- Email: [email protected]
Examining the relationship of perfectionism, depression, and optimism: Testing for mediation and moderation
Personality and Individual Differences 54 (2013) 426–431
Contents lists available at SciVerse ScienceDirect
Personality and Individual Differences journal homepage: www.elsevier.com/locate/paid
Examining the relationship of perfectionism, depression, and optimism: Testing for mediation and moderation Jessica Black ⇑, William M. Reynolds Humboldt State University, 1 Harpst Street, Arcata, CA 95521, United States
a r t i c l e
i n f o
Article history: Received 6 August 2012 Received in revised form 10 October 2012 Accepted 18 October 2012 Available online 13 November 2012 Keywords: Perfectionism Optimism Depression Mediation
a b s t r a c t The present study investigated the relationship between perfectionism and depression, and the mediation/moderation effects of optimism. Participants were126 adults (78% women, mean age = 27 years) who completed an online survey that included the Multidimensional Perfectionism Scale (Frost, Marten, Lahart, & Rosenblate, 1990), the Revised Life Orientation Test (Scheier, Carver, & Bridges, 1994), and the Hamilton Depression Inventory (Reynolds & Kobak, 1995). Maladaptive, adaptive and total perfectionism were examined in separate analyses. In most analyses, the data fit a mediation model. Optimism limited the effect (relationship) of total and maladaptive perfectionism on depression. Adaptive perfectionism was related to depression only through optimism. There was a small gender difference, with indirect effects greater for men. These results are consistent with past research and support the notion that correlations between perfectionism, as a trait, and affective variables must be examined using multivariate models that allow for the determination of complex relationships. Ó 2012 Elsevier Ltd. All rights reserved.
1. Introduction Perfectionism has been generally understood to be associated with negative psychological outcomes (Frost et al., 1990; Hewitt, Flett, Turnbull-Donovan, & Mikail, 1991), although some degree of pursuit of excellence is a necessary component of achievement. Beginning with Hamachek (1978), researchers have sought to identify correlates of perfectionism (Slade & Owens, 1998). Hamachek distinguished between normal and neurotic perfectionism in terms of satisfaction and control: the normal perfectionist derives pleasure from striving for perfection yet knows when to relax standards, while the neurotic perfectionist is driven by fear of failure. Slade and Owens (1998) defined what they called a ‘‘dual-process model of positive and negative perfectionism,’’ according to which positive perfectionism and negative perfectionism can be explained in terms of striving and trying to avoid undesirable results. In a meta-analysis of 15 studies that had used a dualist approach, Stoeber and Otto (2006) concluded that striving was a positive form of perfectionism. They contrasted it with perfectionist concerns, such as an inability to live up to perceived expectations. Research by Stoeber, Uphill, and Hotham (2009) partially clarified the connection between striving and outcome: triathlon participants who scored high on perfectionism and reported high
⇑ Corresponding author. Address: 329 Bayside Rd., Arcata, CA 95521, United States. Tel.: +1 707 633 5362. E-mail address: [email protected] (J. Black). 0191-8869/$ - see front matter Ó 2012 Elsevier Ltd. All rights reserved. http://dx.doi.org/10.1016/j.paid.2012.10.012
achievement goals performed better in their races than those with low achievement goals. Stoeber et al. found that differences between performance approach and avoidance mediated the perfectionism–performance relationship. Researchers have found that correlates of perfectionism are often mediated by other variables. Harris, Pepper, and Maack (2008) used the Multidimensional Perfectionism Scale by Frost and colleagues (MPS; 1990) to examine rumination as a mediator between maladaptive perfectionism and depression. The MPS was designed with six factors; ‘‘personal standards’’ has been used to assess positive perfectionism, while ‘‘concern over mistakes’’ and ‘‘doubts about actions’’ together can reflect negative perfectionism (cf. Harris et al., 2008; Park, Heppner, & Lee, 2010; Sturman, Flett, Hewitt, & Rudolph, 2009). Harris et al. found that rumination mediated the relationship between maladaptive perfectionism and depression, using the MPS subscales of concern over mistakes and doubts about actions. Park et al. (2010) used the same MPS subscales as a measure of maladaptive perfectionism and found that maladaptive coping mediated the relationship between maladaptive perfectionism and distress. This was especially true for men, for whom coping increased more strongly with greater perfectionism than it did for women. Ashby, Dickinson, Gnilka, and Noble (2011) reported that hope mediated the relationship between maladaptive perfectionism and depression in middle school students. Adaptive perfectionism was not significantly correlated to depression, and although it was not considered a mediator, it did lend itself to a statistically significant indirect effects model, as determined by bootstrapping techniques.
J. Black, W.M. Reynolds / Personality and Individual Differences 54 (2013) 426–431
Ashby et al. also tested a moderation model, but found that it did not fit their data on perfectionism, hope, and depression. Hope is closely related to optimism, which is often predictive of positive psychological outcomes (Peterson, 2000). Optimism appears related to perfectionism in that the belief that high goals can be attained is part of seeking their attainment. Peterson proposed two dimensions of optimism, big and little. He associated big optimism with what Scheier and Carver (1992) call dispositional optimism, which their scale, the Revised Life Orientation Test (R-LOT; Scheier et al., 1994), measures. Chang, Chang, and Sanna (2009) outline the complexities of optimism, suggesting that it may have realistic (adaptive) and unrealistic (maladaptive) dimensions. Both optimism and perfectionism have been connected with striving, hope, and depression. Fry and Debats (2009), in a study of predictors of mortality in older adults found perfectionism associated with increased risk of death and dispositional optimism with decreased mortality. Other research (e.g. Bergman, Nyland, & Burns, 2007) implies that adaptive optimism should facilitate striving, encouraging perfectionists to focus on the possibility of achieving excellence rather than on the need to avoid censure for failing to do so. Chang (2009) reported that positive performance perfectionism meant more optimism (r(244) = .22), while negative performance perfectionism meant less optimism (r(244) = .26). We did not, however, find any mediational analysis of the relationship between perfectionism and optimism in the literature. Baron and Kenny (1986) define moderation and mediation models used to test ways in which the relationship between two variables can be affected by third variables. Moderation would indicate an interaction between the independent variable, in this case perfectionism, and the moderator, optimism. Mediation occurs when the relationship between the independent variable and the dependent variable is different when a third variable is included in the model. According to the Baron and Kenny criteria, the direct relationship between two variables, significant when considered alone, approaches zero when measured in the presence of the mediator. An indirect effect can occur even when there is no direct effect (Preacher & Hayes, 2004). In this case, the relationship between independent and dependent variables depends entirely upon the mediator. The measurement of mediation and indirect effects allows some inference of causation, as the independent variable is assumed to cause the mediator, which in turn affects the dependent variable. A moderator, on the other hand, exists independently of the focal predictor, which varies at different levels of the moderator. The current research investigated the degree to which optimism acts as a mediator and/or moderator of the relationship between perfectionism and depressive symptomatology. Because of the possible dual nature of optimism (Chang et al., 2009), we predicted a moderation model. Flett, Blankstein, Hewitt, and Koledin (1992) found gender differences in the correlations between perfectionism and procrastination. In view of Park et al.’s (2010) similar results in a mediation model, we examined potential gender differences in our sample by testing separate models for women and men. We used maladaptive and adaptive perfectionism domains as well as overall perfectionism. Our goal was to better understand the potential role of optimism in the relationship between perfectionism and depression.
2. Method 2.1. Participants Participants were 126 adults (98 women, 28 men) with a mean age of 27.7 (SD = 13.9). Of these participants, 57 were students who
427
completed an online survey through the Psychology Department Research Pool, and 69 were adults who completed the same questionnaire using Survey Monkey. 2.2. Instrumentation The Multidimensional Perfectionism Scale (Frost et al., 1990) is a 35-item questionnaire intended to reflect six aspects underlying the construct. It is considered a reliable measure of general perfectionism (Chang, 2000). In the current investigation the 6-item organization subscale was omitted, resulting in a 29-item scale with an internal consistency reliability of ra = .88. Consistent with previous research (Kawamura, Hunt, Frost, & DiBartolo, 2001; Sturman et al., 2009) we used the concern over mistakes and doubts about actions subscales as a measure of maladaptive perfectionism (ra = .84) and the Personal Standards subscale to represent adaptive perfectionism (ra = .76). The Revised Life Orientation Test (Scheier et al., 1994) was used to assess optimism; its test–retest reliability has proven moderately high (.79 at 28 months), and it has shown acceptable internal consistency. Of the ten statements evaluated on a zero-to-four point scale, three are reverse-coded and four are fillers (ra = .77 for this study). The Hamilton Depression Inventory (HDI; Reynolds & Kobak, 1995) was developed to provide a paper–pencil analog measure to the Hamilton Depression Rating Scale (Hamilton, 1960, 1967) clinical interview. Reynolds and Kobak provided strong evidence for the reliability and validity of three versions of the HDI based on clinical and nonclinical samples. The HDI Short Form, which consists of 15 items, was used in the current study. Reynolds and Kobak reported an internal consistency of .93 for the HDI Short Form in a mixed clinical and nonclinical sample of 357 adults. In the current study ra = .91. 2.3. Procedure An anonymous survey was conducted online through the HSU Psychology Department Research Pool and through Survey Monkey for a total of 2 months. Links to the survey were sent in emails and posted on social networking sites. Participants’ identities were protected by disabling IP address and email tracking features on Survey Monkey. No reward was offered other than class credit available to students responding through the psychology department participation pool; participants were allowed to withdraw at any time without losing credit. 2.4. Data analysis The criteria outlined by Baron and Kenny (1986) provide a conceptual framework to facilitate examination of mediation, but does not include a test of statistical significance for the indirect effect of the mediator (a b; see Fig. 1). Baron and Kenny suggested the Sobel (1982) test as appropriate: it tests the null hypothesis that the indirect effect is zero. However, the Sobel test assumes the distribution of ab is normal, which is seldom the case (Hayes, 2009). Bootstrapping, or the nonparametric resampling of the data set in order to make repeated estimates, makes no such assumptions (Preacher & Hayes, 2008). We present regression statistics as well as results of the Sobel test. We also used bootstrapping (10,000 resamples) to provide bias corrected and accelerated confidence intervals as a test of statistical significance. Following the recommendations of Preacher and Kelley (2011), we present kappa squared, j2, as an effect size that is both standardized and insensitive to sample size. j2 is the ratio of the indirect effect to the maximum indirect possible given the constraints of the design and data. The problem of determining when an effect
428
J. Black, W.M. Reynolds / Personality and Individual Differences 54 (2013) 426–431
Fig. 1. The unmediated path between independent and dependent variables (c) is presented above. Below, optimism is added, resulting in a simple mediation diagram. The indirect effect is c–c0 , or a b.In both cases, as perfectionism increases, so does depression, but when optimism works as a mediator, the relationship is less.
positively correlated with depression, r(124) = .36, 95% C.I.[.21, .50], thus not supporting the test of a moderation model. Depression was negatively correlated with optimism, r(124) = .56, 95% C.I.[.68, .40]. Age demonstrated a very low, nonsignificant correlation with perfectionism, r(124) = .01, 95% C.I.[.17, .15], and optimism scores, r(126) = .09, 95% C.I.[.06, .24], and a low negative correlation with depression, r(124) = .15, 95% C.I.[.30, .01]. Table 1 provides correlations between all scales. Tests of independent means for gender showed statistically nonsignificant differences on total perfectionism, maladaptive perfectionism, adaptive perfectionism, and depression. There was a gender difference in optimism with women (M = 15.8, SD = 3.7) more optimistic than men (M = 14.1, SD = 4.2), t(124) = 2.15, p = .03, d = 0.45. There were no significant differences in means between data collection samples (HSU students versus Survey Monkey; see Table 2). 3.1. Mediation
Table 1 Pearson correlation coefficients between total perfectionism and adaptive and maladaptive dimensions, optimism, depression, and age (N = 126).
1. Total perfectionism 2. Maladaptive perfectionism 3. Adaptive perfectionism 4. Optimism 5. Depression 6. Age ** ***
1
2
– .861***
–
.590*** .302** .364*** .011
.327*** .454*** .521*** .130
3
– .154 .037 .002
4
– .556*** .086
5
Age was initially entered as a covariate in the mediation model. Because it did not have a significant effect, it was thereafter excluded in order to facilitate interpretation and calculation of effect sizes. 3.1.1. Total perfectionism Optimism mediated the effect of perfectionism on depression, with an estimated indirect effect (ab) = 0.014, j2 = .174 (see Table 3 for confidence intervals). For women there was a smaller effect, ab = 0.008, j2 = .112. For men the effect was larger, ab = 0.031, j2 = .358, and the data fit a full mediation model (Baron & Kenny, 1986). Data from the two subsamples, Survey Monkey and the university research pool (HSU), were also tested as two separate models; in both cases significant results ensued (respectively, ab = 0.010, j2 = .106, and ab = 0.015, j2 = .219).
– .154
p < .01. p < .001.
can be considered large is addressed by Preacher and Kelley with reference to Cohen’s (1988) estimates for squared correlation coefficients: .01, .09, and .25 can be considered small, medium, and large respectively. In addition to j2, we also provide the percent of the total effect that is mediated, PM (Table 3 provides complete mediation results). Effect sizes and confidence intervals were generated using MBESS, an R package (Kelley & Lai, 2010). 3. Results Preliminary analyses showed that scores on the MPS and LOTR were normally distributed and scores on the HDI were not: A square root transformation of the HDI data fully corrected skew and kurtosis problems. Total perfectionism scores were negatively correlated with optimism, r(124) = .30, 95% C.I.[.46, .13], and
3.1.2. Maladaptive perfectionism Optimism mediated the relationship between maladaptive perfectionism and depression, ab = 0.029, j2 = .203. As before, for men, optimism fully mediated maladaptive perfectionism, ab = 0.076, j2 = .431. For women the effect was smaller, ab = 0.018, j2 = .141. 3.1.3. Adaptive perfectionism Although the adaptive perfectionism subscale was not significantly correlated with depression, it was positively correlated to optimism, and the possibility of an indirect effect remained (Hayes, 2009). Results of mediation analyses confirmed an indirect effect, ab = .027, j2 = .118. As adaptive perfectionism rose, so did optimism, which related to a decrease in depression (see Fig. 2). When tested separately by gender there was again a difference: In this
Table 2 Means (standard deviations) for total perfectionism (MPS), maladaptive perfectionism (MP), adaptive perfectionism (AP), optimism (LOTR), and depression (HDI).
Total Women Men Cohen’s d – men vs. women t p HSU students Survey Monkey Cohen’s d – HSU vs. Survey Monkey t p *
Untransformed data used for HDI.
MPS
MP
AP
LOTR
HDI*
78.6 (14.4) 78.3 (14.1) 79.8 (15.5) 0.10 0.48 .63 79.0 (16.2) 78.3 (12.7) 0.05 0.26 .70
32.0(7.9) 31.8(7.7) 32.6(8.8) 0.10 0.49 .62 32.5(8.9) 32.0(7.0) 0.06 0.62 .54
23.7(4.5) 23.5(4.3) 24.5(5.2) 0.22 0.97 .33 23.6(4.8) 23.8(4.3) 0.04 0.18 .86
15.4 (3.8) 15.8 (3.7) 14.1 (4.2) 0.45 2.15 .03 15.3 (4.3) 15.6 (3.4) 0.08 0.39 .70
9.8(7.9) 9.6 (6.9) 10.4 (10.9) 0.10 0.47 .64 11.1 (8.1) 8.6 (7.7) 0.32 1.78 .08
429
J. Black, W.M. Reynolds / Personality and Individual Differences 54 (2013) 426–431 Table 3 Mediation Results. Variable
ab
95% C.I. LL
a
b
c
c’
Sobel test z
95% C.I.
LL
UL
LL
UL
0.014 0.008 0.031
0.006 0.002 0.011
0.024 0.018 0.058
0.09* 0.07* 0.12*
0.15* 0.12* 0.27*
0.03* 0.03* 0.03
0.02* 0.02* <0.01
3.07 2.23 2.15
.002 .026 .032
0.44 0.27 1.03
0.19 0.06 0.39
0.81 0.64 4.94
.174 .112 .358
.075 .025 .131
.281 .231 .563
Maladaptive perfectionism Women Men
0.029 0.018 0.076
0.013 0.005 0.039
0.049 0.037 0.140
0.24* 0.19* 0.30*
0.12* 0.10* 0.26*
0.08* 0.08* 0.08*
0.05* 0.06* 0.01
3.77 2.74 2.91
<.001 .006 .004
0.37 0.24 0.90
0.17 0.07 0.45
0.63 0.50 1.98
.203 .141 .431
.104 .048 .231
.310 .260 .625
0.027 0.033 0.008
0.057 0.067 0.087
0.001 0.011 0.087
0.16* 0.22* 0.03
0.17* 0.15* 0.27*
1.68 2.33 0.184
.094 .020 .854
1.94 4.67 0.51
0.63 1.77 35.6
2450 3851 5.40
.118 .144 .045
.015 .054 .000
.224 .262 .135
0.01 0.01 0.02
0.01 0.03 0.02
p
j2
95% C.I.
Total perfectionism Women Men
Adaptive perfectionism Women Men
UL
PM
ab = estimated indirect effect. All confidence intervals generated with bias corrected and accelerated bootstrapping(N = 10,000). p < .05.
*
Fig. 2. The indirect effect of adaptive perfectionism on depression is through optimism.
case, the indirect effect was significant for women, ab = .033, j2 = .144, but not for men.
4. Discussion Perfectionists set high standards for themselves, and the failure to meet these standards often relates to poor psychological outcomes. For persons who manifest strong perfectionistic tendencies, the perception or anticipation of failure could lead to feelings of depression. Perfectionists who have a belief in their ability to achieve their goals should be less likely to experience the same negative consequences. In this manner, optimism should reduce the negative affective consequences of perfectionism. Initially, we expected a moderation model, with persons scoring high on perfectionism and low on optimism reporting greater depression, but persons high on optimism showing less negative outcomes. The current research found a more complex mediation relationship to fit the data. In all the mediation models we tested, with the exception of adaptive perfectionism in men, there was a significant indirect effect. Kappa squared values indicate medium to large effect sizes. For our sample, effects were larger for maladaptive perfectionism than for total perfectionism. They were also larger for men than for women in total perfectionism and maladaptive perfectionism. Although the confidence intervals for men and women overlapped for total perfectionism, they did not for maladaptive perfectionism, suggesting a genuine gender difference. This result is similar to the findings of Park et al. (2010), who reported that maladaptive coping was a greater mediator of the relationship between perfectionism and psychological distress for men. It is interesting that for men in our sample there was complete mediation: the path between the total perfectionism and depression was no longer signif-
icant in the presence of optimism. For women there was partial mediation. It is likely that the mediating effect of optimism was strongest for maladaptive perfectionism because its correlation with optimism was stronger than found for total perfectionism. It needs to be recognized that the total perfectionism score included the maladaptive scale items. The research we reviewed did not compare maladaptive perfectionism to total perfectionism in mediation models, but it did compare it to adaptive perfectionism. Flett, Besser, Davis, and Hewitt (2003) found that self-acceptance mediated the outcome of socially-prescribed (negative) perfectionism but not self-prescribed (positive) perfectionism. They also reported an indirect effect of self-acceptance on other-oriented (negative) perfectionism. Harris et al. (2008) found similar differences between adaptive and maladaptive perfectionism in a mediation context: rumination meant more depression for maladaptive perfectionism, but not for adaptive perfectionism. Our results indicate a different mediation model for adaptive perfectionism. Adaptive perfectionism did not show a statistically significant relationship with depression except through optimism, in a suppression model. Adaptive perfectionism predicted increasing optimism, which in turn related to lower depression (see Fig. 2). Although optimism showed a stronger mediation effect for men than women in total and maladaptive perfectionism, the indirect effect of adaptive perfectionism through optimism was only found for women; for men the path between adaptive perfectionism and optimism was small and statistically insignificant. Because of the limited number of men, wider confidence intervals make it difficult to make inferences. However, the point estimates indicate interesting gender differences that merit future research. Hoyle and Kenny (1999) found little effect of sample size on the path estimates in a Monte Carlo simulation testing samples of N = 25–200. Both kappa squared and PM (percent of total effect mediated) are insensitive to sample size (Preacher & Kelley, 2011). Interestingly, effect sizes for men are much larger for both total and maladaptive perfectionism, and the correlation between optimism and depression is much stronger for men than for women. Nevertheless, the results are necessarily imprecise due to the relatively small number of men in our sample. 4.1. Limitations There is always a question of whether data collected from college students can be generalized to non-academic populations. This study gives some support for generalizability, although the Survey Monkey respondents were not entirely representative, as 78% had at least some college education compared to 55% of adults in the US population (US Census Bureau, 2010). There were no
430
J. Black, W.M. Reynolds / Personality and Individual Differences 54 (2013) 426–431
significant differences in measure means between the data collected through the university research participation pool and that collected through Survey Monkey. Mediation analyses results conducted separately for these two groups were generally comparable for both total and maladaptive perfectionism. The university sample showed larger effect sizes and fit a full mediation model. However, the effect size confidence intervals overlapped so we cannot be positive of a difference in the respective samples. There was no mediation effect for either group when the model was tested with adaptive perfectionism. Overall, there were no appreciable differences between the two data collection subsamples used in our study. There are a number of limitations specific to the sample used in this study. Although the inclusion of data collected online through Survey Monkey compensates for some threats to external validity, the sample was still limited. The sample was largely college educated, and predominantly female. In view of the gender differences found in the present data, further research with a larger sample that is more evenly balanced between men and women is needed. Further limitations are imposed by the self-report method, and by the reliance on the Internet, which restricts participation to those with a connection and the inclination to complete a survey. The comparability of online versus pencil and paper surveys also arises. However, research suggests that online surveys have minimal disadvantages when compared to paper and pencil versions of the same measures (Gosling, Vazire, Srivastava, & John, 2004; Hardre, Crowson, & Xie, 2010). As with all tests of mediation models based on self-report measures given at the same time, inferences about causation are limited. We have tried to control for reverse effects to some extent by using a ‘‘state’’ measure of depression, while both the MPS and LOT-R are believed to measure general ‘‘trait’’ person characteristics. When analyses were conducted with depression as the independent variable and perfectionism as the dependent variable, there was no evidence of mediation. Nevertheless, causation cannot be ascribed. In the present study a moderate relationship (r2 = .31) was found between optimism and depression. When the two were interchanged in the mediation analyses (depression as mediator, optimism as dependent variable), results indicated significant indirect effects despite the fact that optimism is typically considered a trait construct and depression a state condition. Although frequently used as a measure of optimism, the LOT-R is limited in its level of reliability, thus introducing measurement error, which is problematic when conducting mediation analyses. Using optimism as predictor and perfectionism as mediator also produces a mediation model with similar effect sizes. Conceptually, it is plausible that optimism causes positive perfectionism, although it is less intuitive to in the case of negative perfectionism. Further investigation is needed to clarify a coherent causal relationship.
4.2. Recommendations for further research It is advisable to include more constructs in the model, and to use more than one measure of each (Iacobucci, Saldanha, & Deng, 2007). We recommend including a measure of striving and/or maximization, which has been associated with both optimism and perfectionism (Fry & Debats, 2009; Hill, Huelsman, & Araujo, 2010; Stoeber et al., 2009). Stress has been shown to mediate correlates of perfectionism (Chang, 2000, 2006) and could be included in a more complete model. It is clear that perfectionism is a complex construct. Our results indicate that mediational effects of optimism may differ by dimension of perfectionism and by gender. More research is necessary to confirm and explain this finding. Future research on optimism as a
mediator of perfectionism may also benefit from a more reliable assessment of optimism.
References Ashby, J. S., Dickinson, W. L., Gnilka, P. B., & Noble, C. L. (2011). Hope as a mediator and moderator of multidimensional perfectionism and depression in middle school students. Counseling & Development, 89, 131–139. Baron, R. M., & Kenny, D. A. (1986). The moderator–mediator variable distinction in social psychological research: Conceptual, strategic, and statistical considerations. Personality and Social Psychology, 51, 1173–1182. Bergman, A. J., Nyland, J. E., & Burns, L. R. (2007). Correlates with perfectionism and the utility of a dual process model. Personality and Individual Differences, 43, 389–399. Chang, E. C. (2000). Perfectionism as a predictor of positive and negative psychological outcomes: Examining a mediation model in younger and older adults. Counseling Psychology, 47, 18–26. Chang, E. C. (2006). Perfectionism and dimensions of psychological well-being in a college student sample: A test of a stress-mediation model. Social and Clinical Psychology, 25, 1001–1022. Chang, E. C. (2009). An examination of optimism, pessimism, and performance perfectionism as predictors of positive psychological functioning in middleaged adults: Does holding high standards of performance matter beyond generalized outcome expectancies? Cognitive Therapy and Research, 33, 334–344. Chang, E. C., Chang, R., & Sanna, L. J. (2009). Optimism, pessimism, and motivation: Relations to adjustment. Social and Personality Psychology Compass, 3, 494–506. Cohen, J. (1988). Statistical power analysis for the behavioral sciences (2nd ed.). New York, NY: Academic Press. Flett, G., Besser, A., Davis, R. A., & Hewitt, P. L. (2003). Dimensions of perfectionism, unconditional self-acceptance, and depression. Rational–Emotive and Cognitive– Behavior Therapy, 21, 119–193. Flett, G., Blankstein, K., Hewitt, P. L., & Koledin, S. (1992). Components of perfectionism and procrastination in college students. Social Behavior and Personality, 20, 85–94. Frost, R. O., Marten, P., Lahart, C., & Rosenblate, R. (1990). The dimensions of perfectionism. Journal of Cognitive Therapy and Research, 14, 449–468. Fry, P. S., & Debats, D. L. (2009). Perfectionism and the five-factor personality traits as predictors of mortality in older adults. Health Psychology, 14, 513–524. Gosling, S. D., Vazire, S., Srivastava, S., & John, O. P. (2004). Should we trust webbased studies? A comparative analysis of six preconceptions about Internet questionnaires. American Psychologist, 59, 93–104. Hamachek, D. E. (1978). Psychodynamics of normal and neurotic perfectionism. Psychology, 15, 27–33. Hamilton, M. (1960). A rating scale for depression. Neurology, Neurosurgery, and Psychiatry, 23, 56–62. Hamilton, M. (1967). Development of a rating scale for primary depressive illness. British Journal of Social and Clinical Psychology, 6, 278–296. Hardre, P. L., Crowson, H. M., & Xie, K. (2010). Differential effects of web-based and paper-based administration of questionnaire research instruments in authentic context-of-use. Educational Computing Research, 42, 103–133. Harris, P. W., Pepper, C. M., & Maack, D. J. (2008). The relationship between maladaptive perfectionism and depressive symptoms: The mediating role of rumination. Personality and Individual Differences, 44, 150–160. Hayes, A. F. (2009). Beyond Baron and Kenny: Statistical mediation analysis in the new millennium. Communication Monographs, 76, 408–420. Hewitt, P. L., Flett, G. L., Turnbull-Donovan, W., & Mikail, S. F. (1991). The multidimensional perfectionism scale: reliability, validity, and psychometric properties in psychiatric samples. Psychological Assessment: Journal of Consulting and Clinical Psychology, 3, 464–468. Hill, R. W., Huelsman, T. J., & Araujo, G. (2010). Perfectionistic concerns suppress associations between perfectionistic strivings and positive life outcomes. Personality and Individual Differences, 48, 584–589. Hoyle, R. H., & Kenny, D. A. (1999). Sample size, reliability, and tests of statistical mediation. In R. Hoyle (Ed.), Statistical strategies for small sample research. Thousand Oaks, CA: Sage Publications. Iacobucci, D., Saldanha, N., & Deng, X. (2007). A meditation on mediation: Evidence that structural equation models perform better than regressions. Consumer Psychology, 17, 140–154. Kawamura, K. W., Hunt, S. L., Frost, R. O., & DiBartolo, P. M. (2001). Perfectionism, anxiety, and depression: Are the relationships independent? Cognitive Therapy and Research, 25, 291–301. Kelley, K., & Lai, K. (2010). MBESS (Version 3.2.0) [Computer software and manual]. Retrieved from
J. Black, W.M. Reynolds / Personality and Individual Differences 54 (2013) 426–431 Preacher, K. J., & Kelley, K. (2011). Effect size measures for mediation models: Quantitative strategies for communicating indirect effects. Psychological Methods, 16, 93–115. Reynolds, W. M., & Kobak, K. A. (1995). Reliability and validity of the Hamilton Depression Inventory: A paper-and-pencil version of the Hamilton Depression Rating Scale Clinical Interview. Psychological Assessment, 7, 472–483. Scheier, M. F., & Carver, C. S. (1992). Effects of optimism on psychological and physical well-being: Theoretical overview and empirical update. Cognitive Therapy and Research, 16, 201–228. Scheier, M. F., Carver, C. S., & Bridges, M. W. (1994). Distinguishing optimism from neuroticism (and trait anxiety, self-mastery, and self-esteem): A reevaluation of the Life Orientation Test. Personality and Social Psychology, 67, 1063–1078. Slade, P. D., & Owens, R. G. (1998). A dual process model of perfectionism based on reinforcement theory. Behavior Modification, 22, 372–390.
431
Sobel, M. E. (1982). Asymptotic confidence intervals for indirect effects in structural equation models. In S. Leinhardt (Ed.), Sociological methodology (pp. 290–312). Washington, DC: American Sociological Association. Stoeber, J., & Otto, K. (2006). Positive conceptions of perfectionism: Approaches, evidence, challenges. Personality and Social Psychology Review, 10, 295–319. Stoeber, J., Uphill, M. A., & Hotham, S. (2009). Predicting race performance in triathlon: The role of perfectionism, achievement goals, and personal goal setting. Sport and Exercise Psychology, 31, 211–245. Sturman, E. D., Flett, G. L., Hewitt, P. L., & Rudolph, S. G. (2009). Dimensions of perfectionism and self-worth contingencies in depression. Rational–Emotional Cognitive–Behavioral Therapy, 27, 213–231. US Census Bureau. Data on educational attainment. (2010). Retrieved 03.06.2011.
Contents lists available at SciVerse ScienceDirect
Personality and Individual Differences journal homepage: www.elsevier.com/locate/paid
Examining the relationship of perfectionism, depression, and optimism: Testing for mediation and moderation Jessica Black ⇑, William M. Reynolds Humboldt State University, 1 Harpst Street, Arcata, CA 95521, United States
a r t i c l e
i n f o
Article history: Received 6 August 2012 Received in revised form 10 October 2012 Accepted 18 October 2012 Available online 13 November 2012 Keywords: Perfectionism Optimism Depression Mediation
a b s t r a c t The present study investigated the relationship between perfectionism and depression, and the mediation/moderation effects of optimism. Participants were126 adults (78% women, mean age = 27 years) who completed an online survey that included the Multidimensional Perfectionism Scale (Frost, Marten, Lahart, & Rosenblate, 1990), the Revised Life Orientation Test (Scheier, Carver, & Bridges, 1994), and the Hamilton Depression Inventory (Reynolds & Kobak, 1995). Maladaptive, adaptive and total perfectionism were examined in separate analyses. In most analyses, the data fit a mediation model. Optimism limited the effect (relationship) of total and maladaptive perfectionism on depression. Adaptive perfectionism was related to depression only through optimism. There was a small gender difference, with indirect effects greater for men. These results are consistent with past research and support the notion that correlations between perfectionism, as a trait, and affective variables must be examined using multivariate models that allow for the determination of complex relationships. Ó 2012 Elsevier Ltd. All rights reserved.
1. Introduction Perfectionism has been generally understood to be associated with negative psychological outcomes (Frost et al., 1990; Hewitt, Flett, Turnbull-Donovan, & Mikail, 1991), although some degree of pursuit of excellence is a necessary component of achievement. Beginning with Hamachek (1978), researchers have sought to identify correlates of perfectionism (Slade & Owens, 1998). Hamachek distinguished between normal and neurotic perfectionism in terms of satisfaction and control: the normal perfectionist derives pleasure from striving for perfection yet knows when to relax standards, while the neurotic perfectionist is driven by fear of failure. Slade and Owens (1998) defined what they called a ‘‘dual-process model of positive and negative perfectionism,’’ according to which positive perfectionism and negative perfectionism can be explained in terms of striving and trying to avoid undesirable results. In a meta-analysis of 15 studies that had used a dualist approach, Stoeber and Otto (2006) concluded that striving was a positive form of perfectionism. They contrasted it with perfectionist concerns, such as an inability to live up to perceived expectations. Research by Stoeber, Uphill, and Hotham (2009) partially clarified the connection between striving and outcome: triathlon participants who scored high on perfectionism and reported high
⇑ Corresponding author. Address: 329 Bayside Rd., Arcata, CA 95521, United States. Tel.: +1 707 633 5362. E-mail address: [email protected] (J. Black). 0191-8869/$ - see front matter Ó 2012 Elsevier Ltd. All rights reserved. http://dx.doi.org/10.1016/j.paid.2012.10.012
achievement goals performed better in their races than those with low achievement goals. Stoeber et al. found that differences between performance approach and avoidance mediated the perfectionism–performance relationship. Researchers have found that correlates of perfectionism are often mediated by other variables. Harris, Pepper, and Maack (2008) used the Multidimensional Perfectionism Scale by Frost and colleagues (MPS; 1990) to examine rumination as a mediator between maladaptive perfectionism and depression. The MPS was designed with six factors; ‘‘personal standards’’ has been used to assess positive perfectionism, while ‘‘concern over mistakes’’ and ‘‘doubts about actions’’ together can reflect negative perfectionism (cf. Harris et al., 2008; Park, Heppner, & Lee, 2010; Sturman, Flett, Hewitt, & Rudolph, 2009). Harris et al. found that rumination mediated the relationship between maladaptive perfectionism and depression, using the MPS subscales of concern over mistakes and doubts about actions. Park et al. (2010) used the same MPS subscales as a measure of maladaptive perfectionism and found that maladaptive coping mediated the relationship between maladaptive perfectionism and distress. This was especially true for men, for whom coping increased more strongly with greater perfectionism than it did for women. Ashby, Dickinson, Gnilka, and Noble (2011) reported that hope mediated the relationship between maladaptive perfectionism and depression in middle school students. Adaptive perfectionism was not significantly correlated to depression, and although it was not considered a mediator, it did lend itself to a statistically significant indirect effects model, as determined by bootstrapping techniques.
J. Black, W.M. Reynolds / Personality and Individual Differences 54 (2013) 426–431
Ashby et al. also tested a moderation model, but found that it did not fit their data on perfectionism, hope, and depression. Hope is closely related to optimism, which is often predictive of positive psychological outcomes (Peterson, 2000). Optimism appears related to perfectionism in that the belief that high goals can be attained is part of seeking their attainment. Peterson proposed two dimensions of optimism, big and little. He associated big optimism with what Scheier and Carver (1992) call dispositional optimism, which their scale, the Revised Life Orientation Test (R-LOT; Scheier et al., 1994), measures. Chang, Chang, and Sanna (2009) outline the complexities of optimism, suggesting that it may have realistic (adaptive) and unrealistic (maladaptive) dimensions. Both optimism and perfectionism have been connected with striving, hope, and depression. Fry and Debats (2009), in a study of predictors of mortality in older adults found perfectionism associated with increased risk of death and dispositional optimism with decreased mortality. Other research (e.g. Bergman, Nyland, & Burns, 2007) implies that adaptive optimism should facilitate striving, encouraging perfectionists to focus on the possibility of achieving excellence rather than on the need to avoid censure for failing to do so. Chang (2009) reported that positive performance perfectionism meant more optimism (r(244) = .22), while negative performance perfectionism meant less optimism (r(244) = .26). We did not, however, find any mediational analysis of the relationship between perfectionism and optimism in the literature. Baron and Kenny (1986) define moderation and mediation models used to test ways in which the relationship between two variables can be affected by third variables. Moderation would indicate an interaction between the independent variable, in this case perfectionism, and the moderator, optimism. Mediation occurs when the relationship between the independent variable and the dependent variable is different when a third variable is included in the model. According to the Baron and Kenny criteria, the direct relationship between two variables, significant when considered alone, approaches zero when measured in the presence of the mediator. An indirect effect can occur even when there is no direct effect (Preacher & Hayes, 2004). In this case, the relationship between independent and dependent variables depends entirely upon the mediator. The measurement of mediation and indirect effects allows some inference of causation, as the independent variable is assumed to cause the mediator, which in turn affects the dependent variable. A moderator, on the other hand, exists independently of the focal predictor, which varies at different levels of the moderator. The current research investigated the degree to which optimism acts as a mediator and/or moderator of the relationship between perfectionism and depressive symptomatology. Because of the possible dual nature of optimism (Chang et al., 2009), we predicted a moderation model. Flett, Blankstein, Hewitt, and Koledin (1992) found gender differences in the correlations between perfectionism and procrastination. In view of Park et al.’s (2010) similar results in a mediation model, we examined potential gender differences in our sample by testing separate models for women and men. We used maladaptive and adaptive perfectionism domains as well as overall perfectionism. Our goal was to better understand the potential role of optimism in the relationship between perfectionism and depression.
2. Method 2.1. Participants Participants were 126 adults (98 women, 28 men) with a mean age of 27.7 (SD = 13.9). Of these participants, 57 were students who
427
completed an online survey through the Psychology Department Research Pool, and 69 were adults who completed the same questionnaire using Survey Monkey. 2.2. Instrumentation The Multidimensional Perfectionism Scale (Frost et al., 1990) is a 35-item questionnaire intended to reflect six aspects underlying the construct. It is considered a reliable measure of general perfectionism (Chang, 2000). In the current investigation the 6-item organization subscale was omitted, resulting in a 29-item scale with an internal consistency reliability of ra = .88. Consistent with previous research (Kawamura, Hunt, Frost, & DiBartolo, 2001; Sturman et al., 2009) we used the concern over mistakes and doubts about actions subscales as a measure of maladaptive perfectionism (ra = .84) and the Personal Standards subscale to represent adaptive perfectionism (ra = .76). The Revised Life Orientation Test (Scheier et al., 1994) was used to assess optimism; its test–retest reliability has proven moderately high (.79 at 28 months), and it has shown acceptable internal consistency. Of the ten statements evaluated on a zero-to-four point scale, three are reverse-coded and four are fillers (ra = .77 for this study). The Hamilton Depression Inventory (HDI; Reynolds & Kobak, 1995) was developed to provide a paper–pencil analog measure to the Hamilton Depression Rating Scale (Hamilton, 1960, 1967) clinical interview. Reynolds and Kobak provided strong evidence for the reliability and validity of three versions of the HDI based on clinical and nonclinical samples. The HDI Short Form, which consists of 15 items, was used in the current study. Reynolds and Kobak reported an internal consistency of .93 for the HDI Short Form in a mixed clinical and nonclinical sample of 357 adults. In the current study ra = .91. 2.3. Procedure An anonymous survey was conducted online through the HSU Psychology Department Research Pool and through Survey Monkey for a total of 2 months. Links to the survey were sent in emails and posted on social networking sites. Participants’ identities were protected by disabling IP address and email tracking features on Survey Monkey. No reward was offered other than class credit available to students responding through the psychology department participation pool; participants were allowed to withdraw at any time without losing credit. 2.4. Data analysis The criteria outlined by Baron and Kenny (1986) provide a conceptual framework to facilitate examination of mediation, but does not include a test of statistical significance for the indirect effect of the mediator (a b; see Fig. 1). Baron and Kenny suggested the Sobel (1982) test as appropriate: it tests the null hypothesis that the indirect effect is zero. However, the Sobel test assumes the distribution of ab is normal, which is seldom the case (Hayes, 2009). Bootstrapping, or the nonparametric resampling of the data set in order to make repeated estimates, makes no such assumptions (Preacher & Hayes, 2008). We present regression statistics as well as results of the Sobel test. We also used bootstrapping (10,000 resamples) to provide bias corrected and accelerated confidence intervals as a test of statistical significance. Following the recommendations of Preacher and Kelley (2011), we present kappa squared, j2, as an effect size that is both standardized and insensitive to sample size. j2 is the ratio of the indirect effect to the maximum indirect possible given the constraints of the design and data. The problem of determining when an effect
428
J. Black, W.M. Reynolds / Personality and Individual Differences 54 (2013) 426–431
Fig. 1. The unmediated path between independent and dependent variables (c) is presented above. Below, optimism is added, resulting in a simple mediation diagram. The indirect effect is c–c0 , or a b.In both cases, as perfectionism increases, so does depression, but when optimism works as a mediator, the relationship is less.
positively correlated with depression, r(124) = .36, 95% C.I.[.21, .50], thus not supporting the test of a moderation model. Depression was negatively correlated with optimism, r(124) = .56, 95% C.I.[.68, .40]. Age demonstrated a very low, nonsignificant correlation with perfectionism, r(124) = .01, 95% C.I.[.17, .15], and optimism scores, r(126) = .09, 95% C.I.[.06, .24], and a low negative correlation with depression, r(124) = .15, 95% C.I.[.30, .01]. Table 1 provides correlations between all scales. Tests of independent means for gender showed statistically nonsignificant differences on total perfectionism, maladaptive perfectionism, adaptive perfectionism, and depression. There was a gender difference in optimism with women (M = 15.8, SD = 3.7) more optimistic than men (M = 14.1, SD = 4.2), t(124) = 2.15, p = .03, d = 0.45. There were no significant differences in means between data collection samples (HSU students versus Survey Monkey; see Table 2). 3.1. Mediation
Table 1 Pearson correlation coefficients between total perfectionism and adaptive and maladaptive dimensions, optimism, depression, and age (N = 126).
1. Total perfectionism 2. Maladaptive perfectionism 3. Adaptive perfectionism 4. Optimism 5. Depression 6. Age ** ***
1
2
– .861***
–
.590*** .302** .364*** .011
.327*** .454*** .521*** .130
3
– .154 .037 .002
4
– .556*** .086
5
Age was initially entered as a covariate in the mediation model. Because it did not have a significant effect, it was thereafter excluded in order to facilitate interpretation and calculation of effect sizes. 3.1.1. Total perfectionism Optimism mediated the effect of perfectionism on depression, with an estimated indirect effect (ab) = 0.014, j2 = .174 (see Table 3 for confidence intervals). For women there was a smaller effect, ab = 0.008, j2 = .112. For men the effect was larger, ab = 0.031, j2 = .358, and the data fit a full mediation model (Baron & Kenny, 1986). Data from the two subsamples, Survey Monkey and the university research pool (HSU), were also tested as two separate models; in both cases significant results ensued (respectively, ab = 0.010, j2 = .106, and ab = 0.015, j2 = .219).
– .154
p < .01. p < .001.
can be considered large is addressed by Preacher and Kelley with reference to Cohen’s (1988) estimates for squared correlation coefficients: .01, .09, and .25 can be considered small, medium, and large respectively. In addition to j2, we also provide the percent of the total effect that is mediated, PM (Table 3 provides complete mediation results). Effect sizes and confidence intervals were generated using MBESS, an R package (Kelley & Lai, 2010). 3. Results Preliminary analyses showed that scores on the MPS and LOTR were normally distributed and scores on the HDI were not: A square root transformation of the HDI data fully corrected skew and kurtosis problems. Total perfectionism scores were negatively correlated with optimism, r(124) = .30, 95% C.I.[.46, .13], and
3.1.2. Maladaptive perfectionism Optimism mediated the relationship between maladaptive perfectionism and depression, ab = 0.029, j2 = .203. As before, for men, optimism fully mediated maladaptive perfectionism, ab = 0.076, j2 = .431. For women the effect was smaller, ab = 0.018, j2 = .141. 3.1.3. Adaptive perfectionism Although the adaptive perfectionism subscale was not significantly correlated with depression, it was positively correlated to optimism, and the possibility of an indirect effect remained (Hayes, 2009). Results of mediation analyses confirmed an indirect effect, ab = .027, j2 = .118. As adaptive perfectionism rose, so did optimism, which related to a decrease in depression (see Fig. 2). When tested separately by gender there was again a difference: In this
Table 2 Means (standard deviations) for total perfectionism (MPS), maladaptive perfectionism (MP), adaptive perfectionism (AP), optimism (LOTR), and depression (HDI).
Total Women Men Cohen’s d – men vs. women t p HSU students Survey Monkey Cohen’s d – HSU vs. Survey Monkey t p *
Untransformed data used for HDI.
MPS
MP
AP
LOTR
HDI*
78.6 (14.4) 78.3 (14.1) 79.8 (15.5) 0.10 0.48 .63 79.0 (16.2) 78.3 (12.7) 0.05 0.26 .70
32.0(7.9) 31.8(7.7) 32.6(8.8) 0.10 0.49 .62 32.5(8.9) 32.0(7.0) 0.06 0.62 .54
23.7(4.5) 23.5(4.3) 24.5(5.2) 0.22 0.97 .33 23.6(4.8) 23.8(4.3) 0.04 0.18 .86
15.4 (3.8) 15.8 (3.7) 14.1 (4.2) 0.45 2.15 .03 15.3 (4.3) 15.6 (3.4) 0.08 0.39 .70
9.8(7.9) 9.6 (6.9) 10.4 (10.9) 0.10 0.47 .64 11.1 (8.1) 8.6 (7.7) 0.32 1.78 .08
429
J. Black, W.M. Reynolds / Personality and Individual Differences 54 (2013) 426–431 Table 3 Mediation Results. Variable
ab
95% C.I. LL
a
b
c
c’
Sobel test z
95% C.I.
LL
UL
LL
UL
0.014 0.008 0.031
0.006 0.002 0.011
0.024 0.018 0.058
0.09* 0.07* 0.12*
0.15* 0.12* 0.27*
0.03* 0.03* 0.03
0.02* 0.02* <0.01
3.07 2.23 2.15
.002 .026 .032
0.44 0.27 1.03
0.19 0.06 0.39
0.81 0.64 4.94
.174 .112 .358
.075 .025 .131
.281 .231 .563
Maladaptive perfectionism Women Men
0.029 0.018 0.076
0.013 0.005 0.039
0.049 0.037 0.140
0.24* 0.19* 0.30*
0.12* 0.10* 0.26*
0.08* 0.08* 0.08*
0.05* 0.06* 0.01
3.77 2.74 2.91
<.001 .006 .004
0.37 0.24 0.90
0.17 0.07 0.45
0.63 0.50 1.98
.203 .141 .431
.104 .048 .231
.310 .260 .625
0.027 0.033 0.008
0.057 0.067 0.087
0.001 0.011 0.087
0.16* 0.22* 0.03
0.17* 0.15* 0.27*
1.68 2.33 0.184
.094 .020 .854
1.94 4.67 0.51
0.63 1.77 35.6
2450 3851 5.40
.118 .144 .045
.015 .054 .000
.224 .262 .135
0.01 0.01 0.02
0.01 0.03 0.02
p
j2
95% C.I.
Total perfectionism Women Men
Adaptive perfectionism Women Men
UL
PM
ab = estimated indirect effect. All confidence intervals generated with bias corrected and accelerated bootstrapping(N = 10,000). p < .05.
*
Fig. 2. The indirect effect of adaptive perfectionism on depression is through optimism.
case, the indirect effect was significant for women, ab = .033, j2 = .144, but not for men.
4. Discussion Perfectionists set high standards for themselves, and the failure to meet these standards often relates to poor psychological outcomes. For persons who manifest strong perfectionistic tendencies, the perception or anticipation of failure could lead to feelings of depression. Perfectionists who have a belief in their ability to achieve their goals should be less likely to experience the same negative consequences. In this manner, optimism should reduce the negative affective consequences of perfectionism. Initially, we expected a moderation model, with persons scoring high on perfectionism and low on optimism reporting greater depression, but persons high on optimism showing less negative outcomes. The current research found a more complex mediation relationship to fit the data. In all the mediation models we tested, with the exception of adaptive perfectionism in men, there was a significant indirect effect. Kappa squared values indicate medium to large effect sizes. For our sample, effects were larger for maladaptive perfectionism than for total perfectionism. They were also larger for men than for women in total perfectionism and maladaptive perfectionism. Although the confidence intervals for men and women overlapped for total perfectionism, they did not for maladaptive perfectionism, suggesting a genuine gender difference. This result is similar to the findings of Park et al. (2010), who reported that maladaptive coping was a greater mediator of the relationship between perfectionism and psychological distress for men. It is interesting that for men in our sample there was complete mediation: the path between the total perfectionism and depression was no longer signif-
icant in the presence of optimism. For women there was partial mediation. It is likely that the mediating effect of optimism was strongest for maladaptive perfectionism because its correlation with optimism was stronger than found for total perfectionism. It needs to be recognized that the total perfectionism score included the maladaptive scale items. The research we reviewed did not compare maladaptive perfectionism to total perfectionism in mediation models, but it did compare it to adaptive perfectionism. Flett, Besser, Davis, and Hewitt (2003) found that self-acceptance mediated the outcome of socially-prescribed (negative) perfectionism but not self-prescribed (positive) perfectionism. They also reported an indirect effect of self-acceptance on other-oriented (negative) perfectionism. Harris et al. (2008) found similar differences between adaptive and maladaptive perfectionism in a mediation context: rumination meant more depression for maladaptive perfectionism, but not for adaptive perfectionism. Our results indicate a different mediation model for adaptive perfectionism. Adaptive perfectionism did not show a statistically significant relationship with depression except through optimism, in a suppression model. Adaptive perfectionism predicted increasing optimism, which in turn related to lower depression (see Fig. 2). Although optimism showed a stronger mediation effect for men than women in total and maladaptive perfectionism, the indirect effect of adaptive perfectionism through optimism was only found for women; for men the path between adaptive perfectionism and optimism was small and statistically insignificant. Because of the limited number of men, wider confidence intervals make it difficult to make inferences. However, the point estimates indicate interesting gender differences that merit future research. Hoyle and Kenny (1999) found little effect of sample size on the path estimates in a Monte Carlo simulation testing samples of N = 25–200. Both kappa squared and PM (percent of total effect mediated) are insensitive to sample size (Preacher & Kelley, 2011). Interestingly, effect sizes for men are much larger for both total and maladaptive perfectionism, and the correlation between optimism and depression is much stronger for men than for women. Nevertheless, the results are necessarily imprecise due to the relatively small number of men in our sample. 4.1. Limitations There is always a question of whether data collected from college students can be generalized to non-academic populations. This study gives some support for generalizability, although the Survey Monkey respondents were not entirely representative, as 78% had at least some college education compared to 55% of adults in the US population (US Census Bureau, 2010). There were no
430
J. Black, W.M. Reynolds / Personality and Individual Differences 54 (2013) 426–431
significant differences in measure means between the data collected through the university research participation pool and that collected through Survey Monkey. Mediation analyses results conducted separately for these two groups were generally comparable for both total and maladaptive perfectionism. The university sample showed larger effect sizes and fit a full mediation model. However, the effect size confidence intervals overlapped so we cannot be positive of a difference in the respective samples. There was no mediation effect for either group when the model was tested with adaptive perfectionism. Overall, there were no appreciable differences between the two data collection subsamples used in our study. There are a number of limitations specific to the sample used in this study. Although the inclusion of data collected online through Survey Monkey compensates for some threats to external validity, the sample was still limited. The sample was largely college educated, and predominantly female. In view of the gender differences found in the present data, further research with a larger sample that is more evenly balanced between men and women is needed. Further limitations are imposed by the self-report method, and by the reliance on the Internet, which restricts participation to those with a connection and the inclination to complete a survey. The comparability of online versus pencil and paper surveys also arises. However, research suggests that online surveys have minimal disadvantages when compared to paper and pencil versions of the same measures (Gosling, Vazire, Srivastava, & John, 2004; Hardre, Crowson, & Xie, 2010). As with all tests of mediation models based on self-report measures given at the same time, inferences about causation are limited. We have tried to control for reverse effects to some extent by using a ‘‘state’’ measure of depression, while both the MPS and LOT-R are believed to measure general ‘‘trait’’ person characteristics. When analyses were conducted with depression as the independent variable and perfectionism as the dependent variable, there was no evidence of mediation. Nevertheless, causation cannot be ascribed. In the present study a moderate relationship (r2 = .31) was found between optimism and depression. When the two were interchanged in the mediation analyses (depression as mediator, optimism as dependent variable), results indicated significant indirect effects despite the fact that optimism is typically considered a trait construct and depression a state condition. Although frequently used as a measure of optimism, the LOT-R is limited in its level of reliability, thus introducing measurement error, which is problematic when conducting mediation analyses. Using optimism as predictor and perfectionism as mediator also produces a mediation model with similar effect sizes. Conceptually, it is plausible that optimism causes positive perfectionism, although it is less intuitive to in the case of negative perfectionism. Further investigation is needed to clarify a coherent causal relationship.
4.2. Recommendations for further research It is advisable to include more constructs in the model, and to use more than one measure of each (Iacobucci, Saldanha, & Deng, 2007). We recommend including a measure of striving and/or maximization, which has been associated with both optimism and perfectionism (Fry & Debats, 2009; Hill, Huelsman, & Araujo, 2010; Stoeber et al., 2009). Stress has been shown to mediate correlates of perfectionism (Chang, 2000, 2006) and could be included in a more complete model. It is clear that perfectionism is a complex construct. Our results indicate that mediational effects of optimism may differ by dimension of perfectionism and by gender. More research is necessary to confirm and explain this finding. Future research on optimism as a
mediator of perfectionism may also benefit from a more reliable assessment of optimism.
References Ashby, J. S., Dickinson, W. L., Gnilka, P. B., & Noble, C. L. (2011). Hope as a mediator and moderator of multidimensional perfectionism and depression in middle school students. Counseling & Development, 89, 131–139. Baron, R. M., & Kenny, D. A. (1986). The moderator–mediator variable distinction in social psychological research: Conceptual, strategic, and statistical considerations. Personality and Social Psychology, 51, 1173–1182. Bergman, A. J., Nyland, J. E., & Burns, L. R. (2007). Correlates with perfectionism and the utility of a dual process model. Personality and Individual Differences, 43, 389–399. Chang, E. C. (2000). Perfectionism as a predictor of positive and negative psychological outcomes: Examining a mediation model in younger and older adults. Counseling Psychology, 47, 18–26. Chang, E. C. (2006). Perfectionism and dimensions of psychological well-being in a college student sample: A test of a stress-mediation model. Social and Clinical Psychology, 25, 1001–1022. Chang, E. C. (2009). An examination of optimism, pessimism, and performance perfectionism as predictors of positive psychological functioning in middleaged adults: Does holding high standards of performance matter beyond generalized outcome expectancies? Cognitive Therapy and Research, 33, 334–344. Chang, E. C., Chang, R., & Sanna, L. J. (2009). Optimism, pessimism, and motivation: Relations to adjustment. Social and Personality Psychology Compass, 3, 494–506. Cohen, J. (1988). Statistical power analysis for the behavioral sciences (2nd ed.). New York, NY: Academic Press. Flett, G., Besser, A., Davis, R. A., & Hewitt, P. L. (2003). Dimensions of perfectionism, unconditional self-acceptance, and depression. Rational–Emotive and Cognitive– Behavior Therapy, 21, 119–193. Flett, G., Blankstein, K., Hewitt, P. L., & Koledin, S. (1992). Components of perfectionism and procrastination in college students. Social Behavior and Personality, 20, 85–94. Frost, R. O., Marten, P., Lahart, C., & Rosenblate, R. (1990). The dimensions of perfectionism. Journal of Cognitive Therapy and Research, 14, 449–468. Fry, P. S., & Debats, D. L. (2009). Perfectionism and the five-factor personality traits as predictors of mortality in older adults. Health Psychology, 14, 513–524. Gosling, S. D., Vazire, S., Srivastava, S., & John, O. P. (2004). Should we trust webbased studies? A comparative analysis of six preconceptions about Internet questionnaires. American Psychologist, 59, 93–104. Hamachek, D. E. (1978). Psychodynamics of normal and neurotic perfectionism. Psychology, 15, 27–33. Hamilton, M. (1960). A rating scale for depression. Neurology, Neurosurgery, and Psychiatry, 23, 56–62. Hamilton, M. (1967). Development of a rating scale for primary depressive illness. British Journal of Social and Clinical Psychology, 6, 278–296. Hardre, P. L., Crowson, H. M., & Xie, K. (2010). Differential effects of web-based and paper-based administration of questionnaire research instruments in authentic context-of-use. Educational Computing Research, 42, 103–133. Harris, P. W., Pepper, C. M., & Maack, D. J. (2008). The relationship between maladaptive perfectionism and depressive symptoms: The mediating role of rumination. Personality and Individual Differences, 44, 150–160. Hayes, A. F. (2009). Beyond Baron and Kenny: Statistical mediation analysis in the new millennium. Communication Monographs, 76, 408–420. Hewitt, P. L., Flett, G. L., Turnbull-Donovan, W., & Mikail, S. F. (1991). The multidimensional perfectionism scale: reliability, validity, and psychometric properties in psychiatric samples. Psychological Assessment: Journal of Consulting and Clinical Psychology, 3, 464–468. Hill, R. W., Huelsman, T. J., & Araujo, G. (2010). Perfectionistic concerns suppress associations between perfectionistic strivings and positive life outcomes. Personality and Individual Differences, 48, 584–589. Hoyle, R. H., & Kenny, D. A. (1999). Sample size, reliability, and tests of statistical mediation. In R. Hoyle (Ed.), Statistical strategies for small sample research. Thousand Oaks, CA: Sage Publications. Iacobucci, D., Saldanha, N., & Deng, X. (2007). A meditation on mediation: Evidence that structural equation models perform better than regressions. Consumer Psychology, 17, 140–154. Kawamura, K. W., Hunt, S. L., Frost, R. O., & DiBartolo, P. M. (2001). Perfectionism, anxiety, and depression: Are the relationships independent? Cognitive Therapy and Research, 25, 291–301. Kelley, K., & Lai, K. (2010). MBESS (Version 3.2.0) [Computer software and manual]. Retrieved from
J. Black, W.M. Reynolds / Personality and Individual Differences 54 (2013) 426–431 Preacher, K. J., & Kelley, K. (2011). Effect size measures for mediation models: Quantitative strategies for communicating indirect effects. Psychological Methods, 16, 93–115. Reynolds, W. M., & Kobak, K. A. (1995). Reliability and validity of the Hamilton Depression Inventory: A paper-and-pencil version of the Hamilton Depression Rating Scale Clinical Interview. Psychological Assessment, 7, 472–483. Scheier, M. F., & Carver, C. S. (1992). Effects of optimism on psychological and physical well-being: Theoretical overview and empirical update. Cognitive Therapy and Research, 16, 201–228. Scheier, M. F., Carver, C. S., & Bridges, M. W. (1994). Distinguishing optimism from neuroticism (and trait anxiety, self-mastery, and self-esteem): A reevaluation of the Life Orientation Test. Personality and Social Psychology, 67, 1063–1078. Slade, P. D., & Owens, R. G. (1998). A dual process model of perfectionism based on reinforcement theory. Behavior Modification, 22, 372–390.
431
Sobel, M. E. (1982). Asymptotic confidence intervals for indirect effects in structural equation models. In S. Leinhardt (Ed.), Sociological methodology (pp. 290–312). Washington, DC: American Sociological Association. Stoeber, J., & Otto, K. (2006). Positive conceptions of perfectionism: Approaches, evidence, challenges. Personality and Social Psychology Review, 10, 295–319. Stoeber, J., Uphill, M. A., & Hotham, S. (2009). Predicting race performance in triathlon: The role of perfectionism, achievement goals, and personal goal setting. Sport and Exercise Psychology, 31, 211–245. Sturman, E. D., Flett, G. L., Hewitt, P. L., & Rudolph, S. G. (2009). Dimensions of perfectionism and self-worth contingencies in depression. Rational–Emotional Cognitive–Behavioral Therapy, 27, 213–231. US Census Bureau. Data on educational attainment. (2010).