When screening candidates, organizations often give preference to certain applicants on the basis of their familial ties. This “legacy preference,” particularly widespread in college admissions, has been criticized for contributing to inequality and class reproduction. Despite this, studies continue to report that legacies are persistently admitted at higher rates than non-legacies. In this article, we develop a theoretical framework of three distinct sense-making strategies at play when decision-makers screen applicants into their organizations—the meritocratic, material, and diversity logics. We then apply this framework to investigate how legacy preferences either support or undermine each organizational logic using comprehensive data on the population of applicants seeking admission into one elite U.S. college. We find strong support for the material logic at the cost of the other two organizational logics: legacies make better alumni after graduation and have wealthier parents who are materially-positioned to be more generous donors than non-legacy parents. Contrary to the meritocratic logic, we find that legacies are neither more qualified applicants nor better students academically. From a diversity standpoint, legacies are less racially diverse than non-legacies. We conclude with a discussion of our study’s implications for understanding the role of family relationships and nepotism in today’s organizational selection processes.

In response to Nikolaos Ignatiadis and Stefan Wager’s paper on empirical Bayes procedures, this article evaluates the proposed tools for constructing confidence intervals and proposes new areas where the tools may be useful. The tools developed by Ignatiadis and Wager could be useful in the analysis of random experiments, which are frequently used in pharmaceutical trials and in the private sector. These tools could also help better create value-added models in education, which calculate the impact of individual teachers on student performance. Though these applications are promising, there are still issues to solve, such as adapting to unit-specific covariates.