As noted earlier, we expected that depressive symptoms would increase across the years of adolescence and then decline with the transition to early adulthood. To evaluate these predictions, the latent growth curve model with the initial level and two segmental slopes (one for the adolescent years and the other for the early adult years) for depressive symptoms was estimated using covariances and FIML (Arbuckle and Wothke, 1999
). This comprehensive growth curve model showed a reasonably good fit with the data (see ). The CFI was.93 and the RMSEA was .08.
The results showed that the mean initial level of depressive symptoms was 18.32 in 1991. The variation in initial level was significantly different from zero and large (29.75, t = 6.50). This finding indicates a wide range in depressive symptoms for the youth in the study during the 9th grade, with some participants suffering high levels of depressive symptoms while others had no symptoms at all. As expected, the average rate of change during adolescence (adolescent slope) was positive (average rate of change = .10, t = 1.00), but the slope was not significantly different from zero. However, the results also showed that the variation among adolescents in rates of change for depressive symptoms was significantly different from zero (variance in the rate of change = 1.81, t = 4.45), indicating that depressive symptoms likely increased for some adolescents, decreased for others, and remained relatively constant for some from 1991 through 1994. During adolescence, then, we can conclude that at least some adolescents were experiencing growth in depressive symptoms and that, even if there was not evidence for an average upward trend in depressed mood, neither was there evidence for an average decline in mood problems.
The average rate of change during transition to adulthood (slope from 1995 to 2001) was negative and significantly different from zero (average rate of change = -.31, t = -6.21), indicating the predicted average decrease in depressive symptoms during the transition to adulthood. This negative slope for depressive symptoms may partly reflect regression to the mean. That is, adolescents who had lower levels of depressive symptoms in 1995 experienced more of an increase in symptoms from 1995 to 2001 compared to adolescents who had relatively high levels of symptoms in 1995. Results also showed that variation among emerging adults in rates of change for depressive symptoms was significantly different from zero (variance in the rate of change = .53, t = 6.14), indicating that, although there was an average decline in depressed mood, not all youth demonstrated this type of decline from 1995 through 2001. Given this significant variability in the growth parameters for depressed mood of youth, we next evaluated the overall theoretical model for the total sample ().
Early Family Adversity, Youth Depressive Symptom Trajectories, and Social Pathways (Unstandardized coefficients; standard errors in parentheses).
We identified the most parsimonious model that best fits the data. represents this model. The overall fit of this model with the data was quite good ( 2(98) =179 , CFI=.95, RMSEA=.04) and explained 18% of the variance in young adult social status attainment.
The results in indicate that both the adolescent slope and young adult slope for depressive symptoms were negatively associated with the initial level in 1991 ( = -.16, p < .01, and □ = -.07, p < .01, respectively). These negative influences likely indicate regression to the mean; that is, on average adolescents who had lower initial levels of depressive symptoms in 1991 experienced a steeper increase in symptoms from 1991 to 1994 compared to adolescents who had relatively higher initial levels of symptoms. Similarly, adolescents who had lower levels of depressive symptoms in 1991 experienced more of an increase in symptoms from 1995 to 2001 compared to adolescents who had relatively high levels of symptoms in 1991. With regard to change over time, these results make intuitive sense inasmuch as youth with lower initial levels have more room for growth in symptoms in subsequent years.
As expected, the results in show that parent education and family structure (two-parent family status vs. single-parent family status) related negatively to both negative life events (NLE) and parental rejection of children. Both NLE and parental rejection predicted the initial level of depressive symptoms (β = .35, p < .01 and .43, p < .01 respectively). The residual covariance between NLE and parental rejection was .15. Unexpectedly, only NLE predicted an early transition to adulthood (β = .07 p < .05). However, parental education was negatively related to an early transition (β = -.22 p < .05). Together, the family variables and the initial level of depressive symptoms explained 09% of the variance in adolescent early transition.
As shown in , and consistent with theoretical expectations, the initial level of depressive symptoms also predicted early transition to adulthood (β = .04, p < .05). The residual covariance between adolescent early transition events and adolescent slope was negative and not significant. However, consistent with expectations, an early transition predicted growth in depressive symptoms during the transition to adulthood (β = .09, p < .05). More importantly, all of the depressive symptom growth parameters (initial level, adolescent slope, and young adult slope) jeopardized young adult social status attainment (β = -.13, -.66, and -.57, respectively; all ps < .05). Adolescents who experienced a higher level of depressive symptoms at the initial level show lower levels of young adult social status attainment than do adolescents who experienced a lower level of depressive symptoms regardless of subsequent changes. Although the mean adolescent slope is not statistically significant (i.e., the average trend does not differ from zero), adolescents who demonstrated relatively greater rates of increase in the slope for depressive symptoms demonstrated relatively lower levels of young adult social status attainment. Conversely, adolescents who experienced a decline in depressive symptoms during adolescence experienced relatively higher levels of young adult social status attainment.
As a final step in the analysis, we re-estimated the right side of the model in treating social status attainment as an ordered categorical variable using Mplus software (not shown in the figures). We predicted the increase in the odds of social status attainment for a one level increase in each predictor (depressive symptom growth parameters and early transitions). The results were consistent with the previous SEM findings. The log-odds of moving from a given level to any higher category of young adult social status attainment significantly decreased by -.09, -.15, -.88, and -.01 (or odds decreased by 9%, 14%, 59% and 1%) for each one unit increase in the initial level, adolescent slope, transition to adulthood slope, and early transitions, respectively, when all these four predictors were in the model.
Although the average transition to adulthood slope for depressive symptoms is negative, participants who demonstrated relatively less decline or even increases in symptoms of depression during this time demonstrated relatively lower levels of adult social status attainment. That is, youth who experience slower rates of decline in depressive symptoms (a less negative slope) during the transition to adulthood showed lower levels of young adult social status attainment than do youth who experience a faster rate of decline (a more negative slope) in depressive symptoms during the transition to adulthood, regardless of the initial level and previous changes during adolescence. These results are consistent with the prediction that the initial level and growth in depressive symptoms will decrease young adult social status attainment during the early adult years. Support for the theoretical model also came from the finding that an early transition to adulthood had a negative relationship with later young adult social status attainment (β = -.16, p < .05).
We also estimated growth curve models (see ) for males and females separately. Although males and females differed significantly in their initial levels of depressive symptoms (16.70 vs. 19.10, respectively), we did not find significant gender differences in the remaining growth parameters. The adolescent slope was not significantly different from zero for boys (.06) or girls (.11), whereas the transition to adulthood slope was statistically significant for both groups (-.33 and -.31 for boys and girls, respectively). However, the variances of both the adolescent and the transition to adulthood slope parameters were statistically significant for both boys and girls. Although adolescent girls experienced higher average initial levels of depression than did adolescent boys, they did not experience increases or decreases in depressed mood that were significantly different from the increases and decreases experienced by adolescent boys over the study period (these analyses are not shown in the figures).
In addition to the results for the gender-specific growth curves, gender only predicted the initial level of adolescent depressive symptoms in the overall model, with a marginal positive influence (p < .10) on the adolescent slope. We also estimated the overall model in for males and females separately. All of the associations summarized in were consistent for girls and boys. None of the path coefficients differed significantly for boys and girls, indicating that gender did not moderate the associations summarized in .