Measuring Material Hardship among Disabled Women using Latent Class Analysis

Background and Purpose Recently, researchers have begun measuring hardship to augment poverty investigations, due to widespread agreement that official US poverty measures have limitations and new research suggests these limitations are particularly acute for women with disabilities. As compared to other US population subgroups, these women are among the most financially vulnerable: they have exceptionally high rates of income poverty, unemployment, and material hardship. Disagreement remains about how hardship should be conceptualized and measured. One popular method is to count individual hardships (e.g., hunger or eviction), creating an index that assigns each item an ad hoc equal weight; subsequently, an ad hoc threshold on this index defines whether hardship was experienced. These approaches assume that types of hardship are equivalent in their consequences, and that the severity of hardship is linear: as the count goes up, severity does as well. More sophisticated and valid measurement approaches that identify unique non-linear combinations of hardships should be found.

Methods One promising strategy for analyzing hardship that addresses limitations of previous approaches is latent class analysis (LCA). LCA is a form of mixture modeling that defines subpopulations with memberships based on probabilities observed on indicators in the data. LCA defines a latent discrete condition rather than a linear phenomenon and thus can be used to identify unique combinations of hardships. In the present study, we used LCA to define a multidimensional framework of latent classes of food, medical and housing/telephone hardship using 4 indicators each of food insecurity and inadequate health care, 2 measures of housing instability and telephone disconnection from the National Survey of American Families (NSAF). To analyze the application of this methodology, we used the NSAF subsample of disabled women (n=4105), and compared the experience of hardship defined by the LCA with that of the counting approach.

Results We identified three-class solutions for food hardship—consisting of no hardship, “worriers”, and significant hardship classes—and for medical hardship: no hardship, “postponers” of medical services, and the uninsured. Two classes of hardship were found for housing/telephone (no hardship and significant hardship). The LCA method has advantages and disadvantages relative to the count method. The count method either over- or under-estimated hardship, and because it implied that each hardship had equal importance, it did not distinguish between severity of food and medical hardship. However, the LCA overlooked those having only one hardship of each type. Because disabled women were more likely to have one hardship, the LCA may underestimate the extent of hardship among disabled women.

Conclusions and Implications Latent class analysis is a strategy for measuring material hardship that avoids many of the ad hoc decisions inherent to the count approach. Our findings revealed patterns in the types of food and medical hardships experienced, but did not count disabled women having only one hardship of any type. Further inquiry into the hardships experienced by women with disabilities are required in order to inform policymakers interested in efficiently and effectively targeting public resources to reduce need in such vulnerable populations.