With the increasingly competitive nature of college admissions and the less predictable results that many schools are seeing, we thought it would make sense to discuss what we believe is one of the most important and least understood drivers of the admissions process.
Like most parts of society, technology has played a larger role in college admissions over the past several years, a trend we expect will continue as AI becomes increasingly mainstream while most colleges continue to see increasing yield pressure. Because of pressured yield rates, colleges are increasingly turning to advanced enrollment management systems and enrollment consultancies to help them make better decisions about their applicants with an eye to understanding which applicants are most likely to attend their schools. Enrollment management systems and consultants generally use predictive modeling to help schools defend and improve their yield rates by assisting them in selecting the applicants most likely to actually attend their schools.
The statistics regarding the adoption of predictive modeling in higher education may come as a surprise to many. According to Ruffalo Noel Levitz, over half of colleges and universities in the United States utilize statistical modeling to forecast the chances of prospective students enrolling. Furthermore, the American Association of Collegiate Registrars and Admissions Officers (AACRAO) shared survey findings indicating that 75% of colleges (90% of private institutions) take enrollment projections into account when making admission decisions. In brief, through predictive modeling, colleges are actively monitoring a wide range of predictive data sources to continuously evaluate an applicant's likelihood of enrollment, financial capacity, and prospects of graduation.
With the prevalence of predictive modeling, the question that applicants should be asking is what the models are based on. Interestingly, from an admissions standpoint, grades are generally a small part of the equation. While schools seldom disclose the details of their models, the most common data feeds that inform their models are population and demographic data as well as financial aid information. Web page interaction, SMS interaction, social media interactions, school visits, and direct interactions with school representatives are also generally in the models used by schools. Of greater interest is the somewhat frequent inclusion of personality characteristics from the words and linguistic characteristics of essays, emails, and letters of recommendation. Jon Boeckenstedt, the Vice Provost of Enrollment Management at Oregon State, whose web blog we would recommend, calls the vendors working with colleges the ‘admissions industrial complex’ when he discusses their growing influence and the ever-increasing number of data feeds in most models. Some additional examples of some more unique sources of information used in predictive models are Ithaca College looking at the number of posted photographs to determine applicant interest and the University of Denver looking at when students join their online community.
From our standpoint, the existence of predictive models means that while applicants are not always aware of it, the admissions process is far longer than most believe. This is because much of the information used by predictive models is generated long before an application is submitted. We would encourage students to understand the inputs of the predictive models generally used by the colleges they care about and make sure they address them in their interactions with the schools. While our work often allows us to know the specific enrollment management software and predictive models used by schools, the above-mentioned data feeds exist in most models. The same applies to the personality traits that schools value; while we have a good understanding of the specific traits many schools value, there are also general characteristics all schools tend to favor. In the new world of college admissions, students who start earlier and communicate more are generally advantaged.
Comments