Introduction
Perfectionism has been proposed to be a transdiagnostic process across numerous psychological disorders (Egan et al., Reference Egan, Wade and Shafran2011). The Multidimensional Perfectionism Scales (FMPS, Frost et al., Reference Frost, Marten, Lahart and Rosenblate1990; HMPS, Hewitt and Flett, Reference Hewitt and Flett1991) have consistently been found to load on the two perfectionism dimensions of perfectionistic concerns (rumination over mistakes and the belief others expect the individual to be perfect), and perfectionistic strivings (striving towards high standards) (Smith and Saklofske, Reference Smith and Saklofske2017).
Another definition of perfectionism is clinical perfectionism, defined as self-esteem based on striving to meet high standards despite negative consequences (e.g. anxiety and depression) (Shafran et al., Reference Shafran, Cooper and Fairburn2002). The definition of clinical perfectionism emphasises both perfectionistic strivings and concerns as relevant to psychopathology if an individual bases their self-worth on their achievement of high standards (Shafran et al., Reference Shafran, Cooper and Fairburn2002). The cognitive-behavioural model of clinical perfectionism (Shafran et al., Reference Shafran, Cooper and Fairburn2002; Shafran et al., Reference Shafran, Egan and Wade2010) includes a range of processes that maintain clinical perfectionism, including for example, performance checking behaviours, re-setting standards higher following both success and failure, and a number of cognitive biases that the individual engages in (i.e. discounting success and focusing on flaws in performance). The Clinical Perfectionism Questionnaire (CPQ; Fairburn et al., Reference Fairburn, Cooper and Shafran2003) has been used to measure clinical perfectionism. Various studies have found different factor solutions, consisting of single, bi-factor and two factor solutions, consistent with an overall clinical perfectionism construct, in addition to factors congruent with both perfectionistic strivings and perfectionistic concerns (Egan et al., Reference Egan, Shafran, Lee, Fairburn, Cooper, Doll, Palmer and Watson2016; Howell et al., Reference Howell, Anderson, Egan and McEvoy2020; Prior et al., Reference Prior, Erceg-Hurn, Raykos, Egan, Byrne and McEvoy2018; Shu et al., Reference Shu, O’Brien, Watson, Anderson, Wade, Kane, Lampard and Egan2020).
A significant moderate positive association has been found between perfectionistic concerns and symptoms of anxiety, depression, obsessive-compulsive disorder (OCD), and eating disorders in adults (r=.33 to r=.43; e.g. Callaghan et al., Reference Callaghan, Greene, Shafran, Lunn and Egan2023; Stackpole et al., Reference Stackpole, Greene, Bills and Egan2023). Similarly, in adult samples, significant, although smaller correlations have been found between perfectionistic strivings and symptoms of anxiety, depression, OCD and eating disorders (r=.10 to r=.21; e.g. Callaghan et al., Reference Callaghan, Greene, Shafran, Lunn and Egan2023; Stackpole et al., Reference Stackpole, Greene, Bills and Egan2023). Indeed, higher perfectionistic concerns have been found to consistently relate to other correlates of poor mental health, including low self-esteem. Low self-esteem can be defined as a negative view toward oneself and poor self-worth, operationalised as low scores on common measures of self-esteem, such as Rosenberg’s (Reference Rosenberg1965) Self-Esteem Scale (RSES). Moderate to large negative associations have been found between self-esteem and perfectionistic concerns in numerous studies, indicating higher perfectionistic concerns is related to lower self-esteem (e.g. Jayakumar et al., Reference Jayakumar, Sudhir and Mariamma2016; Juwono et al., Reference Juwono, Kun, Demetrovics and Urbàn2022; Moroz and Dunkley, Reference Moroz and Dunkley2015; Murray et al., Reference Murray, Rieger, Karlov and Touyz2013; Piotrowski et al., Reference Piotrowski, Bojanowska, Nowicka and Janasek2023; Raedeke et al., Reference Raedeke, Blom and Kenttä2021). Similarly, meta-analytic evidence of longitudinal studies has demonstrated that low-self-esteem is associated with higher symptoms of anxiety, depression (Sowislo and Orth, Reference Sowislo and Orth2013), and eating disorders (Krauss et al., Reference Krauss, Dapp and Orth2023). The meta-analyses have demonstrated reciprocal effects between self-esteem and psychopathology, where self-esteem both leads to and is reduced by psychopathology (Krauss et al., Reference Krauss, Dapp and Orth2023; Sowislo and Orth, Reference Sowislo and Orth2013).
Unlike the association between perfectionism and psychopathology which has been summarised in numerous meta-analyses (e.g. Bills et al., Reference Bills, Greene, Stackpole and Egan2023; Callaghan et al., Reference Callaghan, Greene, Shafran, Lunn and Egan2023; Limburg et al., Reference Limburg, Watson, Hagger and Egan2017; Lunn et al., Reference Lunn, Greene, Callaghan and Egan2023; Smith et al., Reference Smith, Sherry, Ray, Hewitt and Flett2021; Stackpole et al., Reference Stackpole, Greene, Bills and Egan2023), to date there has been no meta-analysis of the relationship between perfectionism and self-esteem.
It is important to understand processes such as perfectionism which relate to low self-esteem, given the consistent relationship between self-esteem and psychopathology to further understand correlates of low self-esteem (Zeigler-Hill, Reference Zeigler-Hill2011). Furthermore, there are inconsistent associations between perfectionistic strivings and self-esteem, with some studies indicating a non-significant relationship (e.g. Moroz and Dunkley, Reference Moroz and Dunkley2015; Piotrowski et al., Reference Piotrowski, Bojanowska, Nowicka and Janasek2023), some a negative relationship (e.g. Murray et al., Reference Murray, Rieger, Karlov and Touyz2013; Jayakumar et al., Reference Jayakumar, Sudhir and Mariamma2016; Raedeke et al., Reference Raedeke, Blom and Kenttä2021), while others indicate a positive relationship (e.g. Barnes and Caltabianco, Reference Barnes and Caltabiano2017; Chen et al., Reference Chen, Zhong, Cao, Jin, Wang, Ling, Cen, Zhu, Yao, Zheng and Yi2017; Chou et al., Reference Chou, Huang, Lin, Lu, Chiu and Chen2019; Trumpeter et al., Reference Trumpeter, Watson and O’Leary2006).
Given the mixed directions of association and inconsistency in findings, it would be useful to summarise in a meta-analysis the associations between perfectionistic concerns and strivings, and self-esteem to gain clarity on the relationship between perfectionism and self-esteem. Understanding the relationships between perfectionism dimensions and self-esteem has important theoretical implications, particularly for understanding clinical perfectionism, where it is hypothesised that self-esteem is based on striving for high standards (Shafran et al., Reference Shafran, Cooper and Fairburn2002). The cognitive-behavioural approach to self-esteem also suggests that self-esteem and perfectionism are closely related, with the latter being a compensatory strategy to hide low self-esteem (Fennell, Reference Fennell2009). Gaining insight into the associations between the dimensions of perfectionism may provide a direction for future research on how to improve treatments aimed at decreasing perfectionism and improving self-esteem as transdiagnostic processes that maintain psychopathology (Egan et al., Reference Egan, Wade and Shafran2011; Fennell, Reference Fennell2009).
The aim of this systematic review and meta-analysis was to examine the association between perfectionistic strivings, perfectionistic concerns, and self-esteem in adults. We chose to restrict to adult samples rather than include studies across age ranges including children and adolescents given the rationale of developmental differences potentially impacting the results and that associations should be considered separately between youth and adults (Bills et al., Reference Bills, Greene, Stackpole and Egan2023; Lunn et al., Reference Lunn, Greene, Callaghan and Egan2023). Based on previous meta-analyses of the relationship between perfectionism dimensions and psychopathology (Bills et al., Reference Bills, Greene, Stackpole and Egan2023; Callaghan et al., Reference Callaghan, Greene, Shafran, Lunn and Egan2023; Lunn et al., Reference Lunn, Greene, Callaghan and Egan2023; Stackpole et al., Reference Stackpole, Greene, Bills and Egan2023) we predicted there would be a significant moderate negative pooled association between perfectionistic concerns and self-esteem. Given the mixed evidence to date regarding the association between perfectionistic strivings and self-esteem, with some studies finding a negative association, others finding no association, whilst others demonstrating positive correlations, we did not outline a hypothesis for this aim.
Method
The search terms and research protocol were registered on PROSPERO on 23 May 2023, prior to the search (Registration ID: CRD42023425443). The PRISMA guidelines (Page et al., Reference Page, McKenzie, Bossuyt, Boutron, Hoffmann, Mulrow, Shamseer, Tetzlaff, Akl, Brennan, Chou, Glanville, Grimshaw, Hrobjartsson, Lalu, Li, Loder, Mayo-Wilson, McDonald and Moher2021) were adhered to. The research conformed to the Declaration of Helsinki.
Inclusion and exclusion criteria
Studies were judged against the following inclusion criteria: (1) mean age of sample ≥18 years, or a minimum age of 18 in the age range when the mean was not available; (2) included a correlation or convertible effect size between self-esteem and perfectionism; (3) included a standardised measure of perfectionism and self-esteem; (4) reported unadjusted effect sizes; (5) correlational study designs or reported a group comparison; (6) published, quantitative research in peer-reviewed journals reported in English; and (7) from any country, published between 1 January 1990 and 31 May 2023. The rationale for using articles from 1990 onwards was this was the year of publication of the first multi-dimensional measure of perfectionism (FMPS; Frost et al., Reference Frost, Marten, Lahart and Rosenblate1990). Exclusion criteria were grey literature, qualitative studies, dissertations, study protocols, case conferences, and unpublished data.
Search strategy and selection process
A literature search was conducted on 31 May 2023 on PsycINFO, OVID Medline, PsychARTICLES, Scopus, and ProQuest Central. The search terms were decided by S.E. and R.S., and preliminary searches and search strategy were guided by a senior health librarian at Curtin University. The search terms were (perfectionis*) AND (self-esteem OR self-worth OR self-evaluation). Although we were interested in finding studies that were on self-esteem and included a specific measure of measure of self-esteem, the terms self-worth and self-evaluation were added to potentially detect relevant articles where self-esteem may not have been specifically mentioned in an abstract. The senior author (S.E.) trained V.K. (an Honours in psychology student) in how to perform article screening, who screened 100% of articles at title and abstract, and full text levels. A second independent assessor, another Honours student in psychology (S.D.), screened a random 30% of the studies at title and abstract (n=518) and full text (n=59) level. The two assessors (V.K., S.D.) worked independently and were blind to each other’s ratings. Cohen’s kappa coefficient was used to assess inter-rater reliability between the assessors and was calculated in SPSS (version 28; IBM Corporation, 2021). Kappa coefficients were interpreted as: no agreement (κ<0), slight agreement (κ=.01–.20), fair agreement (κ=.21–.40), moderate agreement κ =.41–.60), substantial agreement (κ=.61–.80), or near perfect agreement (κ=.81–.99) (Landis and Koch, Reference Landis and Koch1977). Inter-rater reliability between the first and second reviewers demonstrated substantial agreement at the title and abstract level (κ=.68) and near-perfect agreement at the full-text level (κ=.94). Discrepancies in study eligibility were discussed with S.E. until consensus was reached.
Data extraction and management
Data extraction was performed by V.K. and included Pearson’s correlation coefficients, sample size, study design, population (e.g. clinical or non-clinical), the instrument(s) used to measure self-esteem and perfectionism, the country where the study was conducted, gender, mean age of the sample including standard deviation and age range. In cases where studies did not report a correlation coefficient or the data required to calculate one (i.e. an effect size), V.K. contacted the corresponding author to obtain the data. Five authors were contacted, and two responded. Studies where authors did not respond with the required data were excluded. When extracting data from longitudinal and intervention studies, only the correlations at baseline were obtained to ensure independence of effect sizes. For intervention studies, effect sizes were calculated using the Campbell Collaboration Effect Size Calculator (Wilson, Reference Wilson2001) to transform means and standard deviations into correlation coefficients. The standard error (SE) for the r coefficient for each effect was calculated using the formula (SE=(1–r 2)/√N–3; Gnambs, Reference Gnambs2023).
Risk of bias
The risk of bias was assessed independently by V.K. following training from S.E. with the 14 items from the cross-sectional and observational section of the National Institutes of Health quality assessment tool (National Institutes of Health, 2014). The study design, selection bias, confounding variables, information bias and measurement bias, were rated ‘no’ (criterion not met), ‘yes’ (criterion met), ‘cannot determine’, ‘not applicable’, ‘not reported. An overall score was determined by dividing the number of ‘yes’ responses by the number of appropriate items and multiplying them by 100 to produce a percentage. Scores below 50% indicate ‘poor’ quality, 51–74.99% denote ‘fair’ quality, and 75% and above denote good quality (National Institutes of Health, 2014).
Assessment of heterogeneity
Cochran’s Q and I 2 statistics were calculated to assess the heterogeneity of effect sizes (Higgins et al., Reference Higgins, Thompson, Deeks and Altman2003). Heterogeneity was assessed using Cochran’s Q, with the Hunter-Schmidt method (Higgins et al., Reference Higgins, Thompson, Deeks and Altman2003). I² determined the percentage of variation across studies, where heterogeneity of 25–49% is ‘low’, 50–74% is ‘moderate’, and >75% is ‘high’ (Higgins et al., Reference Higgins, Thompson, Deeks and Altman2003). Generally, it is recommended to explore potential moderators when I² is high (≥75%; Higgins et al., Reference Higgins, Thompson, Deeks and Altman2003). Exploring potential variables can be useful in examining reasons for high heterogeneity (Cuijpers, Reference Cuijpers2016).
Meta-regressions were used where there was high heterogeneity (>75%; Cuijpers, Reference Cuijpers2016; Fu et al., Reference Fu, Gartlehner, Grant, Shamliyan, Sedrakyan, Wilt, Griffith, Oremus, Raina, Ismaila, Santaguida, Lau and Trikalinos2011). Following Deeks et al. (Reference Deeks, Higgins and Altman2023) recommendations, meta-regression can be used to investigate differences for categorical variables, as for subgroup analyses using dummy variables (e.g. 1=non-clinical, 2=clinical) in the meta-regression model, as it is for standard linear regression modelling. The regression coefficient indicated how the effect for each subgroup differed from the nominated reference subgroup and the p-value of each regression co-efficient was examined to indicate whether the differences were statistically significant. Potential sources of heterogeneity, classified as categorical variables, included the mean age of participants (i.e. young adult, 18–24.99 years or adult, 25 years and over), participants’ gender (male or female or mixed), sample type (university sample, community sample, non-clinical (psychological not physical) other or clinical (psychological), and study quality (poor, fair, or good quality). Following Lunn et al. (Reference Lunn, Greene, Callaghan and Egan2023), country of residence was another moderator and was classified as high, low or middle-income (LMIC) based on the Organisation for Economic Co-operation and Development (OECD) country classification system (Guedj, Reference Guedj2022). We did not run perfectionism measure as a moderator as some studies used several different measures which we averaged so that each sample only provided one data point per outcome. Please note that none of the clinical sample studies provided specific effect sizes for perfectionistic concerns or perfectionistic strivings. If a moderator was a significant, we ran subgroup analyses to determine the pooled effects for each category of that moderator.
Data synthesis, sensitivity analysis and assessment of publication bias
The meta-analysis was conducted using JASP (version 0.17.3; JASP Team, 2023). Pooled correlations between perfectionism and self-esteem represent the effect size in this meta-analysis and were estimated using a Hedges random effects model with 95% confidence intervals (Hedges and Olkin, Reference Hedges and Olkin1985). Cohen’s (Reference Cohen1992) conventions were used to interpret the pooled correlations as small (r=.10), medium (r=.30), or large (r≥.50).
Perfectionism outcomes were classified into perfectionistic concerns and perfectionistic strivings following the review of Limburg et al. (Reference Limburg, Watson, Hagger and Egan2017) (Table 1). In cases where correlation coefficients were presented for the subscales or reported for more than one perfectionism measure, the average effect sizes were calculated to determine an overall perfectionistic concerns or strivings correlation with self-esteem. For example, Pearson and Gleaves (Reference Pearson and Gleaves2006) reported two measures of perfectionism: the Almost Perfect Scale-Revised (APS-R; Slaney et al., Reference Slaney, Rice, Mobley, Trippi and Ashby2001) and the FMPS (Frost et al., Reference Frost, Marten, Lahart and Rosenblate1990). The subscales/measures that fit into perfectionistic strivings were averaged for each study (e.g. for the FMPS, the personal standards and organisation subscales were averaged to provide an overall correlation for perfectionistic strivings, the same was done for the high standards and order subscales of the APS-R). Studies were categorised as clinical or non-clinical depending on whether authors stated their sample met the criteria for a diagnosis of a psychological disorder based on the Diagnostic and Statistical Manual of Mental Disorders (DSM-5; American Psychiatric Association, 2013) or the International Classification of Diseases and Related Health Problems (ICD-11; World Health Organisation, 2019). Samples were classified as non-clinical if they did not meet the diagnostic criteria for a psychological disorder.
Table 1. Subscale classification of perfectionism measures into perfectionistic strivings and perfectionistic concerns
Adapted from ‘The Relationship Between Perfectionism and Psychopathology: A Meta-Analysis’ by K. Limburg, H. J. Watson, M. S. Hagger, and S. J. Egan (2017), Journal of Clinical Psychology, 73, 1303 (https://doi.org/10.1002/jclp.22435).
Publication bias was assessed through funnel plot visual inspection and Egger’s test for plot asymmetry (Egger et al., Reference Egger, Davey Smith, Schneider and Minder1997). If Egger’s test was significant (p<.05), plot asymmetry was assumed, indicating potential publication bias concerns. Forest plots were visually inspected for outliers, and case-wise diagnostics were conducted in JASP (JASP Team, 2023) to assess influential studies through Cook’s distance. Sensitivity analyses were conducted without the outliers to determine if removing these cases would meaningfully change the pooled effect sizes.
Results
Study characteristics
There were 83 studies found (Fig. 1) with 88 samples from which 58 effect sizes for perfectionistic strivings and 68 effect sizes for concerns were extracted or computed. Note that some studies/samples did not provide specific effect sizes for either perfectionistic strivings or concerns – instead they include only effect sizes for global perfectionism or another type of perfectionism. These studies are summarised in Table 2 but are not included in the meta-analysis.
Figure 1. Study selection flow chart in accordance with PRISMA 2020 guidelines. n, number of studies. *No valid/standardised measure refers to a questionnaire measure of perfectionism/self-esteem which did not have published evidence of a minimum reliability alpha co-efficient of .70 or above, validity, or other psychometric properties reported in the literature.
Table 2. Study characteristics
OCD, obsessive-compulsive disorder. Dash (—) reflects data not obtained or reported; F (%), percentage of sample that are female; RSES, Rosenberg Self-Esteem Scale; APS-R, Almost Perfect Scale-Revised; BTPS, Big Three Perfectionism Scale; HMPS=Hewitt and Flett Multidimensional Perfectionism Scale; FMPS=Frost Multidimensional Perfectionism Scale; EDI-P, Eating Disorder Inventory-Perfectionism subscale; PANPS, Positive and Negative Perfectionism Scale; CSEI, Coopersmith Self-Esteem Inventory; CPQ, Clinical Perfectionism Questionnaire; DEQ, Depressive Experiences Questionnaire; CSWS, Contingencies of Self-Worth Scale; S-HMPS, Short Multidimensional Perfectionism Scale based on HMPS; DAS-P, Dysfunctional Attitude Scale-Perfectionism subscale; SSES, State Self-Esteem Scale.
a Studies/samples not included in the meta-analysis as they did not include a specific effect size for perfectionistic strivings or concerns.
Studies were conducted between 1991 and 2023 across 20 countries. There were 32,304 participants (approximate mean pooled age=24.66 years). Sample sizes ranged from n=30 to n=4340, with an average sample size of n=367. Samples were predominantly from high-income countries including the USA (n=38), Europe (Belgium, Italy, Portugal, all n=3; Sweden, Germany, Hungary, all n=2, Croatia, Finland, Poland, all n=1), Australia (n=7), the United Kingdom (n=5) and Canada (n=3). There were a small number from high-income nations in Asia (South Korea, n=1; Taiwan, n=2). Low and middle-income countries represented a small number of studies (China, n=5; Lebanon, n=2; Iran, n=3; India, n=2; Turkiye, n=2). Samples were mainly non-clinical (n=82), with only six clinical samples, of which the diagnoses were eating disorders (n=4), OCD (n=1), and mixed psychological disorder diagnoses (n=1). Perfectionism was most commonly measured using the FMPS (n=40), followed by the HMPS (n=18; Hewitt and Flett, Reference Hewitt and Flett1991), APS-R (n=19; Slaney et al., Reference Slaney, Rice, Mobley, Trippi and Ashby2001), and EDI-P (n=10; Garner et al., Reference Garner, Olmstead and Polivy1983) with some studies using more than one measure of perfectionism (see Table 2 for all measures used). The Rosenberg Self-Esteem Scale (RSES; Rosenberg, Reference Rosenberg1965) was the most common measure of self-esteem (n=76).
Risk of bias
The quality rating scores were ‘poor’ (5% of studies), fair (70%) and good (25%) (Table 2). On average, studies were rated as fair (M=68.12%, SD=8.98%), which may be due to the cross-sectional design of most studies, with the majority not justifying their sample size or reporting details of sampling methods.
Publication bias
Egger’s test for plot asymmetry was significant, for perfectionistic concerns (z=2.35, p=.019). However, a trim-and-fill analysis did not impute any studies or alter the effect size, suggesting that there is unlikely to be publication bias. Therefore, asymmetry might be due to other factors. For perfectionistic strivings Egger’s test was not significant (z=–1.01, p=.311), indicating publication bias was not of concern. Visual inspection of funnel plots comprising pooled effect sizes confirmed symmetry for perfectionistic strivings, showing no evidence of publication bias (see Supplementary material for funnel plots).
Pooled correlations between perfectionism dimensions and self-esteem
Pooled correlations between perfectionistic concerns, perfectionistic strivings, and self-esteem are shown in Table 3. Consistent with our prediction, there was a significant, moderate negative pooled correlation (r=–.42) between perfectionistic concerns and self-esteem, indicating higher perfectionistic concerns are related to lower self-esteem. There was a negligible significant, positive pooled correlation between perfectionistic strivings and self-esteem (r=.06), indicating higher perfectionistic strivings are related to higher self-esteem.
Table 3. Summary statistics for pooled correlations between perfectionism dimensions and self-esteem
k, number of data points included (i.e. independent samples); r, pooled association; SE, standard error; CI, confidence interval; LL, lower limit; UL, upper limit; Q, Cochran’s Q; I 2, heterogeneity.
Sensitivity analysis
When assessing perfectionistic strivings, two studies were identified as influential outliers (Besharat, Reference Besharat2009; Moghadam et al., Reference Moghadam, Moghadam and Jahangir2021). Removing these outliers made little difference (r=.06 [.01, .10], p=.009) to the direction or significance of the results; therefore they were retained. Outliers on perfectionistic concerns were identified as Barnes and Caltabiano (Reference Barnes and Caltabiano2017), Chou et al. (Reference Chou, Huang, Lin, Lu, Chiu and Chen2019) and Koivula et al. (Reference Koivula, Hassmen and Fallby2002). Removing these outliers would not have a meaningful impact as the effect size was moved by −.03, and the relationship remained significant (r=−.45 [−.49, −.42], p<.001) hence these studies were also retained.
Heterogeneity
Heterogeneity was high, indicating wide variance in correlations across the studies (I 2≥75%), for perfectionistic concerns and self-esteem, I 2=95.35%; and perfectionistic strivings and self-esteem, I 2=93.22%. Meta-regressions were conducted (Table 4) on the mean age of participants (18–24.99 years and 25 years and older), participant gender (male or female, or mixed), country of residence (high or low- and middle income), the sample type (university or community or other non-clinical), and study quality (poor, fair, or good).
Table 4. Moderator analyses of heterogeneity
a Karatas et al. (Reference Karatas and Tagay2012) was excluded as we could not determine which age category it belonged to from the information provided.
b Reference group is under 25.
c Reference group is high income country.
d Reference group is university sample.
e Reference group is female samples.
f Reference group is poor quality studies. Other non-clinical: non-clinical refers to non-clinical in the sense of psychological disorders rather than physical disorders.
Regarding perfectionistic concerns, only one moderator was significant – gender (Table 4). Specifically, female-only samples had a stronger pooled effect (r=–.55 [–.64, –.45] p <.001, k=10) than mixed-gender samples (r=–.40 [–.45, –.35], p<.001, k=56). However, even after accounting for this moderator the level of heterogeneity was still high (Table 4). There were only two male-only samples that provided an effect size for this association; therefore, we do not provide the pooled effect for male-only samples.
Regarding perfectionistic strivings, again only one moderator was significant – gender. Specifically, inspecting subgroup analyses, mixed gender studies had a negligible positive pooled effect (pooled r=.09 [.04, .14], p<.001, k=48) but female-only had a non-significant pooled effect (pooled r=–.07 [–.23, .09], p=.388, k=8). However, even after accounting for this moderator the level of heterogeneity was still high (Table 4). There were only two male-only effect sizes included in this pooled association, therefore we do not provide a pooled effect for male-only samples.
Discussion
The aim of this meta-analysis was to examine the association between perfectionism dimensions and self-esteem in adults. Consistent with previous meta-analyses on the association between psychopathology and perfectionism in adults (Callaghan et al., Reference Callaghan, Greene, Shafran, Lunn and Egan2023; Limburg et al., Reference Limburg, Watson, Hagger and Egan2017; Stackpole et al., Reference Stackpole, Greene, Bills and Egan2023), perfectionistic concerns had a medium, negative pooled correlation with self-esteem, indicating higher perfectionistic concerns were related to lower self-esteem. Inconsistent with previous meta-analyses on psychopathology and perfectionism in adult samples (Callaghan et al., Reference Callaghan, Greene, Shafran, Lunn and Egan2023; Limburg et al., Reference Limburg, Watson, Hagger and Egan2017; Stackpole et al., Reference Stackpole, Greene, Bills and Egan2023), perfectionistic strivings had a negligible positive pooled correlation with self-esteem, indicating higher perfectionistic strivings are related to higher self-esteem. However, when considering these results, it should be noted that this positive correlation (r=.06) failed to reach Cohen’s (Reference Cohen1992) convention of a small correlation (r≥.10). Our moderators also explained little variance in the imputed effect sizes.
The finding that higher perfectionistic concerns is related to lower self-esteem fits with the wealth of meta-analyses demonstrating that perfectionistic concerns are associated with a range of negative psychological variables and psychopathology (Bills et al., Reference Bills, Greene, Stackpole and Egan2023; Callaghan et al., Reference Callaghan, Greene, Shafran, Lunn and Egan2023; Limburg et al., Reference Limburg, Watson, Hagger and Egan2017; Lunn et al., Reference Lunn, Greene, Callaghan and Egan2023; Smith et al., Reference Smith, Sherry, Ray, Hewitt and Flett2021). Clearly, higher perfectionistic concerns are related to many adverse psychological states, including lower self-esteem. The results also support the tenets of the cognitive-behavioural model of clinical perfectionism (Shafran et al., Reference Shafran, Cooper and Fairburn2002), which proposes that self-esteem is overly dependent on striving to meet demanding standards despite negative effects (e.g. psychopathology). Individuals with lower self-esteem may be more likely to base their self-evaluation on how well they perform when striving to achieve their goals, and a negative interpretation of failure to meet their high standards maintains low self-worth (Shafran et al., Reference Shafran, Cooper and Fairburn2002). Hence the findings of this meta-analysis provide some indirect support for the cognitive-behavioural model of clinical perfectionism (Shafran et al., Reference Shafran, Cooper and Fairburn2002).
We found that higher perfectionistic strivings was related to higher self-esteem, however this was a negligible association and in contrast to previous research on psychopathology and perfectionism in adults (Callaghan et al., Reference Callaghan, Greene, Shafran, Lunn and Egan2023; Limburg et al., Reference Limburg, Watson, Hagger and Egan2017; Stackpole et al., Reference Stackpole, Greene, Bills and Egan2023). It should be noted that this very small association (r=.06) was unlikely to be clinically meaningful. Furthermore, whether these results would be similar in clinical samples is unclear given the lack of clinical samples measuring this outcome in our meta-analysis. Unfortunately, there were insufficient clinical samples to conduct a robust analysis of the differences between clinical and non-clinical samples. It is possible that different relationships may be observed between perfectionistic strivings and self-esteem in clinical samples, in line with large reviews on the association between strivings and psychopathology in adults which had sufficient studies to analyse clinical samples (Callaghan et al., Reference Callaghan, Greene, Shafran, Lunn and Egan2023). Future meta-analyses should aim to classify samples into clinical and non-clinical for moderator analyses, to determine if clinical status makes a difference to the strength or direction of the relationship between perfectionistic strivings and self-esteem. It would also be helpful for more primary studies in clinical samples to be conducted measuring the association between perfectionism and self-esteem. Nonetheless, our findings support other meta-analyses documenting the association between perfectionistic strivings and some positive psychological states such as helpful academic outcomes (Osenk et al., Reference Osenk, Williamson and Wade2020).
Clinical implications
Given the meta-analysis involved cross-sectional data, no direct clinical implications can be drawn from this review. However, there are several directions for future research that may inform future clinical implications. Future research should continue to examine the cognitive-behavioural model of clinical perfectionism (Shafran et al., Reference Shafran, Cooper and Fairburn2002), where self-esteem is based on striving to meet high standards, which is the basis of cognitive behaviour therapy (CBT) for perfectionism, demonstrated to reduce symptoms of anxiety, depression, and eating disorders (Galloway et al., Reference Galloway, Watson, Greene, Shafran and Egan2022; Robinson and Wade, Reference Robinson and Wade2021). A core aim of CBT for perfectionism is to broaden an individual’s self-esteem, so it is less dependent on striving and achievement, whilst simultaneously reducing perfectionistic concerns (e.g. fear of making mistakes and performance-related self-doubt; Egan et al., Reference Egan, Wade, Shafran and Antony2014a; Shafran et al., Reference Shafran, Egan and Wade2018). CBT for perfectionism has been demonstrated to improve self-esteem. For example, Egan et al. (Reference Egan, van Noort, Chee, Kane, Hoiles, Shafran and Wade2014b) and Handley et al. (Reference Handley, Egan, Kane and Rees2015) found CBT for perfectionism resulted in significant increases in self-esteem, which were maintained at 6-month follow-up, concurrent with significant reductions psychopathology. These findings suggest preliminary evidence that interventions addressing perfectionism may have a transdiagnostic impact on reducing low self-esteem and psychopathology. However, further research is required to examine the efficacy of CBT for perfectionism in improving low self-esteem. It is also possible that interventions aimed at reducing low self-esteem such as Fennell’s (Reference Fennell2009) CBT for low self-esteem, may also improve perfectionism in addition to self-esteem. Future research could examine a head-to-head comparison of CBT for perfectionism and CBT for low self-esteem in terms of the relative efficacy on outcomes of perfectionism, self-esteem, and psychopathology to understand the most efficacious approach for clients presenting with low self-worth which is overly dependent on meeting their personally demanding standards.
Limitations
There were several limitations. A major limitation was that the data were cross-sectional. We cannot determine whether perfectionistic concerns and self-esteem are causally related. Future research should aim to conduct further experimental studies such as Hummel et al. (Reference Hummel, Cludius, Woud, Holdenrieder, Mende, Huber, Limburg and Takano2022), where an increase in perfectionistic concerns was found to result in an increase in psychological distress, with the assessment of self-esteem as the outcome from the manipulation of perfectionistic concerns. Another limitation was that grey literature was not included and we only included studies published in English, therefore it is possible that some literature was missed. In addition, while we included numerous databases for the search, inclusion of the Web of Science database may have been useful to locate further articles and would be helpful for future reviews to include as a search database. Furthermore, we did not conduct inter-rater reliability on risk of bias ratings.
Conclusion
The results of this meta-analysis indicate that in adults perfectionistic concerns were moderately related to lower self-esteem. Higher perfectionistic strivings were related to higher self-esteem, although this was a negligible, very small association. It should also be considered that perfectionistic strivings have been found to be associated with psychopathology such as eating disorders (e.g. Bills et al., Reference Bills, Greene, Stackpole and Egan2023; Stackpole et al., Reference Stackpole, Greene, Bills and Egan2023), even when the effect of perfectionistic concerns was controlled (e.g. Limburg et al., Reference Limburg, Watson, Hagger and Egan2017). Similar to recent meta-analyses (Callaghan et al., Reference Callaghan, Greene, Shafran, Lunn and Egan2023), we reiterate that further cross-sectional studies are unlikely to yield significantly new results. The field of perfectionism requires a focus now on experimental and longitudinal studies, and the examination of the causality between dimensions of perfectionism and negative psychological variables, including low self-esteem.
Supplementary material
The supplementary material for this article can be found at https://doi.org/10.1017/S1352465824000249
Data availability statement
The data are available upon request from the corresponding author.
Acknowledgements
None.
Author contributions
Viyona Khossousi: Conceptualization (supporting), Data curation (equal), Formal analysis (supporting), Investigation (lead), Methodology (equal), Project administration (equal), Writing – original draft (equal), Writing – review & editing (supporting); Danyelle Greene: Data curation (lead), Formal analysis (lead), Investigation (equal), Methodology (equal), Writing – original draft (equal), Writing – review & editing (equal); Roz Shafran: Conceptualization (equal), Writing – review & editing (equal); Thomas Callaghan: Conceptualization (supporting), Formal analysis (supporting), Investigation (supporting), Methodology (supporting), Supervision (supporting), Writing – review & editing (supporting); Sheree Dickinson: Investigation (supporting); Sarah Egan: Conceptualization (equal), Methodology (equal), Project administration (equal), Supervision (lead), Writing – original draft (equal), Writing – review & editing (lead).
Financial support
This research received no specific grant from any funding agency, commercial, or not-for-profit sectors.
Competing interests
Sarah Egan and Roz Shafran receive royalties for the books Cognitive-Behavioral Treatment of Perfectionism and Overcoming Perfectionism: A Self-Help Guide Using Cognitive Behavioural Techniques.
Open access
Comments
No Comments have been published for this article.