National Collegiate Honors Council


Date of this Version

Fall 2017


Journal of the National Collegiate Honors Council, Vol. 18, No. 2 (Fall/Winter 2017), pp 213-234.


© Copyright 2017 by the National Collegiate Honors Council


Most enrollment management systems today use historical data to build rough forecasts of what percentage of students will likely accept an offer of enrollment based on historical acceptance rates. While this aggregate forecast method has its uses, we propose that building an enrollment model based on predicting an individual’s likelihood of matriculation can be much more beneficial to an honors director than a historical aggregate forecast. Many complex predictive analytics techniques and specialized software can build such models, but here we show that a basic approach can also be easily accessible to honors directors where a small amount of data collection and basic spreadsheet software allow them to capture most of the benefits without needing the skills of a data scientist. The first step comes in understanding the difference between a forecast and a prediction. A forecast is an estimate of a future event, generally in aggregate form. For example, today I might forecast that our ice cream store will likely sell 1,000 scoops of ice cream based on weather, time of year, day of the week, and regional events—all useful information for staffing and inventory management as well as profitability analysis. Historically, an honors administrator might use this approach to predict the total number of students matriculating to the university or to an individual program. However, with predictive analytics one can acquire even more detail that could be useful in a setting like an honors program where not just the total number of “customers” matter but which ones will create a well-rounded, diverse honors program with students from multiple backgrounds (Siegel). In the ice cream case, a predictive analytics example might predict not just how many total ice cream scoops might be sold but how likely each individual is to buy ice cream. Deeper analysis might predict the type of ice cream, time of day customers might come, and how frequently they might visit the store. Predictive analytics might also lead to prescriptive analytics, where you learn what might be done to persuade someone who was not planning to buy ice cream to do so, e.g., what it might take to change a consumer’s mind so that she will buy ice cream today or how we can we get her to buy two scoops instead of one or to bring a friend. This type of predictive and prescriptive analytics has helped many organizations improve their efficiency and effectiveness (Siegel), and we believe that honors directors can also use it. In this approach, each potential honors student would receive an individualized probability score reflecting his or her likelihood of accepting an offer of admission. This score could still be aggregated into a direct forecast of how many students would likely attend, but it would also show the likelihood that any individual student would attend. The scores could predict how many from a certain group (e.g., science majors or Hispanic students) are likely to attend. This information could help strategically determine scholarship offers as well as the staff’s time commitments to recruitment and follow-up activities.