A Meta-Analysis of the Effects of Social Media Exposure to Upward Comparison Targets on Self-Evaluations and Emotions

ABSTRACT

Social media have become a pervasive part of contemporary culture and are an essential part of the daily lives of an increasing number of people. Its popularity has brought unlimited opportunities to compare oneself with other people. This meta-analysis combined and summarized the findings of previous experimental research, with the aim of generating causal conclusions regarding the effects of exposure to upward comparison targets on self-evaluations and emotions in a social media context. We identified 48 articles involving 7679 participants through a systematic search and entered 118 effect sizes into a multilevel, random-effects meta-analysis. Analyses revealed an overall negative effect of upward social comparison relative to downward comparison and controls on social media users’ self-evaluations and emotions (g = −0.24, p < .001). Specifically, there were significant negative effects of upward comparison on each outcome variable: body image (g = −0.31, p < .001), subjective well-being (g = −0.19, p < .001), mental health (g = −0.21, p < .001) and self-esteem (g = −0.21, p < .001). This meta-analysis indicates that contrast is the dominant response to upward comparison on social media, which results in negative self-evaluations and emotions.

Introduction

The use of social media has grown exponentially in recent years, with over 4.2 billion people worldwide currently accessing social media sites for an average of more than 2 hours per day (Hootsuite & We Are Social, 2021). The carefully curated, idealized self-presentations that a person typically encounters on social media often prompt social comparison processes. Social media allows users to engage in selectively positive self-presentation, resulting in a platform rife with opportunities to compare oneself with seemingly superior others (Wilson et al., 2012). News feeds are filled with exciting social events and attractive photos, whereas unflattering photos, and unhappy days rarely get posted. Consequently, opportunities for downward comparisons that could enhance self-evaluations are limited (Midgley et al., 2020). Expanding upon a systematic review of social comparison by Gerber et al. (2018), this meta-analysis aimed to test social comparison theory by examining the effects of exposure to positive stimuli in the unique context of social media. In recent years, a wealth of research has investigated the effects of engaging in social comparison processes in a social media context. However, a consensus on the effects on users’ emotions and self-evaluations has not yet been reached. With the purpose of providing evidence of causation, the current meta-analysis will be confined to solely experimental studies that have manipulated an independent variable known to effect social comparison processes.

Relevant Theoretical Perspectives

In social comparison theory, Festinger (1954) proposed that humans possess an innate motivation to compare themselves to other people, driven by a need to establish accurate self-appraisals. People acquire valuable information regarding themselves through social comparison, including their abilities, accomplishments, attitudes, and physical attributes. Festinger emphasized that people prefer to rely on objective, nonsocial means to evaluate the self but seek to compare with others when objective benchmarks are unavailable.

Since the introduction of social comparison theory, research has primarily focused on two directions of comparison: upward versus downward. Upward comparison occurs when a person evaluates themselves against someone perceived as superior. This type of comparison is often motivated by self-improvement; however, it commonly results in feelings of envy, anxiety, frustration, and depression (Wheeler & Miyake, 1992). In contrast, a downward social comparison occurs when the comparison target is believed to be inferior to the comparer. Predominantly motivated by self-enhancement, this type of comparison most commonly leads to an enhanced mood and positive self-evaluations (Collins, 1996).

When evaluating oneself relative to a comparison target, there are two possible reactions: assimilation and contrast (Mussweiler, 2003). Assimilation occurs when one’s self-evaluation moves toward one’s evaluation of the comparison target, whereas contrast occurs when one’s self-evaluation moves away from one’s evaluation of the comparison target. Assimilation to an upward comparison target can lead to feelings of inspiration and motivation, while contrast can lead to negative emotions and lowered self-evaluations (Smith, 2000). The selective accessibility model posits that assimilation and contrast occur through rapid cognitive judgments of similarity or dissimilarity to the comparison target (Mussweiler, 2003). A hypothesis of similarity or dissimilarity to the target is formed and a cognitive search of self-knowledge relevant to the comparison domain takes place. If self-knowledge which supports the similarity hypothesis is accessed, a person’s self-evaluations will assimilate toward the comparison target. On the other hand, if self-knowledge supports the dissimilarity hypothesis, self-evaluations are likely to contrast away from the target (Mussweiler, 2001).

Previous Meta-Analysis of Social Comparison Research

In a meta-analysis of over 60 years of social comparison research, Gerber et al. (2018) addressed two longstanding issues in social comparison research: the selection of a comparison target (selection) and the effects of social comparison on self-evaluations and affect (reaction). The results of the selection studies revealed that people have a dominant tendency to compare upwards. Gerber et al. (2018) noted that their findings contrasted with earlier social comparison theories that suggested upward comparisons are avoided due to their detrimental effects on mood and self-esteem (Wills, 1991). The authors speculated that upward comparisons are more commonplace as individuals may not anticipate the self-deflating contrast. They point to the better-than-average effect: a general tendency for individuals to perceive themselves as superior compared to the average person (Brown, 1986). People may look upward to confirm their closeness to successful and attractive others. What follows may involve disappointing evidence that someone is better than themselves.

In the reaction studies, results revealed that the most common response to upward comparison was contrast, which leads to a decrease in the self-evaluations of the individual doing the comparing. Assimilation to the comparison target was not a default response to upward comparison and required special priming. Even when assimilation occurred due to priming, the effect was weak. Upward comparisons led to decreased ability assessments and performance satisfaction and were more potent when the comparison target was a close connection. Surprisingly, there were no significant effects of social comparison on affect, self-esteem or behavior.

Outcomes of Social Comparison on Social Media

Social media may provoke more frequent and extreme upward comparisons due to a tendency for users to present strategically constructed and idealized versions of themselves (Gonzales & Hancock, 2011). People are more likely to post positive rather than negative content on social media and manipulate their content following societal ideals (Haferkamp et al., 2012). The result of such posts is a social environment that almost exclusively provokes upward comparisons (Shim et al., 2016). Recent evidence by Midgley et al. (2020) confirmed that upward comparisons consistently outnumber downward comparisons on social media and that the impact of these comparisons is overwhelmingly negative. Participants in their study browsed their personal social media account and reported the immediate impact on their self-evaluations after viewing each post in their News Feed. To assess the cumulative effects of the posts, participants also reported the impact on their self-esteem and subjective well-being at the end of the browsing session. Results revealed that participants experienced immediate negative effects on their self-evaluations after viewing each post, and that these effects were cumulative and resulted in declines in self-esteem, mood, and life satisfaction. Consistent with the findings of Gerber et al. (2018), other evidence has revealed that contrast is the dominant response to upward comparison on social media (Alfasi, 2019; Hogue & Mills, 2019). Heavy users of social media, compared to infrequent users, are more likely to believe that others are happier, have better lives, and are more successful (Chou & Edge, 2012).

Previous research has investigated several outcomes of social media use including body image, subjective well-being, psychological well-being, self-esteem, and envy. According to sociocultural models of body image, people who repeatedly compare their appearance to others are at risk of developing body image disturbances, including body dissatisfaction and disordered eating (Vartanian & Dey, 2013). Opportunities to compare one’s body with others are abundant on social media, particularly on image-based platforms like Instagram. The advancement of digital technology in the form of filters and photo editing software has encouraged the manipulation of content and fostered the creation of unrealistic and unachievable standards (Haferkamp et al., 2012). Many studies have demonstrated a positive association between social media use and body dissatisfaction, primarily in young women (de Vries et al., 2018; Fardouly & Vartanian, 2016; Holland & Tiggemann, 2016). Relationships have also been found between self-reported Instagram use and body dissatisfaction, drive for thinness, body surveillance and desire for cosmetic surgery (Tiggemann & Miller, 2010). Much of this research is correlational in design though, making it impossible to draw causal conclusions. Although it is possible that Instagram use leads to body image concerns, it is also possible that people with body image concerns are more likely to use Instagram. In recent years, some experimental evidence has revealed that the use of social media leads to increased body dissatisfaction (Livingston et al., 2020; Tiggemann & Anderberg, 2020), yet other experimental studies have found that there is no causal link (Engeln et al., 2020). An experiment conducted by Badillo (2019) concluded that Instagram, but not Facebook, leads to increased body dissatisfaction, with the authors stating that social comparison may be an important mediating factor. Recently, de Valle et al. (2021) conducted meta-analyses of experimental and longitudinal research and revealed that viewing appearance-ideal images had a causal, negative effect on body image. It was considered that the results may be explained by a social comparison process in which viewers felt that they did not meet the standard of the ideal target, and subsequently felt worse about their own appearance. Another recent meta-analysis of experimental research also found that viewing idealized images lead to increased body dissatisfaction among men and women (Fioravanti et al., 2022). In recent years, a body-positivity movement has emerged on social media, involving more inclusive and accessible body ideals. An experiment by Cohen et al. (2019) revealed that brief exposure to body-positive posts was associated with increases in young women’s positive mood, body satisfaction, and body appreciation. However, there was also an increase in self-objectification, resulting from a focus on appearance over other attributes.

In addition to the effects on body image, other research has investigated whether social media has an impact on users’ subjective well-being. Subjective well-being refers to a person’s cognitive and affective evaluations of their life and encompasses three components: life satisfaction, positive affect, and negative affect (Diener et al., 1999). Several studies have identified a relationship between social media use and declines in subjective well-being (Alfasi, 2019; de Vries et al., 2018; Verduyn et al., 2020). For example, Fioravanti et al. (2020) investigated the effect of short-term abstention from Instagram use on subjective well-being. Results revealed that taking a one week break from Instagram was associated with increased life satisfaction and positive affect. The control group that continued to use Instagram experienced decreases in subjective well-being. These results, however, only applied to female participants, as male participants reported no significant effects. The authors theorized that this gender difference may be due to a general tendency for females to place higher importance on appearance relative to males. We further discuss gender differences in social comparison processes below.

In addition to subjective well-being, previous research has investigated the mental health effects of engaging in social comparison. Festinger (1954) posited that people experience adverse mental health effects when they encounter an upward comparison, and a threatened sense of self-worth leads to symptoms of anxiety, frustration, and depression. Several studies have supported this hypothesis and reported a positive association between the intensity of social media use and symptoms of depression and anxiety (Appel et al., 2016; Steers et al., 2014). However, as is the case with most studies in the field, the cross-sectional nature of these studies cannot offer causal insight into the observed findings, and the relationship may be bidirectional. An experimental study by Alfasi (2019) found that participants who browsed their personal Facebook News Feed reported higher levels of depression than participants who browsed a neutral Facebook page that did not contain any social content. Other experimental studies have revealed that participants experienced higher levels of anxiety (Kohler et al., 2021) and depression (Fan et al., 2019) after using social media, compared to participants in control groups. Some longitudinal studies have found no association between social media use and negative mental health outcomes (Coyne et al., 2020; Schemer et al., 2020). Conversely, other longitudinal studies have reported that social media use is negatively associated with mental well-being (McNamee et al., 2021; Thorisdottir et al., 2020).

Prior research has revealed that adverse mental health outcomes may be due to feelings of envy being generated through social media use (Steers et al., 2014). The emotion of envy arises from a feeling of resentment and a desire to possess another person’s superior quality, achievement, or possession (Parrott & Smith, 1993). However, assimilation to an upward comparison on social media can result in benign envy and feelings of inspiration (Park & Baek, 2018). As revealed in the meta-analysis of Gerber et al. (2018), assimilation is not a typical response to upward comparison however, and the default tendency of most people is a contrastive response.

Research regarding the effects of social media on users’ self-esteem has also produced inconsistent results. Some studies have found that actively using social media by sharing content and engaging in conversation can increase self-esteem (Gonzales & Hancock, 2011). However, most research has reported a decline in self-esteem due to social media use (Kross et al., 2013; Vogel et al., 2014). A correlational study by Vogel et al. (2014) revealed that participants who used Facebook more frequently had lower trait self-esteem, and that this relationship was mediated by greater exposure to upward social comparisons. This study was followed by an experiment that examined the causal impact of brief exposure to social media on participants state self-esteem. Participants who viewed the profile page of a fictitious upward comparison target (an individual high in attractiveness, fitness, health, and well-being) reported lower self-esteem than the participants who were assigned to view the profile of a downward comparison target.

Moderation by Age and Gender

Despite substantial evidence that upward social comparison on social media results in adverse outcomes, it is not clear whether these effects are consistent across people. Gender and age are potential moderating factors that could affect how people respond to social comparison targets on social media. It has been suggested that social comparison weakens across the lifespan, with young adults reporting stronger comparison tendencies than older adults (Callan et al., 2015). In body image research, adolescents have been found to be more negatively affected by social comparison than young adults (Yoon et al., 2003). Adolescence is a key developmental period of identity formation (Erikson, 1950), and social comparison may form part of the process of establishing a coherent identity and strong sense of self. Very few studies have examined age differences in social comparison behaviors on social media. A survey of 1,000 Dutch 14–15 year olds revealed that more positive than negative emotions were reported by adolescents when comparing themselves to others on social media (van Driel et al., 2019). Another study found that social comparisons of ability on Instagram triggered adolescents to reflect upon their current and future selves (Noon et al., 2021). In young adults, the same comparisons of ability triggered feelings of self-doubt. Age differences in the assimilation and contrast dynamic may account for the differing results, with adolescents assimilating to the target, and young adults experiencing a contrastive response. In contrast to the findings of Yoon et al. (2003), these findings suggest that young adults are the most negatively affected by social comparison.

Females, in general, have been found to engage in more social comparison behavior than males (Gibbons & Buunk, 1999), perhaps due to a stronger communal and interpersonal orientation (Cross & Madson, 1997). Men are more likely to have independent self-construals and view themselves as autonomous agents, whereas women are more likely to have interdependent self-construals and view themselves as embedded in relationships with others (Kemmelmeier & Oyserman, 2001). Regarding the context of social media, research has shown that females engage in more social comparison, and upward comparisons are especially harmful to females compared to males (Nesi & Prinstein, 2015). Several studies have revealed that females experience greater adverse outcomes after engaging with social media (Fioravanti et al., 2020; Thorisdottir et al., 2020; Twenge & Martin, 2020). These studies involved the general use of social media and did not directly measure social comparison behavior; however, studies have shown that the more time a user spends on social media sites, the more they compare themselves to others (Steers et al., 2014; Vogel et al., 2014). In fact, social comparison appears to be ubiquitous and inevitable in such an environment. One study found that Facebook users make an average of 3 comparisons in 20 minutes (Midgley et al., 2020). This contrasts with the average of one comparison per day that occurs in real-life settings (Wheeler & Miyake, 1992). Overall, these age and gender findings suggest that young people and females may be more affected by upward social media comparisons, and it is a subject that warrants further investigation.

The Current Study

This review of the research literature highlights the increasing concern of the effects of exposure to upward comparison targets on social media. Specifically, research has identified that social media can affect self-evaluations and emotions via the process of social comparison. Given its increasing popularity, it is crucial to gain an understanding of how and when social media exerts a negative impact, and the role of social comparison in producing adverse outcomes.

Comparisons on social media are widespread and occur in multiple domains, with the most prevalent being relevant to appearance/attractiveness, popularity/friendship, and lifestyle/activities (Midgley et al., 2020). Despite these various types of content on social media, studies in the field often focus on appearance comparisons provoked by exposure to attractive and idealized images. As seen in Table SA2 (found in the online supplementary materials: https://osf.io/2ajzb/), this is the case for the majority of studies included in the current meta-analysis.

It was a requirement for inclusion in the meta-analysis that authors attributed the study findings to the processes of social comparison, although variability exists in the way that social comparison was measured. A common method for measuring appearance comparisons is to use the State Appearance Comparison Scale (Tiggemann & McGill, 2004) to directly examine the amount of comparison that participants engaged in. Other studies use pilot testing and manipulation tests to ensure that the target stimuli are evoking the intended comparison process. Finally, some researchers treat social comparison as a spontaneous and unintentional consequence of viewing other’s performances (Gilbert et al., 1995), and believe that it is unlikely that browsing one’s news feed would not result in making comparisons to superior others (Midgley et al., 2020). It is important to note that this meta-analysis does not directly examine the frequency and direction of social comparison processes in the included studies. Comparisons are subjective and without specific measurement we can only surmise that the results are due to the effects of social comparison.

In recent years, several meta-analyses and reviews have been conducted to establish an understanding of the implications of social media use. Reviews examining the effects on body image have found that social media use is associated with body image disturbance (Saiphoo & Vahedi, 2019), and viewing appearance-ideal images on social media has a negative effect on body image (de Valle et al., 2021; Fioravanti et al., 2022). Regarding subjective well-being, a review by Meier and Johnson (2022) found mixed and inconsistent evidence for the claim that social media use involves harmful social comparison processes that negatively affect well-being. Similarly, an umbrella review by Valkenburg (2022) identified that the nine meta-analyses that have been published since 2019 have yielded disagreeing associations between social media use and well-being. In the area of psychological well-being, several reviews have identified an association between symptoms of depression and time spent on social media, or intensity of social media use (Cunningham et al., 2021; Huang, 2017; Yang et al., 2019). A systematic literature review conducted by Krause et al. (2019) revealed that when social comparison takes place on a social media site, users experience decreases in self-esteem. However, receiving positive social feedback was associated with benefits for users’ self-esteem. Most of these reviews have been limited by the research being primarily cross-sectional. Furthermore, there is a tendency to focus on one social media platform and one outcome variable. The current study will be the first meta-analysis to examine solely experimental evidence and look at the effect of social comparison on several outcomes, across multiple social media platforms. We aim to expand on previous reviews by reconciling inconsistent findings and providing new insight into social comparison processes. By combining and summarizing previous research findings, we aim to test the cumulative effects of social comparison across all outcome variables, in addition to testing the effects on each individual variable. Based on past research, we hypothesized the following:

Hypothesis 1:

Exposure to upward comparison targets on social media would lead to a decline in users’ overall self-evaluations and emotions (across all outcome variables)

Hypothesis 2:

Exposure to upward comparison targets on social media would be associated with declines in each outcome variable: body image, subjective well-being, mental health, and self-esteem

Hypothesis 3:

Age would moderate the overall effect of upward social comparison on users’ self-evaluations and emotions, as well as each outcome separately, with young age being associated with increased negative outcomes. Despite mixed findings in previous research, we hypothesize this result due to young age being associated with stronger comparison tendencies.

Hypothesis 4:

Gender would moderate the overall effect of upward social comparison on users’ self-evaluations and emotions, as well as each outcome separately, with females experiencing increased negative outcomes compared to males

Additionally, we investigated the following Research Questions:

Research Question 1: Do studies that involve directly manipulated comparison targets contribute larger effect sizes than studies that evoked indirect social comparison through the general use of social media?

Research Question 2: Do studies that compare upward comparison targets to downward comparison targets contribute larger effect sizes than studies that compare upward comparison targets to neutral content that is designed to not evoke social comparison processes?

Method

This meta-analysis was pre-registered in November 2020 with the International Prospective Register of Systematic Reviews (Registration number: CRD42020219794) and the Open Science Framework (https://doi.org/10.17605/OSF.IO/2UHEK).

Eligibility Criteria for Reaction Studies

We conducted a systematic search to identify relevant articles to be included in the meta-analysis. Studies had to meet the following criteria for inclusion:

1. Reported in English.

2. Experimental in design with the use of random assignment.

3. Between-subjects design with at least two groups.

4. Included self-report outcome measures of body image, subjective well-being, mental health, self-esteem, or envy.

5. Social comparison was evoked via exposure to manipulated upward comparison targets, instructions to scroll through social media feeds, or instructions to use social media as normal.

6. Included a control condition or a downward comparison condition. Studies that only compared an upward comparison to a lateral/less-upward condition were excluded.

7. Social comparison occurred in a social media context. Traditional media studies were excluded.

8. The authors of the studies must have proposed the occurrence of social comparison processes.

9. There were no restrictions applied regarding the social media platform used or the age and gender of the study participants. We excluded studies if they involved clinical populations.

Search Strategy

Several methods were employed to ensure a comprehensive assemblage of literature. The first author conducted searches in August 2020, January 2021, and September 2021 of the following databases: PubMed, PsycINFO, Medline, Web of Science and Google Scholar. ProQuest Dissertations and Theses and Open Grey databases were also searched to identify unpublished literature. We also posted calls for papers requesting unpublished data that met our eligibility criteria on the following listservs: Australian and New Zealand Communication Association, Society of Australasian Social Psychologists, Society for Personality and Social Psychology, Association of Internet Researchers, and International Communication Association.

All database search strategies were comprised of a combination of social media terms (social media, social network* site*, Facebook, Instagram, Twitter, Snapchat, TikTok), and well-being terms (well?being, life satisfaction, affect, self-esteem, body image, envy, depression, anxiety, body satisfaction, appearance satisfaction, self?evaluation, mental health) in addition to the term “social comparison.” We restricted search results to articles published from September 2006 onwards to coincide with when the most popular social media site, Facebook, first allowed public access. The reference lists of eligible studies were also manually searched to identify additional studies.

Study Selection

In total, 1490 potential records were identified through the literature search. After removing duplicates in Endnote, the titles and abstracts of 1139 articles were screened for eligibility. Of these studies, 1013 were excluded for failure to meet the inclusion criteria. Full-text versions of the remaining 126 articles were screened, and a total of 48 studies were identified as eligible to be included in the meta-analysis (see Figure 1).

Figure 1. Flow chart of the study selection process for the meta-analysis.

Type of Dependent Variables

We categorized the outcome measures of interest into four groups:

1. Subjective well-being: included measures of positive affect, negative affect, life satisfaction, happiness, and sadness. Due to insufficient data to form a category, the outcome of envy was also included in the subjective well-being category.

2. Body image: included measures of body dissatisfaction, body satisfaction, appearance satisfaction, facial satisfaction, appearance discrepancies, and body image disturbance.

3. Mental health: included state measures of depression and anxiety.

4. Self-esteem: included measures of state self-esteem.

Meta-Analytic Data Extraction

The following information was extracted from each eligible study: sample size, mean age, the female ratio of participants, and social media platform. Means, standard deviations and the number of participants in each condition (cell sizes) were also extracted to calculate effect sizes for the meta-analyses. Thirteen of the studies did not report the number of participants per condition, therefore cell sizes were calculated based on the total sample size, assuming equal distribution of participants to conditions (Nitschke et al., 2019). When studies involved multiple upward comparison conditions from the same platform (e.g., a Fitspiration condition and a Thinspiration condition), the effect sizes were averaged and treated as completely dependent (correlation of 1). This is the most conservative approach to dealing with dependency between variables (Borenstein et al., 2009). A second coder completed the data extraction independently to ensure reliability in the data extraction process. Inter-rater reliability was calculated using Cohen’s measure of agreement (Cohen’s kappa), which assesses the similarity of ratings while accounting for the possibility of chance (Neuendorf, 2002). The result was perfect agreement, with a kappa value of k = 1.00.

Moderator Coding

Prior research has identified that young people and females are more susceptible to the effects of social comparison, therefore, age and gender were coded as potential moderators. Age was recorded as the mean age of the study sample, while gender was coded as the percentage of participants in the sample identified as female.

Meta-Analytic Statistical Methods

We calculated an effect size for each outcome using Hedges’ g version of the standardized mean difference. Negative values for effect sizes indicated more negative self-evaluations and affect in the upward comparison condition relative to the control or downward comparison condition, while positive values indicated more positive self-evaluations and affect.

The meta-analyses were conducted in R, using the metafor package. As some studies contributed multiple effect sizes, a multilevel, random-effects model with restricted maximum-likelihood (REML) was conducted. A 95% confidence interval was calculated around each effect size and the summary effect. The width of the confidence interval is indicative of the precision of the estimate, with wider intervals indicating less precision (Borenstein et al., 2009).

The heterogeneity of effect sizes across studies was evaluated using Cochran’s Q and I² statistics. Cochran’s Q evaluates whether there is more variation in the effect sizes than expected from sampling error alone. The I² statistic quantifies as a percentage the total variation in study effect sizes due to between-study differences rather than random sampling variation (Borenstein et al., 2009). As we conducted multilevel meta-analyses, we calculated the I² over three levels, splitting the variance into two parts. This resulted in two I² values: one attributable to true effect size differences within clusters, and one attributable to true effect size differences between clusters. Moderator analyses were also conducted to assess for heterogeneity in study effect sizes. The proposed moderators of age, gender and platform type were tested with random-effects, REML meta-regression.

Publication Bias

Potential publication bias was evaluated by visual inspection of a funnel plot and several statistical tests. Kendall’s tau method was used to assess if small studies with negative results were less likely to be published. A significant Kendall’s tau statistic can be interpreted as the presence of publication bias (Begg & Mazumdar, 1994). This statistic, however, may only have moderate power for small meta-analyses (Sterne et al., 2000); therefore, Egger’s regression test was also used. Egger’s regression test is appropriate for small meta-analyses and assesses symmetry of the publication bias funnel plot (Egger et al., 1997). A significant result for Egger’s test indicates asymmetry in the funnel plot, suggesting possible publication bias (Borenstein et al., 2009). The Trim-and-Fill method was also used to test whether any studies needed to be imputed to achieve symmetry in the funnel plot. This method provides a corrected summary effect size that is re-estimated based on evidence of publication bias (Duval & Tweedie, 2000).

Study Quality Assessment

We also assessed the risk of bias and quality of reporting, internal validity, external validity, and power of each study. We adapted the Downs and Black Quality Assessment Checklist (Downs & Black, 1998) to experimental study designs. Table SA1 (in the online supplementary materials) displays the criteria from the original Downs and Black Assessment Checklist and the adapted criteria that we created. A second coder also independently completed the study quality assessment, and disagreements were resolved through discussion. Each study received a score out of 20, with higher scores indicating greater methodological quality.

Results

Overview of Studies

The meta-analysis included 48 studies and 118 effect sizes. The independent samples included data from 7,679 participants (75% females). The mean age of participants across the studies was 22.40 years, with sample means ranging from 15 to 34 years of age. The majority of participants were from the United States of America and Australia. The characteristics of each study appear in Table SA2 in the online supplementary materials.

Overall Effect Size

A random-effects, multilevel meta-analysis was conducted to establish the magnitude of the association between social media use and users’ emotions and self-evaluations. A negative effect size estimate, g, indicates social media use is associated with a contrastive response and lower body image satisfaction, subjective well-being, mental health, and self-esteem in the upward comparison condition relative to the downward/control conditions. Consistent with Hypothesis 1, the pooled effect size estimate revealed a significant, negative effect, g = −0.24, p < .001, 95% CI [−0.29; −0.19]. A forest plot of the effect size estimates and confidence intervals for each study is presented in Figure SA1, in the online supplementary materials.

The Q and I² tests for heterogeneity were significant (Q(47) = 77.47, p = .003, I²_TOTAL = 40.03, I²_{LEVEL 2} = 20.01, I²_{LEVEL 3} = 20.01). A Q value that is higher than k-1 is indicative of substantial between-study heterogeneity (Harrer et al., 2021). The I² test revealed that 20% of the total variation could be attributed to within-study heterogeneity, and 20% can be attributed to between-study heterogeneity. A total I² value below 50% is considered to represent low heterogeneity (Borenstein et al., 2009).

Sub-group meta-analyses were conducted for each outcome variable. Consistent with Hypothesis 2, there was a significant, negative effect for each outcome. The effect was largest on body image, g = −0.31, p <  .001, 95% CI [−0.41; −0.22], followed by self-esteem, g = −0.21, p < .001, 95% CI [−0.31; −0.10], mental health, g = −0.21, p < ;.001, 95% CI [−0.34; −0.09], and subjective well-being, g = −0.19, p < .001, 95% CI [−0.27; −0.11].

Moderator Analyses

Meta-regression was conducted on the overall model, as well as each outcome variable separately, to test the relationship between the effect size estimate and the proposed moderators of age (mean age in study) and gender (% of females in study). For the overall model, there was no significant moderation by age, F(1, 107) = 0.13, p = .71, or gender, F(1, 116) = 0.22, p = .64. Four studies did not report the mean age of the sample and were omitted from the moderator analysis for age. For each outcome variable, there was also no significant moderation by age (body image: F(1, 36) = 0.13, p = .71; self-esteem: F(1, 12) = 0.16, p = .70; mental health: F(1, 6) = 0.01, p = .96; subjective well-being: F(1, 49) = 0.01, p = .96) or gender (body image: F(1, 34) = 0.05, p = .82; self-esteem: F(1, 13) = 0.01, p = .99; mental health: F(1, 8) = 0.01, p = .91; subjective well-being: F(1, 53)  1.56, p = .22).

An additional analysis was conducted to test whether the way in which social comparison was evoked in the studies (directly or indirectly) moderated the overall effect. The results revealed no significant moderation, F(1, 116) = 0.16, p = .69. Finally, a moderator analysis was conducted to test whether there was a difference between studies that compared upward targets to downward targets, versus studies that compared upward targets to neutral content. Results revealed this to be a significant moderator F(1, 116) = 4.34, p = .04. Thirty-six effect sizes involved the comparison of upward targets versus downward targets, and results indicated a small negative effect on self-evaluations and emotions (g = −0.16, p = .006). A larger effect was revealed in the 82 studies that involved upward targets versus neutral content (g = −0.27, p < .0001).

Examination of Publication Bias

Publication bias was evident upon visual inspection of the funnel plot of standard error by Hedges’ g (see Figure 2). The funnel plot appeared asymmetrical, which can be an indication that studies are systematically missing from the meta-analysis as a function of study size or effect size (Sterne et al., 2000). The rank correlation method also suggested an association between the effect size and standard error across studies, another indicator of potential publication bias (Kendall’s tau = −0.27, p = .007). However, results from Egger’s Regression Intercept test indicated no significant asymmetry in the funnel plot, t(46) = 0.33, p = .75. Duval and Tweedie’s Trim-and-Fill analysis suggested that nine studies would need to be imputed to the right of the mean effect size to obtain symmetry. The adjusted mean effect size was estimated as g = −0.18, 95% CI [−0.18; −0.25] with the imputed studies.

Figure 2. Funnel plot displaying relationship between effect size (Hedges’ g) and standard error.

Quality Assessment

We adapted the Downs and Black Quality Assessment Checklist (Downs & Black, 1998) to assess the quality of studies included in the meta-analysis. Each study was scored out of 20, with higher scores indicating greater quality. The overall methodological quality of the studies was M = 15.26, SD = 1.88. Figure 3 presents an overview of the assessment across key domains. The items with the lowest scores were “Power” and “Blinding of Participants,” due to several studies not reporting sufficient information addressing these criteria. We used each study’s quality score in a meta-regression to examine whether the studies’ methodological quality predicted the effect size. The results of the meta-regression indicated that the overall effect size was not affected by study quality, F(1, 116) = 0.13, p = .72.

Figure 3. Bar diagram displaying quality assessment in key domains across studies.

Discussion

In this meta-analysis, we combined and summarized the findings of previous experimental research to determine the effects of exposure to potential upward comparison targets on self-evaluations and emotions in a social media context. We tested four hypotheses. Hypotheses 1 and 2 examined the effects of upward social media comparisons on users’ emotions and self-evaluations. Hypotheses 3 and 4 explored the moderating roles of age and gender in these effects. Finally, our Research Questions examined whether outcomes were affected by the method in which social comparison was evoked in the studies (directly versus indirectly), or the type of control group (downward comparison versus neutral content).

Consistent with Hypothesis 1, the results revealed a small, negative effect of upward social comparison on users’ self-evaluations and emotions. These results support the meta-analytic findings of Gerber et al. (2018) and indicate that contrast is also the dominant response to upward social comparison in the context of social media. Although Gerber et al. (2018) found no social comparison effects on mood and self-esteem, this meta-analysis revealed that exposure to potential upward comparison targets on social media had adverse effects on users’ body image, subjective well-being, mental health, and self-esteem (in support of Hypothesis 2). Exposure to so-called “superior” others on social media leads people to evaluate themselves in a more negative light. As these upward comparisons are often the result of manipulated and idealized self-presentations (Gonzales & Hancock, 2011), we argue that people compare themselves to unrealistic, unrelatable targets.

In conjunction with social comparison theories, the findings of this meta-analysis contribute to a better understanding of social comparison in the context of social media. Festinger (1954) proposed that people engage in social comparison to establish accurate appraisals of the self. The meta-analytic results suggest that when faced with a potential upward comparison on social media, it is likely that users predominantly engage in a comparison process and experience negative appraisals as a result. According to the selective accessibility model, the adverse outcomes evidenced in this study indicate that users are making cognitive judgments of dissimilarity rather than assimilating to the comparison target (Mussweiler, 2001). In support of the findings of Gerber et al. (2018), it is apparent that assimilation and reflection are not dominant human responses to upward comparison.

Contrary to Hypotheses 3 and 4, moderator analyses revealed that the overall effect, as well as the effect for each individual outcome, was not dependent on the age and gender of participants. These findings conflict with the results of previous studies that have reported that females experience greater negative outcomes after engaging with social media than males (Fioravanti et al., 2020; Twenge & Martin, 2020). As previous research has shown that adolescents and young adults are the most affected by social comparison, we hypothesized that younger people would experience negative outcomes of a greater magnitude than older age groups (Noon et al., 2021). The lack of moderation may indicate that all ages and genders are equally affected, or that study limitations exist. There was a general lack of older participants, adolescents, and male representation across the studies, and thus there may have been limited statistical power to detect gender and age effects.

The method by which social comparison was induced across the studies was also not a significant moderator of the overall effect. Some studies evoked the social comparison process by exposing participants to stimuli that featured pre-tested upward comparison targets. A smaller number of studies requested participants to use their social media accounts as usual. Interestingly, using social media in a normal fashion had the same effect as viewing stimuli manipulated to be highly upward in nature. These results support the findings of Midgley et al. (2020), in which upward comparisons were found to consistently outnumber downward comparisons during the browsing of personal social media feeds. Finally, the type of control group was found to be a significant moderator. Surprisingly, studies that compared upward comparison targets to neutral content had larger effects than studies in which the upward comparison condition was compared to a downward comparison condition.

The current study’s findings should be considered in light of some limitations. A main issue was the tendency for studies to operationalize social media use as time spent on social media. Even though research has associated time spent on social media with increased social comparison, a more direct measure of social comparison like the State Appearance Comparison Scale (Tiggemann & McGill, 2004) would provide stronger evidence. Secondly, most studies included in the meta-analysis involved very short-term exposure to the upward comparison targets. Further longitudinal evidence would be beneficial in establishing the effects of engaging in upward comparison regularly and for more extended periods. Secondly, the study samples were relatively homogenous. There was a general lack of representation of males and older people. Furthermore, there was also a lack of representation of participants from non-Western countries. These restricted samples limit the generalizability of the meta-analytic findings, and we recommend that future research involve more diverse samples. Lastly, there were not enough data that could be extracted from the studies to explore the role of individual differences like social comparison orientation. Previous research has suggested that this may play a part in determining how a person responds to potential social comparison targets on social media (Fioravanti et al., 2020; Lim & Yang, 2019).

The limitations notwithstanding, this meta-analysis provides the first causal evidence of the effects of upward social comparison across several social media platforms, including Facebook, Instagram, and WeChat. This evidence offers valuable information to educate the public on the risks involved with social media. With this increased understanding, future research can develop strategies to combat the effects of upward social comparison for the large number of people who engage with these sites.

In conclusion, the current meta-analysis summarizes nearly 15 years of research investigating the effects of social comparison on social media. Results revealed that viewing upward comparison targets is associated with negative effects to users’ emotions and self-evaluations. Specifically, users experienced detrimental effects to their body image, subjective well-being, mental health and self-esteem after exposure to upward comparisons. In support of previous social comparison research in other contexts, the results of the current study suggest that there is a tendency to not assimilate on social media, and it is more commonplace for users to experience the negative repercussions of a contrastive response. Moving forward, it will be of crucial importance to establish methods to combat the negative effects of exposure to upward comparison targets on social media.

Open scholarship

This article has earned the Center for Open Science badges for Open Data, Open Materials and Preregistered. The data and materials are openly accessible at https://osf.io/2ajzb/ and https://doi.org/10.17605/OSF.IO/2UHEK

Disclosure statement

No potential conflict of interest was reported by the authors.

Data availability statement

 The data described in this article are openly available in the Open Science Framework at https://osf.io/2uhek

Supplementary material

Supplemental data for this article can be accessed online at https://doi.org/10.1080/15213269.2023.2180647.

Additional information

Funding

This research did not receive any specific grant from funding agencies in the public, commercial, or not-for-profit sectors.