Dynamic Modeling for Higher Education Planning

Thoughtful planning has never been more critical for higher education institutions. The COVID-19 crisis coupled with the already difficult enrollment and financial challenges facing higher education is leading many institutions to engage in a deeper and more far-reaching planning exercise than has been the case for many years. Institutions must take a deep look at their strategies, plans, and programs, and must consider significant changes, not just minor adjustments (Planning and Budgeting Post COVID-19).

Organizations are complex systems. It is generally difficult to tell a priori the result of making significant changes in strategy or operations. Systems theory—or the systems approach—provides a lens we can use to examine the implications of changes made in any system. Systems theory looks at all systems as comprised of five types of elements (adapted from Ariav and Ginzberg, DSS Design: A Systemic View of Decision Support, Communications of the ACM 28 #10, 1985):

• an environment – the set of entities and conditions outside of the system boundary that affect the system or are affected by it;
• a role, function, or objective that represents the system’s intended impact on its environment, e.g., “production” of educated graduates by a college;
• a set of components, the functional building blocks of the system within its boundary;
• the arrangement of those components, the links among components and between components and the environment; and
• resources, the external elements that are used or consumed by the system as it operates.

Systems thinking requires that we take a holistic view of these five types of elements, how they are structured, and how they interact in order to understand the system and its behavior. A system is an artefact that we create; we define the boundary between system and environment as well as the level of detail of components and arrangement as is appropriate for the issues we are studying and the decisions to be made. The components making up one system might themselves be viewed as systems at a more detailed level.

Colleges and universities are complex, dynamic social systems. Their components and arrangement can be described as collections of interacting subsystems (or feedback loops). The loops can be either reinforcing (positive feedback) or balancing (negative feedback). A simple example of a reinforcing loop involving faculty members might be as follows:

As Faculty Workload increases, the Quality of the Faculty Experience decreases. As the Quality of the Faculty Experience decreases, the Faculty Attrition Rate increases, which decreases the number of Faculty Members, thus increasing the Faculty Workload. Thus, without considering any other changes that may be made, an increase in the Faculty Workload will work through this loop (with some delay) to further increase Faculty Workload, turning this feedback loop into a vicious cycle that can destroy an institution. This is an example of a reinforcing (amplifying) loop.

We can also consider a balancing loop involving faculty members:

In this feedback loop, as Faculty Workload increases, Faculty Hiring will also increase. Increased Faculty Hiring will increase the number of Faculty Members, thus decreasing Faculty Workload. So, in this loop, an increase in Faculty Workload will work through the loop to ultimately reduce Faculty Workload. It is a balancing (self-correcting) loop.

Looking at these two loops, it is apparent that they are not independent. Both the number of Faculty Members and Faculty Workload appear in both loops. Putting the two loops together, we get a more complex faculty subsystem:

Now, an increase in Faculty Workload will tend to further increase Faculty Workload through the mechanism of the reinforcing loop (Attrition) but will tend to decrease Faculty Workload through the action of the balancing loop (Hiring). The outcome of this system will depend on which feedback loop dominates. If Faculty Hiring is robust and quick to respond to an increase in Faculty Workload, the balancing (Hiring) loop will be able to lessen and perhaps overcome any potential negative effects due to the vicious cycle of attrition. But, if Hiring is slow or weak in responding to increases in Faculty Workload, the reinforcing (Attrition) loop may have the dominant effect.

In general, the subsystems comprising an organization, a college or university, for example, are not independent of one another. Changes introduced in one subsystem will lead to reactions in other subsystems, which may eventually propagate through the system and cause a further reaction in the first subsystem. These complex interactions and delays inherent in the structure of the system are what make it so difficult to predict exactly the impact of any change in policy. Jay Forrester, a professor of management and electrical engineering at MIT who developed the field of Systems Dynamics in the late 1950s, referred to this phenomenon as the counter-intuitive nature of social systems.

Systems Dynamics enables us to understand and predict the behavior of complex social systems—e.g., colleges and universities—through feedback modeling. We can build computational models that reflect the structure and dynamic performance of the system, and we can experiment using the model to try out different policies or decisions to understand what their impacts will be. A rigorous Systems Dynamics model can be built one subsystem at a time, and the subsystems can then be integrated into a comprehensive model of the organization. Planning based on this type of dynamic model is superior to that based on traditional, static spreadsheet modeling because the feedback models take into account the effects of numerous loops in the system.

Colleges and universities are, of course, more complex than the simple example of faculty outlined above. Pavlov and Katsamakas have developed a Systems Dynamics model of an undergraduate college and have used the model to explore the impact of various policy changes that could be implemented in the light of declining enrollment, a situation quite common at many small and mid-sized colleges today. Their model structures the college as four interacting feedback loops—a Student loop, a Faculty loop, a Facilities loop, and a Financial loop. They use the model to explore three strategies a college might adopt to deal with declining enrollment:

• a “do nothing” strategy, i.e., taking no specific action to deal with the enrollment decline;
• a “reduce operating costs” strategy that cuts the number of faculty members employed; and
• a “revenue increase” strategy that attempts to attract more students through improvement to the college’s physical facilities.

Using the model, they test all three strategies reflecting different decisions that might be taken by a typical college. The model’s output includes the size of the student body, the size of the faculty, and various financial parameters including the operating surplus (or deficit), expenditures per student, impact on the endowment, and the college’s debt. The model can be run to simulate multiple time periods, thus showing both short- and long-term impacts of the chosen policies. The results of their simulations show some differences among the strategies in the short run (2 years), but substantial differences in the longer run (10 years). In the long run, the “do nothing” strategy results in a smaller student body, high cost per student, and a high operating deficit financed by increasing debt and drawdown of the endowment. The “reduce operating costs” strategy also results in fewer students but shows a smaller deficit financed primarily by the reduction in faculty size. The “revenue increase” strategy results in more students, more faculty members, and more facilities, but at the cost of an ongoing operating deficit and significant debt. Each strategy has a different impact on financial outcomes and likely on the college’s sustainability over time.

Pavlov and Katsamakas’ model is an aggregate or summary level model. It includes only a few key loops and omits many complexities of a real college or university. It treats faculty members as homogeneous, not recognizing any differences across schools, departments, etc. It does not explicitly model administrative functions and their impact on college functioning, nor the interplay of multiple schools and departments which have different operating policies. Nonetheless, experimentation and exploration of various “what if” scenarios with this aggregate model provides a glimpse of how different management decisions would impact critical outcomes. The insights provide a good starting point for a college’s planning efforts. They help identify where the model can be calibrated and elaborated, adding detail to match an institution’s structure and policies, so it can be used for more detailed exploration of alternative strategies and policies. This further exploration can test the sensitivity of various outcome variables—costs, revenues, student enrollment, etc.—to specific policy decisions. A college or university might also build more detailed models to understand the relationships among academic programs—e.g., how enrollments in one unit can affect those in others—or between academic and support programs. This type of dynamic model can help decision makers understand how the parts of the institution will function together and result in outcomes that are often not apparent or easily predictable. With this understanding, decision makers can make far more effective policy choices.

This is part of a series of blog posts meant to outline strategies for dealing with the mid-term and long-term implications of the Coronavirus crisis and our changing higher education environment. Let us know what questions and challenges you have about the future either by leaving a comment below or by contacting our principal consultants directly.