332 STEIGER.ALLEMAND.ROBINS.AND FEND arate nofMIcvite did fcyer the Our third model examined the effects of level and char ceived academic competence on depressive symptoms at age ularity as a time-va with sche time-va and for comp RMSEA=016(CI 1010.0211).Adolescents'percep nce and 001.02(00 C 1 038p. effects.Finally.we tested all these models without the inclusion identical(see Table 4). n depre ssive symptoms at age 16 and at age 35.controlling for ceptually equal sever n th eve ling ver on d not affected by the inclusion of peer popularity as a time-varying RMSEA 031(90%CI[.028.033,and result We next examined the effects of level and change in percei ge 16 and 35 001)and at age 35 (B= .001).Additionally and equal -2578p<001.cH=98 RMSEA020 gina self-es 027 of thai 00D and at age 35 (=00D).For this analysis.we ted at age 35.thus we reduced the depre scale age 16 from earance predicted rolling for age 16 epressive symptoms (se Table 4).The lysis for the two of self-esteer Variable Predictors Dae 3s Without covariates e level atence chanFinally, for our main research aim, we estimated three separate second-order LGM models to test the effects of level and change in self-esteem from age 12 to 16 on depressive symptoms at age 16 and 35. Table 4 and Figure 1 present the results of these analyses. As mentioned above, for global self-esteem, we included peer popularity as a time-varying covariate, and for physical appearance and academic competence, we included BMI and school grades, respectively, as time-varying covariates. We also controlled for the effects of gender. Correlations between the domain-specific self￾esteem of physical appearance and academic indicated that these constructs differ between individuals, revealing only small to medium-sized correlations (see Table 4). We first tested the effects of global self-esteem development on depressive symptoms at age 16 and 35, with peer popularity as a time-varying covariate. The model fit well, 2 (159)
278.94, p .001, CFI
.984, RMSEA
.022 (90% CI [.018, .026]). We found medium-sized effects of self-esteem level on depressive symptoms at age 16 and at age 35 (see Figure 1 and Table 4). Furthermore, self-esteem change had small to medium-sized effect on depressive symptoms at age 16 and at age 35, controlling for age 16 depressive symptoms (see Figure 1 and Table 4). Interest￾ingly, the effect of self-esteem level remained similarly strong even 20 years after the measurement of self-esteem in adolescence, and even though we are predicting age 35 depressive symptoms controlling for age 16 depressive symptoms. These results were not affected by the inclusion of peer popularity as a time-varying covariate. We next examined the effects of level and change in perceived physical appearance on depressive symptoms at age 16 and 35, with BMI as a time-varying covariate. Again, model fit was good, 2 (145)
257.88, p .0001, CFI
.983, RMSEA
.023 (90% CI [.018, .027]). Adolescents’ perceptions of their physical appearance predicted depressive symptoms both at age 16 and at age 35; for depressive symptoms at age 35, we controlled for age 16 depressive symp￾toms (see Table 4). Moreover, change in perceived physical ap￾pearance predicted depressive symptoms at age 16 and 35, con￾trolling for age 16 depressive symptoms (see Table 4). The inclusion of BMI as a covariate did not significantly alter these effects. Our third model examined the effects of level and change in perceived academic competence on depressive symptoms at age 16 and 35, with school grades as a time-varying covariate. Model fit was good, 2 (156)
216.10, p .0001, CFI
.992, RMSEA
.016 (90% CI [.010, .021]). Adolescents’ percep￾tions of their academic competence predicted depressive symp￾toms at age 16 and 35, controlling for age 16 depressive symptoms (see Table 4). Moreover, change in perceived phys￾ical appearance predicted depressive symptoms at age 16 and 35, controlling for age 16 depressive symptoms (see Table 4). The inclusion of school grades did not significantly alter these effects. Finally, we tested all these models without the inclusion of the time-varying covariates in order to see whether results differed in any way. However, all effects remained virtually identical (see Table 4). One possible critique is the fact that two different measures of depression were used at age 16 (T5) and age 35 (T6). In order to address this critique, we extracted the conceptually equal seven items measured at T5 and T6 and additionally ran the same analysis as reported in this article using these overlapping items, revealing very similar results. Our model of self-esteem on de￾pression using only these seven corresponding items again showed a satisfactory model fit, 2 (375)
920.91, p .001, CFI
.947, RMSEA
.031 (90% CI [.028, .033]), and results remained virtually identical. Again, we found medium-sized effects of self￾esteem level on depressive symptoms at age 16 (
.41, p .001) and at age 35 (
.36, p .001). Additionally and equal to our original analysis, self-esteem change had small to medium￾sized effects on depressive symptoms at age 16 (
.40, p .001) and at age 35 (
.21, p .001). For this analysis, we tested the exact same model as our original model, with the only change of excluding those items that had no corresponding items tested at age 35, thus we reduced the depression scale age 16 from 13 to those seven symptoms that were tested in the follow-up survey (T6) at age 35. Analysis for the two domains of self-esteem Table 4 Level and Change in Global Self-Esteem and Domain-Specific Self-Esteem as Predictors of Depression at Age 16 and 35 Variable Predictors Depression age 16a Depression age 35a With covariates Self-esteem level .42


.36


Peer popularity, body mass index, and school grades Self-esteem change .41


.22


Physical appearance level .30


.21


Physical appearance change .19

.12
Academic competence level .42


.15

Academic competence change .31


.18

Without covariates Self-esteem level .41


.35


Self-esteem change .40


.22


Physical appearance level .29


.21


Physical appearance change .20

.11 Academic competence level .43


.15

Academic competence change .31


.15

Note. N
1,527. a Standardized beta coefficients from the latent growth curve models.
p .05.

p .01.


p .001. This document is copyrighted by the American Psychological Association or one of its allied publishers. This article is intended solely for the personal use of the individual user and is not to be disseminated broadly. 332 STEIGER, ALLEMAND, ROBINS, AND FEND