Scheduling and Context

If you haven’t seen Tim Burke’s recent piece about assembling a class schedule for an entire college, it’s worth the read. At the risk of being accused of one-upmanship, though, he’s playing the game on the easy setting. Swarthmore has a pretty consistent enrollment—selective admissions among a surfeit of applicants can do that—and it’s small. It only has one location, and it’s overwhelmingly devoted to in-person classes.

When none of those apply, scheduling is that much harder.

In a community college context, scheduling starts with an estimate both of total enrollment and of the distribution of that enrollment across formats (in-person, hybrid/remote, online). In a multicampus institution, it also involves projecting the distribution of enrollments across locations. In the mid-2000’s, for instance, I could predict with some confidence that any given year’s enrollment would be 2 percent–ish higher than the previous year’s. Over the last five or so years, we’ve had to predict the magnitude of declines. Managing growth is much easier; you add sections when you have the space and staff to add them. Managing decline is trickier.

It’s trickier in part because of the inverse of Say’s law; lack of supply can create lack of demand. Enrollments at a branch location are struggling, and money is tight, so you close some of the smaller sections there. That makes it harder for students to get full schedules there, so they start abandoning what’s left. What starts as fiscal responsibility can quickly lead to a death spiral powered by self-fulfilling predictions. But knowing that, by itself, does not provide the resources to run a lot of small sections. And knowing the difference between priming the pump and throwing good money after bad is as much art as science.

The pandemic complicated matters by accelerating the move to online/remote delivery. Now it’s not enough to know you need to run 20 sections of Intro to Hypothetical Studies in the fall. You also have to predict how many should be in person as opposed to online or hybrid.

Admittedly, asynchronous online classes help avoid some of the cases Burke writes about, in which classes that the same students need crash into each other. That’s helpful. But it can create staffing issues when some faculty in a given department are conspicuously less comfortable and/or effective in one format than the other.

Different colleges handle course allocation differently. At the community colleges at which I’ve worked, the registrar’s office prepares a draft course schedule based on enrollment projections from institutional research. The academic affairs leadership looks through it and makes changes based on strategic directions, known personnel issues or cost. Then a course schedule with blank names goes out to the departments. Each department fills in the faculty names, though each has a slightly different system for doing so. In theory, they all use rotating seniority, but that gets interpreted in various ways. In theory, full-timers go first, but small departments often have superstar adjuncts they desperately want to keep, so they give those adjuncts extra deference. I’ve long advocated for an NFL draft style of course assignment, in which a department would go down the seniority list for everyone’s first course, then do it again for everyone’s second course, and so on. Some do that; others have the most senior fill out their entire schedule first, followed by the next person. That can lead to some pretty annoying schedules for whomever goes last.

As with section allocation, this process was much easier when enrollments were steady or increasing. When they drop quickly, bumping comes into play. I can feel my eye start to twitch at the mere mention of bumping.

When there’s a minimum number of students that most sections have to enroll in order to run, you can expect some sections to come in low as the semester looms. At that point, the administration has to decide when to roll the dice on more students showing up and when to pull the plug. Invariably, some on whom the plug was pulled will take umbrage and argue with great confidence that if only we had given it more time, it would have run. It’s an unresolvable argument, so you just have to make the best good-faith calls you can and accept that they won’t all be right.

Muddying matters a bit, some classes have special claims on running even if they’re small—it’s the only section of a course required for graduation, or it’s the only evening section, or every other section is already full.

One change I made at Brookdale of which I’m proud was moving up the “go/no-go” decision date. When I got there, it was 10 days into the semester. I moved it to about 10 days prior to the semester. That may still sound late, and in some ways, it is, but community college enrollments tend to cluster at the last minute. At least building in a buffer prevented the deeply frustrating scenario of canceling a class after it had first met. And faculty had some certainty before the semester started that their classes would actually run. I would rather have moved it even earlier, but in an environment of declining enrollments, I took what I could get.

The frustrating part of scheduling is that it’s never quite perfect. Cohorts don’t behave like cohorts because of withdrawals, stopouts, transfer credits and changes of major. Last year’s pattern never quite matches this year’s pattern. And there’s a necessary unknowability to some of it, give open admissions and (gestures wildly) the state of the world.

Yes, certain kinds of software could help. Some of this is algorithmic. But those models get you only so far. Managing the many variables when information is partial and emergent and you don’t have the economic buffer for many mistakes is difficult. Bash administration if you want, but these decisions have to be made. Someone has to own them.