Essay content and style are strongly related to household income and SAT scores: Evidence from 60,000 undergraduate applications

Abstract

There is substantial evidence of the relationship between household income and achievement on the standardized tests often required for college admissions, yet little comparable inquiry considers the essays typically required of applicants to selective U.S. colleges and universities. We used a corpus of 240,000 admission essays submitted by 60,000 applicants to the University of California in November 2016 to measure relationships between the content of admission essays, self-reported household income, and SAT scores. We quantified essay content using correlated topic modeling and essay style using Linguistic Inquiry and Word Count. We found that essay content and style had stronger correlations to self-reported household income than did SAT scores and that essays explained much of the variance in SAT scores. This analysis shows that essays encode similar information as the SAT and suggests that college admission protocols should attend to how social class is encoded in non-numerical components of applications.

Can Standardizing Applicant High School and Neighborhood Information Help to Diversify Selective Colleges?

“…One of the challenges to increasing socioeconomic diversity across selective colleges may be that those institutions have historically evaluated applicants on criteria, including college entrance exam scores, extracurricular activities, and personal essays, that often, in part, reflect access to prior opportunities (Alvero et al, 2021; Rosinger et al, 2021). Furthermore, admission officers may lack the tools and resources to easily evaluate applicants in the context of their home, school, and neighborhood environments.…”

Multilingualism and Mismatching: Spanish Language Usage in College Admissions Essays

“…The uniform structure of the format of the college admissions essay, where every applicant has to respond to the same prompts with the same limits on word count for the same deadline, lends itself well to macro analyses of the actual text submitted by students. Recent studies using methods from machine learning have found that computational representations of admissions essays are strong predictors of applicant income [7,8], gender [8], and SAT score [9]. These studies were all similar in their "bag of words" approach to the data, where machine learning models were used to analyze word co-occurrence patterns across all of the documents.…”

“…Self-reported household income (shortened to "income" hereafter) was also provided. Students reporting household incomes below $10,000 were suspected to have done so erroneously and excluded from analyses of income (see [9]). See Table 1 for summary statistics of the data.…”

Holistic Admissions in Higher Education

“…There has been considerable concern around standardized tests in college and graduate admissions, yet there is growing evidence that many other metrics (e.g., letters of recommendation, essays), perceived as more equitable demonstrations of applicants' qualifications, are equally, if not more, prone to influence of family resources and income and can differentially impact applicants of different racial and gender groups (e.g., Alvero et al, 2021;Dalal et al, 2022).…”

“…Recent research shows that essays are not immune from the influence of family income either. Researchers at Stanford University (Alvero et al, 2021) found that essay content is strongly correlated with students' self-reported family income, and the correlation is even stronger than the association between SAT scores and household income. Essay content refers to the topics, diction, grammar, and punctuation in an essay.…”