Informing Federal School Finance Policy with Empirical Evidence

Introduction

This article explains the process behind estimating a National Education Cost Model (NECM) and generating from that model projections of per-pupil costs to achieve 2016 national average outcomes (reading and math grades 3 to 8) across all districts in the United States, from 2019–20 to 2020–21. This article is a follow-up to a preliminary report prepared in 2018 in which we made our first attempts to estimate a National Education Cost Model using methods most often applied to individual states (Baker, Weber, Srikanth, Atzbi, Kim, 2018). Here, we have expanded on this process, to develop a three-step method to take us from an estimated cost model using district level data on approximately 10,000 districts per year from 2009 to 2016, to a simulation of estimated costs for over 13,000 districts for 2019 through 2021. That process involves the following steps to be explained in detail in this article:

• Step 1: Estimating a National Education Cost Model to historical district level panel data with missing data;

• Step 2: Estimating a set of parsimonious weighting factors that approximate the cost model estimates with more easily updated, limited measures, covering all districts;

• Step 3: Developing a formula simulation to apply to current and future district level data, complete panel.

From this formula simulation, we compare the most recent years of district level actual spending reports (fiscal year 2017) to what would be needed for children in each district to have equal opportunity to achieve a given outcome goal. Here, that modest goal, is to raise the national floor to the national average of past years.

Statistical modeling of the type used herein yields estimates. These estimates are imperfect but useful, yest one must be careful not to overinterpret these estimates, or assume them to be exact or perfect targets for the amount of money that must be spent to precisely achieve a selected outcome. The goal of education cost modeling, whether for evaluating equal educational opportunity or for producing adequacy cost estimates, is to establish reasonable guideposts for developing more rational school finance systems. To summarize, the goals and advantages of the approach provided herein are:

• • Cost model estimates provide reasonable marks, where previously there were none.

• • Specifically, they provide estimates related to common outcome goals, which was not previously possible.

• • These marks can guide policy but may not necessarily dictate it.

• • Introducing this evidence into deliberations over a new federal aid formula can help to “bend” public policy, specifically federal aid distribution formulas, in a better direction than if such evidence did not exist or was simply ignored.

• • Ultimately, the goal of introducing rigorous empirical evidence on education costs (tied to outcomes) into formula deliberations is to achieve an end result (from the necessarily political process) that is “less bad than it might otherwise be.”

Statistical cost modeling is the most appropriate method for understanding education costs, cost variation across children and settings, toward achieving a commonly measured outcome goal. Back in 2004, economist Thomas Downes of Tufts explained (in a review of cost analysis methods), that “Given the econometric advances of the last decade, the cost-function approach is the most likely to give accurate estimates of the within-state variation in the spending needed to attain the state’s chosen standard, if the data are available and of a high quality” (p. 9). Significant advances in data quality, statistical computing and econometric techniques since 2004 have improved education cost modeling (Duncombe and Yinger, 2011). The primary objective of this exercise is to better understand the variation in costs toward common, measured outcome goals.

Downes focused on “within-state” variation, because, at the time researchers lacked the ability to compare outcomes of districts across states. With the release and updating of the Stanford Education Data Archive (SEDA), we now have eight years of nationally equated district level reading and math scores, grades 3 to 8. We also have a rich archive of district level fiscal, economic context and student enrollment data in the School Finance Indicators Database (SFID). Finally, two new sources of useful...