You are on page 1of 6

Personality and Individual Differences 172 (2021) 110589

Contents lists available at ScienceDirect

Personality and Individual Differences


journal homepage: www.elsevier.com/locate/paid

Perfectionism and academic burnout: Longitudinal extension of the bifactor


model of perfectionism
Hyunmo Seong a, Sangeun Lee a, Eunbi Chang b, *
a
Korea University, Republic of Korea
b
Kangwon National University, Republic of Korea

A R T I C L E I N F O A B S T R A C T

Keywords: Perfectionism is a personality disposition that is highly relevant in the education domain and related to various
Academic burnout educational and psychological outcomes in students. As students experience challenges and stress in school,
Bifactor model academic burnout is also a critical phenomenon. The current study investigated longitudinal effects of perfec­
Perfectionism
tionism on academic burnout, using a recently advanced bifactor model of perfectionism. With a sample of 336
Adolescents
Longitudinal analysis
secondary school students (7th–12th grades, 63% female) across three waves spanning a six-month period.
Results of the confirmatory analysis and longitudinal invariance tests showed that the bifactor-(S⋅I – 1) model
with the general perfectionism, defined by one item of concerns over mistakes as the reference item, and specific
perfectionism factor better explained the multidimensional structure of perfectionism, relative to the two-factor
model. Specific perfectionistic concerns remained unreliable after controlling for the general factor, while spe­
cific perfectionistic strivings remained as a more reliable factor. The cross-lagged analyses showed that the
general perfectionism predicted longitudinal increases in emotional exhaustion and cynicism. On the contrary,
specific perfectionistic strivings did not predict changes in students’ burnout.

When students fail to cope effectively with academic stress and adaptive outcomes (e.g., Madigan, 2019), but sometimes also have
pressure for academic achievement, they experience academic burnout positive association with maladaptive outcomes (Gotwals et al., 2012).
(Lee et al., 2013). Academic burnout consists of three dimensions: By contrast, perfectionistic concerns refer to discrepancies between
emotional exhaustion, cynical attitude towards learning (i.e., cynicism), actual performance and ideal standards (Slaney et al., 2001), concerns
and decreased efficacy (i.e., inefficacy) (Schaufeli et al., 2002). Aca­ over mistakes, and doubts about one’s performance (Frost et al., 1990).
demic burnout is closely associated with problems in students’ school Their positive associations with maladaptive outcomes are consistently
adaptation and psychological well-being: it is positively associated with reported (e.g., Hill & Curran, 2016).
students’ psychological maladjustment (Lee & Lee, 2018) and school Regarding the association between perfectionism and academic
dropout (Bask & Salmela-Aro, 2013), but negatively associated with burnout, many studies have been conducted and reported consistent
subjective well-being (Raiziene et al., 2014) and performance (May associations: academic burnout is positively associated with perfec­
et al., 2015). Thus, researchers have been interested in factors affecting tionistic concerns and negatively associated with perfectionistic striv­
academic burnout. ings. Hill and Curran’s (2016) meta-analysis of perfectionism and
Perfectionism is a multidimensional personality trait that includes burnout showed that perfectionistic strivings had a small negative or
tendencies of setting and striving for excessively high standards and non-significant association with symptoms of burnout; in contrast,
feeling discrepancy with those standards (Frost et al., 1990; Slaney et al., perfectionistic concerns had medium-to-large positive association with
2001). It has been studied as one of the factors affecting academic symptoms of burnout. However, the relationship of perfectionism
burnout (e.g., Chang et al., 2020; Kljajic et al., 2017). Stoeber and Otto (especially perfectionistic strivings) and burnout showed little differ­
(2006) classified it into two higher-order dimensions such as perfec­ ence when controlling for the overlapping variance between two di­
tionistic strivings and perfectionistic concerns. mensions of perfectionism. That is, there were small-to-medium
Perfectionistic strivings refer to setting high standards (Frost et al., negative association between pure perfectionistic strivings and all
1990; Slaney et al., 2001). They mostly have positive association with burnout symptoms, which are consistent with Stoeber and Otto’s (2006)

* Corresponding author at: Department of Education, College of Education, Kangwon National University, Chuncheon, Republic of Korea.
E-mail address: eunbichang@kangwon.ac.kr (E. Chang).

https://doi.org/10.1016/j.paid.2020.110589
Received 8 September 2020; Received in revised form 27 November 2020; Accepted 8 December 2020
Available online 17 December 2020
0191-8869/© 2020 Elsevier Ltd. All rights reserved.
H. Seong et al. Personality and Individual Differences 172 (2021) 110589

report that the adaptive side of perfectionistic strivings may be clear perfectionistic concerns were measured by concerns over mistakes and
when the correlation between perfectionistic strivings and concerns discrepancy.
were controlled.
Since a substantial construct overlap between the two factors has 1.2.2. Academic burnout
been reported, recent researchers questioned the two higher-order fac­ The Korean version of the Maslach Burnout Inventory-Student Sur­
tor model and suggested a bifactor model of perfectionism. Bifactor vey (MBI-SS; Shin et al., 2011) was used to measure academic burnout.
model is consisted of a general factor and specific factors (e.g., perfec­ The MBI-SS has three subscales: emotional exhaustion (5 items), cyni­
tionistic strivings factor); this model showed a better fit to the data than cism (4 items), and academic inefficacy (6 items). Items were rated
the two-factor model of perfectionism (e.g., Gäde et al., 2017; Howell ranging from 1 (strongly disagree) to 5 (strongly agree). The coefficient
et al., 2020; Smith & Saklofske, 2017). General factor reflects the shared alphas were 0.86, 0.82, 0.82 for emotional exhaustion, cynicism, and
variance of perfectionism subfactors. In contrast, specific factors reflect academic inefficacy, respectively (Shin et al., 2011).
unique variance that general factors do not account for.
So far, research examining the association between perfectionism 1.3. Data analyses
and academic burnout mainly adopted the two-factor model of perfec­
tionism, and does not reflect recent important advancement on the Data were analyzed using Mplus 7.3. First, measurement models of
bifactor model of perfectionism. In addition, as most of the studies (e.g., perfectionism were estimated. Hierarchical two-factor models were
Chang et al., 2020) were cross-sectional and used mediational approach, compared to bifactor models at each time to investigate the factor
it is difficult to infer the direct causal relationship between perfec­ structure of perfectionism. In the two-factor model, each item loaded on
tionism and academic burnout. Lastly, even though burnout is a multi­ their respective first-order factors (i.e., subscale levels), and the four
dimensional phenomenon (Schaufeli et al., 2002), many studies (e.g., first-order factors were expected to measure two correlated second-
Luo et al., 2016) did not examine the association between each order factors (i.e., perfectionistic concerns and perfectionistic striv­
dimension of academic burnout symptoms with perfectionism ings). The bifactor model had two specific factors and a general factor.
separately. However, since a traditional bifactor model is sensible only under the
Therefore, the present study tested the following research questions: assumption that the subscales are interchangeable, Eid et al. (2017)
(a) Will the bifactor model of perfectionism be applied to Korean sam­ suggested the bifactor-(S⋅I – 1) model as one of alternatives. We used the
ple? and (b) what is the longitudinal relationship between multidi­ bifactor-(S⋅I – 1) model in which all items were loaded on the general
mensional perfectionism and academic burnout? Since there is no study perfectionism factor, and each item except the reference item loaded on
adopting the bifactor model of perfectionism for Korean sample, we first one of two specific factors, with the general and specific factors being
examined whether the bifactor model of perfectionism can be applied to orthogonal. The reference item was determined as one that had the
Korean sample, and then examined the associations between perfec­ largest factor loading on concerns over mistake in the two-factor model
tionism and academic burnout by adopting a longitudinal design with in line with previous research using concerns over mistakes as a refer­
three waves at a three-month interval. ence domain (Gäde et al., 2017). Model fit was evaluated using Tucker-
Lewis Index (TLI; values ≥0.90 for acceptable fit), Comparative Fit Index
1. Methods (CFI; values ≥0.90 for acceptable fit), and Root Mean Square Error of
Approximation (RMSEA; values ≤0.08 for acceptable fit) (Hu & Bentler,
1.1. Participants and procedure 1999). For the bifactor model, coefficient omega (ω), omega subscale
(ωS), omega hierarchical (ωH), and omega hierarchical subscale (ωHS)
We recruited 450 7–11th graders from a public middle school and were also calculated to evaluate how much explained variance could be
two public high schools in Gangwon Province, South Korea. They attributed to the general and specific factors (Reise et al., 2013). Sup­
completed three questionnaires at intervals of three-month, spanning a ported models of perfectionism at each wave were combined to evaluate
total of six months. Only completed data with all time points were used longitudinal measurement invariance across time. Invariance was also
in the analysis. The final sample comprised 336 adolescents (213 fe­ tested with three-factor academic burnout models. The configural,
males, 28.6% middle and 71.4% high school students), with a mean-age metric, scalar, and strict invariance were tested using four nested models
of 15.83 (SD = 1.35, range = 13–18). We explained the purpose of the with cutoffs of 0.01 for ΔCFA and 0.015 for ΔRMSEA (Chen, 2007).
study to the students with the help of the teachers, and the students Next, a path model was specified to examine the longitudinal rela­
participated in the study voluntarily and signed the consent forms. They tionship between perfectionism and burnout. Factor scores estimated
received a small gift (about 3 US$) as compensation. from the final invariance models were used to reduce the ratio of pa­
rameters to participants as in previous studies on longitudinal bifactor
1.2. Measures models (e.g., McElroy et al., 2018). The factor score approach can
overcome a drawback of fully latent model that requires a large sample
1.2.1. Perfectionism size (Devlieger & Rosseel, 2017). Chi-square (χ2), TLI, CFI, and RMSEA
To measure perfectionism, subscales of two scales were used: Korean were used to evaluate the fit of alternative cross-lagged models. As there
version of Frost Multidimensional Perfectionism Scale (FMPS; Frost was missing data (≤1%) in our overall dataset, we used the maximum
et al., 1990; Lee & Park, 2011) and Almost Perfect Scale-Revised (APS-R; likelihood estimation with robust standard errors for all the analyses.
Park, 2009; Slaney et al., 2001). Two subscales from the Korean version
of FMPS were used: Concerns over mistakes (8 items) and personal 2. Results
standards (4 items). Doubts about actions subscale was excluded
because it only has two items resulting in questionable internal consis­ 2.1. Measurement models
tency (r = 0.61). All items of the FMPS were rated ranging from 1
(strongly disagree) to 5 (strongly agree). Internal consistencies were 0.82 Preliminary analysis revealed that the data was normally distributed
for concerns over mistakes and 0.73 for personal standards (Lee & Park, (see Supplementary Materials 1). As shown in Table 1, the two-factor
2011). Two subscales from the APS-R were also used: Discrepancy (10 models (Model 1) showed a poor fit. The refined two-factor models
items) and high standards (5 items). The items were scored ranging from with residual correlations (Model 2) still showed a poor fit. The bifactor-
1 (strongly disagree) to 7 (strongly agree). The coefficient alphas were 0.88 (S⋅I – 1) model (Model 3) showed better fit at each time point. Thus, the
for discrepancy and 0.79 for high standards (Park, 2009). Perfectionistic results implied that bifactor model with the item of concerns over mis­
strivings were assessed by personal standards and high standards; takes as the reference best represented the structure of perfectionism.

2
H. Seong et al. Personality and Individual Differences 172 (2021) 110589

Table 1 Table 2. ω values indicated that 96–97% of the total variances were
Model fit indices of perfectionism at each wave. explained by the general and specific factors across time. ωS values that
Models Model fit indices showed the amount of explained variance by the general and the specific
factor were about 0.96 (perfectionistic concerns) and 0.92 (perfection­
χ2 SF df CFI TLI RMSEA
istic strivings) at each time. The proportion of variance in the observed
Time1 scores that was attributed to the general factor was high (ωH range =
M1: Two-factor 863.416*** 1.23 321 0.867 0.854 0.071
M2: Two-factor 738.196*** 1.23 317 0.896 0.885 0.063
0.76–0.79), suggesting that the general factors were dominant sources of
CE permitted variance. ωHS for perfectionistic concerns ranged from 0.23 to 0.24,
M3: Bifactor CE 537.098*** 1.22 294 0.94 0.929 0.05 meaning that only a small amount of variance of perfectionistic concerns
permitted is reliable after controlling for the variance explained by the general
Time2
factor. In contrast, ωHS for perfectionistic strivings ranged from 0.47 to
M1: Two-factor 923.549*** 1.27 321 0.876 0.864 0.075
M2: Two-factor 831.861*** 1.26 317 0.894 0.882 0.07 0.56. Relatively higher values indicated that a significant amount of
CE permitted unique variance could be attributed to specific perfectionistic strivings
M3: Bifactor CE 722.95*** 1.24 294 0.912 0.894 0.066 factor (Reise et al., 2013).
permitted Standardized factor loadings are presented in Table 2. All items were
Time3
M1: Two-factor 1005.96*** 1.43 319 0.87 0.857 0.08
significantly loaded on the general factor across time. However, items
M2: Two-factor 886.683*** 1.42 315 0.892 0.88 0.073 measuring perfectionistic concerns had relatively higher factor loadings
CE permitted on the general factor compared to the items of perfectionistic strivings.
M3: Bifactor CE 806.597*** 1.38 294 0.903 0.884 0.072 Specifically, some items of concerns over mistake had non-significant
permitted
loadings on the specific perfectionistic concerns. Thus, omega co­
Note. CE = Correlated Errors, SF = Satorra-Bentler χ2 scaling correction factor; efficients and factor loadings justified the use of the general perfec­
CFI = Comparative Fit Index; TLI = Tucker-Lewis Index; RMSEA = Root Mean tionism factor and suggest that not perfectionistic concerns but rather
Square Error of Approximation. perfectionistic strivings was the reliable specific factor.
***
p < .001. Additionally, the measurement invariance of the three-factor struc­
ture of academic burnout at each time was also tested (see Supplement
Detailed information can be found in Supplemental Materials 2. Materials 3). The final longitudinal model of academic burnout with
The bifactor-(S⋅I – 1) models at each wave were combined to eval­ equal factor loadings and intercepts was used for cross-lagged path
uate longitudinal measurement invariance. A series of model compari­ analyses.
son revealed that strict longitudinal invariance model, constraining
factor loadings, item intercepts, and residuals to be equal across time, 2.2. Path models
was supported (detailed information can be found in Supplementary
Materials 3). To investigate the longitudinal relations between perfectionism and
The reliability of the final perfectionism bifactor model is shown in academic burnout, factor scores were estimated for the strict-invariance

Table 2
Standardized factor loadings of the final longitudinal perfectionism bifactor model and omega values.
Item Time1 Time2 Time3

General Perfectionistic Perfectionistic General Perfectionistic Perfectionistic General Perfectionistic Perfectionistic


concerns strivings concerns strivings concerns strivings

Perfectionistic CM6 0.79 0.81 0.82


concerns CM1 0.64 0.08 0.66 0.078 0.67 0.08
CM2 0.73 0.06n.s 0.75 0.06n.s 0.76 0.06n.s
CM3 0.77 − 0.06n.s 0.79 − 0.06n.s 0.80 − 0.06n.s
CM4 0.62 0.18 0.64 0.18 0.66 0.18
CM5 0.78 − 0.03n.s 0.79 − 0.03n.s 0.81 − 0.03n.s
CM7 0.60 0.10 0.62 0.10 0.64 0.10
CM8 0.67 0.02n.s 0.69 0.02n.s 0.70 0.02n.s
DIS1 0.59 0.37 0.61 0.37 0.63 0.37
DIS2 0.59 0.47 0.61 0.47 0.62 0.47
DIS3 0.50 0.59 0.52 0.59 0.53 0.59
DIS4 0.59 0.51 0.60 0.51 0.62 0.50
DIS5 0.53 0.61 0.55 0.61 0.56 0.60
DIS6 0.52 0.64 0.53 0.64 0.55 0.63
DIS7 0.58 0.50 0.59 0.51 0.61 0.50
DIS8 0.51 0.68 0.52 0.69 0.54 0.68
DIS9 0.55 0.68 0.56 0.68 0.58 0.67
DIS10 0.52 0.65 0.53 0.65 0.55 0.65
Perfectionistic PS1 0.66 0.30 0.69 0.28 0.71 0.26
strivings PS2 0.36 0.67 0.39 0.64 0.41 0.62
PS3 0.66 0.33 0.69 0.30 0.71 0.29
PS4 0.45 0.63 0.49 0.60 0.51 0.57
HS1 0.45 0.49 0.47 0.46 0.49 0.44
HS2 0.29 0.75 0.31 0.72 0.33 0.71
HS3 0.46 0.71 0.49 0.68 0.52 0.66
HS4 0.41 0.70 0.44 0.67 0.46 0.65
HS5 0.46 0.58 0.49 0.55 0.51 0.53
ω or ωS 0.963 0.951 0.922 0.966 0.956 0.92 0.968 0.958 0.92
ωH or ωHS 0.758 0.244 0.555 0.778 0.236 0.502 0.794 0.225 0.466

Note. n.sFactor loading was nonsignificant. Item CM6 was the reference item. CM, concerns over mistakes; DIS, discrepancy; PS, personal standards; HS, high standards;
ω, coefficient omega; ωS, coefficient omega subscale; ωH, coefficient omega hierarchical; ωHS, coefficient omega hierarchical subscale.

3
H. Seong et al. Personality and Individual Differences 172 (2021) 110589

bifactor-(S⋅I – 1) perfectionism model and the scalar-invariance burnout Confirmatory factor analysis showed an acceptable fit for the bifac­
model. With factor scores, we compared four alternative path models tor-(S⋅I – 1) model across three waves, with one general factor and two
(Kline, 2016). Detailed information regarding factor scores correlations specific factors (perfectionistic strivings and concerns). The results
and model comparisons can be found in Supplementary Materials 4. At a indicated that most of the shared variance was attributable to the gen­
baseline, first- and second-order stability coefficient were estimated, eral factor, and after accounting for this the specific perfectionistic
with within-time covariances set as equal over time. The perfectionism- concerns factor was found to be an unreliable construct. In contrast,
effect model, which adds cross-lagged paths from perfectionism factors specific perfectionistic striving factor remained as a reliable construct,
to burnout factors significantly improved the model fit. The burnout- explaining a substantial amount of variance that is independent of the
effect model, which adds paths from burnout to perfectionism factors general factor. Our results showed that the multidimensional structure
did not show a better model fit than the baseline. The bidirectional of perfectionism could be better explained by the bifactor model than
model, with cross-lagged path from both perfectionism and burnout to the two-factor higher-order model (e.g., Stoeber & Otto, 2006), in line
each other, did not significantly differ from the perfectionism model. with previous studies that favored the bifactor model of perfectionism
Therefore, the more parsimonious perfectionism-effects model was (e.g., Gäde et al., 2017; Howell et al., 2020; Smith & Saklofske, 2017).
chosen as the final model. However, without an explicitly defined conceptual meaning, the
Fig. 1 shows the significant longitudinal paths of the final model. general factor and residualized specific factors might not be easily
Specifically, the general perfectionism factor had significant positive interpretable (Eid et al., 2017). To increase interpretability, Gäde et al.
effects on emotional exhaustion and cynicism. Perfectionistic strivings (2017) implemented the bifactor-(S – 1) model, which sets one facet as a
factors had no significant effects on academic burnout. reference to the general factor and gives conceptual meaning to the
general factor. Our bifactor-(S⋅I – 1) model is based on a similar
3. Discussion approach, and more appropriate for our data as we used different items
from the two perfectionism scales (i.e., FMPS and APS-R), which are not
Although education is a domain where perfectionism and burnout unidimensional, to establish perfectionism dimensions (Eid et al., 2017).
are highly relevant, there has been less research investigating the rela­ Indeed, the factor loading structure of the bifactor model showed that
tionship between the two constructs than in other performance-related different subscales (especially two subscales constituting perfectionistic
domains (Hill & Curran, 2016). Hence, we aimed to examine the lon­ concerns) had somewhat different patterns of loadings on the specific
gitudinal relationship between perfectionism and academic burnout. To factors. This pattern has also emerged in other studies that implemented
incorporate the recent evidence supporting the bifactor structure of different measures to capture overarching perfectionistic strivings and
perfectionism, we evaluated a bifactor model of perfectionism using the concerns dimensions (e.g., Smith & Saklofske, 2017). Given that it is a
FMPS and APS-R. Specifically, we implemented the bifactor-(S⋅I – 1) common practice to measure multidimensional perfectionism by
model by taking one item as a reference item to avoid the problem of implementing more than one instrument (e.g., FMPS, APS-R, and Hewitt
collapsing factors and interpretability of the general factor (cf. Eid et al., and Flett’s (1991) multidimensional perfectionism scale) and combining
2017). In our model, a specific item representing concerns over mistakes several facets, the usage of the bifactor-(S⋅I – 1) model would yield better
was set as a reference item for defining the general factor. Therefore, the outcomes than other bifactor models.
meaning of the general factor can be interpreted based on the reference Further, results from a series of longitudinal measurement invariance
item. tests also suggested that the factor structure of the bifactor-(S⋅I – 1)

Fig. 1. Perfectionism-effects path model.


Note. Only significant longitudinal relations are shown. *p < .05, ***p < .001.

4
H. Seong et al. Personality and Individual Differences 172 (2021) 110589

model was equivalent across three waves. Without such measurement indicators. Future studies with larger sample size would be needed to
invariance, it could be argued that the bifactor structure of perfec­ reproduce our findings utilizing a fully latent model, with lowered
tionism is not stable over time, and hence is not a reliable model. Our standard errors of estimation. Second, the present findings may be
results showed that the bifactor-(S⋅I – 1) structure of perfectionism is limited to the specific time spans examined (three-month between each
stable over at least a 6-month period, which allowed us to explore lon­ time point). Students’ burnout symptoms may vary within time periods
gitudinal development or relations between the bifactor model of as they typically go through periodic experiences from the beginning of
perfectionism and other constructs such as burnout. a new semester to final exams, and from entrance to graduation.
Results of the cross-lagged path model showed that general perfec­ Therefore, whether the longitudinal effects found in the study are
tionism predicted longitudinal increases in emotional exhaustion and applicable to shorter or longer time spans should be examined in future
cynicism after controlling for stability effects (i.e., autoregressive paths), studies.
but not in academic inefficacy. The negative longitudinal effects of the Despite these limitations, the current study has shown that the
general perfectionism on academic burnout seem logical because in our bifactor-(S⋅I – 1) model would better explain the multidimensional
bifactor-(S⋅I – 1) model, the general perfectionism factor represents the structure of perfectionism. Similar to the findings of Gäde et al. (2017),
common variance between perfectionistic concerns and strivings, with we showed that overlaps between perfectionistic concerns and strivings
concerns over mistakes as a reference facet. Characteristics of the gen­ lie at the item level. Thus, grouping or combining of subscales into
eral perfectionism factor could be interpreted in a way one would have higher-order perfectionism dimensions may oversimplify the multidi­
expected for perfectionistic concerns. Our results resonate with the mensional structure of perfectionism. The bifactor-(S⋅I – 1) would be a
findings of Hill and Curran’s (2016) meta-analysis, which reported a useful alternative when different, not unidimensional subscales are used
significant positive relation of perfectionistic concerns with emotional to capture multidimensional perfectionism. Also, we found the delete­
exhaustion and cynicism, but not with academic inefficacy in the edu­ rious effects of the general perfectionism secondary school students.
cation domain. For students, feelings of efficacy might be much more Burnout has been shown to be related to various maladjustive educa­
affected by one’s academic achievement, than one’s perfectionistic trait tional and psychological outcomes. Teachers and mental health pro­
(Damian et al., 2017). Our results showed that general perfectionism fessions in school may benefit from our findings for designing
played the role of personal demands, exhibiting deleterious effects on prevention and intervention strategies targeting students’
students’ suffering and the subsequent development of exhaustion and perfectionism.
cynicism over time (Bakker & Demerouti, 2017).
Magnitudes of the effects of general perfectionism on academic CRediT authorship contribution statement
burnout symptoms may seem rather trivial (β = 0.06– 0.12). However,
in longitudinal autoregressive models, effect sizes of the cross-lagged Hyunmo Seong: Conceptualization, Methodology, Data curation,
paths are often dramatically smaller than those in cross-sectional Formal analysis, Writing – original draft. Sangeun Lee: Visualization,
studies as a function of stability effects (Adachi & Willoughby, 2015). Investigation, Formal analysis, Writing – original draft. Eunbi Chang:
Controlling for the strong stability in dependent variables removes a Resources, Supervision, Writing – original draft, Writing – review &
large portion of variance that was shared with earlier predictor vari­ editing.
ables. Indeed, bivariate correlations between general perfectionism
factor scores (Tn) and emotional exhaustion and cynicism (Tn+1) were Appendix A. Supplementary data
0.46– 0.47 and 0.22– 0.26 respectively (See Supplementary Materials 4).
These bivariate correlations were attenuated by the impact of the strong Supplementary data to this article can be found online at https://doi.
stability effects of burnout factors. Therefore, we argue that the longi­ org/10.1016/j.paid.2020.110589.
tudinal effect sizes of general perfectionism on burnout symptoms are
meaningful. References
By contrast, specific perfectionistic strivings did not predict changes
in burnout after controlling for stability effects. These findings are un­ Adachi, P., & Willoughby, T. (2015). Interpreting effect sizes when controlling for
stability effects in longitudinal autoregressive models: Implications for psychological
expected because it has been claimed that adaptiveness of perfectionistic science. European Journal of Developmental Psychology, 12(1), 116–128. https://doi.
strivings would be clearly demonstrated when shared variance is org/10.1080/17405629.2014.963549.
removed (e.g., Stoeber & Gaudreau, 2017). Our results might be due to Bakker, A. B., & Demerouti, E. (2017). Job demands-resources theory: Taking stock and
looking forward. Journal of Occupational Health Psychology, 22(3), 273–285. https://
rather weak reliabilities of specific perfectionistic factors. We consid­ doi.org/10.1037/ocp0000056.
ered specific perfectionistic strivings as a reliable factor, at least Bask, M., & Salmela-Aro, K. (2013). Burned out to drop out: Exploring the relationship
moderately, based on Reise et al.’s (2013) minimum guideline (i.e., ωHS between school burnout and school dropout. European Journal of Psychology of
Education, 28(2), 511–528. https://doi.org/10.1007/s10212-012-0126-5.
≥ 0.50). However, ωHS for specific perfectionistic strivings did not meet Chang, E., Seong, H., & Lee, S. M. (2020). Exploring the relationship between
with a preferred value (i.e., ωHS ≥ 0.75; Reise et al., 2013). Therefore, perfectionism and academic burnout through achievement goals: A mediation
factor scores for specific perfectionistic strivings might not be meticu­ model. Asia Pacific Education Review, 21, 409–422. https://doi.org/10.1007/s12564-
020-09633-1.
lously reliable. Smith and Saklofske (2017) also raised a question for the
Chen, F. F. (2007). Sensitivity of goodness of fit indexes to lack of measurement
reliability of specific perfectionistic strivings factor and called for invariance. Structural Equation Modeling: A Multidisciplinary Journal, 14(3), 464–504.
caution when interpreting the effects of perfectionistic strivings. Simi­ https://doi.org/10.1080/10705510701301834.
larly, our results suggest using higher-order perfectionism dimensions Damian, L. E., Stoeber, J., Negru-Subtirica, O., & Băban, A. (2017). On the development
of perfectionism: The longitudinal role of academic achievement and academic
without evaluating the bifactor model might be inappropriate. efficacy. Journal of Personality, 85(4), 565–577. https://doi.org/10.1111/
The current study has several limitations. First, we derived factor jopy.12261.
scores of perfectionism and burnout from the measurement models and Devlieger, I., & Rosseel, Y. (2017). Factor score path analysis. Methodology, 13
(Supplement), 31–38. https://doi.org/10.1027/a000001.
used them in the path model. Although a fully latent model would be a Eid, M., Geiser, C., Koch, T., & Heene, M. (2017). Anomalous results in G-factor models:
better choice to test our research question, given its complexity and our Explanations and alternatives. Psychological Methods, 22(3), 541–562. https://
samples size, which are common concerns for longitudinal bifactor psycnet.apa.org/doi/10.1037/met0000083.
Frost, R. O., Marten, P., Lahart, C., & Rosenblate, R. (1990). The dimensions of
models (McElroy et al., 2018), we utilized factor scores. Although using perfectionism. Cognitive Therapy and Research, 14, 449–468. https://doi.org/
factor scores is a common practice in applied research, factor scores may 10.1007/BF01172967.
also be biased (Skrondal & Laake, 2001). In relation to the model Gäde, J. C., Schermelleh-Engel, K., & Klein, A. G. (2017). Disentangling the common
variance of perfectionistic strivings and perfectionistic concerns: A bifactor model of
complexity, it was also challenging to achieve model fit that met with
traditional cut-off criteria because of our large model with items as

5
H. Seong et al. Personality and Individual Differences 172 (2021) 110589

perfectionism. Frontiers in Psychology, 8(160), 1–13. https://doi.org/10.3389/ Madigan, D. J. (2019). A meta-analysis of perfectionism and academic achievement.
fpsyg.2017.00160. Educational Psychology Review, 31, 967–989. https://doi.org/10.1007/s10648-019-
Gotwals, J. K., Stoeber, J., Dunn, J. G., & Stoll, O. (2012). Are perfectionistic strivings in 09484-2.
sport adaptive? A systematic review of confirmatory, contradictory, and mixed May, R. W., Bauer, K. N., & Fincham, F. D. (2015). School burnout: Diminished academic
evidence. Canadian Psychology, 53(4), 263–279. https://psycnet.apa.org/d and cognitive performance. Learning and Individual Differences, 42, 126–131. https://
oi/10.1037/a0030288. doi.org/10.1016/j.lindif.2015.07.015.
Hewitt, P. L., & Flett, G. L. (1991). Perfectionism in the self and social contexts: McElroy, E., Belsky, J., Carragher, N., Fearon, P., & Patalay, P. (2018). Developmental
Conceptualization, assessment, and association with psychopathology. Journal of stability of general and specific factors of psychopathology from early childhood to
Personality and Social Psychology, 60(3), 456–470. https://doi.org/10.1037/0022- adolescence: Dynamic mutualism or p-differentiation? Journal of Child Psychology
3514.60.3.456. and Psychiatry, 59(6), 667–675. https://doi.org/10.1111/jcpp.12849.
Hill, A. P., & Curran, T. (2016). Multidimensional perfectionism and burnout: A meta- Park, H. J. (2009). Validation of the almost perfect scale-revised. The Korean Journal of
analysis. Personality and Social Psychology Review, 20(3), 269–288. https://doi.org/ Counseling and Psychotherapy, 21(1), 131–149.
10.1177/1088868315596286. Raiziene, S., Pilkauskaite-Valickiene, R., & Zukauskiene, R. (2014). School burnout and
Howell, J., Anderson, R., Egan, S., & McEvoy, P. (2020). One factor? Two factor? Bi- subjective well-being: Evidence from cross lagged relations in a 1-year longitudinal
factor? A psychometric evaluation of the Frost Multidimensional Scale and the sample. Procedia-Social and Behavioral Sciences, 116(21), 3254–3258. https://doi.
Clinical Perfectionism Questionnaire. Cognitive Behaviour Therapy, 49(6), 518–530. org/10.1016/j.sbspro.2014.01.743.
https://doi.org/10.1080/16506073.2020.1790645. Reise, S. P., Bonifay, W. E., & Haviland, M. G. (2013). Scoring and modeling
Hu, L. T., & Bentler, P. M. (1999). Cutoff criteria for fit indexes in covariance structure psychological measures in the presence of multidimensionality. Journal of Personality
analysis: Conventional criteria versus new alternatives. Structural Equation Modeling: Assessment, 95(2), 129–140. https://doi.org/10.1080/00223891.2012.725437.
A Multidisciplinary Journal, 6(1), 1–55. https://doi.org/10.1080/ Schaufeli, W. B., Martinez, I. M., Pinto, A., Salanova, M., & Bakker, A. B. (2002). Burnout
10705519909540118. and engagement in university students: A cross-national study. Journal of Cross-
Kline, R. B. (2016). Principles and practice of structural equation modeling (4th ed.). New Cultural Psychology, 33, 464–481. https://doi.org/10.1177/0022022102033005003.
York: Guilford. Shin, H., Puig, A., Lee, J., Lee, J. H., & Lee, S. M. (2011). Cultural validation of the
Kljajic, K., Gaudreau, P., & Franche, V. (2017). An investigation of the 2 × 2 model of Maslach Burnout Inventory for Korean students. Asia Pacific Education Review, 12(4),
perfectionism with burnout, engagement, self-regulation, and academic 633–639. https://doi.org/10.1007/s12564-011-9164-y.
achievement. Learning and Individual Differences, 57, 103–113. https://doi.org/ Skrondal, A., & Laake, P. (2001). Regression among factor scores. Psychometrika, 66(4),
10.1016/j.lindif.2017.06.004. 563–575. https://psycnet.apa.org/doi/10.1007/BF02296196.
Lee, D. G., & Park, H. J. (2011). Cross-cultural validity of the frost multidimensional Slaney, R. B., Rice, K. G., Mobley, M., Trippi, J., & Ashby, J. S. (2001). The revised almost
perfectionism scale in Korea. The Counseling Psychologist, 39(2), 320–345 (https:// perfect scale. Measurement and Evaluation in Counseling and Development, 34(3),
doi.org/10.1177%2F0011000010365910). 130–145. https://doi.org/10.1080/07481756.2002.12069030.
Lee, J., Puig, A., Lea, E., & Lee, S. M. (2013). Age-related differences in academic burnout Smith, M. M., & Saklofske, D. H. (2017). The structure of multidimensional
of Korean adolescents. Psychology in the Schools, 50(10), 1015–1031. https://doi.org/ perfectionism: Support for a bifactor model with a dominant general factor. Journal
10.1002/pits.21723. of Personality Assessment, 99(3), 297–303. https://doi.org/10.1080/
Lee, M. Y., & Lee, S. M. (2018). The effects of psychological maladjustments on 00223891.2016.1208209.
predicting developmental trajectories of academic burnout. School Psychology Stoeber, J., & Gaudreau, P. (2017). The advantages of partialling perfectionistic strivings
International, 39(3), 217–233. https://doi.org/10.1177/0143034318766206. and perfectionistic concerns: Critical issues and recommendations. Personality and
Luo, Y., Wang, Z., Zhang, H., Chen, A., & Quan, S. (2016). The effect of perfectionism on Individual Differences, 104, 379–386. https://doi.org/10.1016/j.paid.2016.08.039.
school burnout among adolescence: The mediator of self-esteem and coping style. Stoeber, J., & Otto, K. (2006). Positive conceptions of perfectionism: Approaches,
Personality and Individual Differences, 88, 202–208. https://doi.org/10.1016/j. evidence, challenges. Personality and Social Psychology Review, 10(4), 295–319.
paid.2015.08.056. https://doi.org/10.1207/s15327957pspr1004_2.

You might also like