In our paper, we examine a dataset of college application essays submitted by Latinx applicants to a selective public university system in 2015. Each essay also includes individual demographic and high school metadata. We limit our analyses to applicants from a large metropolitan area yielding a sample of 13,067 applications. We examine application patterns to system campuses by schools and apply structural topic modeling (STM) to the essay corpus to examine how essay topics at the school-level correlate with application patterns. Lastly, we conduct a focused analysis of 4th quartile SES students at low- and high-SES schools to examine how schools differentially affect the essay content of low-income Latinx students. Structural topic modeling indicated emerging patterns of stratification. For example, high income applicants to the most selective campus were more likely to use topics with terms like “science; biochemistry; and astronomy.” Low-income applicants unlikely to apply to the most selective campus wrote about “mom, dad, and grandma.” Essays of students from 3rd Tier schools were more likely to reference “work, job, and hard.” Fuller analyses will detail how semantic variation in essays correlates with social variation seen in school characteristics and application patterns.