Race and gender disparities in the prevalence and incidence Alzheimer’s Disease and related dementias (ADRDs) are well-documented in national reports: In the U.S., Black people are more likely to have ADRDs than White people, while women make up th...
Race and gender disparities in the prevalence and incidence Alzheimer’s Disease and related dementias (ADRDs) are well-documented in national reports: In the U.S., Black people are more likely to have ADRDs than White people, while women make up the majority of those aged 65 and older with ADRDs. Scholars in the social sciences frequently utilize large, nationally-representative, longitudinal surveys that collect detailed data on cognition and respondents’ lives to identify the influence that modifiable risk factors have on race and gender group differences. However, the ability of these commonly explored individual and contextual variables to fully explain cognition disparities by race and gender are limited. Education, socioeconomic status (SES), stress, and geography fail to fully account for the disparities between Black and White samples and between men and women. A central question endures: What accounts for the remaining race- and gender-based disparities in cognition that social contextual factors cannot explain?The majority of research detailing race and gender disparities in cognition are conducted under the assumption—without demonstration—of measurement invariance, or the essential characteristic that latent variables quantify the same underlying concept across cultures, societies, geographies, and time. Like other assessments that attempt to capture a universal measure of intelligence or aptitude (e.g., IQ test, SATs), cognition assessments in these large, longitudinal surveys may have questions that are unintentionally biased towards certain groups. For example, especially for aging adults, an item that asks people to count backwards from 100 by 7s may be easier for men compared to women because of gendered expectations and institutional steering: men have historically been stereotyped as being “good at math” and encouraged to build that skill, while women have not been encouraged in the same way. The item would be considered biased towards men if, on average, despite both groups having similar cognition, men answered the question correctly and women answered incorrectly. The purpose of the following dissertation is to shift the focus towards the cognition assessment itself, asking to what extent does bias in the measurement of cognition contribute to the persistent disparities in cognition across intersectional race-gender groups.The following dissertation fits into larger historical and current efforts to accurately measure cognition across diverse populations within the United States and cross-nationally. Of particular interest in this study is the measurement of cognition in the Health and Retirement Study (HRS): previous research has established that the cognition assessment in the HRS exhibits measurement invariance separately across race—between non-Hispanic Black, non-Hispanic White, and Hispanic—and gender—between men and women. However, researchers have not examined to what extent measurement invariance holds at the intersection of race and gender in the HRS. In line with the HRS’s efforts to improve the accuracy of the cognition assessment, the following analyses further contemplates measurement invariance by intersectional identity, attending to potentially neglected disparities in the measurement of cognition. In Chapter 2, I perform a multigroup confirmatory factor analysis (MGCFA) of cognition in a cross-sectional sample of the Health and Retirement Study (HRS) across four intersectional race-gender groups, which estimates the presence and degree of bias in the cognition measures. I find evidence for measurement non-invariance, specifically that the underlying cognition factor is not captured equivalently for Black women and White women, as compared to Black men and White men. These findings demonstrate the importance of ensuring measurement invariance for latent variables across intersectional groups.In the broader social sciences and specific field of demography, scholars have endeavored to understand and explain differences in cognition trajectories—and not just in cross-sectional samples—by race and gender in old age. In Chapter 3, using data from 1996 to 2018 in the HRS, I calculate a second-order latent growth model, which permits me to calculate the amount of measurement invariance in cognition over age. In addition, I borrow from demographic research on lifespan variation to conceptualize and hypothesize how variation in cognition trajectories behave across intersectional race-gender groups. Adjusting the measurement model in the latent growth model in the cognition age-trajectories of Black women, Black men, White women, and White men results in a 50% reduction in the gap between Black women and White men. Additionally, I found that there was greater variation in the trajectories of Black women and Black men even when measurement bias was removed, suggesting that White women and men experience greater certainty in their cognition trajectories in old age. This novel modeling strategy demonstrates the importance of assessing and correcting for bias in the measurement of cognition trajectories across race-gender groups and considering the often-overlooked influence of variation in cognition trajectories to race-gender disparities in cognition.Lastly, in Chapter 4, I add to the extant research by removing measurement bias from cognition age-trajectories while simultaneously controlling for three relevant covariates. Using data from Health and Retirement Study (HRS), I calculate an unbiased second-order latent growth model while also controlling for allostatic load to measure individual levels of chronic stress; years of education; and the average years lived in the Southern United States. Education had the largest influence on the level, but not change, of cognition across all four intersectional race-gender groups. When I control for all three variables, the disparities in cognition trajectories between Black women, Black men, and White men are substantively eliminated, while White women’s cognition remains at a comparatively high level. These analyses illustrate the potential explanatory power of measurement invariance analyses for explaining the remaining gaps in cognition trajectories.Taken together, these three empirical studies advance scholarship on measurement invariance in cognition assessments by utilizing an intersectional lens to reveal sizeable bias across race-gender groups that would have gone otherwise overlooked. These analyses consider measurement invariance within an intersectional framework, which is important to examine if we, as a research community, intend to understand the reliability and validity of cognition measurement and improve the lives of aging people in the United States and globally. These results suggest that scholars who study disparities in cognition—or any complex, latent facet of health and well-being—must consider how bias may be contributing to disparities.