile phone use (Ling, 2008). This
perspective explicitly suggests that being
alone in public has more value than commu-
nicating on a mobile phone. Presumably, the
public street provides opportunities for ser-
endipity and exposure to diversity that
would otherwise be missed. There is some
empirical evidence to support this claim
(Hampton et al., 2010), but the logic of this
argument overlooks the value of any com-
munication exchanged over a mobile phone
in public. It could be argued that mobile
phones reduce social isolation in public by
providing people who would otherwise be
Singles with an opportunity for direct inter-
action. When mobile phone use is treated as
a communication tool that brings individuals
who would otherwise be Singles into Withs,
the rate of public isolation decreases on
Chesnut Street from 73.66% to 67.83%, at
Bryant Park from 66.17% to 57.11%, in
Downtown Crossing from 52.88% to
48.95%, and on the Met Steps from 24.15%
to 22.89%.
However, if mobile phone use that takes
place while in the presence of a companion
does transform Withs into Singles
(Humphreys, 2005), we should consider an
additional adjusted isolation rate for those
who are transformed into Singles. If the iso-
lation rate is adjusted to treat as alone those
individuals who were observed with a com-
panion, who had disengaged to talk on a
mobile phone, the rate of isolation fluctuates
only marginally, increasing by 0.18% on
Chestnut Street, 0.46% in Downtown
Crossing, 0.99% on the Met Steps and hav-
ing no change on public isolation around
Bryant Park. Even in the most intrusive of
situations, when used in the presence of a
companion, mobile phone use suppresses
opportunity for public interaction only
Anecdotal impressions of a high rate of
mobile phone use in public spaces may be
related to an increase in the tendency for
more people to linger in public and for
Hampton et al. 1499
at RUTGERS UNIV on April 22, 2015usj.sagepub.comDownloaded from
people on mobile phones to linger more and
for longer periods. The overall proportion
of people lingering is low, relative to the
total number of people in a space, ranging
from 7% of all people at the Met Steps to
3% of people in Downtown Crossing.
However, over the last three decades, there
has been a 57% increase in the proportion
of people lingering at the Met Steps, a 52%
increase in people lingering on Chestnut
Street, a 40% increase in people lingering in
Bryant Park and a 36% decline in people
lingering in Downtown Crossing (we suspect
this reduction in lingering is a result of the
removal of a series of benches from the
area). The likelihood of a mobile phone user
lingering, relative to other people, was 2.87
times higher on Chestnut Street, 3.14 times
higher in Bryant Park and 4.96 times higher
at the Met Steps (all p\0.001). Mobile
phone users were no more or less likely to
linger in Downtown Crossing. The mean lin-
gering time for mobile phone users at Bryant
Park was 175% longer than for those who did
not use a phone (�x = 167.55s; ANOVA,
p\0.001), 183% longer at Downtown
Crossing (�x = 343.20s; ANOVA, p\0.01) and
208% longer at the Met Steps (�x = 345.60s;
ANOVA, p\0.001). There was not a signifi-
cant difference in lingering times for users
and non-users of mobile phones on Chestnut
Our evidence does not support the conclu-
sion that people are more alone in public
spaces than they were in the past. In three of
the four sites observed, the tendency for peo-
ple to spend time in groups was more preva-
lent than it was three decades earlier. The
absence of a trend toward increased public
isolation, or a strong tendency for frequent
disengagement from co-located companions,
does not support speculation that mobile
phone use drives a trend whereby people are
‘alone together’ (Turkle, 2011). The inci-
dence of mobile phone use as a proportion
of pedestrian activity is relatively low and
rarely performed in the presence of groups.
In some public spaces, the tendency for
mobile phone users to linger at a higher rate
than other people may in part explain a per-
ception of higher rate of mobile phone use
in public than what is actually observed.
The context of the place can likely explain
the contrasting variation found between set-
tings in the incidence of isolation. Bryant
Park, the Met Steps and Downtown
Crossing are specific destinations for leisure
and shopping, but Chestnut Street is primar-
ily a transitional space, used when travelling
to and from a workplace. Consistent with
Whyte’s original observations, diverse public
spaces are more likely to host diverse forms
of engagement (Whyte, 2001 [1980]). Some
other observations are also likely driven by
contextual effects. One example is the obser-
vation that the proportion of women present
in Downtown Crossing has declined over
time. Although not conclusive, the observa-
tions of public life on the streets of a shop-
ping district may reflect the broader trend
toward men taking a greater responsibility
for shopping, and possibly, men experien-
cing shopping as a more social activity.
The four public spaces observed for our
analyses represent a small sample of the
types of public spaces in America, and
beyond. In three decades, much more has
changed in the nature of public spaces than
we could hope to capture through our case
studies, including the growth of the
American shopping mall, changes to public
safety (both perceived and absolute) and
increased population diversity. The quality
and availability of public spaces has changed
greatly, this is true in general, and specifi-
cally for the places we observed.
Generalisations from our four case studies
should be made with appropriate caution.
The sociological analysis of video and film
1500 Urban Studies 52(8)
at RUTGERS UNIV on April 22, 2015usj.sagepub.comDownloaded from
brings its own unique challenges in terms of
the reliability of our observations. Concerns
for issues of historical validity that simply
cannot be controlled amplify these concerns
in this study. Nonetheless, it is hard to imag-
ine any other opportunity to conduct a long-
itudinal study of life in public spaces.
Historical comparative research has
repeatedly demonstrated a tendency for peo-
ple to assume that the grass was greener in
the past (Tilly, 1988). The findings of this
rare, longitudinal study of public spaces sug-
gest that assumptions about the changing
nature of public spaces are no different.
From our case studies, there is little evidence
that people today spend more time alone in
public. On the contrary, group participation
in public spaces appears to have increased.
Social mixing between men and women and
public participation by women has also
increased. The most significant change in
public spaces over the past three decades has
been the decline of social isolation experi-
enced by women. The increased presence of
women in public spaces is likely tied to the
increased participation of women in the
labour force and the accompanying tendency
for women to spend more time out of the
home. In contrast to speculation that the
participation of women in the labour force
has driven down engagement in the public
sphere, such as participation in voluntary
associations (McPherson et al., 2006;
Putnam, 2000), our observations suggest
that women’s participation in the workforce
is associated with an increase in other forms
of participation, such as time spent with oth-
ers in public spaces. The increased tendency
for men and women to spend time together
in public is a significant societal shift. It is
part of a historical trend toward friendship
groups that are desegregated by sex and par-
allel shifts away from the tendency to segre-
gate women’s activities in private spaces and
men’s activities in public spaces (Bott, 1955).
New technologies may also support women’s
use of public spaces. Women may be using
mobile phones when alone in public to bal-
ance paid work, unpaid work and ‘net
work’, and as a means to reduce the vulner-
ability associated with being alone in public.
Mobile phone use affords lingering, which
may increase surveillance and public safety.
If privatism and individualism are
increasing in our society, the implications of
these trends on the use of public spaces may
not be as commonly imagined. If, as de
Tocqueville argued, individualism is a
unique quality of life in a democracy, and if
individualism increases with equity, then
counter to concerns that individualism will
lead to a loss of the public, much the oppo-
site may be true. Increased gender equity
has spillover effects that provide new oppor-
tunities for participation in public spaces.
Although contraction within core networks
may come at the expense of diversity within
the core, it may free time to network in more
diverse social settings. Similarly, while it is
common to infer that technologies that
afford activities in the privacy of the home
contribute to a decline in public interactions,
this may not be the case. In general we are
critical of the focus that new technologies
have received as a cause for social change
in public spaces, but are not unconvinced
of their potential for providing new oppor-
tunities for public interaction (Gordon and
de Souza e Silva, 2011). As these technolo-
gies become part of our everyday lives, they
may allow the reorganisation of time for
work, leisure and sociability to accommo-
date a higher level of public participation.
Indeed, other studies have found that indi-
viduals who use a broad range of new
information and communication technolo-
gies spend more time in public and semi-
public spaces (Hampton et al., 2011b).
Contrary to a shift towards social isolation
and spending time alone, the broader trend
in public spaces may be towards more time
Hampton et al. 1501
at RUTGERS UNIV on April 22, 2015usj.sagepub.comDownloaded from
We are grateful to Blaine Beshah, Brett
Bumgarner, Esthér Burke, Florentina Dragulescu,
Kevin Gotkin, Andrew Kener, Kyung Chloe Lee,
Vincent Levy, Paul Sawyers, Lauren Springer and
Julie Xie for assistance with data collection and
coding. We are indebted to the Project for Public
Spaces, especially Steve Davies, Ethan Kent, Fred
Kent, and Kathy Madden. We are also apprecia-
tive of the advice and comments we received from
Pablo Boczkowski, John Cacioppo, Randall
Collins, Paul DiMaggio, Claude Fischer, and
Robert Sampson.
This work was supported in part by the
Annenberg School for Communication at the
University of Pennsylvania.
1. These studies have not replicated the spike in
the number of people with no core ties; they
have found core networks of similar size and
diversity to the 2004 GSS.
2. We report change in proportions rather than
change in absolute numbers. We do this to
minimise error as a result of changes to the
design of the spaces observed. For example,
for Chestnut St, the width of the sidewalk
area was reduced between observation peri-
ods, reporting absolute numbers may under-
report social change as a result of fewer
people being able to occupy the sidewalk at
time two; proportional change is less suscepti-
ble to error of this type.
Blau F, Brummund P and Liu AY-H (2013)
Trends in occupational segregation by gender
1970–2009. Demography 50: 471–492.
Bondi L and Domosh M (1998) On the contours
of public space: A tale of three women. Anti-
pode 30: 270–289.
Bott E (1955) Urban families: Conjugal roles and
social networks. Human Relations 8: 345–383.
Brashears ME (2011) Small networks and high
isolation: A reexamination of American
discussion networks. Social Networks 33:
Cattell V, Dines N, Gesler W, et al. (2008) Min-
gling, observing, and lingering: Everyday pub-
lic spaces and their implications for well-being
and social relations. Health & Place 14:
Comer JS, Mojtabai R and Olfson M (2011)
National trends in the antipsychotic treatment
of psychiatric outpatients with anxiety disor-
ders. American Journal of Psychiatry 168:
Demerath L and Levinger D (2003) The social
qualities of being on foot: A theoretical analy-
sis of pedestrian activity, community, and cul-
ture. City & Community 2: 217–237.
Falk P and Campbell C (1997) The Shopping
Experience. London: Sage.
Fischer C (2009) The 2004 GSS finding of shrun-
ken social networks: An artifact? American
Sociological Review 74: 657–669.
Fishkin JS (2000) Virtual democratic possibilities:
Prospects for internet democracy. Paper Pre-
sented at the Conference Internet, Democracy
and Public Goods. Belo Horizonte, Brazil, 6–10
Gergen KJ (2008) Mobile communication and the
transformation of the democratic process. In:
Katz JE (ed.) Handbook of Mobile Communi-
cation Studies. Cambridge, MA: MIT Press,
pp. 297–310.
Goffman E (1971) Relations in Public. New York:
Basic Books.
Goffman E (1977) The arrangement between the
sexes. Theory and Society 4: 301–331.
Goldin C and Katz LF (2002) The power of the
pill: Oral contraceptives and women’s career
and marriage decisions. Journal of Political
Economy 110: 730–770.
Gordon E and de Souza e Silva A (2011) Net
Locality. Malden, MA: Wiley-Blackwell.
Graham S and Marvin S (1996) Telecommunica-
tions and The City: Electronic Spaces, Urban
Places. London: Routledge.
Habermas J (1989) The Structural Transformation
of the Public Sphere. Cambridge, MA: MIT
Hamilton H (1971) Dimensions of self-designated
opinion leadership and their correlates. Public
Opinion Quarterly 35: 266–274.
1502 Urban Studies 52(8)
at RUTGERS UNIV on April 22, 2015usj.sagepub.comDownloaded from
Hampton KN and Gupta N (2008) Community
and social interaction in the wireless city. New
Media & Society 10: 831–850.
Hampton KN and Ling R (2013) Explaining
communication displacement and large scale
social change in core networks: A cross-
national comparison of why bigger is not
better and less can mean more. Information,
Communication & Society 16: 561–589.
Hampton KN, Goulet LS, Rainie L, et al. (2011a)
Social Networking Sites and Our Lives: How
People’s Trust, Personal Relationships, and
Civic and Political Involvement are Connected
to Their Use of Social Networking Sites and
Other Technologies. Washington, DC: Pew
Hampton KN, Lee CJ and Her EJ (2011b) How
new media afford network diversity: Direct
and mediated access to social capital through
participation in local social settings. New
Media & Society 13: 1031–1049.
Hampton KN, Livio O and Goulet LS (2010)
The social life of wireless urban spaces: Inter-
net use, social networks, and the public realm.
Journal of Communication 60: 701–722.
Hampton KN, Sessions L and Her EJ (2011c)
Core networks, social isolation, and new
media: Internet and mobile phone use, net-
work size, and diversity. Information, Commu-
nication & Society 14: 130–155.
Hannigan J (1998) Fantasy City: Pleasure and
Profit in the Postmodern Metropolis. New
York: Routledge.
Hauser GA (1998) Vernacular dialogue and the
rhetoricality of public opinion. Communica-
tions Monographs 65: 83–107.
Humphreys L (2005) Cellphones in public. New
Media & Society 7: 810–833.
Inglehart RNP (2003) Rising Tide: Gender Equal-
ity and Cultural Change Around the World.
Cambridge, New York: Cambridge University
Ito M, Okabe D and Anderson K (2008) Portable
objects in three global cities. In: Ling R and
Scott C (eds) The Reconstruction of Space and
Time. Edison, NJ: Transaction, pp. 67–88.
Jacobs J (1961) The Death and Life of Great
American Cities. New York: Random House.
Jacobs JA and Gerson K (2001) Overworked indi-
viduals or overworked families? Explaining
trends in work, leisure, and family Time. Work
and Occupations 28: 40–63.
Katz E and Lazarsfeld P (1955) Personal Influ-
ence. Glencoe, IL: Free Press.
Kaufman P, Griffin K, Cohen J, et al. (2010)
Smoking in urban outdoor public places:
Behaviour, experiences, and implications for
public health. Health & Place 16: 961–968.
Klinenberg E (2012) Going Solo. New York:
Kohn M (2004) Brave New Neighborhoods: The
Privatization of Public Space. New York:
Konrath SH, O’Brien EH and Hsing C (2011)
Changes in dispositional empathy in American
college students over time: A meta-Analysis.
Personality and Socal Psychology Review 15:
Kweon B-S, Sullivan WC and Wiley AR (1998)
Green common spaces and the social integra-
tion of inner-city older adults. Environment
and Behavior 30: 832–858.
Ling RS (2008) New Tech, New Ties. Cambridge,
MA: MIT Press.
Lofland L (1973) A World of Strangers. New
York: Basic.
Lofland L (1998) The Public Realm. New York:
Aldine de Gruyter.
Low SM, Taplin D and Scheld S (2005) Rethink-
ing Urban Parks: Public Space & Cultural
Diversity. Austin, TX: University of Texas
McPherson M, Smith-Lovin L and Brashears ME
(2006) Social isolation in America. American
Sociological Review 71: 353–375.
Marcus SC and Olfson M (2010) National trends
in the treatment for depression from 1998 to
2007. Archives of General Psychiatry 67: 1265.
Matias GP, Nicolson NA and Freire T (2011)
Solitude and cortisol: Associations with state
and trait affect in daily life. Biological Psychol-
ogy 86: 314–319.
Otnes C and McGrath MA (2001) Perceptions
and realities of male shopping behavior. Jour-
nal of Retailing 77: 111–137.
Papacharissi Z (2002) The virtual sphere. New
Media Society 4: 9–27.
Peter K and Horn L (2005) Gender Differences in
Participation and Completion of Undergraduate
Education and How They Have Changed Over
Hampton et al. 1503
at RUTGERS UNIV on April 22, 2015usj.sagepub.comDownloaded from
Time. Washington, DC: US Department of
Education, National Center for Education
Popenoe D (1985) Private Pleasure, Public Plight.
New Brunswick, NJ: Transaction.
Putnam R (2000) Bowling Alone. New York:
Simon & Schuster.
Ravelli LJ and Stenglin M (2008) Feeling space:
Interpersonal communication and spatial
semiotics. In: Antos G and Ventola E (eds)
Handbook of Interpersonal Communication.
Berlin: Mouton de Gruyter, pp. 355–394.
Robinson JP (2011) Arts and leisure participation
among IT users: Further evidence of time
enhancement over time displacement. Social
Science Computer Review 29: 470–480.
Sennett R (1977) The Fall of Public Man. New
York: Knopf.
Staats H and Hartig T (2004) Alone or with a
friend: A social context for psychological
restoration and environmental preferences.
Journal of Environmental Psychology 24:
Strauss AL (1961) Images of the American City.
New York: The Free Press of Glencoe.
Tilly C (1988) Misreading, then rereading,
nineteenth-century social change. In: Wellman
B and Berkowitz S (eds) Social Structures: A
Network Approach. Cambridge: Cambridge
University Press, pp. 332–358.
Tocqueville AD (2007 [1835]) Democracy in
America. New York: HarperCollins.
Tönnies F (1887) Community and Society (1957
edition), East Lansing, MI: Michigan State
University Press.
Turkle S (2011) Alone Together. New York: Basic
US Congress Joint Economic Committee (2010)
Women and The Economy 2010. Washington,
DC: United States Congress.
Whyte WH (2001 [1980]) The Social Life of Small
Urban Spaces. New York: Project for Public
Whyte WH (2009 [1988]) City: Rediscovering the
Center. Philadelphia, PA: University of Penn-
sylvania Press.
Wilkes R (2011) Re-thinking the decline in trust:
A comparison of black and white Americans.
Social Science Research 40: 1596–1610.
Wirth L (1938) Urbanism as a way of life. Ameri-
can Journal of Sociology 44: 3–24.
Zukin S (1995) The Cultures of Cities. Cam-
bridge, MA: Blackwell Publishers.
1504 Urban Studies 52(8)
at RUTGERS UNIV on April 22, 2015usj.sagepub.comDownloaded from
View publication statsView publication stats
Available online at
Social media use and well-being: What we know and
what we need to know
Patti M. Valkenburg
Research into the impact of social media use (SMU) on well-
being (e.g., happiness) and ill-being (e.g., depression) has
exploded over the past few years. From 2019 to August 2021,
27 reviews have been published: nine meta-analyses, nine
systematic reviews, and nine narrative reviews, which together
included hundreds of empirical studies. The aim of this um-
brella review is to synthesize the results of these meta-
analyses and reviews. Even though the meta-analyses are
supposed to rely on the same evidence base, they yielded
disagreeing associations with well- and ill-being, especially for
time spent on SM, active SMU, and passive SMU. This um-
brella review explains why their results disagree, summarizes
the gaps in the literature, and ends with recommendations for
future research.
Amsterdam School of Communication Research, University of
Amsterdam, the Netherlands
Corresponding author: Valkenburg, Patti M (
Current Opinion in Psychology 2022, 45:101294
This review comes from a themed issue on Social Media and Well-
Edited by Patti Valkenburg, Ine Beyens, Adrian Meier and Mariek
Vanden Abeele
For a complete overview see the Issue and the Editorial
Available online 20 December 2021
2352-250X/© 2021 The Author(s). Published by Elsevier Ltd. This is an
open access article under the CC BY license (http://creativecommons.
Review, Meta-analysis, Facebook, Instagram, Social media, Mental
health, Well-being, Depression, Idiographic approach, Social compari-
son, Problematic social media use.
It is almost a truism: In the past decade, social media
have become a massive and meaningful part of our daily
existence. Individuals, adults and adolescents alike, use
This research was funded by an NWO Spinoza grant. Many thanks to
Wieneke Rollman for assisting with the literature search, and to Loes
Keijsers for her comments on the first draft.
on average five social media platforms in a comple-
mentary way [1], to interact privately with family
members and friends, and/or to interact publicly with
broader audiences of friends, acquaintances, and col-
leagues [2]. In parallel with this surging social media use
(SMU), research into its potential impact on well-being
(e.g., life satisfaction) and ill-being (e.g., depression)
has also accumulated dramatically [3]. As recent reviews
demonstrate [4e6], the past years have witnessed at
least 300 studies on the impact of SMU on well- and
Together with the exponential increase in empirical
studies, reviews of the impact of SMU on well- and ill-
being have also surged in the past few years. Because
this rapidly expanding research output makes it ever
more difficult for researchers to keep track of it, an up-
to-date umbrella review of this literature is necessary
and important. An umbrella review, also called a meta-
review, is a synthesis of existing reviews [7]. Three
earlier umbrella reviews have focused on the associa-
tions of SMU with well- and ill-being [3,8,9]. One of
these focused on adolescents, thereby excluding reviews
on adults [9], and neither of the two others included the
22 reviews on the effects of SMU on well-/ill-being
published in 2020 and 2021.
In this article, I first outline the search method of this
umbrella review as well as the operational definitions of
SMU, well-being, and ill-being. To assess “what we
know,” I use the meta-analyses to discuss the associa-
tions of seven different types of SMU with well- and ill-
being. The systematic and narrative reviews are used to
complement the meta-analytical results, as well as to
summarize the identified gaps in the literature and the
suggestions for future research. To assess “what we
need to know,” the article ends with some general
conclusions and three additional recommendations for
future research.
Method and operational definitions
Two coders used the same search strategy and terms as
applied in the umbrella review of Valkenburg et al. [9],
except for the search terms related to adolescents (as
the current umbrella review excluded reviews focusing
on adolescents). SMU was operationally defined as the
active (e.g., posting), passive (e.g., browsing), private,
Current Opinion in Psychology 2022, 45:101294
2 Social Media and Well-Being
and public uses of platforms such as Facebook, WeChat,
and WhatsApp. As for the outcomes of SMU, the focus
lied on three well-being components (happiness; life
satisfaction; positive affect) and three ill-being compo-
nents (depressive symptoms/depression; anxiety symp-
toms/anxiety; negative affect). Due to space
restrictions, other components of well-being (e.g.,
eudaimonic well-being) and ill-being (e.g., stress), as
well as risk and resilience factors of well- and ill-being,
such as self-esteem, cyberbullying, and body image
concerns, were not considered.
Results: what we know …
The search yielded 27 reviews: nine meta-analyses
[4,6,10e16], nine systematic reviews [5,17e24], and
nine narrative reviews [25e33] published from January
2019 to August 2021. Except for five meta-analyses,
which included studies on both adolescents and
adults, none of the remaining 22 reviews were included
in the earlier umbrella review of Valkenburg et al. [9].
Seven social media activities
As Table 1 shows, some meta-analyses investigated (a)
general time spent with SM, or (b) time spent with
active and (c) passive SMU. Some (also) focused on
specific behaviors and mechanisms afforded by SM,
including (d) the size of one’s SM network, (e) the in-
tensity of SMU, (f) problematic SMU (i.e., an enduring
preoccupation with SM, reflected in a persistent neglect
of one’s own health and important life areas), and (g)
SM-induced social comparison (the tendency to observe
others to assess how we are looking, thinking, or
behaving in comparison with these others) [34].
Conceptualizations of well-being and ill-being
Two out of the nine meta-analyses [6,12] reversed
different ill-being components (e.g., depression) and
combined them with well-being components (e.g., life
satisfaction) to create an “aggregated well-/ill-being”
outcome. Furthermore, five meta-analyses
[6,10,11,14,15] lumped together components like life
satisfaction and self-esteem to create an “aggregated
well-being” outcome. Likewise, they combined com-
ponents like depression and loneliness to create an
“aggregated ill-being” outcome. However, because
mental health theories agree that a low well-being (e.g.,
low life satisfaction) does not necessary imply a high ill-
being (e.g., suffering from depression) and vice versa
[8,35,36], this umbrella review investigated whether the
three aggregated outcomes, that is, (a) aggregated well-/
ill-being, (b) aggregated well-being and (c) aggregated
ill-being, led to different associations with SMU.
General SMU
The three types of aggregated well- and/or ill-being
outcomes yielded inconsistent associations with gen-
eral SMU (also called time spent using SM, general SNS
Current Opinion in Psychology 2022, 45:101294
use, or the frequency of SM checking). As for aggregated
well-/ill-being, one meta-analysis yielded no association
[6], and another a small positive association with general
SMU [12]. As for aggregated well-being, one meta-
analysis yielded a small negative [6], and another a
small positive association with general SMU [15].
Remarkably though, his latter meta-analysis also re-
ported a small positive association with aggregated ill-
being [15]. Finally, general SMU was consistently asso-
ciated with higher levels of depression/depressive
symptoms [4,6,12,16] and anxiety [6,12], but, again
surprisingly, also with higher happiness levels [12].
Active versus passive social media use
Three meta-analyses compared time spent on active and
passive SMU, again with highly inconsistent results.
Active SMU was not [12] or weakly positively associated
[6] with aggregated well-/ill-being. Furthermore, it was
not [15] or weakly positively associated [6] with aggre-
gated well-being, but not with aggregated ill-being [15].
Passive SMU, in contrast, was not [6] or negatively [12]
associated with aggregated well-/ill-being, but not with
aggregated well-being [6,15] and not with aggregated
ill-being [15]. Yet, both active and passive SMU were
associated with higher levels of depression/depressive
symptoms [6] and anxiety [6].
In all, the meta-analyses yielded scant support for both
the “active SMU hypothesis” and “passive SMU hy-
pothesis,” which respectively argue that active SMU
elicits likes and support, which results in higher well-
being/lower ill-being. And that passive SMU induces
social comparisons and envy, which leads to lower well-
being/higher ill-being [37]. An elaborate explanation of
this lack of support can be found in a review by
Valkenburg et al. [24].
Social comparison
Even though the direct meta-analytic associations of
active and passive SMU with well- and ill-being were
inconsistent, two meta-analyses have addressed one part
of the passive SMU hypothesis, which states that SM-
induced social comparison results in lower well-being/
higher ill-being [33]. Indeed, SM-induced social com-
parison was associated with lower aggregated well-being
and life satisfaction [14] and higher depression [16]. It
must be noted, though, that 78% of SM users report
never feeling worse after comparing themselves to other
users [38], that only a minority of SM users feel envious
while using SM [39], that they more often feel enjoy-
ment [40], and that they sometimes also get inspired
from SM-induced social comparisons [41].
Network size
The results of the two meta-analyses focusing on network
size were rather consistent: The size of one’s SM
network size was associated with higher aggregated well-
Social Media Use and Well-Being: What We Know and What We Need to Know Valkenburg 3
being [11,15], happiness [11], and life satisfaction [11].
It was not [15] or weakly associated [11] with lower
aggregated ill-being, and not with depression [11].
Network size was not related to anxiety [11], but nega-
tively to higher social anxiety [11]. However, this latter
association has mostly been investigated within the social
compensation framework [42], in which social anxiety is
conceptualized as a predictor rather than an outcome of
SMU. Socially anxious people spend more time on SM
[42], but particularly more time on passive SMU [22].
Obviously, expanding one’s network does not occur via
passive but via interactive SMU, which could explain why
socially anxious users tend to have smaller SM networks
than their less socially anxious counterparts.
SM intensity and problematic SMU
Intensity of SMU refers to a mixture of users’ emotional
attachment to SM and the extent to which SMU is in-
tegrated into their lives [4]. It is mostly measured with
(adaptions of) the Facebook Intensity Scale (FIS) [43].
Even though the FIS was not designed as a measure of
problematic SMU, it is highly correlated with prob-
lematic SMU (e.g., b = .57) [44], and in some studies,
intensity of SMU is even included as an indicator of
problematic SMU [19]. It is no surprise, therefore, that,
most meta-analytic effect sizes for intensity of SMU and
problematic SMU are not significantly different [4].
Intensity of SMU was consistently associated with lower
aggregated well-being [6], higher depression/depressive
symptoms [4,6] and higher anxiety [6].
Comparatively, problematic SMU was associated with
lower aggregated well-being [6,10], lower happiness
[10], and life satisfaction [10]. And it was associated
with higher aggregated ill-being [10], depression/
depressive symptoms [4,6,10,13], and anxiety [6,10]. A
likely explanation for the consistent associations of
problematic SMU with well- and ill-being outcomes may
lie in “construct overlap” [4]. After all, it should be no
surprise that well- and ill-being outcomes correlate with
problematic SMU scales consisting of items like “How
often during the last year .” “did you use SM to escape
negative feelings” [45] and “have you become restless or
troubled if you were prohibited from using social
media?” [46].
Identified gaps and directions for future research
Seventeen out of the 27 reviews agreed that the evi-
dence on which their conclusions were based is primarily
cross-sectional and called for longitudinal and/or exper-
imental studies to determine the causal direction of the
effects of SMU [3,5,12,28,29,47], or for research
designed to investigate why and/or for whom SMU is
associated with well- or ill-being [4,5,17,20,31,32].
Other reviews observed an over-reliance on measures of
time spent on SM [4,6,22,28,29] and active and passive
SMU [22,24] at the expense of more fine-grained
measures, such as the purpose of SMU or the type of
communication partners [12,29]. Finally, some reviews
criticized the over-reliance on self-reports [3,5,25,28e
30] and called for more objective measures of SMU,
such as log-based data obtained though screen-time
apps [28,29].
Discussion: what we need to know …
The nine meta-analyses in this umbrella review
disagreed in their conclusions about the associations of
different types of SMU with well-being. This particu-
larly applied to the time-based predictors and not or less
to the other predictors. However, despite these in-
consistencies, all meta-analyses yielded pooled associa-
tions that were mostly small (for the time-based
predictors), occasionally moderate (for problematic
SMU), but never large. The conclusions of the meta-
analyses were largely supported by the narrative and
systematic reviews, which observed comparable gaps in
the literature and provided comparable suggestions for
future research. I end this article with three additional
recommendations for future research.
Recommendation 1: don’t collapse across well- and
ill-being outcomes
Meta-analyses of the effects of SMU can provide
indispensable summaries of the evidence in this vastly
expanding literature [3]. But they can also suffer from
the same shortcomings as any other type of study. An
important shortcoming involves their arbitrary choices
to collapse across distinct well-being and ill-being
components. In fact, the inconsistencies in effect sizes
applied particularly to the six meta-analyses that created
well-being, ill-being, and well-/ill-being composites. As
Table 1 shows, these meta-analyses collapsed across a
great variety of well- and ill-being components in addi-
tion to a range of risk and resilience factors of well- and
ill-being, such as envy, stress, self-esteem, self-harm,
and suicidal ideation.
This lumping together of different well- and ill-being
components and their risk-resilience factors hampers
the validity of the meta-analytic effect sizes for two
reasons. First, ill-being is not simply the flip side of well-
being [36], as demonstrated, for example, by the posi-
tive meta-analytic associations of time spent on SM with
both the well-being component “happiness” and the ill-
being component “depression” [12]. Second, compo-
nents within well- or ill-being composites also led to
different associations with SMU, as confirmed by the
sheer opposite meta-analytic associations of SM network
size with anxiety versus social anxiety [11]. Therefore, a
first crucial step for future research is to avoid lumping
together well- and ill-being components that deserve to
be investigated in their own right [30].
Current Opinion in Psychology 2022, 45:101294
Table 1
Associations of different types of social media use (SMU) with indicators of well-being and ill-being.
Study # Articles Operational definitions of
Main results
et al. (2021)
62 Depressive symptoms r = .11* Time spent on SNS with depressive symptoms
r = .09ns Intensity of SNS use with depressive symptoms
r = .29* Problematic SNS use with depressive symptoms
et al. (2019)
256 Well-/ill-being = aggregate
of anxiety, depression,
loneliness, & eudaimonic,
hedonic, and relational
Hedonic well-
being = aggregate of
happiness, positive affect,
subjective well-being, and
negative affect
r = −.00ns General SMU with well-/Ill-being
r = .13* General SMU with depression
r = .11* General SMU with anxiety
r = −.03* General SMU with (hedonic) well-being
r = .11* Intensity of SMU with well-/ill-being
r = .10* Intensity of SMU with depression
r = .13* Intensity of SMU with anxiety
r = −.16ns Intensity of SMU with (hedonic) well-being
r = .06* Active SMU with well-/ill-being
r = .08* Active SMU with depression
r = .06* Active SMU with anxiety
r = .06* Active SMU with (hedonic) well-being
r = −.03ns Passive SMU with well-/ill-being
r = .07* Passive SMU with depression
r = .21* Passive SMU with anxiety
r = .24ns Passive SMU with (hedonic) well-being
r = −.21* Problematic SMU with well-/ill-being
r = .34* Problematic SMU with depression
r = .32* Problematic SMU with anxiety
r = −.26* Problematic SMU with hedonic well-being
Huang (2020) 123 Well-being = aggregate of
life satisfaction, self-
esteem, happiness, and
positive affect
Ill-being (distress)
= aggregate of
depression, anxiety,
loneliness, and negative
r = −.16* Problematic SMU with well-being
r = −.30* Problematic SMU with happiness
r = −.18* Problematic SMU with positive affect
r = −.11* Problematic SMU with life satisfaction
r = .27* Problematic SMU with ill-being
r = .31* Problematic SMU with depression
r = .30* Problematic SMU with anxiety
Huang (2021) 90 Well-being = aggregate of
life satisfaction, self-
esteem, happiness
Ill-being (distress)
= aggregate of
depression, loneliness,
social anxiety, and
suicidal ideation
r = .08* Network size with well-being
r = .15* Network size with happiness
r = .10* Network size with life satisfaction
r = −.06* Network size with ill-being
r = .01ns Network size with depression
r = .08ns Network size with anxiety
r = −.19* Network size with social anxiety
Liu et al. (2019) 93 Well-/ill-being = aggregate
of life satisfaction,
happiness, self-esteem,
anxiety, depression,
loneliness, and stress
r = −.06* General SNS use with well-/ill-being
r = .09ns General SNS use with life satisfaction
r = .14* General SNS use with happiness
r = .13* General SNS use with depression
r = .10* General SNS use with anxiety
r = .02ns Active SNS use with well-/ill-being
r = −.14* Passive SNS use with well-/ill-being
Vahedi et al. (2021) 55 Depressive symptoms r = .11* General SNS use with depressive symptoms
r = .27* Problematic use with depressive symptoms
Yang et al. (2019) 13 Well-being = aggregate of
life satisfaction, self-
esteem, and
psychological well-being
r = −.20* Facebook social comparison with well-being
r = −.21* Facebook social comparison with life satisfaction
Yin et al. (2019) 63 Well-being = aggregate of
life satisfaction, well-
being, self-esteem, and
positive affect
Ill-being = aggregate of
depression, loneliness,
anxiety, envy, and
negative affect
r = .05* General SNS use with well-being
r = .06* General SNS use with ill-being
r = .13* Network size with well-being
r = −.03ns Network size with ill-being
r = .04ns Active SNS use with well-being
r = .04ns Active SNS use with ill-being
r = −.10ns Passive SNS use with well-being
r = .07ns Passive SNS use with ill-being
4 Social Media and Well-Being
Current Opinion in Psychology 2022, 45:101294
Table 1 (continued)
Study # Articles Operational definitions of
Main results
Yoon et al. (2019) 45 Depression r = .11* Time spent on SNS with depression
r = .10* Frequency of checking SNS with depression
r = .23* Non-directional social comparison with depression
r = .33* Upward social comparison with depression
Notes. *Significant at least at p < .05. Table excludes effects for components of well-being (e.g., eudaimonic well-being) and ill-being (e.g., stress) that do not fit within my operational definitions of well- and ill-being. SNS = Social network sites. Social Media Use and Well-Being: What We Know and What We Need to Know Valkenburg 5 Recommendation 2: we need content-based SM predictors of well- and ill-being The inconsistencies in the associations of the time- based SM predictors may be caused by discrepancies in their operationalizations. For example, in some meta- analyses “general SNS use” referred to time spent on SNS [12], in others to a combination of time spent on SNS and the frequency of checking SNS [6,15], and in yet others it was not defined [13]. Unfortunately though, the time-based predictors not only led to heterogeneity across the meta-analyses but also within the meta-analyses (e.g., I2s ranging from 57% for active SMU [12] to 97% for time spent on SM [4]), which could not or only partly be explained by moder- ators like age and gender. However, in case of consid- erable and (partly) unexplained heterogeneity, meta- analytic effect sizes may not be adequate and reliable [48]. A plausible explanation for the heterogeneity within meta-analyses is that the time-based predictors were operationalized differently in the included empirical studies. This has been confirmed in a recent scoping review, which revealed that of the 40 included survey-based studies, 90% used a unique, self-created operationalization of active and/or passive SMU, which led to a range of inconsistent associations with well- and ill-being components [24]. Yet even though the synchronization of time-based predictors in meta-analyses and empirical research may be a first step, there are also conceptual concerns. Time- based predictors may simply be too coarse to lead to meaningful associations with well- and ill-being com- ponents [30]. Such predictors may be valuable for out- comes like distraction or procrastination, which may be a direct consequence of time spent using SM [49,50]. In addition, they may be valuable when investigating time- based hypotheses, such as the displacement hypothesis, which states that SMU takes away time that could otherwise be spent on activities that are more conducive to well-being than SMU. But since well- and ill-being may be more amenable to the valence of SM in- teractions (cf. humor vs hate, support vs neglect) than to their duration, a second important step for future research is to pay more attention to content-based SMU predictors [24]. Recommendation 3: we need a causal effect heterogeneity paradigm Several reviews have pointed at the need for studies that allow for the investigation of within-person associations of SMU with well-being [24,27]. In recent years, a growing number of such more rigorous studies have appeared [51e56]. Butdagaindmost of these studies found weak average associations with well- and ill-being that were close to zero. What is still too often overlooked in these studies is that such average associations are derived from heterogeneous populations of SM users who differ in how they select and respond to SM [57], a finding that has repeatedly been confirmed in qualita- tive studies [58]. To truly understand the effects of SMU, researchers need to take the next step, that is, adopting a “causal effect heterogeneity” approach [59,60], which enables them to better understand why and how individuals differ in their responses to SMU. To my knowledge, two communication research teams have adopted a causal effect heterogeneity paradigm [50,61], which led to the discovery of striking person- specific effects of SMU on well-being. They found, for example, that about 20% of respondents experienced a negative effect of passive SMU on happiness, 20% a positive effect, and 60% no effect at all [51]. A causal effect heterogeneity paradigm may not only help re- searchers resolve inconsistencies in findings (and repli- cation failures) across studies [60], but it may also help them to arrive at a better understanding of why in- dividuals may or may not be affected by SMU. A causal effect heterogeneity approach, sometimes called an idiographic or person-specific approach, can be applied in experimental designs [59,62], as well as in non-experimental intensive longitudinal designs (e.g., experience sampling studies) [61,63]. The idiographic approach has recently raised concerns among some communication scholars. One of these concerns is that an idiographic approach in non-experimental settings would hinder inferences from an individual to a targeted Current Opinion in Psychology 2022, 45:10129 4 6 Social Media and Well-Being population. Another concern is that this approach hin- ders or even ignores the investigation of moderators to explain differences among subgroups in this targeted population [64,65]. While these are valid concerns, they are well addressed in recent idiographic modeling techniques, such as Dynamic Structural Equation Modeling (DSEM), which combine the strengths of “traditional” methods of analysis (i.e., structural equation modelling and multi- level analysis) with N = 1 time-series analysis [63]. These modeling techniques require the same sizeable samples to generalize to targeted populations as tradi- tional (nomothetic) approaches do. Also, they can be flexibly used to investigate the role of moderators in the person-specific effects of SMU on certain outcomes, see for example [40,66,67]. In fact, an important strength of these modelling techniques is that they allow for the investigation of two types of moderators, (a) trait-like moderators, such as ethnicity and extraversion, and (b) contextual moderators that are assumed to fluctuate within participants, such as their motivations for using SM, as well as their experience of envy or enjoyment during SMU [40]. Idiographic approaches are thus complementing rather than replacing nomothetic approaches. They enable researchers to report aggregated between-person and within-person associations of SMU with well- and ill- being. And in addition to these aggregated statistics, they can demonstrate for how many participants an experimental treatment works or for how many partici- pants certain hypotheses hold [40]. Moreover, as argued by Bryan et al. [60], a causal effect heterogeneity approach can improve interventions and “make them effective for the diverse gamut of populations and con- texts policy must address” (p. 7). Conclusion In sum, in addition to the wealth of valuable suggestions for future research raised in the 27 reviews that have appeared in the past 2.5 years, this umbrella review showed why well-being and ill-being components deserve to be investigated as two separate continuums (see also [8]). In addition, it made a case that we no longer need additional meta-analyses reporting weak aggregate between-person effect sizes of time-based SMU predictors, thereby reiterating the “one-size-fits- all approach” that has long characterized media effects research. Indeed, “we have a bright future before us, and it begins where the average ends.” [ [68], p. 191]. Conflict of interest statement Nothing declared. Current Opinion in Psychology 2022, 45:101294 References Papers of particular interest, published within the period of review, have been highlighted as: * of special interest * * of outstanding interest 1. Waterloo SF, Baumgartner SE, Peter J, Valkenburg PM: Norms of online expressions of emotion: comparing Facebook. Twitter Instagram WhatsApp New Media Soc 2018, 20:1813–1831. 2 * . Bayer JB, Triệu P, Ellison NB: Social media elements, ecolo- gies, and effects. Annu Rev Psychol 2020, 71:471–497. This is an up-to-date conceptual review of computer-mediated communication and social media. It discusses how SM platforms complicate the study of effects, as well as the challenges and oppor- tunities to measure SMU across time. 3 * * . Appel M, Marker C, Gnambs T: Are social media ruining our lives? A review of meta-analytic evidence. Rev Gen Psychol 2020, 24:60–74. This is an earlier umbrella review of the effects of social media, based on meta-analyses on well-being, academic performance, and narcis- ism. It includes a thorough discussion of possible pitfalls of meta- analyses. 4. Cunningham S, Hudson CC, Harkness K: Social media and depression symptoms: a meta-analysis. Res Child Adolesc Psychopathol 2021, 49:241–253. 5. Faelens L, Hoorelbeke K, Cambier R, van Put J, Van de Putte E, De Raedt R, Koster EHW: The relationship between Instagram use and indicators of mental health: a systematic review. Comp Human Behav Rep 2021, 4:100121. 6. Hancock JT, Liu S, X, French M, Luo M, Mieczkowski H: Social media use and psychological well-being: a meta-analysis. In 69th annual international communication association conference, Washington, D.C.; 2019. 7. Aromataris E, Fernandez R, Godfrey CM, Holly C, Khalil H, Tungpunkom P: Summarizing systematic reviews: methodo- logical development, conduct and reporting of an umbrella review approach. JBI Evi Implement 2015, 13:132–140. 8 * * . Meier A, Reinecke L: Computer-mediated communication, social media, and mental health: a conceptual and empirical meta-review. Commun Res 2020. This is another recent umbrella review. It presents the extended two- continua model of mental health, and gives an excellent review and justification for why we cannot aggregate indicators of well-being and ill-being. 9. Valkenburg PM, Meier A, Beyens I: The effects of social media use on adolescents’ mental health: an umbrella review. Curr Opin Psychol 2022, 44:58–68. 10. Huang C: A meta-analysis of the problematic social media use and mental health. Int J Soc Psychiatr 2020. 0020764020978434. 11. Huang C: Correlations of online social network size with well- being and distress: a meta-analysis. Cyberpsychology: J Psychosoc Res Cyberspace 2021, 15. Article 3. 12. Liu D, Baumeister RF, Yang C-c, Hu B: Digital communication media use and psychological well-being: a meta-analysis. J Computer-Mediated Commun 2019, 24:259–274. 13. Vahedi Z, Zannella L: The association between self-reported depressive symptoms and the use of social networking sites (sns): a meta-analysis. Curr Psychol 2021, 40:2174–2189. 14. Yang FR, Wei CF, Tang JH: Effect of Facebook social com- parison on well-being: a meta-analysis. J Internet Technol 2019, 20:1829–1836. 15. Yin X-Q, De Vries DA, Gentile DA, Wang J-L: Cultural back- ground and measurement of usage moderate the association between social networking sites (SNSs) usage and mental health: a meta-analysis. Soc Sci Comput Rev 2019, 37:631–648. Social Media Use and Well-Being: What We Know and What We Need to Know Valkenburg 7 16. Yoon S, Kleinman M, Mertz J, Brannick M: Is social network site usage related to depression? A meta-analysis of Face- book–depression relations. J Affect Disord 2019, 248:65–72. 17. Duradoni M, Innocenti F, Guazzini A: Well-being and social media: a systematic review of Bergen addiction scales. Future Internet 2020, 12:24. 18. Gilmour J, Machin T, Brownlow C, Jeffries C: Facebook-based social support and health: a systematic review. Psychol Popular Media 2020, 9:328–346. 19. Hussain Z, Wegmann E, Yang H, Montag C: Social networks use disorder and associations with depression and anxiety symptoms: a systematic review of recent research in China. Front Psychol 2020, 11:211. 20. Karim F, Oyewande AA, Abdalla LF, Chaudhry Ehsanullah R, Khan S: Social media use and its connection to mental health: a systematic review. Cureus 2020, 12, e8627. 21. Newman L, Stoner C, Spector A: Social networking sites and the experience of older adult users: a systematic review. Ageing Soc 2021, 41:377–402. 22. O’Day EB, Heimberg RG: Social media use, social anxiety, and loneliness: a systematic review. Comp Human Behav Rep 2021, 3:100070. 23. Sharma MK, John N, Sahu M: Influence of social media on mental health: a systematic review. Curr Opin Psychiatr 2020, 33:467–475. 24. Valkenburg PM, van Driel II, Beyens I: The associations of active and passive social media use with well-being: a critical scoping review. New Media & Society 2021, 10.31234/ 25. AlBarashdi HS: Social networking (SNS) addiction among university students: a literature review and research di- rections. J Educ Soc Behav Sci 2020, 33:11–23. 26. Bettmann JE, Anstadt G, Casselman B, Ganesh K: Young adult depression and anxiety linked to social media use: assess- ment and treatment. Clin Soc Work J 2021, 49:368–379. 27. Carboni Jiménez A, Vaillancourt M, Zhu P, Seon Q: Social media and mental health: what we know. McGill J Med 2021, 20. 28 * . Hartanto A, Quek FYX, Tng GYQ, Yong JC: Does social media use increase depressive symptoms? A reverse causation perspective. Front Psychiatr 2021, 12. This is an insightful narrative review that includes an overview of “social media abstinence” studies, in which participants are asked to stay away from social media for varying durations (e.g., days, weeks). It discusses an important pitfall of most of these experiments: the absence of a “placebo control” condition. Such a control condition avoids that expectancy effects of participants in the abstinencer con- dition differ from those in the control condition. Everyone who intends to conduct an social media abstinence experiment should read this review, and its suggested references. 29 * * . Griffioen N, Rooij Mv, Lichtwarck-Aschoff A, Granic I: Toward improved methods in social media research. Technol Mind Behav 2020, 1. This review identifies itself as a narrative review, but is, in fact, one of the best systematic reviews of 94 empirical studies. It draws attention to many important gaps in the literature and provides several important and well-argued suggestions for future research. 30 * . Kross E, Verduyn P, Sheppes G, Costello CK, Jonides J, Ybarra O: Social media and well-being: pitfalls, progress, and next steps. Trends Cognit Sci 2021, 25:55–66. This is an accessible narrative review of the effects of social media use on well-being. It addresses some important pitfalls of the current liter- ature. It acknowledges that it is time to move beyond the active-passive dichotomy, and that a refinement of this dichotomy is necessary. 31. Luo M, Hancock JT: Self-disclosure and social media: moti- vations, mechanisms and psychological well-being. Curr Opin Psychol 2020, 31:110–115. 32. Mou J, Zhu W, Benyocef M, Kim J: Understanding the rela- tionship between social media use and depression: a sys- tematic review. AMCIS 2020 Proceed 2020:15. 33. Verduyn P, Gugushvili N, Massar K, Täht K, Kross E: Social comparison on social networking sites. Curr Opin Psychol 2020, 36:32–37. 34. Festinger L: A theory of social comparison processes. Hum Relat 1954, 7:117–140. 35. Keyes CLM: Mental illness and/or mental health? Investi- gating axioms of the complete state model of health. J Consult Clin Psychol 2005, 73:539–548. 36. Ryff CD, Dienberg Love G, Urry HL, Muller D, Rosenkranz MA, Friedman EM, Davidson RJ, Singer B: Psychological well-being and ill-Being: do they have distinct or mirrored biological correlates? Psychother Psychosom 2006, 75:85–95. 37. Verduyn P, Gugushvili N, Kross E: The impact of social network sites on mental health: distinguishing active from passive use. World Psychiatr 2021, 20:133–134. 38. Burke M, Cheng J, de Gant B: Social comparison and Facebook: feedback positivity and opportunities for comparison, conference on human factos in computing systems. Honolulu, HI, USA: CHI; 2020. 39. Krasnova H, Widjaja T, Buxmann P, Wenninger H, Benbasat I: Why following friends can hurt you: an exploratory investi- gation of the effects of envy on social networking sites among college-age users. Inf Syst Res 2015, 26:585–605. 40. Valkenburg PM, Beyens I, Pouwels JL, van Driel II, Keijsers L: Social media browsing and adolescent well-being: chal- lenging the “passive social media use hypothesis,”. J Computer-Mediated Commun 2022, 27. 41. Meier A, Gilbert A, Börner S, Possler D: Instagram inspiration: how upward comparison on social network sites can contribute to well-being. J Commun 2020, 70:721–743. 42 * . Cheng C, Wang H-Y, Sigerson L, Chau C-L: Do the socially rich get richer? A nuanced perspective on social network site use and online social capital accrual. Psychol Bull 2019, 145: 734–764. This meta-analysis is not included in Table 1, because it does not fit within a social media effects paradigm. It is an excellent theory-based meta-analysis on the the poor-get-richer/rich-get richer hypotheses, which conceptualize social media use as the outcome and extraver- sion, social anxiety, and loneliness as the predictors. 43. Ellison NB, Steinfield C, Lampe C: The benefits of Facebook “friends:” Social capital and college students’ use of online social network sites. J Computer-Mediated Commun 2007, 12: 1143–1168. 44. Xie W, Karan K: Predicting Facebook addiction and state anxiety without Facebook by gender, trait anxiety, Facebook intensity, and different Facebook activities. J Behav Addic 2019, 8:79–87. 45. van den Eijnden RJJM, Lemmens JS, Valkenburg PM: The social media disorder scale. Comput Hum Behav 2016, 61:478–487. 46. Andreassen CS, Pallesen S, Griffiths MD: The relationship be- tween addictive use of social media, narcissism, and self- esteem: findings from a large national survey. Addict Behav 2017, 64:287–293. 47. Faelens L, Hoorelbeke K, Soenens B, Van Gaeveren K, De Marez L, De Raedt R, Koster EHW: Social media use and well- being: a prospective experience-sampling study. Comput Hum Behav 2021, 114:106510. 48. Melsen WG, Bootsma MCJ, Rovers MM, Bonten MJM: The ef- fects of clinical and statistical heterogeneity on the predictive values of results from meta-analyses. Clin Microbiol Infect 2014, 20:123–129. 49. Siebers T, Beyens I, Pouwels JL, Valkenburg PM: Social media and distraction: an experience sampling study among ado- lescents. Media Psychol 2021. 50. Aalbers G, vanden Abeele MM, Hendrickson AT, de Marez L, Keijsers L: Caught in the moment: are there person-specific associations between momentary procrastination and passively measured smartphone use? Mobile Media Commun 2021, 2050157921993896. Current Opinion in Psychology 2022, 45:101294 8 Social Media and Well-Being 51. Beyens I, Pouwels JL, van Driel II, Keijsers L, Valkenburg PM: Social media use and adolescents’ well-being: developing a typology of person-specific effect patterns. Commun Res 2021. 52. Coyne SM, Rogers AA, Zurcher JD, Stockdale L, Booth M: Does time spent using social media impact mental health? An eight year longitudinal study. Comput Hum Behav 2020, 104: 106160. 53. Hall JA, Xing C, Ross EM, Johnson RM: Experimentally manipulating social media abstinence: results of a four-week diary study. Media Psychol 2021:259–275. 54. Jensen M, George MJ, Russell MR, Odgers CL: Young adoles- cents’ digital technology use and mental health symptoms: little evidence of longitudinal or daily linkages. Clin Psychol Sci 2019, 7:1416–1433. 55. Orben A, Dienlin T, Przybylski AK: Social media’s enduring effect on adolescent life satisfaction. Proc Natl Acad Sci Unit States Am 2019, 116:10226–10228. 56. Przybylski AK, Nguyen T-vT, Law W, Weinstein N: Does taking a short break from social media have a positive effect on well- being? Evidence from three preregistered field experiments. J Technol Behav Sci 2021, 6:507–514. 57. Valkenburg PM, Peter J, Walther JB: Media effects: theory and research. Annu Rev Psychol 2016, 67:315–338. 58. Rideout V, Fox S: Digital health practices, social media use, and mental well-being among teens and young adults in the US. San Francisco, CA: HopeLab; 2018. 59 * . Bolger N, Zee K, Rossignac-Milon M, Hassin R: Causal pro- cesses in psychology are heterogeneous. J Exp Psychol Gen 2019, 148:601–618. This paper argues that people differ in their responses to environmental influences, but that most research has ignored this heterogeneity or has treated it as uninteresting error. The paper shows how causal Current Opinion in Psychology 2022, 45:101294 effect heterogeneity can be modeled, and how it presents exciting opportunities for theory and methods. 60 * . Bryan CJ, Tipton E, Yeager DS: Behavioural science is unlikely to change the world without a heterogeneity revolution. Nat Human Behav 2021, 5:980–989. Like Bolger et al. (see previous annotation), this study makes a strong case for why we need a causal effects heterogeneity paradigm in the social and behavioral sciences. 61. Valkenburg PM, Beyens I, Pouwels JL, van Driel II, Keijsers L: Social media and adolescents’ self-esteem: heading for a person-specific media effects paradigm. J Commun 2021, 71: 56–78. 62. Grice JW, Medellin E, Jones I, Horvath S, McDaniel H, O’lansen C, Baker M: Persons as effect sizes. Adv Method Prac Psychol Sci 2020, 3(4):443–455. 63. McNeish D, Hamaker EL: A primer on two-level dynamic structural equation models for intensive longitudinal data in Mplus. Psychol Methods 2020, 25:610–635. 64. Parry D, Fisher J, Mieczkowski H, Sewall C, Davidson BI: Social media and well-being: a methodological perspective. Curr Opin Psychol 2022, 65. Johannes N, Masur P, Vuorre M, Przybylski A: How should we investigate variation in the relation between social media and well- being?. 2021. 66. Valkenburg PM, Pouwels JL, Beyens I, van Driel II, Keijsers L: Adolescents’ experiences on social media and their self- esteem: a person-specific susceptibility perspective. Technol Mind Behav 2021, 2(2). 67. Siebers T, Beyens I, Pouwels JL, Valkenburg PM: Social media are a powerful distractor for the vast majority of adolescents. 2021. 68. Rose T: The end of average: how to succeed in a world that values sameness. New York: Harper Collins; 2016. Social media use and well-being: What we know and what we need to know Introduction Method and operational definitions Results: what we know … Seven social media activities Conceptualizations of well-being and ill-being General SMU Active versus passive social media use Social comparison Network size SM intensity and problematic SMU Identified gaps and directions for future research Discussion: what we need to know … Recommendation 1: don't collapse across well- and ill-being outcomes Recommendation 2: we need content-based SM predictors of well- and ill-being Recommendation 3: we need a causal effect heterogeneity paradigm Conclusion Conflict of interest statement References ARTICLE Windows of developmental sensitivity to social media Amy Orben 1✉, Andrew K. Przybylski2, Sarah-Jayne Blakemore 3,4 & Rogier A. Kievit 5,1 The relationship between social media use and life satisfaction changes across adolescent development. Our analyses of two UK datasets comprising 84,011 participants (10–80 years old) find that the cross-sectional relationship between self-reported estimates of social media use and life satisfaction ratings is most negative in younger adolescents. Furthermore, sex differences in this relationship are only present during this time. Longitudinal analyses of 17,409 participants (10–21 years old) suggest distinct developmental windows of sensitivity to social media in adolescence, when higher estimated social media use predicts a decrease in life satisfaction ratings one year later (and vice-versa: lower estimated social media use predicts an increase in life satisfaction ratings). These windows occur at different ages for males (14–15 and 19 years old) and females (11–13 and 19 years old). Decreases in life satisfaction ratings also predicted subsequent increases in estimated social media use, however, these were not associated with age or sex. OPEN 1 MRC Cognition and Brain Sciences Unit, University of Cambridge, Cambridge, UK. 2 Oxford Internet Institute, University of Oxford, Oxford, UK. 3 Department of Psychology, University of Cambridge, Cambridge, UK. 4 Institute of Cognitive Neuroscience, University College London, London, UK. 5 Cognitive Neuroscience Department, Donders Institute for Brain, Cognition and Behavior, Radboud University Medical Center, Nijmegen, The Netherlands. ✉email: NATURE COMMUNICATIONS | (2022) 13:1649 | | 1 12 3 4 5 6 7 8 9 0 () :,; T echnological innovations have shaped the ways in which we connect with each other1. Yet the recent adoption of social media has fundamentally transformed how humans spend their time, portray themselves and communicate. The repercussions of such changes have induced widespread concern2–5. Yet there is still considerable uncertainty about how social media use relates to well-being. Meta-analyses have iden- tified small or negligible negative links between social media use and well-being6,7, while experimental evidence is mixed8,9. Longitudinal observational studies that have investigated the predictive relationships between social media use and well-being have found that they are either reciprocal10, only present in a certain direction or sex11 or not present at all12. The lack of concrete evidence is an issue routinely highlighted by academics2,3, medical professionals13,14, and policymakers15,16. Much needed progress in understanding how social media use affects well-being could be made by studying the phenomenon through a developmental lens, acknowledging that developmental processes can alter our sensitivity to both the positive and negative impacts of social media17. One such developmental stage is adolescence, which spans 10–24 years18, and represents a period of profound biological, psychological, and social devel- opment. It has been proposed that substantial biological changes in the social brain make adolescence a sensitive period for social development19, self-perception, and social interaction20. Adoles- cence is also a time of cognitive development, especially in domains such as emotional regulation, planning, and response inhibition21. In parallel, most adolescents go through major sociocultural changes and life events such as moves from school to university or work. Such biological, psychological, and social changes magnify the influence of an adolescent’s social environ- ment and make them more attuned to how they are perceived by peers and the broader community. It is therefore plausible that these processes heighten adolescents’ sensitivity to the interactive, communicative, and self-portraying nature of social media22, a technology they use more extensively than other age groups23,24. It is possible that sensitivity to social media does not remain uniformly elevated throughout adolescence, given the diverse and protracted developmental processes experienced during this time19. Periods of increased sensitivity to social media are likely to occur specifically in parallel to relevant developmental changes. The strength of the statistical relationship between social media use and outcomes such as well-being might therefore not only be expected to vary between adolescence and other life stages, but also across adolescent development. To locate such develop- mental windows of sensitivity to social media, it is necessary to explicitly account for the developmental stage in research design and analysis strategy. By extending past work to account for age, as a proxy for development, and to encompass the whole ado- lescent range, this study tests for hypothesized developmental windows of social media sensitivity when stronger links between social media and well-being emerge at specific ages. In examining adolescence, it is additionally important to consider how sex interacts with the developmental stage. While some adolescent developmental processes (whether they be bio- logical, cognitive, or social) are similar in both character and timing across sex, others show variation that needs to be accounted for. For example, females experience pubertal bodily changes earlier than do males, which can provoke further downstream social changes20. Further, female life satisfaction drops earlier in adolescence25,26 and the risk of certain mental health problems such as depression, self-harm, and eating dis- orders is higher in adolescent females than in males27. Research has highlighted differences between males and females in the links between social media use and well-being in adolescence in a small subset of the analyzed data28, or other datasets10,29–33. This study therefore also examines potential sex differences in how social media use relates to well-being across adolescent development. To investigate the existence of developmental windows of sensitivity to social media, we first use cross-sectional data to examine whether adolescence might represent a period during which the association between well-being and social media is different in comparison to other life stages, and if differences between males and females are present during this time. Using longitudinal data, we then test the idea that there exist sex-specific windows of sensitivity to social media during adolescence itself. Results and discussion To address the first research question, we analyzed the UK Understanding Society household panel survey that includes 72,287 10–80-year-old participants surveyed up to seven times each between 2011 and 201834, correlating a single-item life satisfaction measure and participant estimates of how much time they spend using social media on a typical day (raw data plot: Fig. 1; extended plot: Supplementary Fig. 1). We also tested the robustness of these relations using—both linear and quadratic— terms of estimated social media use to predict life satisfaction ratings while adding control variables of log household income, neighbourhood deprivation (measured using the Index of Mul- tiple deprivation), and year of data collection (Supplementary Fig. 2). While it is important to note that responses to, and con- ceptualizations of, life satisfaction might be qualitatively different across the lifespan, our results showed that the relationship between estimated social media use and life satisfaction ratings varied substantially by age. Although the relationship fluctuates to a certain extent across the lifespan, for example, it is more negative in males aged 26–29 years compared to males aged 22–25 years, the most substantial negative relations were found in adolescence (Supplementary Fig. 2). This finding, combined with our developmental interest in adolescence, the fact that social media use is heightened in this age group, and the nature of the data (annual longitudinal social media measures being available only for those aged 10–21), motivated us to focus on adolescence as the age group of interest throughout this study. The relations in the raw data differed when comparing younger (10–15 years) and older adolescents (16–21 years; for more information about these age categories see methods). There was a pronounced inverted U-shaped curve in older adolescence, indi- cating that those who estimated they engaged in very low or very high social media use reported lower life satisfaction ratings than those who estimated that they used between ‘less than an hour’ and ‘1–3 h’ of social media a day (i.e., response options ‘2’ and ‘3’). This pattern in between-person associations, where those participants who use the least or the most social media also report lower well-being ratings, has been previously termed the ‘Goldi- locks hypothesis’ (i.e., the concept that too much and too little digital technology use might be suboptimal)35. Younger adolescents demonstrated a different pattern of between-person associations than older adolescents: the rela- tionship was more linear and showed more prominent differences between males and females (Fig. 1; top). Specifically, there was no evidence in this age range for the ‘Goldilocks hypothesis,’ as those who reported very little social media use did not routinely score lower on life satisfaction than their peers who reported slightly higher social media use. Further, females reporting very high social media use scored substantially lower on life satisfaction than males. This difference between males and females is statis- tically supported by an Akaike weights procedure36, which allows us to quantify evidence ratios between two more models, ARTICLE NATURE COMMUNICATIONS | 2 NATURE COMMUNICATIONS | (2022) 13:1649 | | provided the models are nested. Doing so, we show that models associating estimated social media use and life satisfaction ratings while differentiating for self-reported sex (and controlling for household income, neighbourhood deprivation, and year of data collection) are more likely to be the best models of the data between the ages of 12 and 15 (Akaike weights ratios ranging from 5:1 to over 6,000,000:1 in favour of the model differentiating males and females). In contrast, models differentiating for sex were not more likely to be the best models of the data at other ages (Fig. 1, bottom; Supplementary Fig. 3). These analyses demonstrated that the between-person association linking esti- mated social media use to ratings of life satisfaction was more negative in adolescents compared with most other age groups. Further, they showed that adolescence is unique due to the pro- minent sex differences on the cross-sectional links between esti- mated social media use and ratings of life satisfaction that are not evident at most other ages. We supplemented these cross-sectional analyses by examining the longer life satisfaction questionnaire given to 10–15-year-olds in Understanding Society, and 13–14-year-olds in the Millennium Cohort Study (Fig. 2): this questionnaire asked about satisfaction with appearance, friends, family, school, schoolwork, and life. We found no evidence that a specific sub-component of life satis- faction was the lone driver of the sex differences found in Fig. 1. In Understanding Society, sex differences were found pre- dominately for satisfaction with life and appearance (Akaike weight of model with sex difference compared to the model without sex difference: satisfaction with life 71.9%, appearance 69.1%, family 53.8%, friends 59.1%, school 51.0%, schoolwork 35.0%; the percentage shows how much more likely a model including a sex difference is the best model of the data compared to a model without a sex difference). In the Millennium Cohort Study, the models differentiating for sex were more likely to be the best model of the data for all measures (Akaike weight of model with sex difference compared to model without sex dif- ference: satisfaction with life 100%, appearance 100%, family 100%, friends 99.7%, school 100%, schoolwork 91.2%). Further analyses using a broader range of mental health questionnaires available for these sample participants can be found in Supple- mentary Methods 1, Supplementary Results 1, and Supplemen- tary Figs. 4–6. The limited 10–15-year age range available for these measures did not allow us to compare these adolescents with other age groups. To address our second research question, and locate develop- mental windows of sensitivity to social media that are hypothe- sized to emerge across adolescence, we used methods that test within-person changes and differences over time37,38. While the previous analyses allowed us to examine whether adolescence potentially represents a period of heightened sensitivity to social media in comparison to other age groups, cross-sectional differ- ences cannot be used to map detailed developmental processes. Longitudinal models such as the Random-Intercept Cross Lagged Panel Model (RI-CLPM) need to be used to take into account these dynamic and reciprocating relations between social media use and ratings of life satisfaction10. We estimated these models with robust maximum likelihood to account for deviations from the assumption of multivariate normality and ensure all model comparisons are performed on nested models, and use Full information Maximum Likelihood to account for missingness which may vary systematically across measured covariates and 10 11 12 13 14 15 16 17 18 19 20 21 22−25 26−29 30s 40s 50s 60s 70s 3 4 5 6 7 S a tis fa ct io n w ith L ife 1 2 3 4 5 1 2 3 4 5 1 2 3 4 5 1 2 3 4 5 1 2 3 4 5 1 2 3 4 5 1 2 3 4 5 1 2 3 4 5 1 2 3 4 5 1 2 3 4 5 1 2 3 4 5 1 2 3 4 5 1 2 3 4 5 1 2 3 4 5 1 2 3 4 5 1 2 3 4 5 1 2 3 4 5 1 2 3 4 5 1 2 3 4 5 Social Media Use Δ S e x Fig. 1 Estimated social media use and life satisfaction ratings across the lifespan. Top: The cross-sectional relationship between estimated social media use and a one-item life satisfaction measure for 72,287 UK participants between the age of 10 and 80 years. The results are split by age and self-report sex: females = red, males = blue. The 95% confidence intervals represent the lower and upper Gaussian confidence limits around the mean based on the t-distribution. Bottom: Shading of each rectangle represents calculated AIC weights, i.e., whether a model relating estimated social media use and ratings of life satisfaction that takes into account a possible sex difference is more likely to represent the data than a model that does not take into account sex: darker shade = model with sex differences is more likely. It should be noted that as high levels of social media use are very rare in the youngest and oldest age groups present in the data (e.g., ages 10, 11, and 60+, Supplementary Fig. 1), one cannot evaluate functional form in these groups. Further, as most participants were measured multiple times, more than one data point per participant will appear in this graph. Source data for this figure are provided as a Source Data file. NATURE COMMUNICATIONS | ARTICLE NATURE COMMUNICATIONS | (2022) 13:1649 | | 3 model variables. We applied this modelling framework to data provided by 10–21-year-olds in Understanding Society, who completed estimated social media use and life satisfaction mea- sures annually for up to seven waves (17,409 participants; for additional longitudinal analyses of component life satisfaction measures only available for 10–15-year-olds see Supplementary Fig. 7)39,40. As the outcomes of these longitudinal models depend largely on the time-interval between observations41,42, the annual nature of the data needs to be considered. It is likely that studies on different time frames would show different results and/or reflect distinct mechanisms and processes. Using longitudinal modelling, we can address the question of whether the present use of social media has consequences for future life satisfaction—and vice versa. Do people feel better, or worse, after periods of heightened social media use? Conversely, do people use more, or less, social media after periods of higher life satisfaction? Specifically, the RI-CLPM model allowed us to focus on whether an individual’s deviation from their expected level of a certain variable y (e.g., ratings of life satisfaction) can be predicted from their prior deviation from their expected scores in another variable x (e.g., estimated social media use), while con- trolling for the structural change in y (e.g., ratings of life satis- faction); and vice versa. We fit an initial model that allowed both cross-lagged paths to vary across age and sex (χ2 (434) = 1216.29, p < 0.001, RMSEA = 0.014, [0.013, 0.015], CFI = 0.944, SRMR = 0.072; Full Informational Maximum Likelihood estimation; two- tailed test). The control variables included in this model are time- invariant mean log household income and Index for Multiple Deprivation with freely estimated effects at different ages, to account for the socioeconomic status of both the family and their immediate environment. All RI-CLPM significance tests are two- sided. While the cross-lagged paths were predominantly non- significant, there are specific developmental windows where the data suggests estimated social media use and ratings of life satisfaction do predict each other—and these ages differed for males and females (Fig. 3). Before interpreting these potential windows of sensitivity, we examined whether they were statistically robust via statistical model comparison procedures. A model constraining the path of life satisfaction ratings predicting estimated social media use to be constant across age and sex, while allowing the path of estimated social media use predicting ratings of life satisfaction to vary, was more likely to be the best model for the data compared to other model constraints (Akaike weights procedure: 99.2%, see Sup- plementary Fig. 8; model fit for best fitting model: χ2 (455) = 1234.736, p < 0.001, RMSEA = 0.014, [0.013, 0.015], CFI = 0.945, SRMR = 0.072). We, therefore, examined the paths of this model in more detail (both social media use predicting life satisfaction, split by age and sex, and life satisfaction predicting social media use, not split by 10 11 12 13 13/14 (MCS) 14 15 F e m a le M a le 1 2 3 4 5 1 2 3 4 5 1 2 3 4 5 1 2 3 4 5 2 4 6 8 1 2 3 4 5 1 2 3 4 5 4 5 6 7 4 5 6 7 Social Media Use L ife S a tis fa ct io n S co re School Work (F) Appearance (F) Family (F) Friends (F) School (F) Life (F) School Work (M) Appearance (M) Family (M) Friends (M) School (M) Life (M) Fig. 2 The cross-section relationship between social media use and six different life satisfaction measurements (ages 10–15). The Figure shows the cross-sectional relation between estimated social media use and six life satisfaction measures at ages 10–15 (Understanding Society dataset, US; 10,019 participants and 24,698 measurement occasions) and ages 13–14 (Millennium Cohort Study dataset, MCS; 11,724 participants). Specifically, it displays the cross-sectional correlation between estimated social media use and raw scores of sub-components of life satisfaction (satisfaction with school work, appearance, family, friends, school, and life). The relationships are presented separately for males and females. The 95% confidence intervals represent the lower and upper Gaussian confidence limits around the mean based on the t-distribution. Source data for this figure are provided as a Source Data file. ARTICLE NATURE COMMUNICATIONS | 4 NATURE COMMUNICATIONS | (2022) 13:1649 | | age and sex, Table 1). The effect sizes of both paths were small in magnitude43, something that has been discussed in previous work3,10. However, as dynamic effects can amplify over devel- opmental time44, they are nonetheless worth understanding and addressing with care. The constrained cross-lagged path of life satisfaction ratings predicting estimated social media use was negative, meaning that, for example, if an individual scored lower than their expected value of life satisfaction ratings in one year this predicted a positive change from their expected estimated social media use one year later (or vice versa, b = −0.02 [−0.03, −0.01], se = 0.007, β = −0.02–0.03, p = 0.004). The other cross- lagged path, linking social media use to life satisfaction, only showed a statistically significant negative link at specific ages, depending on sex (Fig. 4). This supports previous findings that the relationship between social media use and life satisfaction is bidirectional in nature10, and provides evidence for the hypoth- esis that the impact of social media on individuals varies depending on how old they are, as well as their sex. We will focus on these differences in the next section, but wish to emphasise that this focus does not mean that the path linking life satisfaction ratings to estimated social media use one year later is unimportant. Specifically, for females, we observed a window of sensitivity to social media between the ages of 11 and 13, when increases in estimated social media use from expected levels predicted a decrease in life satisfaction ratings from expected levels one year later (Fig. 4, top; age 11: b = −0.11 [−0.21, −0.02], se = 0.05, β = −0.09, p = 0.020, age 12: b = −0.14 [−0.22, −0.07], se = 0.04, β = −0.12, p < 0.001, age 13: b = −0.08 [−0.15, −0.01], se = 0.03, β = −0.07, p = 0.019). For males a similar window was in evidence at ages 14 and 15 (Fig. 4, bottom; age 14: b = −0.10 [−0.17, −0.03], se = 0.04, β = −0.10, p = 0.005, age 15: b = -0.18 [−0.29, −0.08], se = 0.05, β = −0.12, p = 0.001). Speculatively, the sex difference in timing suggested that early increases in sensitivity to social media might be due to maturational processes such as puberty, which occur earlier in females compared to males20. A later increase in sensitivity to social media, which was present at age 19 for both sexes, suggested a different underlying process may be present in late adolescence (Fig. 4; females: b = −0.16 [−0.25, −0.07], se = 0.05, β = −0.13, p < 0.001, males: b = −0.16 [−0.26, −0.07], se = 0.05, β = −0.13, p = 0.001). Particularly, the lack of sex differences indicated a process that similarly affects both males and females. Speculatively, this might be related to changes in the social environment such as a move away from home and subsequent disruptions in social networks. However, these explanations cannot be tested directly in this dataset and require further targeted investigation using data containing pubertal and social measurements. The study has multiple limitations that need to be considered. First, to interpret the parameters from our analyses as estimates of causal effects one would need to adopt the following assumptions: (a) there are no time-varying unobserved con- founders that impact the relation between social media use and life satisfaction; (b) the model adequately accounts for unob- served time-invariant confounding through the inclusion of a random intercept; (c) there is no measurement error in the variables; (d) the time interval between studies (one year) is the right length to capture the effects of interest; and (e) the bidir- ectional links estimated by our longitudinal model are linear in nature. Only if these assumptions are met can this observational study be said to capture the causal effects between social media and life satisfaction. Second, the data are self-report and therefore only allow inferences about the impact of self-estimated time on social media, rather than objectively measured social media use. The findings reported here may enable investigation of potential mechanisms of interest, for example, in datasets with pubertal or additional social measurements. One could also carry out more targeted investigations, for example, by examining the mental health measures only completed by select age ranges in the datasets (e.g., ages 10–15, displayed in Supplementary Fig. 4) to understand how they interrelate over the longitudinal time frame. Furthermore, the cross-sectional relation between social media use and life satisfaction ratings showed differences across the whole life span, e.g., in early adulthood and old age. Future work may use a similar approach to investigate interactions in older age groups with a suitably rich sample. This study provides evidence for age- and sex-specific windows of sensitivity to social media use in adolescence. Past research has demonstrated that each life stage exhibits its own unique trajec- tories, goals, and influences19,45–47, and activities can therefore differ from being of no impact, to being adaptive or maladaptive depending on which developmental period one examines46,48. This applies to the activity of using social media, just as it applies to other activities such as exercise or drinking alcohol. While the results support past longitudinal work finding bidirectional influences between social media use and life satisfaction10, it also goes beyond that to suggest that such influences vary, potentially due to concurrent developmental processes. While the windows of sensitivity to social media are prominent in aggregate, they will most probably meaningfully differ across individuals, as each person’s sensitivity is further influenced by a wide range of individual, peer, and environmental dynamics. Additionally, the types of social media use individuals are engaged in will add further variance to this complicated dynamic49. Our work, therefore, opens the door to new Life satisfaction −> Social Media Social Media −> Life satisfaction
10 11 12 13 14 15 16 17 18 19 20 10 11 12 13 14 15 16 17 18 19 20
Sex Female Male
Fig. 3 Results from Random Intercept Cross-Lagged Panel Model (RI-
CLPM) of estimated social media use and life satisfaction for 17,409
participants of the Understanding Society dataset aged 10–21 (52,556
measurement occasions). Results from both cross-lagged paths of a RI-
CLPM where those paths were free to vary across age/sex. Results are
unstandardized and split by path (left: deviations from expected ratings of
life satisfaction at that age predicting deviations from expected social media
use one year later; right: deviations from expected social media use at that
age predicting deviations from expected ratings of life satisfaction one year
later) and sex (female = top/red, male = bottom/blue). The ribbon
represents the 95% Confidence Interval around the point estimate. All tests
are two-sided. Source data for this figure are provided as a Source Data file.
NATURE COMMUNICATIONS | (2022) 13:1649 | | 5
theoretically informed approaches for studying how this
increasingly pervasive technology impacts our population, by
focusing on the within-person dynamics where effects actually
unfold. In particular, an understanding of what neurodevelop-
mental, pubertal, cognitive, and social changes underlie devel-
opmental windows of sensitivity to social media, and how these
are impacted by individual differences, could pave pathways for
targeted interventions that address the negative consequences of
social media while also promoting its positive uses. This will
ultimately enable academic research to help inform critical poli-
cies, interventions, and conversations concerning adolescent well-
being in the digital age.
Ethical approval. The University of Essex Ethics Committee has approved all data
collection on the Understanding Society main study and innovation panel waves,
including asking consent for all data linkages except to health records. Ethical
approval for the Millennium Cohort Study was given by the UK National Health
Service (NHS) London, Northern, Yorkshire, and South-West Research Ethics
Committees (MREC/01/6/19, MREC/03/2/022, 05/MRE02/46, 07/MRE03/32). No
additional ethical approval was needed for this study.
Datasets. The study analyzed the Understanding Society dataset and the Millen-
nium Cohort Study. The Understanding Society dataset is a longitudinal study
following approximately 40,000 British households34. The study sample is designed
to be representative of the UK population50. Started in 2009, its annual waves of
data collection each span two years; we used 7 waves of data from between 2011
and 2018 released in February 2020 (the two first waves were excluded as parts of
the sample were not asked to complete social media related questions). All
household members between 10 and 15 years (whom we label here as ‘younger
adolescents’) filled out a younger adolescent survey, while those 16 and over filled
out an adult survey. 16–21-year-olds (whom we label here as ‘older adolescents’)
further completed a short supplement with additional questions. Participants were
incentivised with an unconditional £10 gift voucher at invitation to the study,
furthermore adults had another £10 incentive if they completed the survey
<5 weeks (if they received a web questionnaire). 10–15-year-olds were incentivised with a voucher. 16-year-olds were entered into a prize draw to win an iPad. Members of households that did not complete the previous wave were given an incentive of £20. Incentivisation procedures changed slightly between waves and are detailed in the Understand Society fieldwork documentation. Oral consent was provided by parents and adolescents. The Millennium Cohort study is a birth cohort study of a sample of around 11,000 young people born between September 2000 and January 200151. The study over-sampled some parts of the population, for example, children from ethnic minorities (e.g., Indian, Pakistani, Bangladeshi, Caribbean, and African in the ethnic minority boost sample). In this study, we only used the wave of data collected in 2015, when the majority of respondents were 13 or 14 years old. This made the two datasets comparable (e.g., in terms of the prevalence and use of social media) to subsections of our Understanding Society sample. Participants were provided with a ‘Participant Pack’ (i.e., leaflet, membership card, key ring, travel card holder, and notebook) as incentivisation. Informed written consent was provided by parents and oral consent was provided by adolescents. Measures Understanding society. For the core analyses of the Understanding Society study, we examined life satisfaction ratings, estimated social media use, age, and sex measures derived from both younger adolescent, older adolescent, and adult sur- veys. We further supplemented these measures with two control variables— household income and Index of Multiple Deprivations—and a range of additional well-being and mental health questionnaires. To measure life satisfaction, younger adolescent survey respondents were asked to respond, “which best describes how you feel about your life as a whole?” (visual analogue scale ranging from 1 = very happy smiley face to 7 = very sad smiley face; written explanation: “1 is completely happy and 7 is not at all happy”; scale reversed so that higher scores indicate higher life satisfaction). They were asked the same question about how they feel about their school work, appearance, family, friends, and school. Adults and older ado- lescents were asked to “please select the answer which you feel best describes how dissatisfied or satisfied you are with the following aspects of your current situa- tion… your life overall” (1 = completely dissatisfied to 7 = completely satisfied). Due to previous work showing that variables containing five or more categories can be treated as continuous with negligible drawbacks, we treated both measures as continuous52. The life satisfaction measures were different in wording and response options for the younger adolescent and older adolescent/adult survey to accommodate age differences, for example by including smileys for young adolescents. An irregularity in Understanding Society fieldwork provided us with an opportunity to test that they were not qualitatively different. Due to a lag in when questionnaires were issued into the field and completed by the participants, 37 16-year-olds mistakenly took the younger adolescent survey while 10 15-year-olds mistakenly took the older Table 1 Results of the best fitting Random-Intercept Cross Lagged Panel Model examining the bidirectional links between life satisfaction and social media use across ages 10–21 years. b (Regression coefficient) Standard error β (Standardized effect size) p Life satisfaction −> social media use
Invariant across age and sex −0.02 [−0.03, −0.01] 0.007 −0.02–0.03 0.004
Social media use −> life satisfaction (female)
Age 10 0.02 [−0.09, 0.12] 0.05 0.01 0.753
Age 11 −0.11 [−0.21, −0.02] 0.05 −0.09 0.020
Age 12 −0.14 [−0.22, −0.07] 0.04 −0.12 0.000
Age 13 −0.08 [−0.15, −0.01] 0.03 −0.07 0.019
Age 14 −0.04 [−0.11, 0.03] 0.03 −0.04 0.215
Age 15 0.01 [−0.07, 0.10] 0.04 0.01 0.784
Age 16 −0.07 [−0.15, 0.01] 0.04 −0.06 0.080
Age 17 0.00 [−0.08, 0.08] 0.04 0.00 0.937
Age 18 0.02 [-0.06, 0.10] 0.04 0.02 0.642
Age 19 −0.16 [−0.25, −0.07] 0.05 −0.13 0.000
Age 20 −0.06 [−0.14, 0.03] 0.04 −0.05 0.181
Social media use −> life satisfaction (male)
Age 10 −0.06 [−0.16, 0.05] 0.05 −0.04 0.291
Age 11 −0.05 [−0.15, 0.04] 0.05 −0.04 0.275
Age 12 0.02[−0.06, 0.09] 0.04 0.01 0.708
Age 13 −0.05[−0.12, 0.03] 0.04 −0.04 0.202
Age 14 −0.10 [−0.17, −0.03] 0.04 −0.10 0.005
Age 15 −0.18 [-0.29, −0.08] 0.05 −0.12 0.001
Age 16 0.02 [−0.05, 0.10] 0.04 0.02 0.551
Age 17 0.01 [−0.07, 0.09] 0.04 0.01 0.806
Age 18 −0.00 [-0.08, 0.08] 0.04 −0.00 0.956
Age 19 −0.16 [−0.26, −0.07] 0.05 −0.13 0.001
Age 20 0.02 [−0.06, 0.11] 0.04 0.02 0.614
The coefficients include 95% confidence intervals around the estimate.
6 NATURE COMMUNICATIONS | (2022) 13:1649 | |
adolescent/adult survey. A two-sided Welch two-sample t-test for unequal
variances comparing the scores of 15-year-olds on both younger adolescent and
older adolescent/adult surveys showed no significance difference (Younger
Adolescent M = 5.64, N = 4095; Older Adolescent/Adult M = 5.25, N = 10; t(3) =
0.33, p = 0.77, normality assumption not met, power to detect only a large
difference of d = 0.89). A two-sided Bayesian two-sample t-test found anecdotal
evidence in favour of the null hypothesis (i.e., no difference) between the two
groups (BF10 = 0.49; variance of normal population: noninformative Jeffreys prior,
standardized effect size: Cauchy prior; BF calculated via Gaussian quadrature). A
second two-sided Welch two-sample t-test for unequal variances further found that
16-year-olds showed no significant difference in both surveys (Younger Adolescent
M = 5.86, N = 37; Older Adolescent/Adult M = 5.49, N = 4703; t(37) = 1.74,
p = 0.09, normality assumption not met, power to detect only a medium difference
of d = 0.46), with a two-sided Bayesian two-sample t-test finding anecdotal
evidence in favour of the null (BF10 = 0.51; variance of normal population:
noninformative Jeffreys prior, standardized effect size: Cauchy prior; BF calculated
via Gaussian quadrature). Further, we used linear regression on data of participants
aged 13-18 years to predict life satisfaction ratings from both age and a categorical
variable indicating the survey type (i.e., younger adolescent or older adolescent/
adult survey measurement). While age significantly predicted life satisfaction
scores, the type of survey measurement did not (age: b = −0.09, se = 0.01,
p < 0.001; survey type: b = 0.05, se = 0.04, p = 0.140; adjusted R-squared = 0.016, two-sided test; assumptions met except normality of residuals). With 24,533 participants included, this test would be highly sensitive (99% power) to extremely small effects (f2 = 0.001; an f2 value of 0.02 is small according to Cohen’s guidelines) at a conventional alpha level of 0.05. A Bayesian approach comparing a regression including age and the survey category as predictors of life satisfaction scores against a regression only including age as a predictor, found the latter model to be a better fit for the data with a Bayes Factor of 11.5 (+/− 1.4%; for priors see Supplementary Methods 2). These analyses, therefore, support the conclusion that the measures are not qualitatively different. We also created cross-sectional plots for additional well-being and mental health questionnaires completed only by the younger adolescent sample (age 10–15 years), these measures can be found in Supplementary Methods 1. Social media use was measured at every wave for younger and older adolescent samples, and every three waves for adults. It entailed two questions in the younger adolescent survey: “Do you have a social media profile or account on any sites or apps?” (1 = Yes, 2 = No) and “How many hours do you spend chatting or interacting with friends through a social web-site or app like that on a normal school day?” (1 = None, 2 = Less than an hour, 3 = 1–3 h, 4 = 4–6 h, 5 = 7 or more hours). Prior to the final two waves, the questions were slightly different: “Do you belong to a social web-site such as Bebo, Facebook or MySpace?” and “How many hours do you spend chatting or interacting with friends through a social web-site like that on a normal school day?”. For adults and older adolescent samples, the questions read: “Do you belong to any social networking web-sites?” (1 = Yes, 2 = No) and “How many hours do you spend chatting or interacting with friends through social web-sites on a normal week day, that is Monday to Friday?” (1 = None, 2 = Less than an hour, 3 = 1–3 h, 4 = 4–6 h, 5 = 7 or more hours). If a participant stated that they do not own a social media account or do not use social media to interact with friends, they were coded as the lowest score of 1; for the rest of the participants, we took their score on the second question, which measures how much time they spent interacting socially online. Self-reported sex was reported annually (“male”, “female”). When first surveyed, adults and older adolescents were asked to report their sex (options: “male” or “female”); subsequently, they were asked to confirm their sex collected in the previous waves. Younger adolescents were asked “are you male or female” (options: “male” or “female”); they were allowed to refuse response. If participants’ report of sex varied between waves they were recorded as NA by the survey F e m a le M a le 10 11 12 13 14 15 16 17 18 19 20 −0.2 0.0 0.2 −0.2 0.0 0.2 Age R e g re ss io n C o e ff ic ie n t Fig. 4 How social media use predicts life satisfaction in longitudinal data (ages 10–21). Results from the cross-lagged path connecting estimated social media use to life satisfaction ratings one year later, estimated through a Random Intercept Cross-Lagged Panel Model of 17,409 participants (52,556 measurement occasions) aged 10–21. Results show how much an individual’s deviation from their expected social media use at a certain age predicted a deviation from their expected life satisfaction ratings one year later (unstandardized estimates). Graph is split by sex (female = top/red, male = bottom/ blue) and the grey boxes indicate those ages where the path became statistically significant (p < 0.05, two-sided test). The thin lines represent the coefficients extracted from 500 bootstrapped versions of the model to visualize uncertainty, dark shaded ribbons represent bootstrapped 95% CIs, light shaded ribbons represent bootstrapped 99% CIs. The other cross-lagged path linking life satisfaction ratings to estimated social media use was constrained not to vary across age/sex and is not shown here. All tests are two-sided. Source data for this figure are provided as a Source Data file. NATURE COMMUNICATIONS | ARTICLE NATURE COMMUNICATIONS | (2022) 13:1649 | | 7 administrators as part of a cumulative sex variable. Twenty-five measurement occasions reported sex as NA, which were too few to garner how their exclusion could have influenced our results. Due to the nature of this item design, we report the responses as “sex” in this manuscript; however, respondents may well have responded according to gender identity or gender assigned at birth based on genitalia as the nature of the questions was ambiguous, especially for younger adolescents. In the waves of data analyzed there was however no opportunity to examine to what extent self-report sex was related to gender identity. Age for both the adolescent samples as well as the adult samples was derived by the data provider using self-reported date of birth and the interview date. Household income was measured using the log monthly total household net income for the household the adolescent belongs to (those households reporting 0 income were changed to 0.1 to allow for log transformation). The mean of all waves of available data pertaining to household income was taken to create the exogenous control variable used in the longitudinal models. Similarly, Index of Multiple Deprivation was derived by taking the Lower Layer Super Output Area (LSOA) of the participant and matching it to the governmentally derived rank on the Index of Multiple Deprivation (IMD; deciles). Again, the mean of IMD rank was taken to create a single exogenous control variable. Millennium cohort study. For the Millennium Cohort Study, we analyzed a variety of mental health and well-being measures filled out by the adolescent or their primary caregiver respondent, a social media question completed by the adolescent respondent and demographic variables including sex and age. To measure social media use, adolescents were asked “On a normal week day during term time, how many hours do you spend on social networking or messaging sites or Apps on the internet such as Facebook, Twitter, and WhatsApp?” (1 = none, 2 = less than half an hour, 3 = half an hour to less than 1 h, 4 = 1 h to less than 2 h, 5 = 2 h to less than 3 h, 6 = 3 h to less than 5 h, 7 = 5 h to less than 7 h, 8 = 7 h or more). Sex and age were taken from the derived variable file of the MCS. The Millennium Cohort Study coded sex at birth: “male”, “female” or “not known/not applicable”, and age was calculated using age at last birthday. Aligned with the Understanding Society well-being measures, adolescents were asked to fill out a well-being questionnaire on a scale from ‘1’ (completely happy) to ‘7’ (not happy at all) to indicate “How do you feel about the following parts of your life”: your school work, the way you look, your family, your friends, the school you go to and your life as a whole. More extended well-being and mental health measured analyzed in the Supplementary can be found in the Supplementary Methods 1. Inclusion criteria. In Understanding Society we excluded a variety of participants from the dataset. First, we excluded people aged over 80, because social media use was very low at higher ages (excluding 14,394 measurement occasions). We also excluded those 7 measurement occasions where a child aged 9 years filled out the younger adolescent survey. We further excluded those participants who completed a questionnaire twice in one age category (we only excluded the second time the questionnaire was filled out in one age category, excluding 9272 measurement occasions) and those whose sex was NA (25 measurement occasions; there were too few datapoints to treat this as a separate category). The latter could be due to them identifying as a different sex in the data collection frame or refusing to answer the question. These exclusions left a sample of 72,287 participants over 7 waves (1 wave = 12,444 participants, 2 waves = 8777 participants, 3 waves = 8794 parti- cipants, 4 waves = 7426 participants, 5 waves = 7099 participants, 6 waves = 9898 participants, 7 waves = 17,849 participants; see Supplementary Table 1 for mea- surement occasions by age and sex). The adolescent (10–21 years) sample used for longitudinal modelling consisted of 17,409 participants and 52,556 measurement occasions (see Supplementary Table 2 for measurement occasions by age and sex). In the Millennium Cohort Study, we excluded those not aged 13 or 14 (160 participants), leaving 11,724 participants (see Supplementary Table 3 for mea- surement occasions by age and sex). There was dropout over time in the Understanding Society longitudinal adolescent sample (see Supplementary Fig. 9 and Supplementary Table 4). Due to the nature of our modelling approach, we were not able to integrate sampling weights into our estimation strategy. This limits the extent to which we can generalize our findings to the whole UK population. Cross-sectional analyses. To examine the cross-sectional relations between esti- mated social media use and life satisfaction ratings across the life span we plotted life satisfaction scores by age and estimated social media use (Fig. 1, top). Fur- thermore, we plotted the amount of estimated social media use by age (Supple- mentary Fig. 1, middle), which showed how at very young and very old age ranges the limited number of high intensity social media users led to the substantial increases in error bars that make interpretation of trends at these ages unfeasible. To test whether the functional forms relating social media use estimates to life satisfaction ratings differ by sex across different ages we used an Akaike Weights procedure36. We used the r-package lavaan to fit different versions of the model (life satisfaction ~ social media + social media2, also including control variables) to the data: a multigroup model that freed both linear and quadratic terms between sexes, a model that freed only the linear or quadratic terms and a model that constrained both terms. While there was a spread between what model was preferred in all ages, at ages 12, 13, and 14 the model constraining both the linear and quadratic term to be equal across sex was rejected (0% weight, Supplementary Fig. 10). When solely comparing a fully freed and fully constrained model we found that models that allowed sex variation in functional form were most favoured between 12 and 15 years (Fig. 1, bottom; Supplementary Fig. 3). In the extension of the cross-sectional analyses, we analyzed a range of questionnaires that were only completed by 10–15-year-olds in the Understanding Society survey and questionnaires completed by 13- and 14-year-olds in the Millennium Cohort Study; additional mental health questionnaires were analyzed in Supplementary Methods 1 and Supplementary Results 1, while we also examined a variety of life satisfaction measures in the main manuscript. For the latter, we plotted each life satisfaction question’s raw scores by social media use and age to examine whether a specific aspect of life satisfaction was more negatively related to estimated social media use (Fig. 2). We further used the Akaike weights procedure detailed above to examine the statistical evidence for sex differences. Longitudinal analyses. To model the data longitudinally we used a Random Intercept Cross-Lagged Panel Model comparison framework37, using the code structure provided by the Unified Framework of Longitudinal Models40. The model was selected due to its focus on within-person effects, without modelling general mean developments that have already been highlighted for both technology use and life satisfaction in adolescence53,54. The RI-CLPM model allowed us to focus on whether an individual’s deviation from their expected level of a certain variable y (e.g., life satisfaction ratings) can be predicted from their prior deviation from their expected scores in another variable x (e.g., estimated social media use), while controlling for the structural change in y (e.g., life satisfaction ratings); and vice versa. We added two control variables, average log household income and the Index for Multiple Deprivation across all waves of data available for each parti- cipant, to account for the socioeconomic status of both the family and their immediate environment. We did not include time-varying control variables, such as income, at every wave or year of data collection because the model could not be fitted with the level of missingness present in the data. We tested a variety of model constraints that force parameters to be equal across ages and sex, all of which were rejected: constraining the covariance of the residuals of latent factors after age 10 and constraining the residual variance of both social media and life satisfaction after age 10 (χ2(63) = 2023, p < 0.001); constraining the regression of the observed variables onto both mean IMD and mean log household income (χ2(92) = 178, p < 0.001); constraining the within- person carry-over effect for both social media estimates and life satisfaction ratings (also known as the within-person autoregression, positive carry-over means that a person who scores higher than their expected score is more likely to also score higher than their expected score in the following year; χ2(42) = 218, p < 0.001). A model constraining the cross-lagged paths of interest also suffered a significant drop in model fit: χ2(42) = 77.7, p = 0.001. Having set up our core model, we first estimated a ‘free model’ where we allowed both cross-lagged paths (a deviation in social media use predicting a deviation in life satisfaction one year later; and vice versa) to vary across age and sex. The model was fit using robust Maximum Likelihood (MLR) to account for deviations from multivariate normality, and robust Huber-White Standard errors, and missing data were accounted for by Full-Information Maximum Likelihood (FIML) estimation55. All RI-CLPM parameter tests are two-sided Walds tests. We note here that FIML cannot guarantee to give unbiased estimates with missing exogenous variables, i.e. our control variables, as those are assumed to be measured without error. The model fit the data well: χ2 (434) = 1216.29, p < 0.001, RMSEA = 0.014, [0.013, 0.015], CFI = 0.944, SRMR = 0.072. We extracted the value of the cross-lagged paths by age and sex, and plotted them in Fig. 3. We also fit this model to extended life satisfaction data collected for 10–15-year-olds and presented in Supplementary Fig. 7. We then used model comparison and Akaike weights to examine whether models that constrained one or two of the cross-lagged paths (social media use predicting life satisfaction and life satisfaction predicting social media use) to be constant across age and sex fit better than the initial freed model. The model that constrained only life satisfaction predicting social media use did not fit less well than a completely freed model (χ2(21) = 27.8, p = 0.15), while one that constrained only life satisfaction predicting social media use fit less well than the freed model (χ2(21) = 49.5, p < 0.001). The Akaike weights procedure showed that the model constraining only life satisfaction predicting social media use was more likely to be the best model for the data (99.2%), while the model freeing both (0.7%), the model constraining social media use predicting life satisfaction (0.0%) and the model constraining both (0.1%) were not (Supplementary Fig. 8). Finding that the cross-lagged path from life satisfaction predicting social media use can be constrained across sex and age without loss of model fit, we, therefore, fit a second model with this constraint. The model fit was acceptable (χ2 (455) = 1,234.74, p < 0.001, RMSEA = 0.01, [0.01, 0.02], CFI = 0.95, SRMR = 0.07). The model’s carry-over effect paths of life satisfaction (average unstandardized b: male = 0.179, female = 0.229; average standardized β: male = 0.178, female = 0.220; average SE: male = 0.048, female = 0.044) and social media use (average unstandardized b: male = 0.327, female = 0.373; average standardized β: male = 0.315, female = 0.359; average SE: male = 0.039, female = 0.035) were predominantly positive, suggesting that individuals who scored higher on life ARTICLE NATURE COMMUNICATIONS | 8 NATURE COMMUNICATIONS | (2022) 13:1649 | | satisfaction or social media use than expected in one year were also more likely to score higher than expected on life satisfaction or social media use one year later. The cross-lagged paths are reported in the main paper. We further ran 500 bootstrapped samples of this model to examine the uncertainty around our estimates. We then plotted the value of the cross-lagged path of social media use predicting life satisfaction by age and sex in Fig. 4. Reporting summary. Further information on research design is available in the Nature Research Reporting Summary linked to this article. Data availability The Understanding Society and Millennium Cohort Study data used in this study have been deposited in the UK Data Service database. They are accessible after registration with the UK Data Service and completion of an End User Agreement (Understanding Society: University of Essex, Institute for Social and Economic Research, NatCen Social Research, Kantar Public. (2019). Understanding Society: Waves 1-9, 2009-2018 and Harmonized BHPS: Waves 1-18, 1991-2009. [data collection]. 12th Edition. UK Data Service. SN: 6614, Millennium Cohort Study: University of London, Institute of Education, Centre for Longitudinal Studies. (2020). Millennium Cohort Study: Sixth Survey, 2015. [data collection]. 6th Edition. UK Data Service. SN: 8156). Some of the data used in our study might not be accessible to users outside the UK. Please check the terms and conditions for each of the datasets prior to use on the UK Data Service. The Special License data necessary to calculate the Index of Multiple Deprivation in Understanding Society are available after an approval procedure through the UK Data Service. The figure and table data generated in this study are provided in the Source Data file. Amy Orben accessed all datasets. Understanding Society is funded by the Economic and Social Research Council and various Government Departments, with scientific leadership by the Institute for Social and Economic Research, University of Essex, and survey delivery by NatCen Social Research and Kantar Public. The Millennium Cohort Study is funded by grants from the UK Economic and Social Research Council. Source data are provided with this paper. Code availability The code for all analysis and data cleaning are available on the Open Science Framework: Received: 12 January 2021; Accepted: 1 March 2022; References 1. Grimes, T., Anderson, J. A. & Bergen, L. Media Violence and Aggression: Science and Ideology (SAGE, 2008). 2. Bell, V., Bishop, D. V. M. & Przybylski, A. K. The debate over digital technology and young people. BMJ (Clin. Res. ed.) 351, h3064 (2015). 3. Orben, A. & Przybylski, A. K. The association between adolescent well-being and digital technology use. Nat. Hum. Behav. 3, 173–182 (2019). 4. Odgers, C. L. & Jensen, M. R. Annual Research Review: Adolescent mental health in the digital age: Facts, fears, and future directions. J. Child Psychol. Psychiatry 61, 336–348 (2020). 5. Twenge, J. M., Joiner, T. E., Rogers, M. L. & Martin, G. N. Increases in depressive symptoms, suicide-related outcomes, and suicide rates among U.S. adolescents after 2010 and links to increased new media screen time. Clin. Psychol. Sci. 6, 3–17 (2017). 6. Hancock, J., Liu, S. X., French, M., Luo, M. & Mieczkowski, H. Social Media Use and Psychological Well-Being: A Meta-Analysis (International Communications Association, 2019). 7. Appel, M., Marker, C. & Gnambs, T. Are social media ruining our lives? A review of meta-analytic evidence. Rev. Gen. Psychol. 24, 60–74 (2020). 8. Vanman, E. J., Baker, R. & Tobin, S. J. The burden of online friends: The effects of giving up Facebook on stress and well-being. J. Soc. Psychol. 158, 496–508 (2018). 9. Allcott, H. et al. The Welfare Effects of Social Media. site/allcott/research. (2019). 10. Orben, A., Dienlin, T. & Przybylski, A. K. Social media’s enduring effect on adolescent life satisfaction. Proc. Natl Acad. Sci. USA 116, 10226–10228 (2019). 11. Heffer, T., Good, M., Daly, O., MacDonell, E. & Willoughby, T. The longitudinal association between social-media use and depressive symptoms among adolescents and young adults: An empirical reply to Twenge et al. (2018). Clin. Psychological Sci. 7, 462–470 (2019). 12. Jensen, M., George, M. J., Russell, M. R. & Odgers, C. L. Young adolescents’ digital technology use and mental health symptoms: Little evidence of longitudinal or daily linkages. Clin. Psychol. Sci. 2167702619859336 (2019). 13. Davies, S. C., Atherton, F., Calderwood, C. & McBride, M. United Kingdom Chief Medical Officers’ commentary on ‘Screen-based activities and children and young people’s mental health and psychosocial wellbeing: a systematic map of reviews’. Department of Health and Social Care. government/uploads/system/uploads/attachment_data/file/777026/UK_CMO_ commentary_on_screentime_and_social_media_map_of_reviews.pdf (2019). 14. Viner, R., Davie, M. & Firth, A. The health impacts of screen time: a guide for clinicians and parents. rcpch_screen_time_guide_-_final.pdf?msclkid=ccac1ebca61811ecae3b67 a507b94d20 (2019). 15. Dickson, K. et al. Screen-based activities and children and young people’s mental health and psychosocial wellbeing: A systematic map of reviews. http:// (2018). 16. Hawkes, N. CMO report is unable to shed light on impact of screen time and social media on children’s health. BMJ 364, l643 (2019). 17. Valkenburg, P. M. & Peter, J. The differential susceptibility to media effects model: Differential susceptibility to media effects model. J. Commun. 63, 221–243 (2013). 18. Sawyer, S. M., Azzopardi, P. S., Wickremarathne, D. & Patton, G. C. The age of adolescence. Lancet Child Adolesc. Health 2, 223–228 (2018). 19. Blakemore, S.-J. & Mills, K. L. Is adolescence a sensitive period for sociocultural processing? Annu. Rev. Psychol. 65, 187–207 (2014). 20. Pfeifer, J. H. & Allen, N. B. Puberty initiates cascading relationships between neurodevelopmental, social, and internalizing processes across adolescence. Biol. Psychiatry (2020). 21. Andrews, J. L., Ahmed, S. P. & Blakemore, S.-J. Navigating the social environment in adolescence: The role of social brain development. Biol. Psychiatry (2020). 22. Crone, E. A. & Konijn, E. A. Media use and brain development during adolescence. Nat. Commun. 9, 1–10 (2018). 23. Ofcom. Children and parents: media use and attitudes report 2018. 1–17 childrens/children-and-parents-media-use-and-attitudes-report-2018 (2019). 24. Pew Research Centre. A quarter of Americans are online almost constantly | Pew Research Center. about-three-in-ten-u-s-adults-say-they-are-almost-constantly-online/? msclkid=e6109554a61811ec875c44156f6f50c7 (2018). 25. Campbell, O., Bann, D. & Patalay, P. The gender gap in adolescent mental health: a cross-national investigation of 566,827 adolescents across 73 countries. (2020). 26. Orben, A., Lucas, R. E., Fuhrmann, D. & Kievit, R. Trajectories of adolescent life satisfaction. (2020). 27. Salk, R. H., Hyde, J. S. & Abramson, L. Y. Gender differences in depression in representative national samples: Meta-analyses of diagnoses and symptoms. Psychol. Bull. 143, 783–822 (2017). 28. Booker, C. L., Kelly, Y. J. & Sacker, A. Gender differences in the associations between age trends of social media interaction and well-being among 10–15 year olds in the UK. BMC Public Health 18, 321 (2018). 29. Kreski, N. et al. Social media use and depressive symptoms among United States adolescents. J. Adolescent Health j.jadohealth.2020.07.006 (2020). 30. Nesi, J. & Prinstein, M. J. Using social media for social comparison and feedback-seeking: Gender and popularity moderate associations with depressive symptoms. J. Abnorm Child Psychol. 43, 1427–1438 (2015). 31. Viner, R. M. et al. Roles of cyberbullying, sleep, and physical activity in mediating the effects of social media use on mental health and wellbeing among young people in England: a secondary analysis of longitudinal data. Lancet Child Adolesc. Health 3, 685–696 (2019). 32. Twenge, J. M. & Martin, G. N. Gender differences in associations between digital media use and psychological well-being: Evidence from three large datasets. J. Adolescence 79, 91–102 (2020). 33. Kelly, Y., Zilanawala, A., Booker, C. & Sacker, A. Social media use and adolescent mental health: Findings from the UK millennium cohort study. EClinicalMedicine 6, 59–68 (2019). 34. University of Essex, Institute for Social and Economic Research. Data from “Understanding Society: Waves 1-8, 2009–2017 and Harmonised BHPS: Waves 1-18, 1991–2009.” UK Data Service SN-6614-12 (2018). 35. Przybylski, A. K. & Weinstein, N. A large-scale test of the Goldilocks hypothesis. Psychol. Sci. 28, 204–215 (2017). 36. Wagenmakers, E.-J. & Farrell, S. AIC model selection using Akaike weights. Psychonomic Bull. Rev. 11, 192–196 (2004). 37. Hamaker, E. L., Kuiper, R. M. & Grasman, R. P. P. P. A critique of the cross- lagged panel model. Psychol. Methods 20, 102–116 (2015). 38. Kievit, R. A., Frankenhuis, W. E., Waldorp, L. J. & Borsboom, D. Simpson’s paradox in psychological science: A practical guide. Front. Psychol. 4, 513 (2013). NATURE COMMUNICATIONS | ARTICLE NATURE COMMUNICATIONS | (2022) 13:1649 | | 9 39. Rosseel, Y. lavaan: An R package for structural equation modeling. J. Stat. Softw. 48, 1–36 (2012). 40. Usami, S., Murayama, K. & Hamaker, E. L. A unified framework of longitudinal models to examine reciprocal relations. Psychol. Methods 24, 637–657 (2019). 41. Gollob, H. F. & Reichardt, C. S. Taking account of time lags in causal models. Child Dev. 58, 80 (1987). 42. Dormann, C. & Griffin, M. A. Optimal time lags in panel studies. Psychol. Methods 20, 489–505 (2015). 43. Gignac, G. E. & Szodorai, E. T. Effect size guidelines for individual differences researchers. Personal. Individ. Differences 102, 74–78 (2016). 44. Van Der Maas, H. L. J. et al. A dynamical model of general intelligence: The positive manifold of intelligence by mutualism. Psychol. Rev. 113, 842–861 (2006). 45. Arnett, J. J. Emerging adulthood: A theory of development from the late teens through the twenties. Am. Psychologist 55, 469–480 (2000). 46. Baltes, P. B., Staudinger, U. M. & Lindenberger, U. Lifespan psycchology: Theory and application to intellectual functioning. Annu. Rev. Psychol. 50, 471–507 (1999). 47. Lachman, M. E. Development in midlife. Annu. Rev. Psychol. 55, 305–331 (2004). 48. Baltes, M. M. & Carstensen, L. L. The process of successful ageing. Ageing Soc. 16, 397–422 (1996). 49. Kross, E. et al. Facebook use predicts declines in subjective well-being in young adults’. Edited by Cédric Sueur. PLoS One 8, e69841 (2013). 50. Lynn, P. Sample design for Understanding Society https://www. 2009-01.pdf (2009). 51. University of London, Institute for Education, C. for L. S. Millennium Cohort Study: Sixth Survey, 2015. SN: 8156 2 (2017). 52. Rhemtulla, M., Brosseau-Liard, P. É. & Savalei, V. When can categorical variables be treated as continuous? A comparison of robust continuous and categorical SEM estimation methods under suboptimal conditions. Psychol. Methods 17, 354–373 (2012). 53. Ofcom. Children and parents: media use and attitudes report 2018. https:// children-and-parents-media-use-and-attitudes-report-2018 (2019). 54. Orben, A., Lucas, R. E., Fuhrmann, D. & Kievit, R. Trajectories of adolescent life satisfaction. (2020). 55. Enders, C. K. A primer on maximum likelihood algorithms available for use with missing data. Struct. Equ. Modeling: A Multidiscip. J. 8, 128–141 (2001). Acknowledgements This research was supported by a College Research Fellowship from Emmanuel College, University of Cambridge (A.O.), UK Medical Research Council MRC SWAG/ 076.G101400 (A.O.), the UK Economic and Social Research Council ES/T008709/1 (A.O. and A.K.P.), the Huo Family Foundation (A.K.P.), Wellcome Trust WT107496/ Z/15/Z (S.J.B.), the Jacobs Foundation (S.J.B.), the Wellspring Foundation (S.J.B.) the University of Cambridge (S.J.B.) and the UK Medical Research Council SUAG/047 G101400 (R.A.K). Author contributions Conceptualization: A.O., A.K.P., S.J.B., and R.A.K. Data curation: A.O. Formal analysis: A.O. and R.A.K. Software: A.O. and R.A.K. Supervision: A.K.P., S.J.B., and R.A.K. Visualization: A.O. Writing - original draft preparation: A.O. Writing - review & editing: A.O., A.K.P., S.J.B., and R.A.K. Competing interests The authors declare no competing interests. Additional information Supplementary information The online version contains supplementary material available at Correspondence and requests for materials should be addressed to Amy Orben. Peer review information Nature Communications thanks Candice Odgers and the other, anonymous, reviewers for their contribution to the peer review of this work. Peer reviewer reports are available. Reprints and permission information is available at Publisher’s note Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations. Open Access This article is licensed under a Creative Commons Attribution 4.0 International License, which permits use, sharing, adaptation, distribution and reproduction in any medium or format, as long as you give appropriate credit to the original author(s) and the source, provide a link to the Creative Commons license, and indicate if changes were made. The images or other third party material in this article are included in the article’s Creative Commons license, unless indicated otherwise in a credit line to the material. If material is not included in the article’s Creative Commons license and your intended use is not permitted by statutory regulation or exceeds the permitted use, you will need to obtain permission directly from the copyright holder. To view a copy of this license, visit licenses/by/4.0/. © The Author(s) 2022 ARTICLE NATURE COMMUNICATIONS | 10 NATURE COMMUNICATIONS | (2022) 13:1649 | | Windows of developmental sensitivity to social media Results and discussion Methods Ethical approval Datasets Measures Understanding society Millennium cohort study Inclusion criteria Cross-sectional analyses Longitudinal analyses Reporting summary Data availability Code availability References Acknowledgements Author contributions Competing interests Additional information

Place your order
(550 words)

Approximate price: $22

Calculate the price of your order

550 words
We'll send you the first draft for approval by September 11, 2018 at 10:52 AM
Total price:
The price is based on these factors:
Academic level
Number of pages
Basic features
  • Free title page and bibliography
  • Unlimited revisions
  • Plagiarism-free guarantee
  • Money-back guarantee
  • 24/7 support
On-demand options
  • Writer’s samples
  • Part-by-part delivery
  • Overnight delivery
  • Copies of used sources
  • Expert Proofreading
Paper format
  • 275 words per page
  • 12 pt Arial/Times New Roman
  • Double line spacing
  • Any citation style (APA, MLA, Chicago/Turabian, Harvard)

Our guarantees

Delivering a high-quality product at a reasonable price is not enough anymore.
That’s why we have developed 5 beneficial guarantees that will make your experience with our service enjoyable, easy, and safe.

Money-back guarantee

You have to be 100% sure of the quality of your product to give a money-back guarantee. This describes us perfectly. Make sure that this guarantee is totally transparent.

Read more

Zero-plagiarism guarantee

Each paper is composed from scratch, according to your instructions. It is then checked by our plagiarism-detection software. There is no gap where plagiarism could squeeze in.

Read more

Free-revision policy

Thanks to our free revisions, there is no way for you to be unsatisfied. We will work on your paper until you are completely happy with the result.

Read more

Privacy policy

Your email is safe, as we store it according to international data protection rules. Your bank details are secure, as we use only reliable payment systems.

Read more

Fair-cooperation guarantee

By sending us your money, you buy the service we provide. Check out our terms and conditions if you prefer business talks to be laid out in official language.

Read more