Economic Evaluation in Education. Henry M. Levin
think about adding the word per after cost to clarify what is meant. This definition is more apt because we typically do not know how effective the intervention is (strictly speaking, the intervention’s output). Number of students served—the cost per student—is unlikely to be a satisfactory measure of effectiveness. Lower cost per student is likely to be associated with lower quality and hence lower effectiveness. Indeed, effectiveness is what the analyst is often seeking to identify in relation to costs as an explicit cost-effectiveness (CE) analysis. Importantly, in the cost analysis framework, total cost is the cost for the entire population served by the program or education (e.g., a school of 1,000 students), average cost is the cost per unit of education provided (e.g., each student within the school), and marginal cost is the cost per extra unit provided (e.g., from the 999th to the 1,000th student). When referring to costs of education, it is typically cost of provision that is intended.
The terms costs and expenditures should be distinguished: They are not interchangeable. Expenditures typically refer only to dollar outlays by a specific group. For example, commentators often refer to increases in the “cost of college” when they actually mean the “price of college paid by enrollees” or “what students spent on college.” Tuition prices paid by students are not the cost of college. The cost includes all the resources used to provide college and, because of public subsidies and charitable support, student tuition fees do not pay for all these resources (Desrochers & Hurlburt, 2016). In cost analysis, we are explicitly interested in all resources, regardless of where the dollars come from or if any money was transacted. In fact, cost analysis explicitly distinguishes between costs and financing—that is, between resources used and who pays for them.
An important cost concept for educational evaluation is incremental cost. As discussed in Chapter 2, CE and benefit-cost (BC) analyses evaluate programs relative to baseline, counterfactual, or business-as-usual conditions. Incremental cost refers to the relative difference in costs (Gray, Clarke, Wolstenholme, & Wordsworth, 2011, p. 13). (Incremental cost is not the same as marginal cost because we do not know how output has expanded.) In many cases, an intervention might be a supplement or add-on to regular schooling or college programs; incremental costs are then the costs of the program. But there are other cases. One is where the intervention involves a redistribution of resources such that intervention students get extra lessons in writing but fewer lessons in math; incremental costs might be very low in this case or even negative (if writing classes cost less than math classes). Another case is where students receive progressively more intensive interventions. So, if the analyst estimates the cost of a mentoring program at $1,000 per student and an after-school program with a similar objective at $1,500 per student, the incremental cost of the after-school program is $500. Note that in each case, we assume that each student receives some regular schooling, so the cost is genuinely incremental over regular schooling. Later on, we compare incremental costs with effects to derive an incremental cost-effectiveness ratio (ICER).
Sometimes, the term cost is used to describe economic losses or burdens. For example, a policy discussion might draw attention to the “cost of school failure” or the “cost of juvenile delinquency.” This usage may be confusing. These are not costs in the sense of resources expended to provide an educational intervention. Instead, they are burdens or losses caused by not implementing effective educational interventions. The so-called cost of juvenile delinquency is what is spent on the juvenile justice system, not what is spent on educational programs to reduce delinquency. As such, this “cost” is better understood as the economic benefits from overcoming school failure or reducing juvenile delinquency. (Even then, it is sometimes an inaccurate measure of benefits because it includes both preventive and remedial expenditures; see the discussion of defensive expenditures in Chapter 9.)
3.3. Costs and the Theory of Change
The analyst must identify, and then calculate, all resources involved in “making the intervention work.” In practice, identifying all resources requires detective work and a thorough understanding of the intervention. To help acquire this understanding, it is helpful to think about the costs of an intervention in relation to its theory of change (for an example with a resilience program for youth, see Crowley, Jones, Greenberg, Feinberg, & Spoth, 2012).
The primary cost will be the actual delivery or implementation of the program within its educational setting. For example, a school might purchase a new math curriculum, so the costs will be all the resources required to teach children using this new curriculum. It should include all the preparatory resources needed to ensure that the program can be implemented, such as training of teachers in the new curriculum and changes to the school schedule.
This implementation cost must be related to the resources required for the alternative or counterfactual program. The alternative might entail no resources, but this must be established by the researcher. For example, a pull-out reading program for struggling third-grade students will require resources, but the students remaining in class are now in a smaller group and may receive more personalized instruction. Also, many struggling readers do already receive additional learning supports, and these may not be available if the student is pulled out of the classroom. The researcher must measure the resource use that is additive, beyond what would have otherwise been received.
Strictly, we think of costs as the costs to provide the program or implement the policy and to make sure it works. We do this so that a policymaker can understand what resources are actually needed for implementation. However, there are often other costs associated with educational interventions.
Induced costs are those arising from behavioral change after an intervention has been implemented. These induced costs should be examined in relation to the theory of change: They are the changes in resource use associated with the mediators of an intervention (Bowden, Shand, Belfield, Wang, & Levin, 2016; Chandra, Jena, & Skinner, 2011; Weiss, Bloom, & Brock, 2014). For instance, career academies are intended to help students get better jobs (Maxwell & Rubin, 2000). We can estimate the incremental cost of a career academy over an alternative youth training program; this would be the direct implementation cost. However, career academies often encourage students to stay in school longer, after which they get even better jobs. Whereas the delivery cost is the amount of resources needed to implement a career academy education, the induced cost is the cost of staying in school longer. If we are interested in the full cost of career academies, we should account for these induced costs. (These induced costs are sometimes referred to as indirect or mediated costs; for BC analysis, they are better thought of as negative benefits; see Chapter 9.)
In fact, when viewed in the context of their theory of change, many interventions explicitly anticipate these induced costs. Often, interventions begin with a diagnosis of educational need or a foundational treatment, which then leads to further services. These additional services are the mediator through which the program is effective.
For example, interventions to change remedial education lead to new college pathways; a new placement test might reassign students to fewer remedial classes, saving on resources as well as boosting college completion rates (see Scott-Clayton, Crosta, & Belfield, 2014). One prominent example is simplification of the FAFSA college aid application process (Bettinger, Long, Oreopoulos, & Sanbonmatsu, 2012); this simplification helped students apply for college and so induced them to attend and make progress in college. Student counseling and informational interventions work in the same way (Castleman, Page, & Schooley, 2014), and financial incentives in the first semester boost credit-taking in later semesters (Barrow, Richburg-Hayes, Rouse, & Brock, 2014).
Also, much of the research literature on school choice should be appraised in light of its induced costs: giving families a choice of school is likely to mean they will choose a school with a different—increased—level of school resources (Hsieh & Urquiola, 2006). These interventions affect students directly, but they also place students on a different educational or developmental pathway—and this new pathway has resource consequences. Finally, we can think of interventions that induce new resources outside the school.