Background
Work is largely a social endeavour [
1], and consequently many stressful experiences at work are social in nature [
2,
3]. As will be elaborated below, there are many different concepts that reflect social stressors (e.g. aggression, incivility, or abusive supervision). However, in line with stress-as-offense-to-self theory [
4], it can be argued that these different concepts have an important characteristic in common: They represent a threat to the self in terms of two strongly associated aspects of self-esteem, that is, social self-esteem (one’s perception of how one is evaluated by significant others) and personal self-esteem (one’s self-evaluation). This threat arises from feeling devalued (Leary & Allen [
5] refer to “relational devaluation”), which violates the basic human need to belong ([
6]; see also the need for relatedness in self-determination theory [
7]). Relational devaluation may induce stress not only if directed against an employee as a person. Rather, people’s social roles, including their occupational roles, often become part of their identity [
1] and therefore, part of their selves [
4,
8‐
10]. It follows that a threat to the self may be induced by savaging occupational roles; such threats can be regarded as “identity-relevant stressors” [
11].
Relational devaluation frequently leads to attempts to enhance and protect one’s self-esteem [
12‐
14]; unless these are successful, emotional distress in the form of “dysphoric reactions are likely, such as depression, anxiety, jealousy, and hurt feelings” ([
6] , p. 4), but anger-related reactions, somatic complaints, and other manifestations of low health and well-being are also typical [
4]. The association of social stressors with health and well-being is the focus of the current meta-analysis, with a special emphasis on the question of the extent to which social stressors represent a common core construct (relational devaluation) or different constructs.
Social stressors are an issue of common interest in society. The prevalence of social stressors at work is noticeable. For example, the Swiss stress study [
15] showed that 22% of the sample reported to have been exposed to social stressors within the previous 12 months.
In the UK [
16] the number of cases with work-related stress, depression and anxiety in 2018/19 was 602,000 in total, corresponding to a prevalence rate of 1800 per 100,000 employees. The number of days lost per employee was estimated at 21.2 per year, corresponding to a loss of 12.8 million working days. Social stressors (violence, threats and bullying) accounted for 13% of the work-related stressors reported.
The European working conditions survey [
17] showed that within 1 month prior to the survey 12% of employees experienced verbal abuse, 2% unwanted sexual attention, 6% humiliating behaviour and 4% threats. The percentage of employees reporting at least one adverse social behaviour differed between approximately 3 and 27% depending on the country. Those stressors were named as main reasons for staff turnover and absenteeism. Thus, empirical data indicate that social stressors represent a problem that is worth being taken seriously.
Although social stressors in terms of relational devaluation are rather common [
18,
19], for a long time social stressors at work have received relatively little attention compared to stressors included in prominent models of stress research, such as workload. The first self-report scales on social stressors were introduced in German by Frese and Zapf [
20], and in English as late as 1998 by Spector and Jex [
21]; interpersonal conflict scale]. More recently, however, various research traditions have developed that focus on constructs akin to relational devaluation. Three approaches can be distinguished. One approach focuses on specific types of social stressors such as bullying [
22] or abusive supervision [
23,
24]; for an overview see [
25]. A second group focuses on specific outcomes of social stressors such as aggression [
26]. A third approach focuses on the argument that all kinds of social stressors may be subsumed under one term. For example, Bowling and Beehr [
27] integrated different mistreatment variables under the term
workplace harassment (abusive supervision, bullying, emotional abuse, generalised workplace abuse, incivility, interpersonal conflict, mobbing, social undermining, victimisation, workplace aggression), emanating from the assumption that all mistreatment variables refer to the same overall construct. Similarly, Hershcovis [
28] conducted a meta-analysis in which she included different stressors in the category of
workplace aggression, taking into account different moderator variables such as intent, intensity, or frequency. Both [
27,
28] found effects similar in size, but there is also evidence for differences. For example, when looking at the source of aggression, Hershcovis and Barling [
26] found strongest effects for aggression by supervisors, as compared to colleagues, on employee outcomes such as job satisfaction, affective commitment, intent to turnover, or psychological distress. Overall, it is not clear to what extent different stressors have different or similar effect sizes. Because most of the studies have focused on one specific stressor and different outcomes, or on different stressor constructs and one outcome, it is difficult to see the big picture. An important step towards integration is the paper by Hershcovis [
28], who meta-analysed associations of five stressor constructs (incivility, supervisor aggression, bullying, interpersonal conflict, and social undermining) with various outcomes. Her results point more towards similarities than towards differences. However, a common core and specific differences may well exist simultaneously, as demonstrated by Baillien, Escartin, Gross, and Zapf [
29], who showed that bullying/mobbing shares many features with other conflicts yet has very specific characteristics as well.
In the current meta-analysis, we aim to show commonalities and differences in the associations between various social stressors and a) well-being (emotional, physical, mental, general, burnout), b) behaviour (turnover intention, absenteeism, organisational citizenship behaviour, performance, counterproductive work behaviour), and c) attitudes (commitment, satisfaction). Our aim is to test whether results by Hershcovis [
28] and Bowling and Beehr [
27], which did not yield much support for specific patterns, can be supported using a comprehensive data set, or whether relations between stressors and outcomes are different, so that it is not justified to assume that all social stressors belong to an overall construct. This issue is important because it has implications for future research. If our results point to a predominance of commonalities among social stressors, research should more strongly focus on differentiating situational characteristics [
29] when measuring social stress at work. If our results point to differences in stressor–outcome relations for different social stressors, the emphasis should be on identifying the most influential stressors and the most prominent outcomes for these different stressors. Thus, our results can lead to implications for work redesign and for interventions regarding social communication. At this point, we do not know of any meta-analysis comparing such a high number of social stressors and outcomes. Our intent is to broaden the knowledge about the relations between social stressors at work and various outcomes by expanding the number of potential associations.
Social stressors
Sonnentag and Frese ([
3], p. 562) defined social stressors in terms of “poor social interactions with direct supervisors, coworkers, and others”. Such social stressors convey devaluating social messages in a direct way. In addition, social stressors may also emanate from conditions at work. Thus, people can send “indirect social messages” through causing stress for others by being negligent (illegitimate stressors [
30];), or by assigning tasks that people think should not have to be done or should be done by someone else (illegitimate tasks [
31]).
In the literature, many different concepts of social stressors can be found; they include some key distinguishing features, but they also have considerable definitional, conceptual, and measurement overlap [
28,
32,
33]. The most frequently used concepts of social stressors can be found in the supplementary file
1. Those behaviours can occur in varying intensity, duration or frequency, directedness, and intention [
25,
28,
29,
34], which we refer to as situational characteristics. Besides these distinguishing situational characteristics, we contend that all social stressors have an important main feature in common: they are perceived as relational devaluation [
5].
Consequences of social stressors
The experience of relational devaluation through social stressors at work has been linked to various detrimental outcomes. In their meta-analysis, Hershcovis and Barling [
26] categorised outcomes into three broad groups: health-related outcomes, behaviour, and attitudes. As the health-related outcomes typically refer to psychological and physical symptoms rather than to diagnosed illness, we will refer to these outcomes in terms of well-being in this article.
The following evidence is derived from several meta-analyses and reviews.
Well-being
Well-being is a multifaceted construct that can be addressed from physical, emotional, psychological, and mental perspectives [
35]. Physical well-being is impaired by the experience of social stress at work [
22,
26,
36]. Social stressors also influence emotional well-being in terms of high-arousal negative emotions (e.g. anxiety) [
22,
27] and low-arousal negative emotions (e.g. depression) [
22,
26,
37]. The impact of social stress on psychological and general well-being has been shown in terms of burnout [
26,
27,
38] and general well-being [
24,
28]. Further, the influence on mental health problems has been well established [
22,
39,
40].
Behavioural outcomes (including behavioural intentions)
Bowling and Beehr’s [
27] meta-analysis revealed that experiencing social stressors (summarised as harassment) increased turnover intentions. Similarly, increased absenteeism has been found [
22]. Furthermore, task performance [
23,
24,
26,
27] as well as organisational citizenship behaviour (OCB) [
23] decreased by experiencing relational devaluation. On the other hand, counterproductive work behaviour (CWB), which describes a behaviour that negatively influences the productivity of an organization and/or its employees (e.g. withdrawing effort, leaving early, but also sabotage [
41]) increased when experiencing destructive leadership [
23,
24] or workplace harassment [
27].
Attitudinal outcomes
Social stressors can affect attitudes such as commitment and satisfaction, as shown for workplace harassment [
27], workplace bullying [
22], and destructive leadership [
24].
Table
1 displays recent findings from meta-analytic investigations.
Table 1
Integration of meta-analytic findings on social stress at work
Well-Being | Emotion: high arousal – negative | | | | | | | |
Well-Being | Emotion: low arousal – negative | .21* | | | .24* | | .18* | .38* |
Well-Being | physical | | | | .15* | .15* | .20* | .17* |
Well-Being | mental | | | | .25* | .30*-.31* | .19* | .19* |
Well-Being | Burnout | .32* | | .31* | .30* | | .25* | .31* |
Well-Being | general | | .01–.37* | .27* | | | | |
Behaviour | Turnover intention | | .22*-.34* | | .26* | .30* | .20* | .15* |
Behaviour | Absenteeism | | | | | | | |
Behaviour | Organisational Citizenship Behaviour (OCB) | .21* | | | | | | |
Behaviour | Performance | .17* | .18* − .22* | | .15* | | .07* | |
Behaviour | Counterproductive Work Behaviour (CWB) | .31*-.48* | .29*-.40* | | .29*-.34* | | .25*-.38* | .18*-.24* |
Attitude | Commitment | .23* | .19*-.28* | | .24* | .24*-.26* | .17* | .07* |
Attitude | Job satisfaction | .31* | .27*-.35* | | .32* | .34*-.35* | .20* | .12* |
Attitude | Life satisfaction | | | | | | | |
Category | Facet | Interpersonal conflict (Hershcovis, 2011) | Interpersonal conflict (Nixon, Mazzola, Bauer, Krueger, & Spector, 2011) | Incivility (Hershcovis, 2011) | Role ambiguity (Alcaron, 2011) | Role ambiguity (Nixon, Mazzola, Bauer, Krueger & Spector, 2011) | Role ambiguity (Schmidt, Roesler, Kusserow, & Rau, 2014) | Role conflict (Nixon, Mazzola, Bauer, Krueger, & Spector, 2011) | Role conflict (Schmidt, Roesler, Kusserow, & Rau, 2014) | Role conflict (Alcaron, 2011) |
Well-Being | Emotion: high arousal – negative | | | | | | | | | |
Well-Being | Emotion: low arousal – negative | | | | | | .28* | | .29* | |
Well-Being | physical | .16* | .22* | .17* | | .15* | | .27* | | |
Well-Being | mental | .35* | | .33* | | | | | | |
Well-Being | Burnout | | | | .24*-.26* | | | | | .29*-.42* |
Well-Being | general | | | | | | | | | |
Behaviour | Turnover intention | .33* | | .36* | | | | | | |
Behaviour | Absenteeism | | | | | | | | | |
Behaviour | Organisational Citizenship Behaviour (OCB) | | | | | | | | | |
Behaviour | Performance | | | | | | | | | |
Behaviour | Counterproductive Work Behaviour (CWB) | | | | | | | | | |
Attitude | Commitment | .21* | | .31* | | | | | | |
Attitude | Job satisfaction | .29* | | .40* | | | | | | |
Attitude | Life satisfaction | | | | | | | | | |
Category | Facette | Workplace discrimination (Dhanani, Beus, & Joseph, 2018) | Discrimination (Jones, Peddie, Gilrane, King, & Gray, 2016) | Workplace bullying (Nielsen, Indregard, & Øverland, 2016) | Workplace bullying (Nielsen & Einarsen, 2012) | Bullying (Hershcovis, 2011) | Workplace harassment (Bowling & Beehr, 2006) | Work harassment (Sojo, Wood, & Genat, 2016) | Sexual harassment (Sojo, Wood, & Genat, 2016) |
Well-Being | Emotion: high arousal – negative | | | | .27* | | .25* | | |
Well-Being | Emotion: low arousal – negative | | | | .34* | | .28* | | |
Well-Being | physical | .19*a | .13* | | .10–.28* | .32* | .25* | | .17* |
Well-Being | mental | .29*a | .25* | | .20*-.34* | .40* | | .37* | .27* |
Well-Being | Burnout | | | | .27* | | .33* | | |
Well-Being | general | | | | .31*-.37* | | | | .23* |
Behaviour | Turnover intention | | | | .28* | .35* | .29* | | |
Behaviour | Absenteeism | | | .13* | .11*-.12* | | .06 | | |
Behaviour | Organisational Citizenship Behaviour (OCB) | | | | | | .02 | | |
Behaviour | Performance | | | | .12 | | .06 | | |
Behaviour | Counterproductive Work Behaviour (CWB) | | | | | | .30* | | |
Attitude | Commitment | | | | .19* | | .30* | | |
Attitude | Job satisfaction | | | | .22* | .39* | .32* | | |
Attitude | Life satisfaction | | | | | | .18* | | |
As the table shows, most of the meta-analytic studies have investigated the effect of a specific social stressor on different outcomes. While effects within studies do differ, the results reflect a certain pattern.
Some outcome variables show significant associations across stressors rather consistently. These are negative emotions (both high and low arousal), mental well-being, burnout, general well-being, turnover intention, and attitudinal outcomes (commitment, job satisfaction). An exception is the correlation of .01 for destructive leadership and general well-being, but that is based on k = 1 with an N of 67. These outcomes reflect mainly psychological well-being and job/organization-related attitudes/intentions.
Another group of outcomes shows associations that differ more strongly depending on the specific stressors investigated, although, with one exception, all correlations are significant. These are physical well-being, OCB, performance, and counterproductive work behaviour, which seem to differ depending on the specific stressor. Note that behavioural outcomes dominate this category.
Finally, for absenteeism and life satisfaction, only few effects have been reported, and these were rather small.
From these meta-analyses, we can conclude that experiencing social stressors is associated with well-being, behaviour, and attitudes in the expected direction. There are commonalities and differences, the former relating mostly to psychological reactions, the latter to physical reactions and behaviour. The reason for this pattern might be that affective reactions represent rather direct effects, whereas behaviour depends on many additional mechanisms, such as forming intentions and encountering circumstances that allow for behaviours such as counterproductive work behaviour. It is debatable, however, whether the differences in these data arise from real differences between the social-stressor variables. To investigate differences and commonalities in associations between specific stressors and outcomes, we therefore will discriminate among social stressor-variables as well as outcomes in our meta-analysis. We will use a comprehensive data set with sufficient power to detect possible difference.
Discussion
Within the present meta-analysis, 557 studies comprising 640 different samples were analysed with respect to the association of social stressors at work with well-being and health-related outcomes. We found an overall relation of
r = −.30, representing a moderate correlation. To deal with the heterogeneity of the results, we grouped the outcome variables into the three categories well-being, behaviour, and attitudes. All three categories were affected by social stressors at work, but their effects differed significantly. Furthermore, social stressors differed significantly when analysed as a moderator within the overall mean effect. Due to the comprehensive sample including various social stressor constructs, a notable amount of heterogeneity arose. Within this variance, the outcome categories well-being, behaviour, and attitudes were able to explain a significant amount of variance. Nevertheless, when comparing the size of the correlation effects of the three groups, they are comparable with regard to Cohen’s conventions [
61].
The overall mean correlation of social stressors and health- and well-being-related outcomes supports the existing evidence [
22,
27,
28]. The effects differed by both type of social stressor and outcome category. Differences concerning outcome variables (well-being, behaviour, attitudes) show a clear pattern, with attitudinal outcomes such as commitment and job satisfaction showing the highest effects. This also supports earlier research [
28]. In addition to commitment and job satisfaction, emotional exhaustion (burnout) and counterproductive work behaviour also yielded moderate effect sizes. Surprisingly, there was a comparatively low effect for absenteeism, which might be due to fewer samples contributing to this effect. However, Nielsen and Einarsen [
22] and Nielsen, Indregard, and Øverland [
62] found effects of bullying on absenteeism comparable to the one we found. A problem with studies on absenteeism is that it is often not clear whether only voluntary absenteeism is included, or also absenteeism due to sickness [
63]. Physical outcomes show comparatively low associations with stressors in our study, and also in research on stress at work on the whole [
64]. In general, all outcome categories reached significance and are therefore associated with social stressors at work. The fact that attitudinal outcomes are concerned most might show the influence on short-term outcomes. It is conceivable that after an attitude has developed it might lead to behavioural intentions or acts, for instance in terms of revenge [
65,
66], implying that an individual is more likely to engage in negative behaviours as compensation tit for tat [
67]); however, engaging in such behaviours requires additional steps and decisions, implying lower associations as compared to attitudes.
When looking at the mean effects of the different social stressor constructs, we can also see a variety of effects indicated by a significant amount of heterogeneity; however, while statistically significant, the differences between effect sizes are not very large. Lack of justice shows the strongest effects, followed by incivility and mobbing and bullying. Unfortunately, stereotype threat, identity threat, illegitimate tasks, undermining, perceived victimisation, and hostility were represented by comparatively few samples. It is therefore difficult to draw conclusions regarding these variables. Overall, the variance in effect sizes between different social stressor concepts is smaller as compared to those of the outcomes. Put differently, the effects of different social stressors are more similar than the general effect of social stress on different outcomes.
To be able to draw more specific conclusions regarding distinct relations, we calculated cross-correlations, which showed certain patterns. First, there is one group of outcome variables for which the effects of distinct social stressors are not different. This group contains negative emotions, both high arousal and low arousal, burnout, absenteeism, OCB, and performance. Second, we have a group of outcomes for which the type of social stressor matters with respect to the effect size. This category includes physical, mental, and general well-being, turnover, CWB, commitment, and job satisfaction. This indicates that there are indeed distinct patterns for social stressors and strain relations, and that it might be important to distinguish the kind of devaluation experienced with regard to the outcome of interest. It seems striking that lack of justice, supervisor mistreatment, and mobbing/bullying are especially important for attitudes. All three constructs are characterised by an imbalance of power, which underscores the importance of this aspect. Furthermore, harassment, social exclusion, and interpersonal conflicts seem to be important especially when measuring counterproductive work behaviour. What also striked us is that incivility shows comparatively high effects on all outcomes. Even though it is defined as a low-intensity behaviour of ambiguous intent [
67], it is associated with health and well-being to a much higher extent than more obviously threatening stressors (e.g. mistreatment, aggression, and violence), which might be due to its higher occurrence compared to high-intensity behaviours.
Range restriction regarding specific social stressors
When looking only at the strength of the social stressor concepts with sufficient sample sizes across different outcomes (Table
4), we see that most of them are similar in size, suggesting the presence of an overall construct. The largest effect sizes concern lack of justice (
r = −.33), incivility (
r = −.32), and mobbing/bullying (
r = .32), representing moderate effect sizes. Most of the other concepts yield effect sizes ranging between .24 and .29, representing small to moderate effects. Smaller effect sizes are found only for sexual mistreatment (
r = −.19), physical violence (
r = −.17), and undermining (
r = .18), the latter being represented by a fairly small number of studies. It is striking that both sexual mistreatment and physical violence not only express relational devaluation but also threaten and violate a person’s physical integrity. Both concepts have a very high intensity, and they likely occur less often compared to lower-intensity behaviour, such as incivility. Consequently, they might show restricted variance when compared to stressors occurring more often. This would explain their lower effect size despite their high intensity. Additionally, sexual mistreatment might especially suffer from underreporting due to victims not reporting each incident that might have happened. This would further support our assumption of restricted variance in those concepts. Rare reporting, be it due to rare occurrence and/or to underreporting, induces skewness in the data and reduces the maximum correlation possible. If that is taken into account, the values obtained for variables that are reported infrequently may well be erroneous in suggesting a low association. More specifically, in such cases low-correlation coefficients can be associated with high relative risks associated with the occurrence of rare social stressors (cf. the correlation of
r = .10 between smoking and lung cancer, which implies a relative risk of 11) ([
68] , pp. 53–54).
The similarity in effect sizes across different social stressors and different outcomes suggests that there is, indeed, a common core to social stressors; we feel that the term “relational devaluation” describes this core very well. Differences would then mainly be due not to a different nature of the different stressors but to additional characteristics mentioned in the literature. The most prominent examples of such additional characteristics are duration/repetitiveness, intensity, and power differential [
25,
28,
29,
34]. Such distinguishing features might be responsible for the heterogeneity found, as well as for the restriction of variance in high-intensity/low-frequency constructs, such as physical violence. As an example, offensive behaviours such as ridiculing, excluding, and insulting, which characterise mobbing, may well occur as social stressors yet not be characterised as mobbing if they are infrequent, experienced over short periods of time (e.g. at times in which a superior is stressed), and addressing not a specific target but whoever happens to be around [
69,
70]. Thus, it is the specific additional characteristics of social stressors in a given context, rather than their intrinsic qualities, that are likely responsible for differences in effect sizes. The validity of this conclusion can be determined, however, only if these additional characteristics are assessed simultaneously; only then is it possible to determine if such characteristics play a decisive role in predicting outcomes—and if they do, in which constellations. We therefore suggest that future studies assess as many of these characteristics as possible.
Following Berry and colleagues [
71], we were able to show that the existing empirical data show patterns, but the differences found were predominantly on the outcome level. This means we mainly found differences in effect sizes depending on the measured outcomes. Our data indicate that overall social stressors representing relational devaluation should be taken into account as a serious threat to well-being and health. For organisations, it should be interesting that lack of justice, supervisor mistreatment, and mobbing/bullying show especially high associations. Note that all three of these are typically characterised by a high power differential. All three of them can be targeted by organisational policies to prevent them or to raise awareness. Supervisor trainings can be tools to optimize leadership behaviour and to sensitise leaders for employee needs. Furthermore, the fact that attitudes might be affected most easily by social stress at work yields a good diagnostic tool for organisations. If they consistently ask about their employees’ satisfaction, they will have a good indicator for interpersonal as well as climate problems. Because we found that all investigated social stressor concepts show an impact, it might be appropriate for organisations to use more general measurement instruments to screen for relational devaluation. In case of specific problems within the organisation, more specified measurement instruments could be used, depending on the situational characteristics. Our results indicate that low-intensity but high-frequency behaviours (e.g. incivility) should not be underestimated. At the same time, the comparatively low associations with outcome variables of high-intensity but low-frequency behaviours, such as physical aggression or sexual offenses, should not be mistaken to imply a low impact, as low frequency and underreporting restrict variance, and thus the maximum correlation that can be obtained.
For future research, it is important to investigate long-term effects. It would be interesting to see if the effects change during long-term assessments, with stronger effects for well-being-related outcomes due to the consistency of social stressors over time (versus adaptation). Furthermore, it would be interesting to have a closer look at our finding of differences in terms of outcomes but much smaller differences between social stressor constructs. Why are the effects of social stressors different within some outcomes and not within others? Moreover, the consideration of moderating situational variables, such as power balance, source of stress (customer, colleague, and supervisor), duration, and intensity, seems to be promising. Further, gender and ethnicity might play a role as well. As the European working conditions survey [
17] pointed out, depending on the adverse social behaviour investigated, there seem to be differences relating to gender. For example, more men are targeted by threatening behaviour. Following, it could be promising to explore the effects for men and woman separately, which is not often done in existing research. As has been mentioned in the introduction the numbers of adverse social behaviour differ a lot depending on the country, which might point to cultural differences. A completely new field in the area of social stressors is arising with research on human–machine interactions [
72]. So-called “hybrid teams” consisting of at least one human and one machine agent can also be expected to be exposed to social stress, for instance when machine agents take over tasks typically carried out by a human agent. It seems promising to focus more strongly on this research and draw comparisons between human and machine agents.
Limitations
The present meta-analysis has some limitations. First, we did not include studies published in books, dissertations, or “grey” literature, such as research reports. Second, we included a broad variety of different types of social stressors as well as well-being- and health-related outcomes to be able to compare them. Therefore, we had to face a highly heterogeneous data set. We tried to deal with this issue by grouping social stressors as well as outcomes in categories derived from the literature. Third, due to our strategy for cross-calculations, we randomly selected one effect per study to ensure interpretable results. Therefore, not all evidence has been integrated within the cross-calculations. However, due to the random selection there should be no bias. Fourth, the effects were mainly cross-sectional, preventing us from drawing causal conclusions and conclusions regarding long-term effects. Investigating long-term effects certainly seems warranted. As mentioned above, we did not include research on human–machine interactions but focused on human interactions exclusively. Further, effects could not be analysed separately for men and women, as such analyses are seldom reported in the literature, and using the percentage of males/females as a moderator would not be more than a very rough proxy.
On the other hand, our study has several strengths. First, we included a broad variety of different social stressors as well as well-being- and health-related outcomes to be able to compare them. By building classifications and testing for potential moderators, we tackled the existing heterogeneity. Due to our cross-correlations, we were able to show unique patterns that have not been shown before. Second, we included 557 studies representing 640 different samples, which is a comparatively high number. We do not know of any meta-analytic study in the field integrating such a great amount of data. This should result in high power to find results (although it also entails the danger of chance findings, which we dealt with by applying a strict criterion for statistical significance). Third, we applied meta-analytic calculations to quantify our findings.
Publisher’s Note
Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations.