Follow-up to UC Irvine Mess - The Problem of Predicting Yield

[Scipio]

LA Times:

http://www.latimes.com/local/lanow/la-me-uc-enrollment-predict-20170807-story.html

Excerpt:

"Not that the carefully crafted forecasting model can account for all teen decisions. When Whittingham asked students why they chose UC Santa Cruz, one told her “oxygen levels,” and another said the campus was close to Seattle.

“One of the things you can’t lose sight of is that literally we’re predicting the behavior of 17- and 18-year-olds,” she said. “I can’t predict my own son’s behavior.”

[lookingforward]

Good article, though short. Part of me wants to blame enrollment managers; after all, they’re paid the big bucks to work this issue. Another part wants to blame the schools for, eg, not tracking, much earlier, the current students’ intentions to reenroll, maybe slowing down admits closer to the final commit date. Or offering 2nd semester start dates or gap years.

[ucbalumnus]

From the article:

@Teresa Watanabe and Rosanna Xia in Los Angeles Times

At Santa Cruz, officials used to predict group behavior. Now their model looks at each applicant, calculating the likelihood that a student will come based on such variables as their grades and test scores, zip codes, intended majors and campus visits.

Those considered more likely to attend UC Santa Cruz generally live closer to campus, visit more frequently before deciding and have a weighted grade point average that is somewhere around 3.9. Students planning to major in engineering, film and computational media are more likely to attend than those focused on music or feminist studies.

Surprising that UCSC did not predict yield based on these factors before its 2015 overenrollment incident. When UC Statfinder was up and publicly available (with data from the 1990s to 2009), anyone could use it to see that yield to a given campus was higher for lower stat admits (who may have seen that campus as a “reach”) and lower for higher stat admits (who were more likely to have several other choices).

[bluebayou]

At Santa Cruz, officials used to predict group behavior.

What the heck does this mean? Adcoms sitting around singing Kumbaya?

[coolweather]

Maybe waiting list is the solution.

[lookingforward]

Not kumbaya, but some magic insight found in some, they’d like us to believe, science. Enrollment managers, not adcoms. You know it happens through all aspects of our lives. Predictions about behavior based on variables.

St Marys may be part of a group sharing these folks or contract out. But a UC making this mistake is surprising. As is apparently being blindsided, not prepared.

[Ynotgo]

USC has a certificate program in enrollment management. https://cerpp.usc.edu/certificate-program/
Maybe more colleges should sign up for the program.

[ucbalumnus]

https://www.universityofcalifornia.edu/infocenter/freshman-admissions-summary

On this page, you can select “Freshman Yield Rate”, a specific campus, and “HS GPA”. You will see that yield goes down as HS GPA rises.

If UCSC ignored such information (and how other applicant characteristics correlate to yield) before 2015, that seems very surprising.