Forecasting Enrollment in New Programs

It’s tough to make predictions, especially about the future. - Yogi Berra

New programs introduce complexity to your enrollment forecast. By definition, they do not have past history to rely on for a prediction – what is an enrollment professional to do? Here’s how to handle them in stride:

The most common ways of handling new programs utilize some of the following methods:

• Infer enrollment based on a comparable program, either at the same institution or another one (if there is sufficient information in the public domain). For example, if you’re introducing a Bachelor of Computer Science focused on data analytics, you could infer behaviour either from a standard Bachelor of Computer Science, or Statistics, or a hybrid of the two.
• Assume enrollment based on a series of rules. This often works better for programs that are highly structured. In this example, you could say that 90% of students in first year will enroll in all four data analytics courses required (even in a highly structured program, you’ll never have 100% of students taking the required courses at the “right” time). Then, for their fifth course, they will choose one course from a list of four courses. In the absence of better information, you can assume they’ll split somewhat evenly, with 25% of the students taking Course A, 25% Course B, and so on.
• Base enrollment on the behaviour of similar student profiles. This method projects course enrollment data by comparing students with similar profiles to the incoming class. Here you might say that Profile Student A enrolls in 3 required courses plus 3 electives in first year. _Profile Student B_enrolls in 5 required courses plus 3 electives in first year. Map these profiles onto the new set of requirements and to create projection assumptions.
• Use government or institutional funding. This works better for programs that are funded based on a specific number of students enrolled, such as in Nursing programs. If you know how much funding will be available, you can reverse engineer how many students will be enrolled in the related programs and courses.
• Use human intuition. This works better when partnering with the academic unit - the faculty who designed the program likely have an intrinsic sense of how popular each course will be. The academic unit’s estimates could be incorporated as a form of manual forecast.

All of these methods assume there is no data available. As that assumption changes, your model should learn from new incoming data. For example, by about 2 terms into a new program, you should have most of an academic year’s worth of robust data about student enrollment patterns. At that point, your estimated methods from above are no longer valid, and you should be using the actual data to inform your plans. You can also use the actual data to enhance the above methods for years further into the future. This new data can help inform how accurate your assumptions were.

The methods discussed above almost always have a wider confidence band than more robust forms of prediction. You want to be on the lookout for common pitfalls, such as:

• Programs that double in size. I once had a program that quickly went from zero students to representing over half of the university and consequently I was able to spot a formula error prior to distributing results. Have a colleague review your formulas if you can.
• Programs with no students. This suggests there is a mistake in your input data, and the model is assuming either no intakes or no retention.
• Enrollment that doesn’t align with the expectations of the faculty. This may suggest a gap in your modelling, or gaps in assumptions from the faculty.

How does your institution model new programs?