Table 3-2. The cost of living adjustments change our view of the magnitude of differences in revenue between states, but within-state Theils do not change appreciably. The cost-adjusted, between-state Theils are 20 to 40 percent lower than the unadjusted Theils. Once we account for differences in costs, we find that differences in revenue between states accounts for 53 to 60 percent of the total disparity in per-pupil resources in the United States; when we do not adjust for cost differences, between-state inequality accounts for 66 percent of total inequality.13
We now turn to the impact of reform on low-household-income districts. One of the key goals of the long legal battle over education finance has been to sever the link between the ability to pay for education (as measured in terms of wealth or income) and actual spending. For the most part the literature has focused on the impact of court decisions on low- and high- spending districts. This would be the right direction if spending and income were perfectly correlated. Here we offer some direct evidence on the effect of the courts on districts where the ability to pay for education is limited.
To address these issues we matched our district level resources with district level social and economic data from the 1970, 1980, and 1990 decennial census. The U.S. Bureau of the Census provides a mapping from school district boundaries to census block groups and tracts. We use the district summaries from the 1970 census (Special Fifth Count Summary Tapes) from the Bureau of Census (1973), the 1980 census (Summary Tape File 3F, School Districts Machine Readable Data file) from the Bureau of the Census (1982b), and the 1990 census (School District Data Book CD-ROM) from the National Center for Education Statistics (1994a). For each district in our sample, we calculate a number of important variables, including real household median income and the proportion of the district population that is black. We then match the 1970 census data with 1972 expenditure data, the 1980 census data with 1982 expenditure data, and the 1990 census data with 1992 expenditure data.
To examine the effect of the courts on the link between income and spending, we classify districts by the within-state, pupil-weighted quartile in median family income. Next, for each state/year/quartile group, we calculate the average student-weighted, per-pupil revenues from local and state sources. We then use these values as dependent variables in a series of econometric models. Let Ysqit be the average per-pupil revenues from source s in income quartile q for state i in time t. The basic structure of these models is similar to the ones we employed in previous analysis, where
The variable Dit is one of our two court-mandated reform variables, µqi and ηqt are the quartile-specific fixed state and year effects, respectively, and εsqit is a random error. The fixed-effect model is particularly appropriate in this context