It is tempting to think that assessing the costs of an intervention is the easy part, committee member David Weimer observed, but calculating costs beyond the expenditures listed in a program budget can be difficult. Henry Levin and Clive Belfield, respectively, provided an overview of what is required and illustrated some of the issues.
RESOURCES AND COSTS OF REPLICATION
Levin indicated that studies of early childhood education rarely measure the associated costs thoroughly or accurately. First, many studies rely on budget figures, which are usually developed prior to actual expenditures, are not necessarily corrected after the fact, and rarely account for the true costs of the resources involved. Although there is fairly broad consensus among economists about how costs should be measured to obtain accurate results, that standard is seldom met in the early childhood context.
Ideally, Levin explained, there are five steps to measuring cost accurately:
Specify dimensions of quality.
Identify resource requirements to meet goals for each dimension.
Assess market and shadow costs for each resource.
Aggregate for total and obtain average and marginal cost.
Allocate cost burden among government support, private support, and client costs.
A comprehensive list of the aspects of the program that contribute to its quality might include the time children spend in the program (e.g., hours per day, days per week, weeks per year), the personnel ratios, the range of services supplied, facilities and materials, and so forth. Each program has its own characteristics; even when replicating a successful model, the goal is not generally to stamp out identical centers. Thus, the question of tradeoffs among certain quality features and costs arises from the beginning. Levin explained that there are various ways to document how priorities are established in the program design and replication process. This may include direct observation, in-depth interviews with the staff to ascertain what aspects of the inputs are critical, and review of program design requirements and other archival materials.
To identify the resources necessary to meet the target level of quality for selected program features, one would begin by identifying any known market prices (e.g., for staff salaries, one of the largest costs). However, current market prices may understate the long-term cost, if there is likely to be a large expansion of demand for the needed resource, or overstate it, so one must also calculate a shadow price. (Shadow prices are discussed in greater detail in Chapter 5.) Standard economic criteria should be used to calculate opportunity costs for participants, including those due to prospective market changes. Shadow prices also need to be calculated for any required resource for which there is not a competitive market equivalent. With all of this information, one can calculate total costs for a given enrollment goal, as well as the marginal costs that would be applicable if the program were to grow.
Sensitivity analysis—procedures to investigate the effects of various possible changes in the parameters—are important at this stage of the process, although Levin noted that this is rarely done in cost analysis. He advocated setting up confidence intervals for the cost estimates, as well as varying the quality dimensions to identify the tradeoffs and cost implications. The cost analysis will be most useful to decision makers if they can explore the cost feasibility with different budget or enrollment constraints. Levin closed with the general observation that a detailed, accurate picture of costs is just as important as a sophisticated picture of effects.
AN EXAMPLE: THE ABBOTT PRESCHOOL PROGRAM
Clive Belfield began by noting that he is often asked how much high-quality early childhood education costs. His response is that it is the wrong question—that one doesn’t evaluate an investment solely in terms
The Abbott Preschool Program
A 1998 ruling of the New Jersey Supreme Court required the state to provide full-day preschool for 3- and 4-year-old children living in the 31 Abbott school districts, which are high-poverty urban districts located throughout the state (the Abbott rulings mandated numerous other educational measures as well). The court set quality standards that include qualified teachers (a state-certified teacher and an assistant in every class) and small class sizes (15maximum); a developmentally appropriate curriculum aligned with the state’s K-12 content standards; and the provision of social and health services, transportation, and support for students with limited English proficiency or disabilities. Currently there are more than 600 centers in the program, serving 38,000 children in public school or private settings or through Head Start programs.
SOURCES: Education Law Center (2007); Belfield (2009).
of how much it costs, but in terms of what is the optimal investment. He illustrated this proposition with a cost analysis of the New Jersey Abbott Preschool Program, a large-scale program in which legally mandated state standards are being implemented in a variety of settings (see Box 3-1).
The New Jersey program offered several benefits in terms of available data. In general, the program’s administrative data are of high quality; line-item budget information is available for every center, as well as reports from the independent quality assurance inspections (the program uses the Early Childhood Environment Rating Scale, Revised Edition, ECERS-R;1 see FPG Child Development Institute, 2009b). Nevertheless, some data elements are missing. For example, the program is not a full school day and does not run during the summer, but no budget data were available for the wraparound care (for the portion of the day not occupied by the program) or for the summer program. These components are funded separately, in ways that are not uniform across centers. The data do not include capital grants or parent and other nonmarket resources, nor do they include transportation costs or costs for special education services for students who need them.
Another issue is what Belfield described as contaminated resources. Funding for preschool may come from a variety of sources, including
state allocations, child development block grant funds, welfare funds, and so forth. Many centers may be subsidized indirectly through the use of facilities or the sharing of administrative or management staff from the public education system. There are also an unknown number of younger children who are not eligible to participate for free but whose parents pay for them to participate. Furthermore, the costs for the public centers are available at the district level, whereas center-level costs are available for the private ones—but all the quality measures are at the classroom level.
Another set of problems arises with the external validity of the data, or means of interpreting it in the context of other programs given the lack of information on the costs of designing the program, state administration costs, or the costs to deploy and evaluate the program. Belfield also did not have information about the secondary labor costs to parents participating in the program (resulting from time away from work or changes in employment decisions, for example). Teacher pay also presents challenges—he did not have data on their benefits, pensions, or training costs. The program is growing rapidly, so short-run operating costs may not be the same as long-term marginal costs. Finally, the cost-of-education index (Taylor and Fowler, 2006), which many researchers use to compare wages and other education resources across geographic areas, was developed for K-12 education. Many factors that are different in a preschool context (e.g., large percentage of part-time workers, different credentialing requirements) limited its usefulness for Belfield’s analysis of the Abbott Preschool Program.
There are several different kinds of questions one might want to answer using cost analysis, Belfield pointed out. One is to calculate the net present value, or the value of the benefits minus the value of the costs. Another is to establish links between quality and resources invested—to determine whether spending more money on specific program features will yield higher quality. Are economies of scale likely if the program is expanded? Can economies of scope be achieved, if more services are offered? Do costs vary depending on whether the programs are run through the public school system or through private providers? How do costs vary as a result of differences in local labor markets?
Four methods of cost estimating are used in litigation related to K-12 education, in which states have been charged to calculate the cost of an adequate or equitable education: (1) developing a cost function model, (2) applying data from other successful programs, (3) using an evidence-based template, and (4) asking a panel of experts to make judgments. Belfield applied three of these in his cost analysis of the Abbott Preschool Program (he deemed the professional judgment panel insufficiently reliable). He developed separate estimates for public and private centers because many of the assumptions were different for these two settings
(particularly teacher pay scales). The program in both settings operates six hours per day, nine months of the year, and Belfield noted that across the board, education costs about 20 percent more in New Jersey than in other states. His results are shown in Table 3-1.
Belfield concluded the following from his analysis:
Higher quality does cost more—he found that raising the quality by one unit on the ECERS scale increased costs by about 2 percent.
Costs were higher in private centers for several reasons. Private centers have higher facility costs, and they do not generally benefit from the cross-subsidies available to programs in the public school system. They also pay more for teachers, primarily the assistant teachers who are at the lower end of the salary scale.
A weak link exists between average costs and scale. Higher enrollments and multiple centers did not have much effect on costs, and costs did not vary much by district size.
He cautioned that other studies have produced somewhat different results, and in response to a question he suggested that his estimates might be capturing as little as two-thirds of the theoretical total opportunity cost of the program. Costs that are not reflected include parental resources (primarily time they spend supporting their children’s experiences), costs to the state for administering and evaluating the program, transportation costs, the tax burden of raising the money to fund the program, and capital costs.
Belfield suggested several areas in need of research. More detail about possible economies of scale or scope would be useful, since expanding successful programs is a policy priority. Teacher salaries are the largest cost for preschool programs, so more understanding of the labor market
TABLE 3-1 Cost per Child of the Abbott Preschool Program for 2008-2009
would be useful. The burden of funding is another area that has not been thoroughly explored. For example, Belfield pointed out that a dollar of funding from one source may not have the same policy implications as a dollar of funding from another. And, he noted, ways of making cost estimating easier, such as more accurate “plug-in” numbers (estimates for particular costs that can be used to streamline cost analysis) or standardized discount rates, would encourage decision makers to take a more detailed look at costs.