Do UC Berkeley readers score applications differently?

BBerkeley's admissions readers can differ sharply in how they rate applications, and California's State Auditor documented it in report 2019-113, a 2020 review of admissions at several UC campuses. Most coverage of that report focused on applicants admitted through side doors. For families applying today, the more useful finding was about ordinary readers doing ordinary work, the people who will open next year's files.

Which reader opens a given file is a matter of chance, and the audit suggests that chance is larger than anyone in the process likes to admit. The numbers are worth seeing in full.

Differences among Berkeley reader ratings

The auditor compared the ratings of three actual Berkeley readers. One gave Strongly Recommend to 35 percent of the applications they read. Another most often chose Recommend. The third gave Do Not Recommend to 80 percent. The report concluded that applicants' ratings were highly dependent on who evaluated them, and that the same concern applied at UCLA and UC San Diego. It also found that Berkeley and UCLA lacked clear criteria for choosing whom to admit among applicants with similar ratings.

The audit described individual files that make the point sharply. A low-income student at a disadvantaged high school, in the top nine percent of the class and recommended by both readers, was denied. A staff member's child and a donor's child with two Do Not Recommend ratings each were admitted. The improper admissions drew the headlines. The variance is what touches every applicant.

A generous reader may infer what you meant, and the reader who declines four files in five will not.

Why one reader matters so much

Berkeley's own 2005 Hout report found that the read score, the average of two readers' scores, completely determined about 89 percent of admission decisions in the cycle it studied. The auditor's UCLA figures show how far a single reader moves the outcome: in UCLA's College of Letters and Science, a file rated 2 by both readers had a 93 percent chance of admission, and a 2 paired with a 3 had 31 percent. Each UC campus reads independently, and the pattern was common to the campuses reviewed.

Write for the harshest likely reader

You can't choose your reader. You can make the file hard to misread. Put your strongest evidence where a hurried reader will see it: in the first lines of each PIQ, in the opening words of each activity description, and in plain statements of context. Let the same strengths appear in more than one place, so a reader who skims the activities meets them again in an essay. Don't rely on implication. A generous reader may infer what you meant, and the reader who declines four files in five will not.