Socioeconomic Status, Depressive Symptoms, and Academic Achievement: Meta-Analytic Structural Equation Modeling with Individual Participant Data

• In Event: Leveraging Innovative Methods in Meta-Analysis to Advance Developmental Theory

Abstract

Socioeconomic status (SES) is one of the most highly researched constructs in developmental science, yet important questions underly how to best model it. Decades of empirical evidence indicate that SES explains variation in developmental outcomes such as depressive symptoms and academic achievement (Korous et al., 2018; Harwell et al., 2017). The scholarly discussion on how to operationalize SES has primarily centered on which components to include (Diemer et al., 2013) whereas less attention has been given to the nature of their relations. That is, are developmental relations with SES always in the same direction or does the direction of association change at different levels of SES? This question is important because the majority of research on SES has overemphasized a more is always better model (linear) instead of considering the risk at both ends of the extreme (quadratic), despite accumulating evidence supporting the latter (Luthar et al. 2019). In this study, we conduct a meta-analysis using individual participant data (IPD) to examine two research questions: 1) Does a nonmonotonic model (quadratic) of the relations between components of SES (i.e., income, education, and occupation status), depressive symptoms, and academic achievement (grade point average) fit better than a monotonic model (linear)? and 2) Is the magnitude of relation moderated by developmental period (childhood, adolescence, young adulthood, adulthood, older adulthood), sex (male, female), or race/ethnicity (e.g., Black, Latino)? We hypothesize that there will be more support for the nonmonotonic model. Our moderation analyses are exploratory.

The data used in this study were identified from a larger IPD study on racial/ethnic disparities in depressive symptoms (Causadias et al., 2018). These datasets were nationally representative of the U.S. population, identified from the Inter-university Consortium for Political and Social Research, that included measures of depressive symptoms. From this pool of datasets, we identified 61 for inclusion in our analyses. These represent 24 unique survey series and over 1,000,000 participants. The mean age of the samples included ranged from 4 to 90 years old.

To address our aims, we use a one-stage meta-analytic structural equation model (MASEM) to average across correlations (Jak & Cheung, 2019). We accomplish this in two steps. First, we extract correlations for specific groups (e.g., Black adolescent males, Latina adult females) from each dataset. Then, using one-stage MASEM, we fit path models (Figure 1) to the pooled correlation (10X10) matrix. This method allows us to estimate the contribution of each component of SES while controlling for the others, and it allows us to compare the amount of variability explained between our two models. Notably, we can include datasets regardless if they are missing some of the correlations that make up the pooled correlation matrix (e.g., correlation between income and academic achievement). This method also allows us to examine if the path parameters in Figure 1 are larger or smaller given the developmental period, sex, or race/ethnicity. Our findings will have the potential to advance theoretical understanding of the risk and promotive effects of SES and its role in development.

Authors

• Kevin Michael Korous, GDIT
  Presenting Author

• Karina Cahill, Arizona State University
  Non-Presenting Author

• Longfeng Li, Pennsylvania State University, University Park
  Non-Presenting Author

• José M. Causadias, Arizona State University
  Non-Presenting Author