About this poster

---

Panel information
Panel 5. Identity

Abstract

---

This presentation applies Critical Quantitative (CQ) methodology and MIMIC modeling to demonstrate their affordances for anti-racist and justice-oriented measurement in developmental science. Consistent with the conference theme, this presentation operates from the perspective that white supremacy and bias permeate many of the measures we use – and so aims to provide a broader methodological perspective (CQ) and a specific methodological approach (MIMICs) to help identify, attenuate, and/or eliminate biased and/or racist items in the measures developmentalists commonly use. [CQ is guided by five principles foundation, goals, parity, subjectivity, self-reflexivity, not detailed here for space reasons (AUTHORS, 2023), is a critical approach to quantitative methodology and research distinct from CRiTQuant because it is flexible regarding the critical theory(ies) that animate its methodological approach. MIMIC, or Multiple Indicator and MultIple Causes models, afford powerful claims about measurement - particularly biased items – and are relatively simple to specify and test.]

To achieve these aims, this tutorial-oriented presentation illustrates how even in a measure designed to capture the interpersonal discrimination that Black people face, the Everyday Discrimination Scale (Sternthal, Slopens & Williams, 2011), applying a CQ perspective and MIMIC modeling illuminated that two of the scale’s five items under-measure racism for Black respondents.

MIMIC modeling analyses proceeded in three steps; data and code are available via this anonymous OSF repository. In the first step, a confirmatory factor analysis (CFA) model was applied to a subsample of 3,660 participants, aged 32-42, who identified as Black (N = 863, 23.57%) or as white (N = 2797, 76.42%). These data came from AddHealth, Wave V, from 2016-2018 (Harris, 2013). CFA results are depicted in Table 1; essentially the model fit well and each item loading was strong and statistically significant.

In the second MIMIC step, an exogeneous covariate (where 0 = white and 1 = Black respondents) predicted the
Everyday Discrimination latent variable. As depicted in the Figure (topmost “Black” box in the figure) and as expected, Black respondents had a significantly higher (β = .09) latent mean than white respondents (albeit one would expect a larger mean difference). This regression has the effect of adjusting the latent mean between white and Black respondents to be similar, before testing for item bias in step three.

In the third MIMIC step, the “poorer service” and “afraid” observed items are regressed onto the same Black exogeneous covariate. Each of these paths represent a formal test for DIF, or differential item functioning. (These tests of DIF are equivalent to tests of scalar invariance (e.g., intercepts) in a multiple-group measurement invariance approach.) Black respondents reported poorer service in restaurants/stores more frequently (𝛽 = .20, p < .001) and reported that other people more frequently act afraid of them (𝛽 = .11, p < .001), indicating DIF in these two items. The direction of effect indicates that these items are biased in that they under-measure everyday discrimination for Black respondents - surprising in a measure explicitly developed to capture the discriminatory experiences that Black people face (Sternthal et al., 2011).

This application of a broader CQ perspective and the specific MIMIC approach showed potential for anti-racist and/or justice-oriented measurement in developmental science, by identifying items that under-estimate discrimination for Black people. Because of well-established links between racialized discrimination and health outcomes among people of color (Sternthal et al., 2011), the under-measurement of racialized discrimination in this scale may bias estimates of the relations between discrimination and health among Black respondents. Because the CQ perspective and MIMIC modeling approach may identify biased and/or racism in the measures that developmental scientists employ, this approach holds promise in facilitating a more justice-oriented and anti-racist developmental science.

Author information

Author	Role
Matthew A. Diemer, University of Michigan	Presenting author
Michael B. Frisby, Georgia State University, United States	Non-presenting author
Aixa D. Marchand, University of Illinois, United States	Non-presenting author

⇦ Back to session