The more I learn about colleges admissions, the less certain I am that colleges really are all that good about predicting their so-called “yield.” As a parent, I had just assumed that the admissions office leveraged historic trend data along with statistical modeling/predictive algorithms to project how many admits would enroll.
But looking over data for just the last 3 cycles, I see that year over year yield rates can vary significantly, even for well-established colleges that admit large freshmen classes each year. Anyone have any insight or knowledge about how yield modeling actually works inside an admissions office?
I don’t have any inside knowledge about yield management or algorithms that colleges use but I will say that the last three cycles are probably not good indicators given the situation with Covid. Starting in 2020 with effects being felt thru this current cycle there’s been a lot of upheaval because students took gap years and leaves of absence and that affected the class sizes and yields for the subsequent incoming classes.
In addition, in a more long-standing situation, the common app continually creates issues since students can apply to so many schools making it more difficult for schools to ascertain the likelihood a student is actually going to attend.
The last 3 cycles had unpredictability due to COVID-19 and related issues.
Colleges presumably assign a yield probability to each admit based on its model taking into account whether the student is “overqualified” (less likely to yield) or is reaching (more likely to yield) for the college or specific major, probably various demographic factors (SES, region, gender, race/ethnicity, etc.), and financial aid related characteristics. The college may also choose to influence the admit with scholarships or preferential packaging of financial aid (how much $ is needed to raise the admit’s yield change by X%?).
This does not mean that their models will be accurate, so colleges have incentive to use ED more (to lock down a greater portion of students early) and use waitlists more (admitting slightly fewer to avoid overenrollment if an overyield occurs, but being able to admit from the waitlist if actual yield does not fill the class).
I agree covid has messed with the models and hoping they are back on track this year.
Many colleges hire external consulting companies to create and manage the predictive analytic models. You can go on the websites of those companies and sometime find white papers and even videos that discuss the process. One LAC that I’m familiar with had 24 variables in their model (calculated an individual yield for each applicant).
The yield model is just one of several models used in the process. Once their is a yield estimate, that value feeds into the net revenue model (net rev = tuition revenues less fin aid) to see where that is based on predicted yield.
If that result is short of the net revenue goal, then need aware schools start taking out admitted students with need and replacing them with students with less or no need need until they hit the target. There was a NYT article a few years ago that shared some of these details of a similar process at Trinity College.
Most enrollment mgmt peeps are working towards a net revenue goal. Some might also have a yield goal too, but net revenue trumps that.
Episode 308 of Your College Bound Kid has a good explanation of the basic approaches to yield modeling. Apparently there are dynamic models (based on a student’s demonstrated interest or lack thereof) and static models (based on demographics), and different schools blend them in various ways. One school indicated off-the-record that it purchased one model of each type for the same cycle and intends to compare results. Another school admitted that one of its models was off by 20% last year.
Which type of model does the student’s application strength relative to the admit pool factor into? It is (or should be) common knowledge that “overqualified” students are more likely to have other attractive admission or scholarship offers and therefore less likely to yield than students for whom the college is a reach.
Schools have been getting this wrong for decades - Tufts over-accepted in the mid 70s and had to put up students in a hotel for a few years.
The common app may have been the first disruptor of this millennium, but TO policies -especially post-covid, have really thrown everyone for a loop.
Many schools have responded by accepting more of the class in binding rounds (and making it easier for family to calculate expectated FA).
They have always maintained models and add things into it like demonstrated interest. The WL now plays a pretty big role at many in filling out the class.
It’s been several weeks since I listened to the podcast, but in my recollection, the strength of grades/scores relative to the pool would likely be weighted more heavily in static models, as it’s a background fact about the student rather than an action taken by the student. But it sounds like different models mix and match dozens of factors, so it’s hard to say how any school will weigh anything in a particular year.
One of the interesting things about the podcast to me was the divergence from the typical CC view of yield protection. In these forums, you often see a more straightforward take along the lines of “Student A has stats well beyond the published range for School X, so of course they were waitlisted. Yield protection.” Based on information the podcaster was able to glean from industry connections on both sides of the fence, it sounds like the reality is more complicated. For example, if a student’s high school/neighborhood/region, or family income level, or choice of major (or more likely all of these combined) is similar to those of other students who have chosen to enroll in the past, that may counterbalance lower stats and/or lack of demonstrated interest to some extent. Conversely, if these (and many other) background factors in aggregate indicate a low likelihood of enrollment based on past results, that may well be a bigger factor in deciding to waitlist or deny than just perceived likelihood of this particular student choosing a “better” school.
As with any proxy measure, proxy measures of applicants’ interest become less reliable as people try to game them, so colleges may change what proxy measures they use if they believe that people are gaming the ones they have been using.
Pretty much anything describing a correlation to a group is profiling that group.
Surely this is an ongoing battle for AOs at every school where interest is considered. I would guess AOs probably get a better sense for this over time, but a determined student who rigorously checks all the right boxes may get full credit for doing so. That’s one good reason to use models with lots of varied factors in play.
Sure, you could call it that.
It’s become a financial and/or competitive imperative at many schools. For those who are interested in attending such schools, this is going to be part of the process, for better or worse. I doubt the colleges themselves really love it, especially when a model proves to be inaccurate.
Remember also students who didn’t get in, or who are considering Tulane: there is the IDT (interdivisional transfer) from the Tulane School of Professional Advancement to NTC with fulfilling the 6 core requirements:
Foreign Language
Math
Science
English (English 1010 compulsory)
Social Science
Humanities
With a C (or C-) or better
At least two (2) semesters in SoPA (School of Professional Advancement)
At least 18 Tulane credits
Upon completion of such, you will require a deans letter of approval from SoPA, just like going from NTC to SoPA, and you will have completed IDT to becoming a full fledged Tulanian!
Google: Tulane interdivisional transfer, it’ll show up under the School of Professional Advancement.
Perfect way to circumvent the normal freshman undergraduate route if you have a year or three of a break from school, and last I saw, you need a 2.0 GPA to get in. Not necessarily the 1-3 years break from school. Also, you don’t need to break from school, all you have to do is go into the SoPA and work on the transfer process.
They really know your background outside of what YOU tell them.
Good luck everyone, especially if you’re financially hurting, SoPA may be the PERFECT way to start!