Most universities have offices of academic assessment or variations thereof. Within these, the goal is to analyze the success of their current cohorts in an attempt to boost student retention, graduation, and placement rates. One of the greatest concerns in regards to academic persistence is what is called "retention rate", which is defined as whether or not the student returned to the institution after the first year. Universities like to examine this factor because the ideal goal is to have more students be retained. This often causes universities to classify students as "high-risk", "at-risk", and "traditional".
But what are the predictive factors behind retention rate? Universities attempt to implement First Year Experience surveys to gauge student probability on retention, but self-assessment on likelihood may not always be the best method to predict student retention. Instead, using data taken from the Department of Education's College Scorecard 2021-2022, as well as a 2016 case study done in the University of Texas System, we can attempt to find possible trends for student retention. The goal, of course, would be to encourage universities to look at the greatest causal factors and implement programs to mitigate risk.
In all businesses, predictive modeling is often used to make business decisions based on trends. The goal is always to make enhancements to current systems by finding new causal factors and adjust accordingly based on probability. Businesses can then provide better services to their clients and develop more means to engage with them: stronger communication, enhance models, and create a more enriching experience.
The possibilities are endless when it comes to variability in academic persistence. Using predictive analytics to pinpoint the students that need the most help is based on a variety of factors, but there were a few that I knowingly overlooked. For example, the College Scorecard used only took the "most recent" dataset, but it technically goes back as far as far as 1996. Perhaps year matters more? Also, I combined L4 schools (that is, institutions that are less than four years) with the regular four-year institutions. There is a likelihood that this institution difference affects student retention. When comparing "loans", I had a tendency to loop in those receiving Pell Grants with those receiving loans in general. It is more than likely that the two affect retention differently. Institutional biases themselves are a matter as well, such as "self-supplied scores/rates", as well as lack of values to begin with. It'd be interesting to seperate the values by what is supplied if completion is used instead, such as by demographics or by those who received PELL grants.
The Texas case study is also not recent, and it's likely trends have changed in regards to the strength academic achievement has on retention.
Technical Documentation: https://collegescorecard.ed.gov/assets/InstitutionDataDocumentation.pdf
US Department of Education: https://collegescorecard.ed.gov/data/