Yu and Elwert Develop Approach to Better Understand Causes Behind Group Disparities

Profile pictures of Ang Yu and Felix Elwert
Ang Yu and Felix Elwert

Understanding roots of social inequalities

Social scientists have long worked to better understand the roots of societal inequalities between groups. For example, why do members of one group on average do better in school, or why does another group have a lower average income? There is a large literature that has tried to find the sources of inequality and there are formal mathematical frameworks going back to the 1950s that attempt to identify these sources.

With support from an Institute for Diversity Science Seed Grant, sociologists Ang Yu and Felix Elwert have developed what is perhaps one of the most significant new statistical approaches to study of inequality in the last 50 years. Until now, scientists have focused on two mechanisms, by which some “treatment” — for example a vaccine, a college education, or a tax cut – might increase or decrease inequalities between two groups. The first is that members of one group get more of the treatment  for example, they get more education. The second is that members of one group benefit more from the treatment — for example, they benefit more from the education they get.

Yu and Elwert uncover “new” driver of inequalities

But Yu and Elwert have uncovered a third mechanism that can be a driver of inequalities between groups: differences in who within each group receives treatment. If group members who benefit from the treatment most were more likely to receive the treatment in one of the two groups, then there will be a disparity between the two groups even if (a) the same percentage of members in the two groups received the treatment and (b) the members of both groups benefit from treatment to the same extent on average.

Consider the following example: Imagine you find a difference in COVID mortality between two groups, which could be two ethnic/racial groups or two US states. This difference could be due to one or more of the following reasons: (1) a larger percentage of members of one group received the COVID vaccine, (2) the COVID vaccine was more effective in one group than in the other, and (3) the COVID vaccine was distributed differently in the two groups so that those who benefited the most from the vaccine — e.g., the elderly and those with preexisting conditions — were more likely to get the vaccine in one of the two groups but not in the other.

As another example, let’s say that two groups on average both receive the same education and benefit from that education at the same level. The average effect of education is the same for both groups, but in each group there are some people who benefit from education and some who don’t. If in one group more people who don’t benefit from education receive that “treatment,” that would drive differences between the two groups.

Approach allows causal claims about factors contributing to inequalities

Identifying this third mechanism is just the beginning, however. Yu and Elwert also developed statistical methods that allow them to estimate the impact of each of the three mechanisms described above using empirical data. Put another way, these new methods allow researchers to make causal claims about the distinct mechanisms by which a particular “treatment” — an intervention, a policy, a societal phenomenon — contributes to a specific observed inequality between groups.

As part of this effort, Yu developed an open-source statistical software package that is freely available to other researchers. To date, this software package has been downloaded over 6,000 times, reflecting the significant contribution that this IDS-supported research has for the study of inequality and disparities.