Figure 1. Undergraduate enrollment by degree-level AI exposure

Enrollment counts indexed to 100 = academic year 2017

Vertical line at ChatGPT's release data (11/30/2022).
Exposure measure is based on Eloundou et al. (2023) GPT4 beta administrative exposure measure. Exposure is mapped onto degrees using the typical degree-to-occupation flows in the ACS across 2009-2019 for 22- to-27-year-olds, weighted by the share of graduates holding each occupation in a given year. Quintiles of AI-exposure shown as 5 (Most Exposed), 4 (Highly Exposed), 3 (Moderately Exposed), 2 (Slightly Exposed), 1 (Least Exposed).
Source: IPUMS ACS samples 2009-2019, Eloundou et al. (2023), National Student Clearinghouse Research Center. Get the data