In 1993, the courts threw out Missouri's school funding system. The judge in the case declared that because of huge gaps in spending, public schools ranged from "golden" to "god-awful." So the state created a new funding formula that was supposed to be more equitable. Still, over the years, many educators and lawmakers complained that disparities still existed. The education reporters at the St. Louis Post-Dispatch decided to tackle the school funding issue last fall. We knew it was going to come up for debate in the next legislative session. Meanwhile, across the Mississippi River, Illinois lawmakers also grappled with the issue. So we wanted to do two separate analyses, one for each state.
Here was the perfect opportunity for me to test my skills as the education team's CAR specialist. I had used Microsoft Excel and Access extensively and recently learned SPSS. Still, I would suffer plenty of headaches and emit several primal screams before our three-part series ran, after about three months of work.
We wanted to know if school districts spent about the same amount on each child. Sure, one could just get official spending figures for each district and compare. But those numbers don't tell the whole story. Just as it costs more to live in Chicago than in Peoria, some districts have to pay more for teachers and resources based on their location. Plus, children with disabilities or other special needs cost more to educate. So, we wanted to compare districts by taking those differences into account.
I called various school finance experts for help and struck gold with the National Center for Education Statistics. A researcher in the NCES school finance division pointed me to a bunch of studies and sources. I learned about something called the geographic cost of education index, which lets you adjust for cost differences by school district. The index is based mostly upon salaries for teachers with comparable experience, plus other costs.
Through NCES I also found a way to adjust for differences in student costs. Generally, children with disabilities cost 2.3 times as much to educate, while poor children and children with limited English cost about 1.3 times as much. So if a district has 1,000 students and 200 of them are in special education, it's the equivalent of educating 1,260 "regular" students. That is, 800 + (200x2.3) = 1,260.
Getting the data
To do this project, I pulled data from several sources.
I downloaded the cost-of-education index from the NCES Web site.
The Illinois education department e-mailed me spreadsheets with finance information, student enrollment, and the number of special education, limited-English and poor students in each district. I defined "poor" students as those who get free or low-cost meals through the federal school lunch program.
Luckily, officials at the Missouri and Illinois education departments not only provided data willingly, but also helped me make sense of the numbers.
Missouri's education department e-mailed me a monster (16.8 megabytes) text file with detailed financial information. For reasons discussed later, I needed Missouri school district spending broken down in ultra-specific categories.
I pulled student information for Missouri districts from an Access database I had already created.
It seems that on every CAR project, I inevitably run into more problems than I ever expected. This one proved no different. For starters, we had to determine how to count students. Illinois gives districts money based on student attendance. Missouri uses "eligible pupils," which is like attendance, but with some differences. We picked student enrollment as our measure, because even if kids don't come to school every day, the district still has to pay for their teachers, textbooks, utilities, etc. Plus, poor districts tend to have lower attendance, thereby inflating the amount spent per pupil. We also had to define expenditures. We used operating expenditures, which includes costs to run the district, but not capital costs. To calculate Missouri district spending based on enrollment instead of "eligible pupils" (love that education jargon), we had to add some spending categories and subtract others. That's why we needed detailed spending files. I spent two weeks just importing and cleaning those files. Luckily, officials at the Missouri and Illinois education departments not only provided data willingly, but also helped me make sense of the numbers and financial terms.
Number Crunching
I imported the various spreadsheets and tables into Microsoft Access, then joined them so I had a single table for each state. From there, I divided each district's operating expenditures by its cost-of-education index to get adjusted expenditures. Next, I adjusted student enrollment using the weights for special education, limited-English and poor students. Then I divided the adjusted expenditures by adjusted enrollment to get adjusted spending per pupil.
Actual expenditures = Adjusted expenditures Cost-of-education index
Regular students + 2.3*special ed students) + (1.3* students) + (1.3*limited-English students) = Adjusted enrollment
Adjusted expenditures = Adjusted spending per pupil Adjusted enrollment
So you might have two districts that spend $5,000 per student in actual dollars. But factor in location and student need, and one district could have more "buying power" than the other - say, $5,200 per student vs. $4,600 per student.
Now that we had adjusted spending for every district, I turned to SPSS for further analysis. I calculated the minimum, maximum and average spending per student for each state, in both actual and adjusted dollars, weighting for district size. One school finance expert suggested we also look at the coefficient of variation, which is the standard deviation divided by the average spending per pupil. This is the range that about two-thirds of students fall within.
In both Missouri and Illinois, the gap between the highest- and lowest-spending districts shrank using adjusted dollars, but not entirely. This table summarizes the results for Missouri:
Spending per pupil for Missouri school districts, 1998-99
Actual dollars / Adjusted dollars Minimum: $3,500 / $3,250 Maximum: $10,500 / $9,500 Average: $5,400 / $4,400 Coeff. of variation: 20% / 15%
In actual dollars, about two-thirds of Missouri students are in districts that spend within 20 percent of the average - that is, from $4,320 to $6,480 per child. In adjusted dollars, those students are within 15 percent of the average - from $3,740 to $5,060.
Answers
Ok, so we know the gaps still exist. Why? For answers, I ran some regressions in SPSS of school spending against district poverty rates, tax rates, property tax bases, and other variables.
We found that local tax rates explained about a third of the variation in Missouri district spending. The property tax base explained about a third of the variation in Illinois. In our story, we explained that Missouri rewards communities willing to tax themselves, while in Illinois, what a district spends depends more upon the property wealth of the community.
We also found that in Missouri, poor children weren't necessarily trapped in the lowest-spending districts. But in Illinois, the lowest-spending districts had more poor students, even after adjusting for student need. Small districts in both states tended to spend more because transportation, administration and other costs eat more of the budget.
I also tried to see if differences in spending had any effect on test scores, but I didn't find a strong relationship. Perhaps the relationship does exist, but one would need to break down spending into more specific categories - say, dollars spent on classroom instruction. Plus, because test scores can vary from school to school, next time I might try analyzing on the school level, not the district level.
Another issue we ran into was how to report our findings. Because I defined pupils differently, the actual spending did not match district figures. Would readers get the distinction? In the end, we decided to group districts as high, medium and low spenders based on actual spending, then on adjusted spending.
Advice
Here's some advice if you attempt a similar project:
Follow what I call the Rule of 3.7 - Estimate the time you think the project will take, then multiply by 3.7. Or more.
Be shameless in consulting with experts. Get their advice on how to design your analysis. What you're doing has been done before, so you can learn about pitfalls and caveats up front.
Get those experts to review your findings.
Work with graphics editors early and often. This is complicated stuff, and you should present your findings in a way that makes sense to readers but doesn't overwhelm them. CAR directors probably know this already, but as a reporter doing CAR, I made this crucial discovery too late.
Once you've analyzed the numbers, get into the classroom. So one district spends $5,000 per child and another spends $10,000. How does that play out in the classroom? Two education reporters spent several days in schools getting vivid examples to illustrate the spending gaps.
Holly Hacker can be reached by e-mail at hhacker@post-dispatch.com