Would weighted-student funding enhance intra-district equity in Texas? A simulation using DEA

Abstract

We use data envelopment analysis to model the educational production function, and then explore how a shift to weighted student funding using the student weights embedded in the Texas School Finance Formula would alter the allocation of inputs and potential outputs. School outputs are measured as value-added reading and math scores on standard achievement tests. We find that if school districts allocated their resources efficiently, then they would not allocate their resources to campuses according to the funding model weights. Policies that promote greater efficiency would also enhance equity in educational outcomes.

Keywords:

• weighted student funding
• outcomes equality
• distance functions

Notes

1 For a more complete description of school funding formulas see Verstegen and Jordan (2009), which is the source of the information about formula weights for student need. See Baker and Duncombe (2004) for a discussion of scale adjustments. See Taylor and Fowler (2006) for a discussion of formula adjustments for higher labour cost.

2 Because they have access to a different educational technology, open enrolment charter schools have been excluded.

3 Students for whom the prior test score was missing are treated as one of the groups. This is equivalent to assuming that all students with missing pre-test data had the state average pre-test score.

4 Details of this model are described in an Appendix that can be found at http://cstl-hcb.semo.edu/bweber.

5 Ruggiero et al (2002) provide an alternative method of incorporating environmental variables into DEA.

6 According to the Resource Allocation Handbook for the fiscal year 2014–2015, published by the Houston Independent School District, the weights used in distributing resources are broken down as follows: Special Education: 0.15, State Compensatory Education (50% free/reduced lunch and 50% at-risk): 0.15, Gifted and Talented: 0.12, Vocational Education (CATE): 0.35, Bilingual/ELL (English Language Learner): 0.10, Homeless: 0.05, Refugee: 0.05.

7 Pagan and Ullah (1999) discuss bootstrapping of kernel distributions.

8 For the random variable x the Gini coefficient is calculated as G=1−((2)/(N−1))(N−(∑_i=1^Ni × x_i)/(∑_i=1^Nx_i)) where i is the rank of x_i. The Theil index is calculated as