Dear Author,

Thank you for using BluPencil, the online Proofing System.

While this PDF is auto-generated on the fly, the changes made here are saved as comments
or annotations, and figures may be of low resolution. This version will not be as-is published
online nor printed. Once the corrections are registered on BluPencil, the same will be validated
against the journal-specific styles by our editors, updated, and then only the final typeset
PDF will be generated for publication.

Thank you,
Team BluPencil
 Professional Psychology: Research and Practice
© 2024 American Psychological Association
ISSN: 0735-7028 [Link]

There Is No Evidence That Time Spent on Social Media Is Correlated With

Adolescent Mental Health Problems: Findings From a Meta-analysis
Christopher J. Ferguson1, Linda K. Kaye2, Dawn Branley-Bell3, and Patrick Markey4
1
Department of Psychology, Stenson University
2
Edge Hill University
3
Northumbria University
4
Villanova University

The issue of whether social media use does or does not influence youth internalizing mental health disorders
(e.g., anxiety, depression) remains a pressing concern for policymakers, parents, and psychologists.
Widespread claims suggest potentially harmful effects of social media use on youth. This was investigated
in a meta-analysis of 46 studies of youth social media use and mental health. Results indicated that the
current pool of research is unable to support claims of harmful effects for social media use on youth
internalizing disorders. Some types of methodological weaknesses, such as evident demand characteristics
and lack of preregistration, remain common in this area. It is recommended that caution is issued when
attributing mental health harm to social media use as the current evidence cannot support this.

Public Significance Statement

Policymakers, parents and health care professionals continue to worry whether social media use
contributes to mental health problems among youth. The present study finds that the evidence for such
beliefs is lacking, and social media use does not predict mental health problems in youth. It is not
unreasonable for parents to ask questions about children’s social media use, however at present, parents
may be misled by unsupportable rhetoric from policymakers and some professional guilds to believe that
the evidence for harm is greater than it is. Policymakers and professional guilds need to adopt more
cautious reporting standards when discussing social concerns for which evidence is weak.

Keywords: social media, youth, adolescents, mental health, suicide

Concerns about young people’s mental health have become (Orben, 2020). Moral panic refers to significant societal fear that
increasingly prevalent over the last decade. In response, commen- takes place around a perceived threat and a specific responsible
tators have been prone to attribute the reasons for this increase to agent or technology—but where the threat is exaggerated or
technological innovations that occurred concurrently during the misplaced (S. Cohen, 2011). Moral panics have been witnessed in
same period, such as social media. The tendency to blame perceived relation to the introduction of radio, television, and video games
social ills on new technology is often associated with moral panic (Orben, 2020).

[Link] Linda K. Kaye was named a contributor to “An Open

Susan Jane Simonian served as action editor. Letter to Mr. Mark Zuckerberg: A Global Call to Act Now on Child and
Christopher J. Ferguson [Link] Adolescent Mental Health Science" ([Link]
CHRISTOPHER J. FERGUSON (PhD) is a professor of psychology at Stetson letter-to-mark-zuckerberg/) published in December 2021. She has also
University in Orlando Florida. He researches media effects, video games, and contributed to proceedings of the U.K. Government Online Safety Bill
race and policing. ([Link]
LINDA K. KAYE (PhD) is an Associate head of Department in Christopher J. Ferguson played an equal role in conceptualization, data
Psychology. Her research interests include online worlds, emoji, and curation, formal analysis, project administration, writing–original draft,
social media. and writing–review and editing. Linda K. Kaye played an equal role in
DAWN BRANLEY-BELL (PhD) is associate professor of Cyberpsychology & conceptualization, data curation, writing–original draft, and writing–review
Director of the Psychology and Communication Technology Lab at and editing. Dawn Branley-Bell played an equal role in conceptualization,
Northumbria University. She is particularly interested in online behaviors writing–original draft, and writing–review and editing. Patrick Markey played
related to eating disorders and self-harm. an equal role in data curation, formal analysis, writing–original draft, and
PATRICK MARKEY (PhD) is a professor of psychology at Villanova writing–review and editing.
University. His research interests include video games, media, body image, CORRESPONDENCE CONCERNING THIS ARTICLE should be addressed to
and personality development. Christopher J. Ferguson, Department of Psychology, Stetson University,
Data are available on Open Science Framework and can be accessed at DeLand, FL 32729 United States. Email: cjfergus@[Link]

Template Version: 14 Jun 2024 ▪ 12:21 pm IST PRO-2023-0960_blupencil ▪ 18 July 2024 ▪ 1:42 pm IST
2 FERGUSON, KAYE, BRANLEY-BELL, AND MARKEY

A prominent sociotechnological innovation that dominated there are well-known issues regarding best research practices that can
the last decade was the advancement of social communicational falsely inflate effect sizes (Drummond et al., 2020).
technology, specifically the proliferation of social media sites into
our daily lives (Ofcom, 2022). This has led to the suggestion that Issue 1: Time Spent Using as a Measure of Social
social media may sometimes be used as a “scapegoat” to mask more Media Exposure
complex explanations for poor mental health in youth (Sharp, 2021).
Relatedly, there has been a significant public and academic debate There have been numerous theoretical explanations for why social
regarding the role of social media use on mental health, especially in media use might relate to mental (and physical) health variables.
relation to children and young people. Of course, a historical pattern Many of these tend toward the notion of time or volume being the
of focusing blame on technology (see Bowman, 2016) does not root of the explanation, in which time spent using social media is time
necessarily mean that the current technology, social media, does not displaced from engaging in other, presumably healthier, pursuits
have an actual pernicious effect on youth mental health. It is the (see recent review by Hall & Liu, 2022). While this makes some
hope that rigorous science can help policymakers and parents conceptual sense for physical health variables such as sleep quantity
answer this important question. (i.e., we cannot use social media while concurrently being asleep),
Discussion surrounding new technologies and mental health is it is perhaps less clear how this provides explanatory value for
commonplace in both public and academic debate and underpins mental health variables. However, despite this, it is commonplace for
numerous health and social policy agendas (House of Commons research in this area (and the “screen-time” topic more generally) to
Science and Technology Select Committee, 2019; U.K. Parliament, obtain measures of volume as the primary, and often sole measure, of
2023; US Department of Health and Human Services, 2023).1 social media use. A recent meta-analysis has illustrated that when
Specifically concerning academic debate, the issue of social media controlling for other relevant factors (e.g., best practice in research,
use and mental health remains polarized. While some researchers age, family environment), the “blunt” instrument of amount of screen
contend that social media presents a significant public health risk to time was not found to be significantly related to mental health
children and young people (e.g., Alonzo et al., 2021; Brailovskaia variables for young people (Ferguson et al., 2022); it is noted that this
et al., 2023; Haidt, 2020; Keles et al., 2020; Twenge & Campbell, analysis examined screen-time broadly and including young adults,
2019; Twenge et al., 2018), others express caution regarding the rather than social media use specifically for adolescents. The current
quality and consistency of the available evidence and argue that meta-analysis differs in several important ways. First, the time frame
there is limited scope to establish the causality or universality of is longer (2012–2022). Second, the 2022 meta-analysis focused on
mental health outcomes associated with social media use and indeed screen-time broadly, and some critiques offered suggested that it was
screen use more generally (e.g., Ferguson et al., 2022; Kaye et al., less focused on youth and social media use specifically. This meta-
2020; Odgers & Jensen, 2020; Orben et al., 2019, 2022; Valkenberg analysis focuses specifically on social media use. Third, this meta-
et al., 2022; Vuorre et al., 2021). Within these discussions, the analysis focuses specifically on youth aged 18 and younger. There
literature broadly tends to focus on mental health variables such as was an overlap of seven studies (15%) with a previous study of screen
depressive symptoms (Cunningham et al., 2021; (Keles et al., 2019) time more broadly (Ferguson et al., 2022).
), anxiety ( (Keles et al., 2019) ), loneliness (O’Day & Heimberg, We propose that it makes little sense to use blunt measures of
2021), self-esteem ( (Rounsefell et al., 2020) ), body image and screen time as a predictor for mental health and that any research
disordered eating (Marks et al., 2020), and suicide ideation in this space needs to be more nuanced in nature—in particular
(Twenge, 2020), and these often exist alongside related discussion paying attention to the type of usage (e.g., content viewed, type of
about physical well-being effects, such as sleep quality and quantity engagement) and how individual differences may influence potential
(Alonzo et al., 2021) and physical activity engagement impacts for users.
(Brailovskaia et al., 2023; Shimoga et al., 2019).
Issue 2: The Limits of Self-Report
Methodological Concerns As noted, many researchers adopt measures of social media
This topic is notoriously difficult to navigate when drawing volume—how they do so is often based mainly on their own
conclusions about how social media use might relate to mental health. interpretation due to the lack of consistent, validated measures.
This is largely because the literature presents mixed findings and Researchers typically request users’ time or frequency of social media
highly disparate perspectives. We note that the existing literature on use based on an average estimate or retrospectively from a recent
social media use and mental health is limited by several known time-frame reference point. This typically involves asking questions
conceptual and methodological shortcomings. These fall into several such as: “How many hours did you spend on social media last week?”
specific areas. First, measures of social media use often tend to focus However, previous research has shown that the reference point used
solely on time spent on social media rather than how time is spent on for this brings about discrepancies in the accuracy of these estimates
social media as a way to conceptualize social media use. That is, compared with objective log data or screen-time data (Ernala et al.,
scholars have yet to provide any consistent measurement (or indeed 2020; Parry et al., 2021). Namely, Ernala et al. (2020) found that the
conceptualization) about what “social media use” might actually reference point used when asking people to report their Facebook use
mean (Trifiro & Gerson, 2019). Second, studies often rely upon self-
1
report, which can be notoriously unreliable as a measure of any We focus our review of policy on the United States and the United
Kingdom, as we are most familiar with these, recognizing that other countries
type of screen use (Mahalingham et al., 2023). Third, studies rarely may follow myriad other policy approaches. We also note here that are our
consider differences across platforms and overrely on vague concepts literature review highlights key studies and is not intended to be a systematic
of “addiction” that remain controversial (Satchell et al., 2021). Last, literature review.

Template Version: 14 Jun 2024 ▪ 12:21 pm IST PRO-2023-0960_blupencil ▪ 18 July 2024 ▪ 1:42 pm IST
 FINDINGS FROM A META-ANALYSIS 3

made a difference in how accurate people were in their reports. People and mental health concerns inevitably incorporate a wide range
were found to be most accurate (albeit still not to an acceptable of elements. Both the constructs of social media use and mental
degree) when asked to report how much time per day, on average, health incorporate heterogeneous components, and some caution is
they use Facebook when multiple-choice options were available for warranted in discussing them as if they were unitary constructs with
them to select. As such, when the research literature has no consistent clear boundaries.
measurement for social media volume, discrepancies in findings are
likely to be attributed to measurement inconsistencies.
Related to this is the concern that self-report may be unreliable, The Present Study
owing both to poor memory and perhaps the social desirability of
Moral panic theory (Bowman, 2016; S. Cohen, 2011) posits that,
underreporting sedentary—or otherwise stigmatized—screen activi-
among other stakeholders, social sciences often play a critical role in
ties. For instance, Prince et al. (2020) found that self-report measures
enforcing the panic, particularly during early stages. This may lead
tend to perform poorly, in comparison to time diaries or devices that
to selective attention to studies that support the panic and incuriosity
directly measure screen use or other sedentary activities. In fact, a
or dismissal of studies that do not. It can also lead to overstatement
recent study by Mahalingham et al. (2023) found no relationship at all
of weak effect sizes or misapplication of poor-quality studies to real-
between self-reported social media use and objective usage measures.
life phenomena. Furthermore, it may lead to the poor use of meta-
It is possible, as such, that self-report studies may be particularly
prone to returning inaccurate and/or biased results. analysis, for example, where there is an overreliance on bivariate
correlations rather than controlled standardized regression coeffi-
cients. An “average effect size wins” approach may effectively put a
Issue 3: Failure to Distinguish Between Platforms and thumb on the scale in favor of the panic hypothesis.
Misuse of the “Addiction” Framework Clearly, there are critical conceptual and methodological issues that
A further issue is that while self-report tools have been used persist and will continue to adversely affect the quality of scientific
to measure aspects of social media (e.g., Jenkins-Guarnieri et al., evidence available on the associations between social media use and
2013; van den Eijnden et al., 2016), use can vary across platforms. mental health. It is important to note that evidence of social media
Unfortunately, most researchers either do not distinguish between effects on adverse mental health does not necessarily mean that
platforms in measurement tools or might focus on use in respect to positive effects cannot exist (e.g., social support; Branley & Covey,
only specific platforms (e.g., Facebook), which are unlikely to apply 2017; Sendra et al., 2020) and vice versa. We are sympathetic to some
or be consistent with other platforms. To further confound the area, a concerns relating to social media (e.g., age appropriateness, profit-
disproportionate amount of research tends to adopt an “addiction” or driven attention economy, use of data and algorithms, negative and
“problematic use” perspective of social media, whereby measure- extreme content). There may indeed be harms that exist, but we
ment relates to users’ attitudes about their (problematic) social remain vigilant to the nature of the scientific evidence that explores
media use. Equally, these types of instruments have brought about these. Of primary interest, we are committed to the need for high-
debate, specifically regarding their validity (Satchell et al., 2021), quality science to explore relevant concerns, particularly regarding
accuracy (H. Shaw et al., 2020), and researchers’ inconsistent use of mental health and well-being impacts. Further, we are also committed
scoring methods (Connolly et al., 2021). These issues are not unique to the need for this high-quality science to be the basis for informing
to social media research but also plague the wider topic of policy guidelines and social technology innovation surrounding
behavioral addictions (Billieux et al., 2015). mental health and social media use. Our concerns are that current
policy priorities are being influenced by an evidence base that
primarily consist of studies with significant methodological limita-
Issue 4: Best Research Practices
tions or failings. As such, our meta-analysis sought to establish the
Across media studies some poor practices have been identified nature of the existing literature base with two overarching aims:
that are prone to inflating effect sizes. For instance, the use of
unstandardized and poorly validated measures can increase effect 1. To establish the effect size between social media use and
sizes (Drummond et al., 2020). It is also important to control for mental health variables in adolescents. As observed, during
theoretically relevant control variables such as age, gender, and, in a period of moral panic, a potential arises that public
longitudinal studies, Time 1 outcome variable. For screen-time discourse may differ from the magnitude of effect observed
measures, using the same respondent for the predictor and outcome in published studies. Having a clearer understanding of that
variable, and close pairing of questions related to the predictor and magnitude of effect may help inform public discourse and
outcome, can create demand characteristics and artificially inflate policymaking.
effect sizes (Ferguson et al., 2022). Thus, for social media studies, it
2. To establish the prevalence of best practice in studies in
can be worth considering whether these issues impact effect sizes in
the existing literature and their impact on observed effect
empirical studies. One recommendation is to preregister studies to
sizes. As observed in our literature review, methodological
help reduce concerns about artificially inflated effect sizes due to
concerns ranging from overreliance on bivariate effects,
questionable researcher practices (Orben & Przybylski, 2019);
self-report surveys, and so forth may impact effect sizes
however, despite being widely encouraged as a good practice, at the
leading to over- or underconfidence in results.
moment, the majority of studies are not preregistered.
Some of these issues, particularly Issues 1 and 2, arguably
fall under concerns related to constructs and construct validity
(e.g., De Boeck et al., 2023). Both issues related to social media use

Template Version: 14 Jun 2024 ▪ 12:21 pm IST PRO-2023-0960_blupencil ▪ 18 July 2024 ▪ 1:42 pm IST
4 FERGUSON, KAYE, BRANLEY-BELL, AND MARKEY

Method control variables) available in each study or effects based on

experimental results (F value, t test, etc.). Raw interrater correlations
Open Science Practices
(r) between the recorded effect sizes was = .99. Kappa reliability for
A preregistered plan was completed, which outlined the search absolute agreement was .78.
strategy, criteria for inclusion and exclusion, and analysis plan. This Jamovi4 was used to calculate a random-effects mean effect size,
can be accessed at [Link] Data generated from the as well as to calculate risks of publication bias including basic funnel
meta-analysis can be accessed at [Link] This includes plot analysis, Egger’s regression, trim and fill, p-curve, and
the following information: full citation, sample size, effect size, best p-uniform. We used a restricted maximum-likelihood model with
research practices analyses, and moderator variables. A PRISMA Fisher’s r-to-z transformation. Random-effects models were used.
diagram for our search is available at [Link] A list of Given the high power of meta-analysis, almost all meta-analyses
included studies can be accessed at [Link] are “statistically significant.” Nonetheless, many small effects may
be statistical artifacts due to methodological issues such as demand
Selection of Studies characteristics or single responder bias. Consistent with the
recommendations of Orben and Przybylski (2019), we considered
We searched on APA PsycInfo and Medline using the terms an effect size of r = .10 as the minimum for practical significance.
(“Social Media” or “Facebook” or “Instagram” or “Twitter” or Ferguson and Heene (2021) also provided documentation for how
“snapchat” or “social networking” or “TikTok”) and (“depression” or effects below this threshold are unable to be distinguished from
“anxiety” or “loneliness” or “suicide” or “mental health” or “mental noise due to methodological precision issues.
well*” or “mental illness” or “mental well-being” or “psycho-
logical well-being”) and (“youth” or “teen*” or “adoles*”) as subject
searches. We limited our search to studies from 10 years old to the Best Research Practice Analysis
date of the search. The search took place in September 2022.
To enable us to assess the relevance of studies, we identified that To analyze the prevalence of best research practices adopted
they should meet the following inclusion criteria: include a measure within the literature and test whether this moderated the observed
of social media or experimental comparison of social media with a effect sizes, we utilized the following criteria (based on the criteria
control condition,2 have a sample only including participants used by Ferguson et al., 2022), from which a numeric score could be
between the ages of 12 and 18 and include sufficient information calculated and used in moderation analysis.
from which we could calculate an effect size “r.”3 We note a few Correlational/cross-sectional studies were given credit (1 point
samples with age ranges that were slightly younger (n = 5). each) for the following best research practices:
Deviating slightly from the preregistration, we decided to retain • used a standardized outcome measure
these, but this decision was made prior to examining the data.
Regarding social media time, most studies included self-report • used clinically validated measures (e.g., Child Behavior
surveys of time spent on social media, time diaries, or electronic Checklist) and, for social media use, those that were
recording systems. Most studies focused on platforms ranging from objective rather than subjective measures (e.g., screen-time
Facebook, Twitter, Instagram, Snapchat, and so forth, although apps or log data, time diaries being superior to estimates of
some also included video or streaming platforms such as YouTube use based on self-report.)
or Twitch.
Regarding outcomes, most outcomes were self-report indices of • used more than one type of respondent (e.g., parent
internalizing symptoms. These included inventories of depression and child)
and anxiety but also mental wellness, self-esteem, and satisfaction
• included distractor questionnaires to reduce demand
with life more broadly. Few studies employed clinical cutoffs or
characteristics
official diagnoses, so it may be best to think of the outcome as a
general cluster of mental wellness rather than clinical disorders. • controlled gender, age, and family environment (family
A PRISMA chart for our search is provided at the link in the Open conflict, stress, academic pressure from parents). For
Science Practices section. Our search ultimately netted 55 studies longitudinal studies, Time 1/baseline/pretest outcome vari-
on social media use and youth mental health. However, nine were able was also controlled.
subsequently found to be missing important data needed to calculate
an effect size, and either authors did not respond to requests for data • preregistered the analysis plan
or we were that informed the data were unavailable. This resulted
in a final pool of 46 studies. Between them, allowing for different
2
effects for boys and girls in some studies, these articles included To reduce noise in the data, we focused on time spent on social media
and so excluded studies that measured motivations for using social media or
79 total effect sizes. for what purposes social media was used and studies measuring “problematic
social media use.”
3
“r” is used here to denote the most controlled/conservative effect size
Effect Size Extraction and Calculation from each study, which in most cases (but not all) were standardized
Two authors extracted the effect sizes from each article from regression coefficients.
4
This does represent a slight deviation from our preregistration that
which we calculated interrater reliability. Effect size was calculated mentioned using comprehensive meta-analysis and shinyapps for calcula-
as a standardized regression coefficient (betas) based on the most tions. We have switched to jamovi during this time, which was unrelated to
conservative value (i.e., employing the most theoretically relevant the results of the meta-analysis.

Template Version: 14 Jun 2024 ▪ 12:21 pm IST PRO-2023-0960_blupencil ▪ 18 July 2024 ▪ 1:42 pm IST
 FINDINGS FROM A META-ANALYSIS 5

Experimental studies were given credit (one point each) for the significant heterogeneity between studies suggesting the potential
following best research practices: impact of moderators.
A table of all studies with the types of social media included is
• used a standardized outcome measure available at [Link] However, it is emphasized that
• used a clinically validated measure these social media forms were typically mentioned in questions or
prompts, and data for them were not typically collected or analyzed
• used a closely matched control condition differing only in separately.
independent variable-related content
• used distractor tasks to reduce demand characteristics Sex Differences
• included queries for hypothesis guessing A key potential moderator is sex, as it may be the case that
girls have more vulnerability to social media effects than boys
• preregistered the analysis plan (Twenge, 2020). As such, we considered biological sex differences
This allowed us to calculate a score that could be tested for in effect size. Initial results using a mixed-effects model (k =
potential moderator effects with effect size. This score allowed us to 56) were nonsignificant ( p = .074); however, resilience testing
examine whether study quality was associated with either increased suggested that model estimator had an impact on p value with
or decreased effect size, thus allowing us to understand how p values ranging from <.001 (Hunter-Schmidt method) to .111
methodological noise might be impacting research results. (Sidik-Jonkman method). An examination of effect sizes revealed
that effect sizes for girls were slightly larger (β = .075) than for
boys (β = .044) but that both effects fell below the threshold
Citation Bias Analysis for evidence.
Citation bias occurs when study authors only cite articles
supporting their hypotheses, failing to inform readers of
Other Moderator Analyses
inconsistencies in a research area. Previous meta-analyses have
often identified citation bias as a predictor of inflated effect sizes, Contrary to other meta-analyses (e.g., Ferguson et al., 2022), best
suggesting that citation bias may be one indicator of researcher research practices were not a continuous moderator of effect size
expectancy effects (e.g., Ferguson, 2015). In cases where the (though, as with biological sex, resiliency analysis suggested that
literature review included no citations that conflicted with the this depended on the model used) nor was the study year. The age of
authors’ hypotheses, they were coded as having citation bias. the participants in the sample was also nonsignificant ( p = .053).
However, if a article acknowledged at least one research study Citation bias proved not to be a moderator ( p = .319) and study type
or article conflicting with the authors’ hypotheses, they were not (correlational vs. longitudinal) was also nonsignificant ( p = .067).6
coded as having bias. The effect sizes for longitudinal studies were slightly smaller (β =
.044) than that of correlational studies (β = .072), albeit once again
Moderator Analysis all effect sizes were below the threshold for evidentiary value.
Contrary to our expectations, the use of self-report versus time
Several moderators were considered as part of this study. First, as diaries and other objective methods did not prove to be a moderator
indicated above, both best research practices and the presence of ( p = .430) though the type of data set (bespoke vs. large national vs.
citation bias were considered moderators. Second, some data sets, dissertation) did ( p = .043). In this case, bespoke data sets (β = .041)
such as “Monitoring the Future,” have produced multiple articles, and dissertations (β = .045) had smaller effect sizes than did national
often from different author groups and sometimes giving conflicting data sets (β = .067).
results. Reusing such data sets across multiple articles may give
undue weight to the methods in those studies. In the present study,
we sought to address this by considering the data set as a potential Best Research Practices
moderator, particularly examining whether effect sizes differed The utilization of best research practices varied. Some, such as
between large multiuse national data sets versus bespoke data sets the use of standardized and well-validated measures of mental
used in some individual studies that were not repeated.5 Third, the health, were very common among reported effect sizes (approxi-
age of the study’s participants and differences between boys and mately 95% and 92% of studies, respectively). The use of basic
girls were also considered possible moderators, as was the study controls for gender, age, family environment, and Time 1 outcomes
year and whether the study employed self-report data or other in longitudinal studies was also common (64%) though not as
methods such as time diaries.
5
This differs from the intent in our preregistration where we had hoped to
Results extract a single effect size from each data set. However, many of the debates
in this area considered the most appropriate way to do so. As such, we
Table 1 presents the results for all analyses. Figure 1 presents a decided to include all articles but include potential multiuse data sets as a
funnel plot for all studies. An analysis of all effect sizes suggested moderator.
6
that the mean effect of social media on mental health across studies There was only one experimental study; thus, this was not included in the
analysis.
was near zero (β = .061) and below our threshold for evidentiary
value. This suggests that observed effects are indistinguishable from
statistical noise (Ferguson & Heene, 2021). However, there also was

Template Version: 14 Jun 2024 ▪ 12:21 pm IST PRO-2023-0960_blupencil ▪ 18 July 2024 ▪ 1:42 pm IST
6 FERGUSON, KAYE, BRANLEY-BELL, AND MARKEY

Table 1
Meta-analytic Results of Social Media and Mental Health Outcomes

Effect size k β 95% CI Homogeneity test I2 τ Publication bias?

2
All studies 79 .061 [.047, .075] X (78) = 4404.45, p < .001 98.8 .055 No
Biological sex
2
Male 27 .044 [.025, .062] X (26) = 164.79, p < .001 94.7 .040 No
Female 29 .075 [.050, .101] X2(28) = 388.10, p < .001 97.9 .063 No
Study type
Correlational 48 .072 [.05, .090] X2(47) = 4167.83, p < .001 99.3 .057 No
Longitudinal 30 .044 [.023, .066] X2(29) = 169.54, p < .001 83.6 .049 No
Data set
Bespoke 21 .044 [.012, .070] X2(12) = 167.61, p < .001 89.2 .046 No
National survey 53 .067 [.050, .084] X2(52) = 99.21, p < .001 99.2 .058 Yes
Dissertation 5 .045 [.016, .074] X2(4) = 1.81, p = .770 57.2 0 No
Note. k = number of studies; β = pooled effect size estimate; CI = confidence interval; I2 = heterogeneity statistic.

much as expected. By contrast, the use of multiple respondents Discussion

(19%), distractor questions or tasks (0% reported), preregistra-
tion (5%), or careful querying for hypothesis guessing (1%)7 We explored the existing literature on social media use and
were very rare. Citation bias was present for 15% of reported impact on adolescent mental health. We contend this to be important
effect sizes. to gauge the quality of the existing evidence, especially as this
research often informs public policy, health guidelines, and social
technology innovation relating to user well-being on social media.
Publication Bias It also has the potential to fuel moral panic if results are overstated or
misconstrued. As such, our meta-analysis addressed two overarch-
Evidence generally suggested an absence of publication bias in ing aims:
this area. However, this result is cautioned on the observation that
publication bias measures tend to be underpowered, particularly • to establish the effect size between social media use and
concerning large data sets with small effect sizes wherein p values of mental health variables in adolescents
.05 are easily surpassed, making bias more challenging to detect with
• to establish the prevalence of best research practices in
measures dependent upon p values. Evidence from Egger’s regression
studies in the existing literature and their impact on observed
(p = .021) and trim and fill (missing studies: four) suggested potential
effect sizes
for publication bias in this meta-analytic data set.

Figure 1
Funnel Plot for All Studies

Template Version: 14 Jun 2024 ▪ 12:21 pm IST PRO-2023-0960_blupencil ▪ 18 July 2024 ▪ 1:42 pm IST
 FINDINGS FROM A META-ANALYSIS 7

Overall Findings the public statements of some scholars, advocates, and policy-
makers. This pattern of discrepancy between research results and
Overall, our findings indicate that the current research literature is
exaggerations of evidence in support of a panic theory is consistent
unable to provide strong evidence for a clinically relevant link
with other media fields that have gone through periods of moral
between time spent on social media and mental health issues in youth.
panic, such as those related to television (Freedman, 2002) or violent
This was true for both boys and girls and across both correlational and
video games (Ferguson, 2015).
longitudinal studies. As such, we observe that public statements
We believe it is incumbent upon researchers to be more aware
coded toward warning language, such as those provided by the
of this historical pattern. Indeed, though we understand that tea-
American Psychological Association (2023) or the US Department of
ching time is limited, it may be worth specifically including it in
Health and Human Services (2023), are not faithful to the research
curriculum related to social psychology, developmental psychology,
data as it currently stands. To be fair, the American Psychological
media psychology, and cyberpsychology specifically. Academic-
Association (2023) did note “Using social media is not inherently
fueled moral panics have significant capacity to distract societies
beneficial or harmful to young people” (p. 3), though its narrative
from more pressing issues, giving this matter at least some degree of
generally covers evidence for harm, playing less attention to critiques
urgency. Given that this pattern repeats with each new technology, it
of that view or null studies.8 The US Department of Health and
is clear that researchers have not learned to take this historical view
Human Services (2023) also briefly mentions benefits of social
and become more cautious in advancing theories related to the
media use, but neither report outlines how they searched for studies,
pernicious impacts of new technologies.
suggesting that their coverage of research may be selective and
One thing that is evident is that in some cases, different scholars are
nonsystematic. Put more plainly, the current concerns about social
examining the same data and coming to very different conclusions.
media appear consistent with a pattern of overstatement by scholarly
This, too, is expected under moral panic theory. Part of the issue is
and governmental groups; this may be related to moral panic
that social science lacks a clear rigor in regard to evaluating effect
(Bowman, 2016).
sizes. Thus, when large sample studies find statistically significant
Compared with other research topics that have experienced
results with weak effect sizes, it can be possible for scholars to (in
moral panic (e.g., Ferguson, 2015), our meta-analysis suggested two
good faith) ignore warnings about the overinterpretation of weak
interesting differences. First, citation bias was not a predictor of effect
effect sizes (e.g., Wilkinson & Task Force on Statistical Inference,
sizes nor were best research practices. It appears that even advocates
American Psychological Association, Science Directorate, 1999), so
of the technology harm approach (e.g., Twenge, 2020) faithfully
long as statistical significance has been achieved (which it almost
acknowledge inconsistent research and, for that, deserve considerable
always is in large sample studies, no matter the variables; see
credit. On the contrary, the issue of best research practices is mixed.
Some best research practices, such as the use of well-validated Ferguson & Heene, 2021). Thus, beliefs in a harm hypothesis can be
measures or control variables, appear to be more common than in maintained, even if the evidence is negligible, particularly when there
other areas, such as media violence (Freedman, 2002) or body is social pressure and incentive to do so.
dissatisfaction and media (Want, 2014). However, other best research In noting that a cycle of moral panic has emerged among policy-
practices, such as preregistration, were rare, and the area remains makers and professional guilds9 such as the American Psychological
dependent on studies with high-demand characteristics and few Association and American Psychiatric Association, we do not mean to
checks for unreliable or mischievous responding. In this sense, our imply that parents are wrong to have questions regarding the potential
best research practice approach may have lacked much by way of impacts of social media use. Nor are scientists wrong to study
variances as both strengths and weaknesses of the area are spread hypotheses about potential relationships between social media use and
rather evenly. mental health (as indeed we study authors do ourselves). However,
Thus, the main issue for this field of research appears to be a failure parents arguably have been let down by societal leaders and, as often
to accurately communicate two key things: between-study incon- as social media is (rightly) criticized for misinformation, so too many
sistencies and overall weak effect sizes that are largely universal societal leaders (whether policymakers or mental health experts) have
between studies. In this sense, this area appears to have fallen misinformed parents regarding effects of social media use on young
for a common problem in research psychology of overinterpreting people. There may be legitimate concerns regarding some aspects of
statistically significant but near-zero effect sizes in large samples. social media, particularly the poor quality of information that youth
Though these problems have been known for some time (e.g., J. are exposed to. However, it is difficult to offer concrete solutions to
Cohen, 1994; Wilkinson & Task Force on Statistical Inference, these legitimate concerns until those who speak on behalf of science
American Psychological Association, Science Directorate, 1999),
7
they persist particularly when narratives of “harm” have political and It is possible that queries for hypothesis guessing or distractor
tasks may have been used in some studies but went unreported in the
social purchase during periods of moral panic. manuscripts.
We note that this work extends the findings of Ferguson et al. 8
See also the November 2023 letter from the American Psychiatric
(2022). This current meta-analysis clarifies that negligible evidence Association to the Department of Commerce: [Link]
for effects also extends to the specific concerns related to social tattachment/8861f2e3-2fbc-4d8b-9a26-b5e49dd7eaab/APA-Letter-NTIA-
media use among youth. [Link]
9
We are aware that such organizations may object to being referred to as
“guilds,” perhaps preferring to viewing themselves as having complex roles
including professional guilds but also science organizations and publishing
Moral Panic houses. However, given the long-standing concerns about how well such
organizations function as science organizations, not limited to this issue (e.g.,
We observe that there is a divergence between the quality, O’Donohue & Dyslin, 1996), we are confident in using the term guild rather
consistency, and effect magnitude of research results in this field and than “science organization.”

Template Version: 14 Jun 2024 ▪ 12:21 pm IST PRO-2023-0960_blupencil ▪ 18 July 2024 ▪ 1:42 pm IST
8 FERGUSON, KAYE, BRANLEY-BELL, AND MARKEY

(even if these are professional guilds, not science organizations) are impacts in youth. It is our observation that such efforts by the
not guilty of providing significant misinformation on the technology government, to the extent that they rely on such claims, are out of sorts
at the core of the concern. with the current available data that do not suggest any currently
support claims of “harm” by social media for youth wellness. We are
cautious about the fact that public policymaking and discussions
Future Directions and Policy Implications
are unfolding without the robust backing of empirical research. This
Social media use may arguably be better understood not just from leads to a situation where the consequences, especially regarding
“how much” (i.e., volume and/or frequency of use) but also through government regulation of media under the pretext of shielding from
understanding the what, how, and why behind users’ use and “harm,” remain largely uncertain.11
behaviors on such platforms (Kaye, 2022). Namely, what type of
content users are accessing or engaging in (what), the types of
behaviors or interactions in which they engage (how), and why they Limitations
might be using social media at a given point in time (why). As We note some limitations of our research. One limitation that
such, we recognize that our inferences regarding the lack of strong extends to any meta-analytic research includes the fact that the
evidence about the relationships between social media use and quality of the research is determined by the quality of studies they
mental health variables are based on the way scholars have incorporate. As previously noted, this topic is plagued by a range
operationalized measures of these constructs which might, in some of methodological issues such as demand characteristics, lack of
cases, not fully capture the complexities of the social media behavior preregistration, and few reliability checks, and so it is likely that
or engagement. For example, recent research has noted the relevance effect sizes are inflated by these issues, which is equally attributable
of understanding specific interactions and behaviors when measur- to our findings reported here. However, promising advancements in
ing social media use (Meier & Reinecke, 2021; Trifiro & Gerson, the area are beginning to overcome some of these methodological
2019). Furthermore, recent findings have established that these shortcomings. Namely, while the literature has historically been
may bring about differential impacts on psychosocial functioning overreliant on cross-sectional designs (which fail to capture causal
(D. J. Shaw et al., 2022; Valkenberg et al., 2022). For example, effects) and retrospective self-report measures of social media use
when objectively measuring social media-related behavior, more (which can lead to inaccurate and ill-defined measurements), more
interactive behaviors relative to more passive or reactive ones are recent research is making use of more advanced study approaches
associated with greater feelings of social connectedness and social such as experience sampling to better capture person-specific factors
capital (D. J. Shaw et al., 2022). As such, research that exclusively in social media use over time and context (see Valkenberg et al.,
measures the volume of social media use is failing to capture key 2022). These will be better positioned to offer a more authoritative
nuances that might otherwise elucidate varying relationships account of the nuances of these issues and explore the possibility that
between social media use and mental health. there is no linear or uniform relationship between social media use
Acknowledging nuances in user behavior and influence on and mental health, with outcomes instead being based on a complex
outcomes indicates that approaches that focus on banning or and diverse range of factors. Although samples vary regarding their
restricting social media by age are unlikely to be of value. First, there representativeness of a wide range of ethnicities and cultures, more
is little empirical support for such approaches, and they may backfire studies with underrepresented groups would be helpful, as ethnic
insofar as limiting youth ability to adjust to this technology when differences in outcome may be possible. An additional limitation
parental and teacher influences on good practices might be most was that we observed a low level of variance within the best practice
influential. Banning approaches may also create a forbidden fruit analysis, which limited our ability to provide a full exploration of
phenomenon that can make access more enticing; may result in users these potential effects. Best practices were either almost ubiquitous
hiding access from their parents, guardians, or teachers (Kerr & (e.g., standardization) or entirely lacking (e.g., preregistration). We
Stattin, 2003); or may impact on positive outcomes related to usage dutifully report our best practices analysis as it was preregistered
such as peer support and/or mental health recovery information but observe that our results are inevitably impacted by this lack of
(e.g., Branley & Covey, 2017). Approaches that focus on education, variance. We note that controlling for best practice typically reduces
media literacy, and helping adolescents develop good practices on observed effect sizes (Ferguson et al., 2022), so the current findings
social media are likely to be more constructive. Just as abstinence might not fully represent the potential magnitude of these expected
approaches in sex education often fail to reduce teen pregnancy, effects. Last, we acknowledge that by focusing our search on APA
abstinence approaches to social media may likewise be less than PsycInfo and Medline, it is possible that studies from other fields
productive. may have been missed.
Interestingly, since this meta-analysis was originally conducted, One possible explanation for the lack of findings may be that all
41 state attorney generals in the United States have sued the company youth have been exposed to social media and we may be looking at a
Meta (of Facebook and Instagram), claiming that the social media
sites do harm to minors (see (Lima-Strong & Nix, 2023) ). This case 10
For example, Florida Attorney General Ashley Moody released a
rests upon multiple concerns, not only mental health and “addiction” statement: “Meta has gone unchecked for too long, and our children are
but also related to privacy concerns. However, mental health claims suffering the consequences of these unlawful practices. Today, I took action
appear central to the case.10 We do not seek to dismiss the possibility to stop Meta from targeting minors with addictive features to keep them
that some harms outside the realm of internalizing disorders studied in online for hours, collecting their data and other unlawful actions that harm
teens’ mental health,” (from: [Link]
the current analysis may be possible. However, it is interesting to note attorney-general-moody-takes-legal-action-against-meta-protect-children).
that our meta-analysis does not suggest that the empirical literature 11
We observe that censorship efforts often come in the guide of protecting
supports a relationship between social media use and mental health one “vulnerable” group or another.

Template Version: 14 Jun 2024 ▪ 12:21 pm IST PRO-2023-0960_blupencil ▪ 18 July 2024 ▪ 1:42 pm IST
 FINDINGS FROM A META-ANALYSIS 9

cohort effect that is difficult to detect using between-subjects designs Branley, D. B., & Covey, J. (2017). Pro-ana versus pro-recovery: A content
due to saturation. However, we find this argument unconvincing analytic comparison of social media users’ communication about eating
for several reasons. First, most critics of social media companies, disorders on Twitter and Tumblr. Frontiers in Psychology, 8, Article 1356.
including professional guilds and policymakers, do rely on between- [Link]
subjects studies, albeit we observe often in exaggerated or selective Cohen, J. (1994). The Earth is round (p <. 05). American Psychologist,
49(12), 997–1003. [Link]
ways consistent with past patterns of media moral panics such as for
Cohen, S. (2011). Folk devils and moral panics (1st ed.). Routledge. https://
video games (Bowman, 2016). It is unscientific to rely on such studies
[Link]/10.4324/9780203828250
when convenient but dismiss them when not. Naturally, youth will Connolly, T., Atherton, G., Cross, L., & Kaye, L. K. (2021). The Wild West
vary among each how much time is spent using social media. If of measurement: Exploring problematic technology use cut off scores and
simple exposure to social media is a key issue of concern as has been their relation to psychosocial and behavioural outcomes in adolescence.
expressed by policymakers, professional guilds, and some scholars, Computers in Human Behavior, 125, Article 106965. [Link]
then it is not unreasonable to see some appreciable relationship .1016/[Link].2021.106965
between time spent on social media and mental health concerns. Cunningham, S., Hudson, C. C., & Harkness, K. (2021). Social media
Second, neither cohort comparison studies (Ferguson, 2021; Vuorre and depression symptoms: A meta-analysis. Research on Child and
et al., 2021), nor time-series analysis (Padmanathan et al., 2020), Adolescent Psychopathology, 49(2), 241–253. [Link]
nor, for that matter, studies of functional brain organization (Miller s10802-020-00715-7
De Boeck, P., Pek, J., Walton, K., Wegener, D. T., Turner, B. M., Andersen,
et al., 2023) support the concern that social media use has had an
B. L., Beauchaine, T. P., Lecavalier, L., Myung, J. I., & Petty, R. E. (2023).
appreciable adverse impact on youth well-being. Thus, arguments Questioning psychological constructs: Current issues and proposed
about a cohort saturation effect appear to be unwarranted at present. changes. Psychological Inquiry, 34(4), 239–257. [Link]
1047840X.2023.2274429
Drummond, A., Sauer, J. D., & Ferguson, C. J. (2020). Do longitudinal
Conclusion studies support long-term relationships between aggressive game play and
Our findings illuminate that based on the analyzed literature, the youth aggressive behaviour? A meta-analytic examination. Royal Society
observed effect size for the relationship between social media use and Open Science, 7(7), Article 200373. [Link]
Ernala, S. K., Burke, M., Leavitt, A., & Ellison, N. B. (2020). How well
adolescent mental health is below the evidentiary threshold and is
do people report time spent on Facebook? [Conference session]. CHI
as likely due to methodological noise as any actual effect. Although Conference on Human Factors in Computing Systems, Honolulu, HI,
some best research practices are widespread in the area (i.e., well- USA, April 2020. [Link]
validated outcomes measures and employment of control variables), Ferguson, C. J. (2015). Do angry birds make for angry children? A meta-
others are not (i.e., controls for demand characteristics, preregistra- analysis of video game Influences on children’s and adolescents’
tion, reliability checks) and do not meet standard expectations to be aggression, mental health, prosocial behavior and academic performance.
considered high-quality research. This is somewhat concerning given Perspectives on Psychological Science, 10(5), 646–666. [Link]
that many of the lower quality studies are those which are drawn upon .1177/1745691615592234
as an evidence base for informing policy and practice surrounding Ferguson, C. J. (2021). Links between screen use and depressive symptoms
social media and mental health. We emphasize that improvements in adolescents over 16 years: Is there evidence for increased harm?
must be made to ensure that the future scientific evidence base is of Developmental Science, 24(1), Article e13008. [Link]
sc.13008
sufficient quality to ensure that our understanding of the potential
Ferguson, C. J., & Heene, M. (2021). Providing a lower-bound estimate
risks and benefits has been explored appropriately and rigorously. for psychology’s “crud factor”: The case of aggression. Professional
Psychology: Research and Practice, 52(6), 620–626. [Link]
.1037/pro0000386
References Ferguson, C. J., Kaye, L. K., Branley-Bell, D., Markey, P., Ivory, J. D.,
Alonzo, R., Hussain, J., Stranges, S., & Anderson, K. K. (2021). Interplay Klisinan, D., Elson, M., Smyth, M., Hogg, J. L., McDonnell, D., Nichols,
between social media use, sleep quality, and mental health in youth: A D., Siddiqui, S., Gregerson, M., & Wilson, J. (2022). Like this meta-
systematic review. Sleep Medicine Reviews, 56, Article 101414. https:// analysis: Screen media and mental health. Professional Psychology:
[Link]/10.1016/[Link].2020.101414 Research and Practice, 53(2), 205–214. [Link]
American Psychological Association. (2023). Health advisory on social 0000426
media use in adolescence. [Link] Freedman, J. (2002). Media violence and its effect on aggression: Assessing
rnet/health-advisory-adolescent-social-media-use the scientific evidence. University of Toronto Press. [Link]
Billieux, J., Schimmenti, A., Khazaal, Y., Maurage, P., & Heeren, A. (2015). .3138/9781442627512
Are we overpathologizing everyday life? A tenable blueprint for Haidt, J. (2020). Digital technology under scrutiny: A guilty verdict. Nature,
behavioral addiction research. Journal of Behavioral Addictions, 4(3), 578(7794), 226–227. [Link]
119–123. [Link] Hall, J. A., & Liu, D. (2022). Social media use, social displacement, and
Bowman, N. D. (2016). The rise (and refinement) of moral panic. In R. well-being. Current Opinion in Psychology, 46, Article 101339. https://
Kowert & T. Quandt (Eds.), The video game debate: Unravelling the [Link]/10.1016/[Link].2022.101339
physical, social, and psychological effects of digital games (pp. 22–38). House of Commons Science and Technology Select Committee. (2019).
Routledge/Taylor & Francis Group. Impact of social media and screen-use on young people’s health (pp. 1–92).
Brailovskaia, J., Swarlik, V. J., Grethe, G. A., Schillack, H., & Margraf, J. [Link]
(2023). Experimental longitudinal evidence for causal role of social media [Link]
use and physical activity in COVID-19 burden and mental health. Journal Jenkins-Guarnieri, M. A., Wright, S. L., & Johnson, B. (2013). Development
of Public Health, 31(11), 1885–1898. [Link] and validation of a social media use integration scale. Psychology of
022-01751-x Popular Media Culture, 2(1), 38–50. [Link]

Template Version: 14 Jun 2024 ▪ 12:21 pm IST PRO-2023-0960_blupencil ▪ 18 July 2024 ▪ 1:42 pm IST
10 FERGUSON, KAYE, BRANLEY-BELL, AND MARKEY

Kaye, L. K. (2022). Issues and debates in cyberpsychology. Open of Sciences of the United States of America, 116(21), 10226–10228.
University Press. [Link]
Kaye, L. K., Orben, A., Ellis, D. A., Hunter, S. C., & Houghton, S. (2020). The Orben, A., & Przybylski, A. K. (2019). The association between adolescent
conceptual and methodological mayhem of “screen-time.” International well-being and digital technology use. Nature Human Behaviour, 3(2),
Journal of Environmental Research and Public Health, 17(10), Article 173–182. [Link]
3661. [Link] Orben, A., Przybylski, A. K., Blakemore, S. J., & Kievit, R. A.
Keles, B., McCrae, N., & Grealish, A. (2019). A systematic review: the (2022). Windows of developmental sensitivity to social media. Nature
influence of social media on depression, anxiety and psychological distress Communications, 13(1), Article 1649. [Link]
in adolescents. International Journal of Adolescence and Youth, 25(1), 022-29296-3
79–93. [Link] Padmanathan, P., Bould, H., Winstone, L., Moran, P., & Gunnell, D. (2020).
Keles, B., McCrae, N., & Grealish, A. (2019). A systematic review: the Social media use, economic recession and income inequality in relation to
influence of social media on depression, anxiety and psychological distress trends in youth suicide in high-income countries: A time trends analysis.
in adolescents. International Journal of Adolescence and Youth, 25(1), Journal of Affective Disorders, 275, 58–65. [Link]
79–93. [Link] .2020.05.057
Keles, B., McCrae, N., & Grealish, A. (2020). A systematic review: The Parry, D. A., Davidson, B. I., Sewall, C. J. R., Fisher, J. T., Mieczkowski, H.,
influence of social media on depression, anxiety and psychological distress & Quintana, D. S. (2021). A systematic review and meta-analysis of
in adolescents. International Journal of Adolescence and Youth, 25(1), discrepancies between logged and self-reported digital media use. Nature
79–93. [Link] Human Behaviour, 5(11), 1535–1547. [Link]
Kerr, M., & Stattin, H. (2003). Parenting of adolescents: Action or reaction? 021-01117-5
In A. C. Crouter & A. Booth (Eds.), Children’s influence on family Prince, S. A., Cardilli, L., Reed, J. L., Saunders, T. J., Kite, C., Douillette, K.,
dynamics: The neglected side of family relationships (pp. 121–151). Fournier, K., & Buckley, J. P. (2020). A comparison of self-reported and
Lawrence Erlbaum. device measured sedentary behaviour in adults: A systematic review and
Lima, C., & Nix, N. (2013). 41 states sue Meta, claiming Instagram, meta-analysis. The International Journal of Behavioral Nutrition and
Facebook are addictive, harm kids. The Washington Post. [Link] Physical Activity, 17(1), Article 31. [Link]
[Link]/technology/2023/10/24/meta-lawsuit-facebook-insta 00938-3
gram-children-mental-health/ Rounsefell, K., Gibson, S., McLean, S., Blair, M., Molenaar, A., Brennan,
Lima-Strong & Nix (2023). People of the State of California v. Meta L., Truby, H., & McCaffrey, T. A. (2020). Social media, body image and
Platforms Inc. food choices in healthy young adults: A mixed methods systematic review.
Mahalingham, T., McEvoy, P., & Clarke, P. (2023). Assessing the validity of Nutrition & Dietetics: The Journal of the Dietitians Association of
self-report social media use: Evidence of no relationship with objective Australia, 77(1), 19–40. [Link]
smartphone use. Computers in Human Behavior, 140, Article 107567. Rounsefell, K., Gibson, S., McLean, S., Blair, M., Molenaar, A., Brennan,
[Link] L., Truby, H., McCaffrey, T.A. (2020) Social media, body image and food
Marks, R. J., De Foe, A., & Collett, J. (2020). The pursuit of wellness: Social choices in healthy young adults: A mixed methods systematic review.
media, body image and eating disorders. Children and Youth Services Nutrition & dietetics, 77(1),19-40. [Link]
Review, 119, Article 105659. [Link] 0080.12581
.105659 Satchell, L. P., Fido, D., Harper, C. A., Shaw, H., Davidson, B., Ellis, D. A.,
Meier, A., & Reinecke, L. (2021). Computer-mediated communication, Hart, C. M., Jalil, R., Bartoli, A. J., Kaye, L. K., Lancaster, G. L. J., &
Social media, and mental health: A conceptual and empirical meta-review. Pavetich, M. (2021). Development of an Offline-Friend Addiction
Communication Research, 48(8), 1182–1209. [Link] Questionnaire (O-FAQ): Are most people really social addicts? Behavior
0093650220958224 Research Methods, 53(3), 1097–1106. [Link]
Miller, J., Mills, K. L., Vuorre, M., Orben, A., & Przybylski, A. K. (2023). 01462-9
Impact of digital screen media activity on functional brain organization in Sendra, A., Farré, J., & Vaagan, R. W. (2020). Seeking, sharing and co-
late childhood: Evidence from the ABCD study. Cortex, 169, 290–308. creating: A systematic review of the relation between social support
[Link] theory, social media use and chronic diseases. Social Theory & Health,
O’Day, E. B., & Heimberg, R. G. (2021). Social media use, social 18(4), 317–339. [Link]
anxiety, and loneliness: A systematic review. Computers in Human Sharp, A. (2021) Social media: The answer or the scapegoat behind teen
Behavior Reports, 3, Article 100070. [Link] mental health challenges? Family Perspectives, 2(2), Article 9. https://
.2021.100070 [Link]/familyperspectives/vol2/iss2/9
O’Donohue, W., & Dyslin, C. (1996). Abortion, boxing and Zionism: Shaw, D. J., Kaye, L. K., Ngombe, N., Kessler, K., & Pennington, C. R.
Politics and the APA. New Ideas in Psychology, 14(1), 1–10. https:// (2022). It’s not what you do, it’s the way that you do it: An experimental
[Link]/10.1016/0732-118X(95)00025-C task delineates among passive, reactive and interactive styles of behaviour
Odgers, C. L., & Jensen, M. R. (2020). Annual Research Review: Adolescent on social networking sites. PLOS ONE, 17(12), Article e0276765. https://
mental health in the digital age: Facts, fears, and future directions. Journal [Link]/10.1371/[Link].0276765
of Child Psychology and Psychiatry, and Allied Disciplines, 61(3), 336– Shaw, H., Ellis, D., Geyer, K., Davidson, B., Ziegler, F., & Smith, A. (2020).
348. [Link] Quantifying smartphone “use”: Choice of measurement impacts relation-
Ofcom. (2022). Children and parents: Media use and attitudes report 2022. ships between “usage” and health. Technology, Mind, and Behavior, 1(2).
[Link] [Link]
[Link] Shimoga, S. V., Erlyana, E., & Rebello, V. (2019). Associations of social
Orben, A. (2020). The Sisyphean cycle of technology panics. Perspectives media use with physical activity and sleep adequacy among adolescents:
on Psychological Science, 15(5), 1143–1157. [Link] Cross-sectional survey. Journal of Medical Internet Research, 21(6),
1745691620919372 Article e14290. [Link]
Orben, A., Dienlin, T., & Przybylski, A. K. (2019). Social media’s enduring Trifiro, B. M., & Gerson, J. (2019). Social media usage patterns: Research
effect on adolescent life satisfaction. Proceedings of the National Academy note regarding the lack of universal validated measures for active and

Template Version: 14 Jun 2024 ▪ 12:21 pm IST PRO-2023-0960_blupencil ▪ 18 July 2024 ▪ 1:42 pm IST
 FINDINGS FROM A META-ANALYSIS 11

passive use. Social Media + Society, 5(2). [Link] Valkenburg, P. M., van Driel, I. I., & Beyens, I. (2022). The associations of
305119848743 active and passive social media use with well-being: A critical scoping
Twenge, J. M. (2020). Increases in depression, self-harm, and suicide among review. New Media & Society, 24(2), 530–549. [Link]
U.S. adolescents after 2012 and links to technology use: Possible 14614448211065425
mechanisms. Psychiatric Research and Clinical Practice, 2(1), 19–25. van den Eijnden, R. J. J. M., Lemmens, J. S., & Valkenburg, P. M.
[Link] (2016). The social media disorder scale. Computers in Human Behavior,
Twenge, J. M., & Campbell, W. K. (2019). Media use is linked to lower 61, 478–487. [Link]
psychological well-being: Evidence from three datasets. Psychiatric Vuorre, M., Orben, A., & Przybylski, A. K. (2021). There is no evidence that
Quarterly, 90(2), 311–331. [Link] associations between adolescents’ digital technology engagement and
Twenge, J. M., Joiner, T. E., Rogers, M. L., & Martin, G. N. (2018). mental health problems have increased. Clinical Psychological Science,
Increases in depressive symptoms, suicide-related outcomes, and suicide 9(5), 823–835. [Link]
rates among U.S. adolescents after 2010 and links to increased new media Want, S. C. (2014). Three questions regarding the ecological validity of
screen time. Clinical Psychological Science, 6(1), 3–17. [Link] experimental research on the impact of viewing thin-ideal media images.
.1177/2167702617723376 Basic and Applied Social Psychology, 36(1), 27–34. [Link]
U. K. Parliament (2023). Online safety bill. [Link] .1080/01973533.2013.856783
lls/3137 Wilkinson, L., & Task Force on Statistical Inference, American Psychological
U.S. Department of Health and Human Services. (2023). Social media Association, Science Directorate. (1999). Statistical methods in psycholog-
policies. Retrieved May 5, 2023, from [Link] ical journals: Guidelines and explanations. American Psychologist, 54(8),
media/policies/[Link] 594–604. [Link]
Valkenburg, P. M., Beyens, I., Meier, A., & Vanden Abeele, M. M. P.
(2022). Advancing our understanding of the associations between social
media use and well-being. Current Opinion in Psychology, 47, Article
101357. [Link]
Valkenburg, P. M., Meier, A., & Beyens, I. (2022). Social media use and its Received September 22, 2023
impact on adolescent mental health: An umbrella review of the evidence. Revision received May 13, 2024
Current Opinion in Psychology, 44, 58–68. [Link] Accepted May 23, 2024 ▪
psyc.2021.08.017

Template Version: 14 Jun 2024 ▪ 12:21 pm IST PRO-2023-0960_blupencil ▪ 18 July 2024 ▪ 1:42 pm IST