In a previous post, I provided estimates of the savings in special education costs from a high-quality universal pre-k program. I stated that these cost savings started out at about 4% of the pre-k program’s costs in kindergarten. That is, in the program’s second year, the cost savings from the first-year cohort would cover about 4% of the annual costs of serving the second-year cohort. These cost savings result from estimated cost savings due to reduced special education assignments of universal pre-k attendees in kindergarten.

These cost savings then grow over time. In the program’s 3^rd year, there are savings from reduced special education assignments for both the first-year cohort, which will be in first grade, and the second-year cohort, which will be in kindergarten.  These cost savings in the 3^rd year will cover 8% (=4% from first-year cohort in first grade, plus 4% from second-year cohort in kindergarten) of the pre-k program’s costs in the third year.

Cost savings will continue to grow until all grades from kindergarten through 12^th grade have students who have participated in the universal pre-k program. The cost savings will then stabilize at 48% of the annual program costs of the universal pre-k program. The cost savings are a little less than 4% times 13 years because there is some out-migration out of state over this time period.

My estimates of special education cost savings are quite conservative estimates for a universal pre-k program.  Could the immediate special education cost savings be higher for some early childhood programs? Yes, they could be higher.  What could lead to higher special education cost savings?

(1)    Special education cost savings will be higher if the early childhood program’s percentage effects on special education usage are higher. For example, these special education percentage effects might be higher for targeted pre-k programs compared to universal pre-k programs.  The baseline level of special education assignments may be higher for the more disadvantaged children that would be in a targeted pre-k program. Furthermore, pre-k programs may have somewhat greater effects in general for disadvantaged students compared to more advantaged students.

(2)    Special education cost savings will be higher if special education costs per special education student are higher. Some states and school districts may have higher special education costs per student. Also, the trend in special education costs per student is upwards, so over time we would expect special education cost savings to become greater.

(3)    Special education cost savings will be higher if the baseline rate of special education assignment is higher. Some states and school districts may have higher rates of special education assignments. Furthermore, these special education assignment rates may be growing over time.

(4)    Special education cost savings will be higher if the early childhood program affects weekly hours per special education student as well as assignment rates. Most previous analyses of special education cost savings have only looked at how the early childhood program affects special education assignment rates. But early childhood programs may also affect the average number of hours that students are in special education, which will also affect costs.

(5)    Special education cost savings may be higher in present value terms and political significance if the cost savings are skewed towards the immediate years after participation in the early childhood program, for example if these cost savings are greater in kindergarten and first grade than in 11^th and 12^th grade. Many previous analyses of special education cost savings have assumed that the cost savings are uniform across all 13 grade levels from kindergarten to 12^th grade.

For example, suppose I redo my special education cost savings analysis for a targeted program. Then I conclude that the next-year cost savings for kindergarten would be 11%. (Note on assumptions: I adjust my numbers by assuming only the most disadvantaged group participates, but I continue to assume that some participants in the targeted pre-k program are displaced from other pre-k programs.) Such a program would pay for itself by the time the first-year cohort is in 10^th grade.

In addition, I assumed annual special education costs per special education student were around $10,000, with about 90% of these costs paid locally. A recent study in Montgomery County Maryland estimated special education costs in that school district per special education student, over and above general education costs, of a little over $16,000 per special education student.  Suppose we modify my simulation of special education costs to reflect an assumption of special education costs of this magnitude.  Then the special education cost savings in the year after the targeted program is started, when the first-year cohort is in kindergarten, would be 17% of the program’s annual costs. The program would pay for itself in special education cost savings by the time the first-year cohort is in 6^th grade.

The Montgomery County report also estimates that some pre-k programs may not only reduce special education assignments, but also reduce hours in special education for those students in special education.  The report estimates, in its Table 12, that full-day Head Start reduces the percentage of students in special education in kindergarten from 24% to 11%. This is similar to the magnitude of reduction of special education assignments that I assume from a targeted program. But in addition, full-day Head Start reduces the weekly hours in special education of special education students from 9.8 hours to 3. 7 hours.

I can’t use these weekly hours reductions to generate cost savings, because I don’t know how the costs of special education vary with weekly hours. Presumably there are both fixed and variable costs of serving special education students, and I don’t know their relative magnitude. Also, these weekly hours reductions may be related to the types of special education assignments that students have, and I do not know how pre-k programs affect the types of special education assignments and what their relative costs are.

However, the bottom line is that these weekly hours reductions may yield even further cost savings, beyond what is estimated in most studies.

I should note that all these estimates of special education cost savings assume that as early childhood programs expand, the criteria used by the K-12 system to assign students to special education remain the same. If the K-12 system responds to the reduced “need” for special education services of entering kindergartners by expanding special education participation of less needy students, then these cost savings will not be realized.

What does this all mean for policymakers and researchers? I think this points to the need for much more current research on immediate cost savings from reduced special education usage.

This research could be done most rigorously by a regression discontinuity analysis. Such a regression discontinuity analysis would require looking at both entering pre-k students, and entering kindergartners who have participated in a year of pre-k.  Rather than a full-blown evaluation of special education eligibility for all these students, which might raise some ethical issues, I would suggest that the students all be subject to similar tests of literacy and math skills, and social and emotional behavior, along with any other short tests that might be thought to predict special education participation. These test scores would then be used to predict special education participation for the entering kindergartners. We could then predict what the special education participation would have been for the entering pre-k students if they had instead been entering kindergarten.  The analysis would control for student age, and would be expected to show a “discontinuous” jump down in predicted special education participation for students just old enough to have participated for a year in pre-k, versus similar students just below the age cutoff.

These analyses should be careful to estimate the true incremental costs of special education, and should be careful to include any variation in special education hours as well as special education participation. In addition, these analyses should be validated by data showing that as the pre-k program has expanded, that we observe trends in aggregate special education participation and usage in kindergarten that are at least roughly consistent with the estimated effects of the pre-k program expansion.